Cost-Benefit Analysis of Badger and Cattle Management
SE3117
Report of Phase II: June 2005 to May 2007
Authors:GC Smith, D Wilkinson (CSL)S Rushton, M Shirley (Newcastle University)RM Bennett, ID McFarlane (Reading University)
1
Table of Contents1. Summary..............................................................................................................6
1.1 Executive Summary......................................................................................................61.2 Summary.......................................................................................................................6
2. Introduction........................................................................................................83. Model.................................................................................................................11
3.1 Economics...................................................................................................................113.1.1 Introduction..........................................................................................................113.1.2 Costs of TB if no action taken.............................................................................123.1.3 Pre-movement testing..........................................................................................123.1.4 Costs of implementing culling or vaccination strategies.....................................133.1.5 Cattle costs...........................................................................................................143.1.6 VLA costs............................................................................................................143.1.7 Savings resulting from culling or vaccination strategies.....................................15
3.2 Model Descriptions.....................................................................................................153.2.1 Purpose................................................................................................................153.2.2 State variables and scales.....................................................................................153.2.3 Process overview and scheduling........................................................................173.2.4 Design concepts...................................................................................................173.2.5 Initialisation.........................................................................................................193.2.6. Input....................................................................................................................193.2.7. Submodels...........................................................................................................203.2.8 Perturbation Processes.........................................................................................26
4. Verification........................................................................................................284.1 Badger-to-Cattle Transmission Rates.........................................................................284.2 Cattle Mortality Rates.................................................................................................284.3 Cattle Herd Breakdown Rate......................................................................................284.4 Sensitivity Analyses....................................................................................................284.5 Cross-Model Validation..............................................................................................294.6 Verification Overview................................................................................................32
5. Validation..........................................................................................................325.1 Badger Validation.......................................................................................................325.2 Cattle Validation.........................................................................................................32
5.2.1 Tests and reactors................................................................................................325.2.2 Reactors per Cattle Herd Breakdown..................................................................33
5.3 Economic Validation..................................................................................................345.3.1 Costs of TB breakdowns......................................................................................345.3.2 Annual costs of TB breakdowns..........................................................................35
5.4 RBCT Validation........................................................................................................356. Methods............................................................................................................37
6.1 Control and Management Options Tested..................................................................376.1.1 Default settings and assumptions.........................................................................376.1.2 Main Scenarios....................................................................................................38
7. Results..............................................................................................................407.1 Sensitivity Analysis....................................................................................................40
7.1.1 Cattle skin test sensitivity....................................................................................407.1.2 Cattle test-interval................................................................................................407.1.3 Disease transmission rates...................................................................................437.1.4 Pre-movement testing – sensitivity to scale.........................................................447.1.5 Perturbation..........................................................................................................45
7.2 Population Management.............................................................................................487.2.1 Badger population................................................................................................48
2
7.2.2 Prevalence of TB in badgers................................................................................497.2.3 Number of TB-infected badgers..........................................................................517.2.4 Parish types..........................................................................................................527.2.5 Cattle herd prevalence.........................................................................................527.2.6 Cattle herd breakdown rate..................................................................................537.2.7 Varying control area............................................................................................547.2.8 Varying farm compliance....................................................................................597.2.9 Varying control rates...........................................................................................627.2.10 No-Immigration.................................................................................................667.2.11 Gamma-Interferon.............................................................................................67
7.3 Economic Modelling..................................................................................................687.3.1 Pre-movement testing – economic analysis.........................................................717.3.2 Economic benefit distribution – scenario 1.........................................................727.3.3 Economic comparison of different scenarios......................................................747.3.4 No Immigration – economic analysis..................................................................787.3.5 Gamma-Interferon testing – economic analysis..................................................78
7.4 Sensitivity Analysis of Economic Output...................................................................797.5 Perturbation – Alternative Options.............................................................................80
8. Discussion........................................................................................................828.1 Verification and Validation........................................................................................838.2 Cattle Testing..............................................................................................................848.3 Badger Culling............................................................................................................858.4 Badger Vaccination....................................................................................................868.5 Economics...................................................................................................................86
9. Conclusions......................................................................................................8710. References......................................................................................................88Appendix A – Economic Parameters.................................................................91Appendix B – Farm Costs...................................................................................92Appendix C – Cattle Mortality.............................................................................94Appendix D – Cattle Herd Breakdown Rates (historic).....................................96Appendix E – Parameter Values (CSL model)...................................................97
E.1 Default Settings and Parameter Values used in the CSL model................................97E.1.1 Temporal Settings:..............................................................................................97E.1.2 Spatial Settings:...................................................................................................97E.1.3 Badger Settings:..................................................................................................98E.1.4 Cattle Settings:....................................................................................................99
Appendix F – Parameter Bounds (CSL model Sensitivity Analysis).............104Appendix G – Sensitivity Analysis Output.......................................................109Appendix H – Cattle Testing Flowchart............................................................117Appendix I – Test Interval Algorithm................................................................118Appendix J – Badger Movement Flowchart.....................................................119Appendix K – Model Assumptions...................................................................120
K.1 Model Assumptions.................................................................................................120K.1.1 Temporal Assumptions.....................................................................................120K.1.2 Spatial Assumptions.........................................................................................120K.1.3 Badger Assumptions.........................................................................................120K.1.4 Cattle Assumptions...........................................................................................120K.1.5 Economic Assumptions....................................................................................121K.1.6 Badger Control Assumptions............................................................................121
Appendix L – Glossary.........................................................................................122
3
List of Figures and Tables
Figure 3.1.1. A diagrammatic representation of the cost-benefit analysis (CBA) incorporated into the CSL bovine tuberculosis (TB) model....................................11
Figure 4.5.1 Cross-validation of the models......................................................29Figure 5.2.1.1. Number of reactors plotted against number of annual cattle TB tests.......................................................................................................................32Figure 5.2.2.1. Frequency distributions of number of reactors per CHB. A comparison of real and simulated data..............................................................33Figure 5.3.1.1. Scatter distribution of cost of each individual CHB in 2005, plotted against the number of confirmed new incidents (CNIs)......................34Figure 5.3.2.1. Distribution of the simulated annual cost (£) of all CHBs, multiplied up from the modelled grid area to the whole T1 and T2 areas.......34Figure 5.4.1. RBCT validation of the NU model.................................................35Table 5.4.1 A comparison of the RBCT CHB rate with simulated data............36Table 6.1.2. Details of the 7 badger control scenarios.....................................38Figure 7.1.1.1 Cattle skin test sensitivity...........................................................39Figure 7.1.2.1 Effect of changing parish test-interval dependent on CHB rate................................................................................................................................40Figure 7.1.2.2 Effect of changing the routine cattle-test-interval.....................41Figure 7.1.3.1 Badger contribution to the CHB rate..........................................42Figure 7.1.4.1 Effects of extending the area over which PrMT is applied.......43Table 7.1.5.1. Change in simulated CHB rate following control.......................44Table 7.1.5.2. Change in simulated CHB rate following control.......................45Figure 7.1.5.1 Illustration of the spatial effect of culling in the NU model......46Figure 7.2.1.1. The percentage of badger groups subjected to culling for different rates of farm compliance.....................................................................47Figure 7.2.1.2 Scenario 1: badger social group size.........................................48Figure 7.2.2.1 Scenario 1: badger TB prevalence.............................................49Figure 7.2.2.2 Example output from the NU model...........................................50Figure 7.2.3.1 Scenario 1: Number of infected badgers...................................50Figure 7.2.4.1 Scenario 1: Parish test frequency..............................................51Figure 7.2.5.1 Scenario 1: Cattle herd prevalence............................................52Figure 7.2.6.1 Scenario 1: CHB rate...................................................................53Table 7.2.7.1 Varying control area: Badger social group size.........................54Table 7.2.7.2 Varying control area: Number of infected badgers....................54Table 7.2.7.3 Varying control area: Badger prevalence....................................55Figure 7.2.7.1 Varying control area: Badger prevalence..................................55Table 7.2.7.4 Varying control area: Mean CHB rate..........................................57
4
Figure 7.2.7.2 Varying control area: Mean CHB rate.........................................57Table 7.2.8.1 Varying compliance: Badger social group size..........................58Table 7.2.8.2 Varying compliance: Number of infected badgers.....................59Table 7.2.8.3 Varying compliance: Badger prevalence.....................................59Figure 7.2.8.1 Varying compliance: Badger prevalence...................................59Table 7.2.8.4 Varying compliance: CHB rate.....................................................60Table 7.2.9.1 Varying control rates: Badger social group size........................61Table 7.2.9.2 Varying control rates: Number of infected badgers...................61Figure 7.2.9.1 Varying control rates: Number of infected badgers..................62Table 7.2.9.3 Varying control rates: Badger prevalence...................................63Table 7.2.9.4 Varying control rates: CHB rate...................................................63Figure 7.2.9.2 Varying control rates: CHB rate..................................................64Figure 7.2.10.1 No immigration: Badger social group size..............................65Figure 7.2.10.2 No immigration: CHB rate.........................................................66Figure 7.2.11.1 Gamma-interferon: CHB rate.....................................................66Figure 7.3.1. Cash flow by method of badger control.......................................68Table 7.3.1. NPV for badger control...................................................................68Figure 7.3.2. Cash flow to Defra for badger control..........................................69Figure 7.3.3. Cash flow to industry for badger control.....................................69Figure 7.3.2.1 Net Benefit: Scenario 1................................................................71Table 7.3.3.1 Percentage chance of economic loss..........................................73Figure 7.3.3.1 Discounted Net Benefit: shooting and vaccination..................74Figure 7.3.3.2 Discounted Net Benefits: Defra and Industry............................75Figure 7.3.4.1 Discounted Net Benefit: No immigration...................................77Figure 7.3.5.1 Discounted Net Benefit: gamma-interferon...............................78Table 7.4.1. Economic sensitivity to culling and vaccination..........................79Figure 7.5.1. Badger prevalence.........................................................................80Table 7.5.1 Discounted net benefits of control with reduced perturbation.. . .80Table 7.5.2 Percentage economic loss..................................................................81
5
1. Summary
1.1 Executive Summary1. The aim of this project was to construct a simulation model of badgers, cattle and
bovine tuberculosis epidemiology, and to evaluate the costs and benefits of selected badger and cattle management scenarios, as an aid to policy development.
2. This involved constructing a generic spatial model and running two comparative simulations (one with the management scenario(s) and one without). The economic costs of both simulations are then compared and the Net Present Value recorded. This is repeated 100 times for each scenario, the results are then ordered and presented in histogram form.
3. In addition, a GIS version of the model was produced. Both models produced similar output and both were validated against the results of the RBCT. The GIS version was also created to allow simulations in specific landscapes (e.g. using local farm structure and badger density) for future analysis of management scenarios in specific geographical areas.
4. Badger culling is known to disrupt the behaviour of surviving badgers and this has been the hypothesised reason for an increase in cattle herd breakdowns in the area surrounding the badger culling. The changes to badger movement and disease transmission, as a result of culling, were simulated in both models using the best data available. This led to a predicted increase in herd breakdowns following badger culling, and this must be regarded as additional evidence for the effect of social perturbation on increased incidence of bovine tuberculosis in cattle.
5. The models predicted that badger culling does not result in an overall decrease in the number of herd breakdowns, due to social perturbation. Economic analysis of the outputs therefore suggested that badger culling is not economically viable unless badger immigration can be prevented.
6. The model predictions suggest that the observed differences between the RBCT and the Irish Four Areas Study are a consequence of badger immigration.
7. These analyses were undertaken with the best available data. Further analysis may improve the accuracy of the model predictions, particularly with regard to temporal changes in badger social perturbation, but are unlikely to fundamentally change them.
8. Simulations of cattle management suggest that Pre-Movement Testing may be economically viable, that routine skin testing could be further optimised in terms of area and frequency, and the any reduction in the cattle skin test sensitivity (as a result of poor implementation) could dramatically increase herd breakdown rates.
9. We therefore recommend that further fieldwork be conducted to improve the parameterisation of badger social perturbation. We also recommend that further simulation be performed on cattle testing and badger vaccination.
1.2 SummaryThis is a substantial and comprehensive technical project report, detailing the development, construction, verification and validation of a large complex, stochastic simulation model, with the potential to inform Defra policy on the likely effects of different management strategies on the incidence of bovine TB in cattle. The research project was overseen by a steering group, who commented and advised as the work progressed. Much of the detailed report deals with comments raised during these discussions, and a full description of the model and the parameterisation is given to allow a thorough peer review of the process.
6
Whilst the full report would be very useful to a reviewer or another modeller, it would make more sense for others to read particular sections of interest.
The overall aim of this project was to construct a simulation model of badgers, cattle and bovine tuberculosis epidemiology, and to evaluate the costs and benefits associated with different badger and cattle management scenarios to produce a research model capable of answering policy questions related to badger and cattle management. This involved extending a previously constructed and published spatial model of badger TB epidemiology, constructing a cattle layer and adding economic components. The model was run using two comparative simulations (one with the management scenario(s) and one without). The economic costs of both simulated scenarios were then compared and the discounted Net Present Value recorded. This was repeated 100 times for each scenario and the results were ordered and presented in histogram form. The economic cost of each compartment of the model was evaluated on an individual animal basis. For example, the cost of a herd breakdown was calculated by assigning a cost to each cow slaughtered (using research results from a previous study of the costs associated with TB breakdowns on cattle farms), where the number of cattle slaughtered depends upon the detection of infection in the simulated herd.
A GIS version of the model was produced by Newcastle University to allow simulations in specific landscapes (e.g. using local farm structure and badger density) for future analysis of management scenarios in specific geographical areas. Both models produced similar output and both were validated against field results, including the results of the RBCT. Model verification and validation is a vital component of producing a useful model. Since the two models produced similar output and a full and independent sensitivity analysis gave very similar results we have a strong evidence base that the models are ‘correct’. Both models were then subjected to validation against field data (e.g. badger social group size, overall costs, number of cattle per herd breakdown, spatial herd breakdown rates seen in the RBCT, etc). This validation procedure allowed us to adjust the models to closely simulate reality by adjusting parameters for which there were limited data (e.g. the spatial extent of social perturbation). The models performed very well against the validation data sets, suggesting that they provide an adequate although simplified representation of the underlying biological and spatial processes in the badger population and its interaction with livestock.
The models constructed here are based on a vast amount of data and expertise, and the similarity of the output from the two models is extremely encouraging. The authors believe that the constructed models are the most accurate available to date (although there is always room for improvement). The output from these models will be presented at scientific meetings and published in peer-reviewed journals to ensure the models and results are subjected to scrutiny.
A variety of badger and cattle management scenarios were simulated, as suggested by Defra and the overall economic output is summarised below. For further detail on how these scenarios affect disease epidemiology in the badger, it is necessary to read Section 7. Pre-movement testing (PrMT) of cattle was more likely than not to give an economic benefit, and all other forms of TB management assumed that PrMT was already in place. All badger control methods are costly to implement. As a result of social perturbation (particularly that caused by badgers immigrating into the culled area) no method of badger culling was reliably economically beneficial. Badger vaccination, as simulated here, had the highest probability of achieving a net economic benefit. The benefit or loss with vaccination was generally much lower than with any badger culling strategy. The most
7
important factor leading to the economic loss of badger culling was the immigration of individual badgers (infected or otherwise) from outside the culled area. In one example (for gassing) the presence of perturbation led to only 2% of simulations giving an economic benefit, while preventing immigration completely led to 95% of simulations achieving a benefit. Thus, reducing immigration and ensuring compliance is above some (as yet unknown) minimum threshold are both necessary to ensure that badger culling is potentially viable.
Limited investigations of cattle management suggest there is some potential for further improvement in TB management. Current cattle testing may be improved by optimising the test frequency against herd breakdown rate, optimising the geographical area over which test frequency is decided and limited use of the skin test combined with the gamma interferon test. The sensitivity of the current skin test should definitely not be reduced in field use as this can lead to a substantial increase in cattle TB. Approaches to ensure that the test is applied rigorously and to the highest standards are advised.
This model has been verified and validated to available data at the time of construction. It is capable of simulating most badger management strategies, many cattle management strategies and combined management strategies. However, without further data and programming, it is not yet able to simulate specific farm types (e.g. finishing farms) or simulate changes to on-farm biosecurity. For a full list of assumptions used in the model, see Appendix K Whilst the model is validated to the RBCT research results at the end of the trial, further analysis is required to take account of changes in herd breakdown rate following the cessation of culling. It is noted that, from analysis of the RBCT results after the completion of this model (Jenkins, Woodroffe & Donnelly, 2008), the assumptions used in the model on the temporal aspects of social perturbation no longer produce herd breakdown rates consistent with the available data, and the model will be further updated to take account of this.
2. IntroductionThis report summarises more than two years of work in model construction, model verification and validation and the output of suggested scenarios for badger culling and some cattle management. The modelling is building on a series of previously published scientific papers, with the main development in this project of building a cattle-farming simulation in addition to the badger simulation, and using the latest available data on the badger’s behavioural responses to culling. The report is detailed and quite technical in parts, but the main conclusions are based on the overall costs and benefits of different management scenarios. These economic outputs are presented in Sections 7.3.5 to 7.3.9. Model validation, by simulating the effect of the RBCT is presented in Section 5.4. These two sections and the discussion/conclusions are the most important sections to read if you wish to avoid the technical aspects of the modelling.
The number of cattle herd bovine tuberculosis (TB) breakdowns has been increasing in recent years. Bovine tuberculosis, caused by Mycobacterium bovis, is a serious disease of cattle, and can act as a zoonotic disease (Smith et al., 2004). The European badger is often infected with TB (Delahay et al., 1998), and there is now conclusive evidence that they are responsible for transmission of the disease to cattle (Independent Scientific Group, 2007). Previous strategies have all involved badger culling to reduce the incidence of cattle-herd breakdown (CHB) and such culling had not previously been scientifically shown to be beneficial (Krebs et al., 1998). Theory suggested that some culling strategies may actually increase the incidence of cattle TB (Swinton et al., 1997), through a process known as
8
social perturbation. As this project was being performed, results were being published from the Randomised Badger Culling Trial (RBCT), which began in 1998. Results from the RBCT indicated that reactive badger culling leads to an increase in cattle herd breakdowns (Donnelly et al., 2003), and that large-scale proactive badger culling can reduce CHB in the area culled, but increase it in the immediately surrounding area (Donnelly et al., 2006). Both these effects are believed to be caused by social perturbation of the badger (Woodroffe et al., 2006).
The government is, increasingly, relying on evidence-based policies. For some policies, such as badger and cattle management, it is very difficult, politically complex, expensive and time consuming to obtain field-based evidence. This project modelled a variety of different badger and cattle management strategies and determined their overall cost or benefit to society, in relation to the consequent change in cattle herd breakdown rates. These outputs, along with their intrinsic uncertainties and assumptions, can be used to help inform policy decisions on methods of badger and cattle management to be employed, potential spatial configuration for mixed strategies and the costs and benefits associated with these strategies.
The CSL previously produced a simulation model of TB in the badger (Smith et al., 2001a; Smith et al., 2001b; Wilkinson et al., 2004), which included a simple representation of cattle herds. This original model was the first to give realistic levels of disease (spatial and temporal) without serious population suppression (Smith et al., 1995), and it can be used to simulate any badger control policy: culling, vaccination, and proactive or reactive removal. The University of Reading completed a project to assess the economic impacts of TB and alternative control polices (Defra report SE3112). The CSL model was used to simulate the badger culling patterns seen in the RBCT, and the economics was bolted onto the output in Phase I of the project (Smith et al., 2007). Here, the Reading CBA model was integrated into the CSL model, so that output from the CSL model included disease prevalence, cattle herd breakdowns and the overall costs and benefits. This approach ensured that the full amount of uncertainty and variability within the simulation model, and within the CBA can be included. All assumptions and limiting constraints are reported below, and badger and cattle management strategies were evaluated to demonstrate the applicability of this combined system.
As an added benefit, the validity of the output was improved by the construction of an independent, GIS-based version of the model by Newcastle University. This version included all the badger and cattle parameters as in the CSL model. This GIS version also allows simulation of specific geographical areas, which is helpful in taking a generic management strategy, and evaluating it for specific areas. This allows for the additional validation of simulating the RBCT. The GIS version includes additional mechanistic aspects of the combined badger/cattle epidemiology, but not the costs and benefits.
This project had six objectives, which together with the proposed approaches are given below. All these objectives have been met, with this report fulfilling objective six. Scientific papers are in preparation and will be finalised after this report is completed.
The proposed approaches for each objective, as given in the project proposal, were:1. Integrate the Reading CBA within the CSL model.
The CSL model will be re-written so that the costs and benefits for each simulation may be calculated. Phase 1 of this project utilised output from the CSL model as input to the Reading CBA. The integration of the CBA within the CSL model will allow the output to include the variability and uncertainty inherent in the model, by putting confidence limits
9
and probability distributions on all output values. The completed badger/TB model will then be validated against the results of the pilot project. This work will involve both Reading and CSL. The full costs of cattle testing will be included as the cattle layer is formulated (Objective 3). This will allow the full benefits in terms of reduced cattle infection due to onward transmission, and reduced cattle testing frequency (if breakdowns levels permit) to be calculated.
2. Analyse all available badger data, update the CSL model and validate the results.
The results of current work on sensitivity analysis (including structural sensitivity) of the CSL and Newcastle models will be used to prioritise parameters for further analysis. For all sensitive parameters, updated analysis will be performed on the Woodchester Park data. All suitable and available data obtained from other sources (e.g. the RBCT and Irish four-areas trial) will also be analysed to improve parameter estimates (e.g. immigration rates). The model can be validated against results relating to badger demography from the RBCT, Irish trial, Woodchester Park and other historical data. Further validation will occur after the cattle layer has been completed. The work will be performed primarily by CSL.
3. Evaluate the approach for, and program, the additional cattle simulation within the CSL model.
A series of meetings will be held to define the cattle model structure and required parameters. The output from these meetings will be a flow-chart of the cattle layer (spatial configuration of farms, categories of cattle farms, cattle movement, disease epidemiology and cattle testing). This flow-chart will be distributed to contractors and the steering group for comment and agreement. Following this, parameters will be estimated, and the cattle layer will be programmed. Data will be obtained from the June 2004 census, VETNET and Cattle Tracing Scheme. This will include the costs and benefits for TB within the cattle layer. A further round of model validation will take place, which investigates the cattle herd breakdown rate. The meetings will involve all project personnel, and external experts and reviewers will also be invited to attend and contribute.
4. Produce a GIS version of the badger/cattle model.During construction of the cattle layer into the CSL model, an independent version will be completed by Newcastle University. Newcastle already has a GIS version of the current CSL model, but without any cattle. Therefore this objective will be completed by updating the badger model with any revised parameters from objective 2, and adding a new cattle layer following the protocol developed in objective 3. Sensitivity analysis and validation will be performed in a similar manner to objective 2. The GIS model will be used to project the badger epidemiology and herd breakdown for each of the proactive RBCT squares to validate this version of the model.
5. Evaluate badger and cattle management strategies as identified during the project.
Badger control strategies and costs and benefits of control will be imported from the pilot project, subject to any required changes. Cattle management strategies will be identified following consultation with Defra and the VLA. Management strategies tested below were prioritised during steering group meetings and all suggested strategies were evaluated. The CSL model can be used to test general strategies, while the Newcastle model can be used to simulate specific geographical areas.
6. Produce lay and scientific reports on the outcomes and potential future steps.An extensive lay report will be produced for Defra. This will highlight the full list of underlying assumptions and areas where further progress can be made to improve the models. The results will also be written up for at least one scientific paper for submission to a peer-reviewed journal. This report is the first part of this objective.
10
3. Model
3.1 Economics
3.1.1 IntroductionTo complete objective (1) it was necessary to include all the economic parameters, along with their known variation, within the CSL model. This Section of the report describes the economic parameters included in the integrated model (see Figure 3.1.1) and the assumptions associated with those parameters.
The CSL TB model produces simulations of a number of key variables that can be fed into a partial cost-benefit analysis (CBA) in terms of cattle herd breakdowns and areas or numbers of animals (cattle or badgers) affected by particular control strategies. The cost-benefit analysis part of the model can then estimate the costs associated with a particular TB control strategy and weigh these against the benefits (which are measured in terms of the costs avoided due to the estimated number of herd breakdowns prevented).In terms of TB in cattle, the key variables of the CSL model that are fed into the CBA include:
the number of cattle herd breakdowns,whether breakdown farms are dairy or beef,the number of reactors per breakdown.
Figure 3.1.1. A diagrammatic representation of the cost-benefit analysis (CBA) incorporated into the CSL bovine tuberculosis (TB) model
The cost-benefit part of the model places a value (cost) on these outputs by drawing on distributions of the costs of a TB breakdown to farms derived from a previous survey by Reading University of 151 cattle farms (Bennett & Cooke, 2006; Defra, 2006b). These costs include those relating to slaughter of reactors and associated animals, costs of
11
CBA(NP Benefit)
CSL Model:-Areas surveyed-Badgers culled-Badger population-Cattle Herd Breakdowns
Survey of breakdown farms
Costs of TB Costs of TB testing to farms breakdown to farms
Costs of surveying and badger culling
(from Defra Wildlife Unit, RBCT etc.)
Govt. costs of ‘test andslaughter’ or other policy
(veterinary costs of testing etc.)
isolation of reactors, costs of movement restrictions and other costs to farm businesses identified in the survey. Sampling from these distributions adds a further element of variation to the model’s simulations.
3.1.2 Costs of TB if no action takenThe cost of control of TB includes regular tuberculin testing (when testing is at one to four yearly intervals), compulsory slaughter of reactors and close contacts, and other farm costs such as isolation and movement restrictions. National disease prevalence has been rising, as have the costs. Defra report expenditure of £80M for TB surveillance and testing, control and research in 2006/07. The equivalent costs were £88M in 2003/04, £90M in 2004/05 and £99M in 2005/6. The recent rise in costs has been associated with an increase in the spread of TB, whereas the simulation model considers a fixed area.
The dynamic CSL model proceeds from a set of initial conditions to an equilibrium state, and persists in that state until the introduction of a control strategy. When the (stochastic) model is in the equilibrium state the annual costs of TB are predicted to be in the range £65M to £75M. These costs include:
routine testing costs for farmers and for Defra VLA costs of culturing samples to confirm outbreaks farm costs of isolation of reactors and inconclusive reactors (IRs)s costs of movement restrictions the value of slaughtered animals
but not the Defra costs of research into the disease. In the estimation of cost savings attributable to each of the control strategies, the average ‘without control’ annual cost is taken as the baseline against which to assess the predicted benefit of each control strategy. This baseline assumption may be a conservative one in that it does not consider a situation of an increasing number of TB herd breakdowns per unit area over time. However, the Cattle Herd Breakdown (CHB) rate per unit farm has been relatively constant in recent years (see section 4.3 and Appendix D), so the increasing cost is related more to the increasing area of annually tested parishes.
The State Veterinary Service (SVS, now Animal Health) costs of routine testing of herds and testing associated with a cattle herd TB breakdown are included as part of the cost to Defra of managing the disease. A formula lies behind the published SVS charges for TB tests, and parameters for the formula were used to estimate testing costs currently paid by Defra. The costs incurred at VLA of culturing samples to confirm a breakdown, including cost of transporting samples, are also included in the Defra costs.
3.1.3 Pre-movement testingPre-movement testing of cattle over 15 months old was introduced in England in March 2006, and has applied to cattle over 42 days old since March 2007 (Defra, 2007). Cattle moving out of a 1 or 2 yearly tested herd must have tested negative to a TB test within 60 days prior to movement, and pre-movement tests must be arranged and paid for by the herd owner. However, routine TB surveillance tests paid for by the Government qualify as pre-movement tests, if animals are moved within 60 days after that test.
These regulations have been introduced in advance of any badger culling or vaccination strategy. The CSL model simulated the effects of pre-movement testing on disease prevalence separately and this is reported below. The disease prevalence was permitted to regain equilibrium in the CSL model after introduction of pre-movement testing was
12
simulated and before another strategy was introduced. Thus the economic consequences of any further measures to control TB in cattle are assessed without being affected by the consequences of pre-movement testing.
The model simulated the numbers of cattle tested for TB each year in particular areas and the economic element of the model then attached costs for farms (on-farm costs of testing taken from the Reading survey of breakdown farms) and costs to government of testing (from SVS and VLA data) including veterinary time and costs of laboratory tests.
3.1.4 Costs of implementing culling or vaccination strategiesBadger culling costs per unit area for surveying badger setts, performing a cull according to a given protocol and costs per badger culled (derived from the Defra Wildlife Unit, RBCT data and other sources) are applied to the simulations to estimate the costs associated with a badger culling measure across a particular area. To assess the possible benefit of badger culling or vaccination, various assumptions have been made concerning the component set up and operating costs.
Component costs:Set up costsIt is assumed that farmers/landowners pay the whole cost of set up for badger culling or vaccination. Prior to culling, there is a one-off cost for equipment, as detailed in Appendix A.
Ecological monitoring costsInformation on which to base Defra’s administration costs associated with badger culling or vaccination is not available. The assumption has been made that Defra will incur an initial cost in issuing permissions for badger culling or instructions for badger vaccination, and a further cost for assessing results at the end of the culling or vaccination period. A figure of £1000/km2 for ‘ecological monitoring’ has been assigned to both of these costs, pending further clarification.
Farm survey costFor every year in which badger culling or vaccination occurs, there is assumed to be a requirement for a survey of badger setts requiring two days of farm labour per km2, based on experience from the Randomised Badger Culling Trial (RBCT).
Culling costsFarmers/landowners are assumed to pay the whole cost of badger culling or vaccination. The cost of operating cage traps, with despatch of badgers caught and carcass disposal, is known from the RBCT. There is also some information about costs of operating restraints from work in Northern Ireland. For the other methods, and for vaccination, various informed assumptions have been made, as outlined in Appendix A.
Vaccination costsFor the hypothetical strategy of using an oral vaccine for badgers, costs are estimated for two types of vaccine pellet, based on farm labour costed at current farm labour rates and estimated vaccine costs.
13
3.1.5 Cattle costsBeef cattle slaughter valuesDetailed information on the cost of a TB breakdown in beef herds was obtained from the Reading Farm Survey (de Lisle, Buddle & Collins, 2007). This detailed data, updated for beef market value, is all included in the lognormal distribution of farm cost per slaughtered animal shown in Appendix B.
Dairy cattle slaughter valuesDetailed information on the different pattern of cost of a TB breakdown in dairy herds was likewise obtained from the Reading Farm Survey. This data was similarly updated for inflation in the value of dairy cattle, and another lognormal distribution used to represent the range of costs, also shown in Appendix B.
Tuberculin testThe model simulates specific numbers of cattle tested per farm during each simulation. It is thus possible to estimate the SVS (Animal Health) test cost per herd, which varies with herd size according to a tariff published by SVS. The SVS charges for this testing are a major part of the annual cost to Defra for control of TB.
Gamma Interferon (gIFN) testThe model included optional use of a gIFN test for TB made at the same time as tuberculin testing. Costs are included for taking blood samples, transporting them to VLA, and subsequent analysis at VLA. These costs are as for the tests reported in the Gamma Interferon Field Trial (VLA, 2006). An assumption has been made that no additional farm labour is required if both tests are made at the same time.
Farm management costThere is a farm management cost associated with each test, on which information is available from the Reading Farm Survey. This is one of the components of the overall cost of the disease to the farms responding to that survey. A small allowance for farm labour was estimated, and incorporated in the overall farm cost, as indicated in Section 3.1.2.
Movement restriction costsThe Reading Farm Survey also collected data on daily farm costs of movement restrictions during an outbreak. These costs are higher for beef farms than for dairy farms. The model calculates the duration of movement restrictions for each type of farm in each simulation, enabling estimation of total cost to farms of movement restrictions in the outcome of each simulation.
Compensation paymentsDefra compensation rates for animals that are compulsorily slaughtered vary from time to time. Since these are transfer payments, they do not affect the overall economic outcome of the control methods assessed. As an approximate way to distinguish the annual cash flows for Defra and for farms, median compensation values for beef and dairy cattle are included in the calculations, where this breakdown of cost is presented.
3.1.6 VLA costsEvery time a tuberculin test returns a positive result, samples are sent to VLA for culturing to confirm the presence of TB. Costs for this work have been obtained directly from VLA (private communication). These costs have also been included as part of the cost to Defra of control of TB.
14
3.1.7 Savings resulting from culling or vaccination strategiesThe cost-benefit part of the model produces estimates of the Net Present Value (NPV) of particular TB control strategies according to different scenarios. These different scenarios incorporate different specifications as to the assumptions contained in the analyses. The NPV estimates apply a discount rate (the default is the current Treasury rate) to the annual flows of costs and benefits that are simulated and calculated over whatever time period is being considered (e.g. 15 years). These costs and benefits are:
NPV = discounted [farm level costs of cattle herd breakdowns + farm level costs of TB testing + government veterinary costs of TB testing + government administration etc. costs of TB + costs of additional TB control measures]
The highest NPV is the preferred TB strategy or scenario (other things being equal). Measures to control TB in cattle or badgers will incur costs, so reducing the NPV. However, if a control measure is successful these costs may be mitigated and perhaps outweighed by the benefits of control in terms of lower costs associated with reduced cattle herd breakdowns/reactor animals, lower costs of testing (for example if reduction in TB means that farms move to a status having a longer interval between routine tests) and lower costs of government administration on TB.
3.2 Model DescriptionsThe model descriptions are detailed below following the standard protocol of Grim et al, (2006).
3.2.1 PurposeThe overall purpose of the models is to provide evidence to the UK government to assist the decision-making process aimed at controlling and reducing the rates of TB in UK cattle herds. To ensure quality of work and customer confidence, two models were developed using the same basic methodology described below, but using different spatial approaches. The two models were cross-validated against each other.
The Central Science Laboratory model (CSL model) is an individual-based model with an integrated cost-benefit analysis module, produced by Reading University, designed to compare the costs of a range of cattle management and badger control strategies with a “no-control” option. The badger control options modelled were shooting, trapping, snaring, gassing, and vaccination. The CSL model is a stochastic spatial model, which does not attempt to simulate the detail of any existing landscape.
The Newcastle University model (NU model) is an individual-based model with explicit spatial form, so can simulate badgers, cattle and TB in a landscape structure matching closely to one in the real world. As one of the validation processes, the NU model was used to simulate the spatial configuration and badger group structure of the Randomised Badger Culling Trial (RBCT) areas, and the effects of culling on TB herd breakdowns incidence were compared.
3.2.2 State variables and scalesThe models consists of three distinct interacting components; a badger population dynamics component, a cattle herd dynamics component, and a management component.
15
The badger component consists of a population modelled at the level of the individual. All individual badgers belong to a social group, but no processes act on the group as an entity.
The cattle component consists of a population modelled at the level of the individual and at the level of the herd.
The management component consists of a number of state variables indicating which management procedures are imposed in each iteration of the simulation. The state variables of management are: presence or absence of pre-movement TB-testing of cattle; frequency of routine cattle TB-testing; test sensitivity; test specificity; and method and efficacy of badger removal operations.
Parameters and the default values used in the models are tabulated in Appendix E.
CSL model:Badger numbers are recorded in a matrix characterised by social group identifier, sex, age (juvenile/1st year, yearling/2nd year, adult/3rd year plus), and TB status (healthy, infected, infectious, super-infectious, vaccinated).
Similarly, cattle numbers are recorded in a matrix categorised by herd identifier, sex, age (30 x 2-month categories with last category also used for older cattle), and TB status (healthy, infected, infectious, super-infectious or anergic). Herds are categorised as beef or dairy, so parameter values can be varied according to herd-type.
The model operates on a two-month time step, and runs for 19 years with badgers only, a further 100 years with both cattle and badgers and their interactions, then five years with badger control options implemented, and then a further 11 years without any badger control. The initial 119 years before badger control is a period for the population and disease dynamics to settle in the model, and is necessarily long as various processes are introduced sequentially.
Three separate scales of model grid were implemented: (a) 20km by 20km (400km2), (b) 32km by 32km (1024km2), and (c) 40km by 40km (1600km2). In all three the resolution of the grid was set to 200m by 200m squares, and the whole area was wrapped round on itself to form a torus in order to prevent edge-effects. The above three scales were chosen to allow three options of badger control (approx 100km2, 300km2, or 400km2) to give a similar ratio of controlled to uncontrolled land (23%, 27% and 23% respectively).
A network of parish areas was defined over the model grid, which, as in real life, formed the scale as a basis of both cattle management options (such as TB-test frequency) and badger control options (such as culling or vaccination).
Both the parish layer and the badger social-group layer were spatially defined by tessellating around randomly placed points to give a specified density. The cattle grazing areas were spatially defined by spiralling around randomly placed points to give a range of grazing areas at a specified density. So badger social groups were always modelled as contiguous with other groups, whereas cattle grazing areas were sometimes contiguous with each other. A new spatial configuration and initial populations (badger and cattle) were created for each iteration.
NU model:
16
The individual badgers are characterised by: a badger identity number; age; sex; social group identity; and TB status.
Individual cattle are characterised by: age; sex; herd identity number; and individual TB status.
The model operates on a two-month time step, and runs for fifteen years. The model simulated a 20km x 20km square region; which contains a mean of 302.1 badger social groups, and 311.8 cattle herds over 100 simulations of the model; this corresponds to 0.75 badger social per km2
and cattle farm density of 0.78 per km2, similar to the CSL model.
3.2.3 Process overview and schedulingCSL modelFirst the grid is populated with badgers and cattle, seeded with TB infection and allowed to stabilise. Thereafter, within each time step (six times per year, or else as otherwise stated) the following processes occur in the following order:
a) Births of badgers (time-step 1 only = February)b) Ageing of badgers (time-step 1 only = February)c) Births of cattled) Mortality of badgerse) Mortality of cattle (sent to slaughter)f) Dispersal of badgersg) Control of badgers (e.g. culling/vaccination) (years 120 to 124 inclusive, and time-
step 3 only = June)h) Social perturbation of badgersi) Movement of cattle (farm to farm) including pre-movement testing of cattle, if
includedj) Spread of TB infection – first badger-to-badger, then cattle-to-cattle, badger-to-
cattle, and cattle-to-badgerk) TB disease progression in badgersl) TB disease progression in cattlem) Cattle testing (routine testing of cattle for TB, if due for herd)n) Ageing of cattleo) Analyse and output data (time-step 6 only – end of each year)
NU ModelWithin each year and time step, 8 modules are processed in the badger component in the following order: birth, mortality, dispersal, badger control, social perturbation, disease transmission (badger to badger), disease transmission (cattle to badger) and TB infection
Within each year and time step, 8 modules are processed in the cattle component in the following order: birth, mortality, movement, herd management, disease transmission (cattle to cattle), disease transmission (badger to cattle), TB infection and TB testing
3.2.4 Design conceptsCollectives (badger component): all badgers are organised into social groups. Each social group is characterised by its own state variables: a social group identity number; carrying capacity; geographical location; and the identity of those social groups with which it shares a border. Transmission of TB within a social group is assumed to be higher than
17
transmission between social groups. The carrying capacity of each social group is achieved by limiting the number of breeding females.
Collectives (cattle component): all cattle are organised into herds, and herd management procedures operate simultaneously on all members of the herd. Herd state variables consist of: a herd identity number; field identity number of the current field occupied (NU model), or grazing area identity (CSL model); ideal size of herd; type of cattle-herding enterprise (dairy, beef, or mixed); presence or absence of movement restriction orders; cattle isolation order status; test-identified herd TB status; actual herd TB status; and current testing interval. In the NU model, in addition to the herd state variables, each field used by the cattle herds are characterised by: field identity number; geographical location; herd identity number; and the identity of those fields with which it shares a common border (neighbouring herd). Each herd utilises two or more fields throughout the year (but only one at a time); and each field is unique to any given herd. In the CSL model each grazing area is characterised by grazing identity number, and the identity of all grazing areas with which it shares a common border (neighbouring herd). For both models transmission of bovine TB within a herd is assumed to be higher than transmission between contiguous herds. Each herd has an ideal size; each time-step excess cattle are moved out, and deficiencies made up for by moving cattle in (simulating cattle market movements).
Interaction: there are three main types of interaction in the model, which revolve around the transmission of bovine TB between individuals. Firstly, infected badgers can pass TB to uninfected members of the same social group, or to badgers in neighbouring social groups. Secondly, infected cows can pass TB to uninfected cows of the same herd, or to cows of a contiguous neighbouring herd. Thirdly, the badger component and the cattle component can pass TB between them. Any herd whose current grazing area overlaps the territory of a badger social group can potentially pass bovine TB to the badgers. Additionally, any badger social group whose territory overlaps the grazing area currently used by a cattle herd can potentially pass bovine TB to the cattle. A fourth interaction is that the parish with the highest rate of cattle herd breakdown (CHB – TB confirmed by test or post mortem) and several adjoining parishes are identified and badger control can be performed in those areas.
Emergence: the dynamics of TB within the badger social groups and cattle herds emerge from the population dynamics of the individuals, and any badger control carried out, but the life cycle of individuals is entirely represented by empirical rules describing birth, mortality, and dispersal / movement rates. In the CSL model the costs of any strategy emerge from a combination of the costs of CHBs that might occur, the costs of cattle management including testing, movement restrictions and isolating cattle, and the costs of applying the badger control.
Stochasticity: all demographic parameters are interpreted as probabilities, or else are drawn from empirical probability distributions. This was done to include demographic variation in the spatial and temporal dimensions. In the CSL model the landscape structure (e.g. shapes of badger territories and grazing areas and which are neighbours to which) are also determined stochastically, albeit the densities being kept to specified values.
Observation: for cross-validation between the two models, two key variables were compared: TB prevalence in badgers; and mean CHBs per farm. For the main model analysis the following aggregated variables were recorded:
a) mean badger group sizeb) mean number of infected badgers per badger group
18
c) badger TB prevalenced) mean cattle herd size, beef and dairye) cattle TB prevalencef) herd TB prevalenceg) number of parishes in each cattle test-interval categoryh) mean number of TB test reactors (conclusive and inconclusive) per farmi) mean CHBs (confirmed) per farmj) badger contribution to the CHB rate – the proportion of CHBs initially caused by
badger-to-cattle disease transmission as opposed to cattle-to-cattle transmissionk) mean movement-restriction-months per farm (partial and whole herd)l) total costs/benefits of modelled management strategies (discounted Net Present
Values – NPVs) were calculated across the simulated grid. A NPV was output for each of 100 model iterations, enabling display of a distribution of NPVs (cost-benefit risk distribution).
m) NPVs separated for Industry (farmers) and Public (Defra) purses.
3.2.5 InitialisationCSL ModelThe badger component was initialised with specified distributions of age, sex, and social group size, then the model run for 19 years to allow the badger population dynamics to settle. In year 20 the cattle component was initialised with specified distributions of age, sex, herd size and herd-type (dairy/beef), and both the badger and cattle components were seeded with TB. In year 50, and annually thereafter, parishes were allowed to change test-interval status according to their CHB history, using the same rules that are applied in real life. From year 100 pre-movement testing was done for all areas of T1 and T2 status (yearly and two-yearly testing-interval status). In year 120, each badger-control option (including a no-control option) was simulated separately, for 5 years, each control being run from identical starting conditions in year 120, in terms of badger and cattle populations, TB status, spatial structure etc. Each scenario was run for a further 11 years after badger control finished.
NU ModelThe badger component was initialised with specified distributions of age, sex, social group size, and TB prevalence. These were then simulated for 99 years with no badger control to develop mature populations with a stable age structure. In year 100, the cattle component was introduced to the simulation, initialised with specified distributions of age, sex, herd size, and TB prevalence. Badger and cattle interactions were then simulated for 20 years with no badger control to develop a stable interaction between the two components of the model. In year 120, badger control was introduced. Each control scenario was simulated using the same initial population (that from year 119), and applied for five years of control followed by ten years of no control, for each replicate simulation.
3.2.6. InputInput to the model, in terms of badger and cattle populations, TB prevalence and any emergent spatial structure, was provided by simulation of the components until the population and disease dynamics had stabilised prior to introduction of badger control scenarios (see Initialisation, above). There are no “collected data” inputs apart from parameter values (see Appendix E).
19
3.2.7. SubmodelsCSL submodels
1. Creation of badger territories A specified number of badger territories are created: main setts are placed randomly across the grid, and grid squares allocated to closest main sett to give fully contiguous tessellated badger groups. The grid is treated as a torus so there are no edges. The total number of territories added to the grid (see Appendix E.1.2) gives an average territory size of 1.33km2: giving a territory density of 0.75km-2.
2. Creation of Farms A specified number of farms are added to the grid to produce a realistic farm density of 0.78km-2. Farm “centroids” are added to the grid at random, and allocated as a farm type (beef, dairy, mixed, X1, X2, X3, X4) stochastically to give the correct proportions of farm-types. X1-X4 represent farms that also have other stock and hence are allocated as beef or dairy, but with proportionately smaller grazing areas for cattle. Potential farmland for grazing allocation is determined by tessellation from the centroids.
3. Creation of grazing areas The grazing area for each farm is determined according to preset proportions of grazing land for each farm-type (see Appendix E.1.4). Sufficient grid squares are marked as grazing land by using a spiralling algorithm starting from a random grid square within the farmland. Thus, all grazing area is contiguous within a farm.
4. Definition of “near-farms” The distance from each farm’s grazing area centroid to every other is calculated, sorted, and stored in a matrix for later use in the “Move Cattle” procedure, so that farms buy cattle from nearby farms in preference to distant farms.
5. Creation of parishes A specified number of parishes are created (see Appendix E.1.2) by randomly allocating grid-squares as parish centroids, then tessellating around each centroid. The mean simulated parish size is 13.3km2.
6. Definition of neighbours This procedure determines the badger neighbours of each badger territory, the cattle herd neighbours of cattle herds (grazing areas contiguous), and which cattle herds overlap which badger territories. This allows between-group TB transmission to be simulated. Badger-to-cattle and cattle-to-badger transmission only occurs where the grazing and the badger territory overlap (as opposed to being simply adjacent).
7. Addition of badgers Badgers are added to each badger territory at the start of year one, with some stochastic options (see Appendix E.1.3), to give a stable mean badger group size of about 7.5 adult badgers per group, as measured at the end of December.
8. Addition of cattle Individual herd size is calculated from the grazing area on each farm, and a stocking density taken at random from the stocking-rate distribution (see Appendix E.1.4). Cattle are then added stochastically, using probabilities based on the profile of ages/sexes for the herd type. For simplicity, all cattle are initially allocated to ages equivalent to the first time-step of each year. Cattle are added to each farm at the start of year 1, but kept static till year 20. The simulated stable mean herd sizes are about 44 head for beef, and about 86 head for dairy.
9. Births of badgers In the model female badgers give birth in the first time step of each year, which is equivalent to January + February. The number of females that breed in any one badger group is determined probabilistically (see Appendix E.1.3), although this is limited by the number of 2+ yr-old females, and the carrying capacity of the group.
20
The breeding probability for the first female is fixed, but the probabilities of the 2nd/3rd/4th are higher for groups with fewer badgers present (linear relationship). Litter sizes are also determined probabilistically (see Appendix E.1.3), mean litter size is 2.94, and the cub male:female ratio is 1:1.
10. Ageing of badgers This occurs in the first time step of each year within the “birth of badgers” procedure. All badgers are aged by one year immediately after the birth routine, but just before the new cubs are added to the main population array.
11. Mortality of badgers Badger mortality rates (Wilkinson et al., 2000), are dependent on sex, age, and health status (see Appendix E.1.3), and are adjusted linearly to give lower mortality rates for smaller groups. The mortality rates are applied to individual badgers probabilistically.
12. Dispersal of badgers Dispersal probabilities are sex-dependent (see Appendix E.1.3), but are not related to age or season. The dispersal routine occurs every time-step. Badgers disperse only as far as their neighbouring group, and tend to move to a group with fewer badgers if one is available. Badgers are not allowed to disperse twice in one time-step.
13. Social perturbation of badgers This procedure moves badgers to fill vacancies (see flowchart, Appendix J). It occurs whether or not badger control is being simulated, but obviously these perturbation movements are more frequent immediately following badger removal. Sexes are checked independently, groups that already have two of a sex would not receive a third, and the donor group must also have at least three more badgers of that sex than the recipient group. Badgers are moved shorter distances in preference, and a badger is not allowed to make two moves within the same time-step. Note: for details of the simulated “perturbation effect”, see “Transmission of TB – badger to badger” description below and also section 3.2.8.
14. Seeding of TB in badgers At the start of year 20, each badger group is given a high probability to have one badger of random sex and age to be infected with TB. Each selected badger during this seeding process is given a TB status of “infected”.
15. Seeding of TB in cattle At the start of year 20 about 10% of farms are chosen at random, and in each of these a single cow of random sex and age is transferred from healthy to infected disease status.
16. Set timing for annual cattle-test At the start of year 20, each herd is allocated a random time-step (between 1 and 6) to determine when it will be due for its annual TB-testing programme.
17. Birth of cattle All female cattle aged over 22 months give birth to one calf annually, on their birthday. The sex ratio of the calves is set at 1:1, the sex being determined probabilistically. Over the age of 60 months, since all cattle still alive remain in that age category, births are determined probabilistically each time-step, using a 2-monthly birth rate. Births are applied to Main herds and Isolated herds in the same way.
18. Mortality of cattle Mortality only applies to cattle going to slaughter. For simplicity, natural mortality on the farm is not modelled. Each cow is categorised by age into 6-month periods, and mortality rates applied probabilistically. The mortality rates are dependent on herd-type, sex, and age, but independent of health status (see Appendix E.1.4).
21
Mortality rates were calculated from the CTS data through construction of life-tables (Appendix C). If a cow going to routine slaughter is infectious or super-infectious (not just infected) a probability of TB detection at slaughter is applied (see Appendix E.1.4), and if TB is detected, movement restriction and testing is triggered at the farm of origin. Mortality is applied to Main herds and Isolated herds in the same way.
19. Movement of Cattle If herd-size (sum of main + any isolated herd) is smaller or larger than the ideal size (according to stocking density) extra cattle are moved on or off the farm. All spare cattle for all farms are initially moved into a holding stock (market equivalent), with males moving off a dairy farm as priority, and females off a beef farm, but otherwise picked at random (i.e. independent of age and health-status). Extra animals are then chosen at random to send to “market” to ensure that 40% of cattle move each year. Cattle are then moved from “market” to farms that are short of cattle, females moving to dairy, and males to beef as first priority, then proximity to donor farm as the second priority. Following all within-grid movements, if more cattle are needed, they are added into the grid, and if there are cattle left in the “market”, they are removed to simulate movement from T1 to T3 and T4 areas. Each cow being moved into the grid is given a probability of being infectious with TB, calculated from the proportion of Britain that is T1, and the cattle TB prevalence in the model at the time of the move. The number of infected cattle being moved outside the grid is used to calculate the extra CHBs that could be caused by those movements (e.g. from T1 to T4 areas). After year 99, if pre-movement testing (PrMT) is switched on, all cattle in T1 and T2 areas are tested before movement (see “Test Cattle pre-movement”, below), and if any test from a farm is positive, movement is not allowed and standard testing procedures are triggered.
20. Transmission of TB – badger to badger Each infectious badger has a chance of infecting every contact, both within-group, and between-group (neighbours). Transmission rates are set higher for super-infectious badgers, and between-group rates are set to 5% of within group rates (see Appendix E and Smith et al., 2001b). During years of badger control, between-group infection rates are recalculated (see “Setting Transmission Rates” below) to give higher TB transmission rates in/around the control area (i.e. a higher probability of between-group contacts). This is to simulate the “perturbation effect” of badger culling.
21. Transmission of TB – cattle to cattle Each infectious cow has a chance of infecting every contact, both within-group, and between-group (neighbours). Probabilities of transmission are currently the same for infectious and super-infectious; and between herd rates are set to 5% of within herd rates (see Appendix E.1.4). Transmission rates from beef cattle are set to about twice the value of dairy cattle (Munroe & Dohoo, 1999). Unconfirmed reactors that are separated from the main herd (put together with other unconfirmed reactors in an isolated field on the farm) are classified as “isolated herds” in the model (herds isolated as part of the TB control procedures). Such isolated cattle in the model are able to transmit TB infection to each other within the isolated herd, but do not transmit TB infection to any of the healthy cattle in the farm’s main (non-isolated) herd, or to any cattle on neighbouring farms. However, it is assumed that badgers still have access to the field holding the isolated cattle, so TB transmission is still able to occur between cattle and badgers.
22. Transmission of TB – badger to cattle Each infectious badger has a chance of infecting every contact cow that grazes on
22
land shared by the badger (i.e. where the badger territory and the grazing land overlap). Only super-infectious badgers are given a transmission rate greater than zero (see Appendix E.1.3).
23. Transmission of TB – cattle to badger Each infectious cow has a chance of infecting every contact badger where the badger territory overlaps the grazing land. Only super-infectious cows are given a transmission rate greater than zero, and it is set to the same transmission rate as from badger to cow (see Appendix E.1.4).
24. Setting transmission rates (including perturbation effect) Transmission rates are applied stochastically, and default values are listed in Appendix E.1.3. Special rates, however, are applied to those badger groups subjected to culling, and to their immediate neighbours. This is to simulate higher contact rates during a period of social perturbation as a result of the culling. This perturbation effect is simulated during the whole period of culling (five years), and for three years after the last cull (i.e. a total of eight years). Wherever and whenever the perturbation effect is applied, all badger-to-badger between-group transmission rates are increased to equal the within-group rates. Badger-to-cattle, cattle-to-badger, and cattle-to-cattle rates are not adjusted.
25. Disease progression in badgers Badgers with TB are given the chance of transferring from one TB-status to another, according to pre-set probabilities (see Appendix E.1.3). A badger can only make one such change per time-step. Disease progression is from infected to infectious to super-infectious. Infectious badgers also have a probability of reverting back to the infected stage, but if a badger becomes super-infectious, it stays in that state till death. A newly infectious badger does not itself have the chance to infect another badger or cow until the following time-step.
26. Disease progression in cattle Cattle with TB are given the chance of transferring from one TB-status to another, according to pre-set probabilities (see Appendix E.1.4). Disease progression is from infected to infectious to anergic. Infected cows also have a possibility of transferring straight to the anergic state in one time-step. Infectious and super-infectious cattle are not able to revert to a lower disease state, and if a cow becomes anergic, it stays in that state till death. A newly infectious cow does not itself have the chance to infect another cow or badger until the following time-step.
27. Testing of Cattle This procedure simulates both routine testing and TB-triggered testing, and is recorded as a cost to Defra. A countdown system is used to trigger the routine “whole-herd” tests for each farm at the appropriate time-step, and a different countdown for “partial-herd” tests if there are isolated cattle. For the whole-herd test both Main and Isolated cattle are tested. Every cow is tested using probabilities to determine whether it will be a Reactor or Inconclusive (see Appendix E.1.4). These test probabilities are dependent on cow health-status and test-type (standard or severe interpretation). Inconclusive (unconfirmed) reactors are modelled to simulate the processes that would occur in the field (isolation, movement restrictions, test follow-ups) including economic costs. Inconclusives are isolated, but any individuals testing Inconclusive for the third time running are classed as Reactors. Any Reactors are slaughtered and subject to post mortem examination. It is assumed that all infected reactor cows will be confirmed at post mortem, flagging up a confirmed CHB and triggering movement restrictions on contiguous herds, and their testing in the next time-step. Test results are analysed on a herd-basis, and a herd’s test-status and next test requirement stored. When a test is positive a series of procedures are brought into effect, simulating the veterinary procedures that are
23
used in the field. A summary of this process is described in a flowchart in appendix H. If tests are negative, and appropriate, isolated cattle rejoin the main herd.
28. Test cattle – pre-movement After pre-movement testing has been switched on, this procedure is called whenever cattle are about to be moved, and all cattle over a specified age are tested. The costs of this testing is recorded as a cost to the farmer (Industry). Only the cattle about to be moved are tested, and in the first year, of pre-movement testing only cattle of age 16 months or older, and from the second year onwards only cattle aged 2 months or older. The actual values used in the field are 15 months for the first year and 6 weeks from the second year onwards, but the values of 2 months and 16 months are used in the model to fit in with the model’s 2-month time-step. If any animals react positive to the pre-movement TB test, all animals of that herd are stopped from moving, and a series of procedures are brought into effect, simulating the veterinary procedures that are used in the field. A summary of this process is described in a flowchart in appendix H. If all tests are negative then the cattle are allowed to move (to market). Pre-movement testing is not applied to cattle moving straight to slaughter.
29. Ageing of cattle Every two-month time step all cattle are aged by two months, except those already aged to the maximum category of 30 (60 months = 5 yrs), which simply stay in that category till death.
30. Switch Test Intervals This procedure allows farms in a parish to switch their test-interval status according to their CHB rate history. It is applied at the end of each year, from year 50 onwards, and the average number of herds that have had a breakdown within the previous six years is calculated for each parish. This determines what Test Interval all the farms in a parish should be (T1 to T4) (see Appendix E.1.1 for details of the algorithm). In the model, a parish can only change by one category in any one year (e.g. T1 to T2, but not T1 to T3). A new test month is calculated for each farm at the end of each year, dependent on any test interval change and when the next test was due, and this new test month applies immediately.
31. Apply badger control Each control method is applied for five years from year 120 to year 124 inclusive, starting with identical conditions to the no-control option within that simulation at the start of year 120, to give a fair comparison between the control methods. The badger groups to be controlled are determined by selecting the parish with the highest CHB density (from years 117 to 119) and selecting the appropriate number of parishes located contiguously around it. If there are choices, the parishes with the highest CHB densities are selected. A proportion of farms are excluded at random from having badger control, to simulate non-compliance. If 10% or more of a badger territory overlaps with a parish selected for badger control, that badger group is marked for control. Control of the selected badger groups is applied stochastically at the specified control rate for the method (see Section 6.1.1 for control rates simulated) every third time-step (equivalent of May/June) once per year for the five years.
32. Save data Output parameters are calculated at the end of each year, including badger population, badger TB (number and prevalence), cattle herd breakdown rates, costs of each control (Net Present Values) and control benefits compared with no-control. Most outputs are summarised across the simulations, but a discounted net benefit value is output for each simulation. All output data for each set of simulations is saved in one Excel file.
24
NU submodelsThe following submodels describe procedures in the badger and cattle components of the model that require more explanation than simple stochastic determination of birth, death, etc. Except where indicated below, the submodels of the NU model had identical processes to those of the CSL submodels. All data used were identical to the CSL model, with the exception of disease transmission parameters, since these were adjusted to ensure the output values were similar to the field data. These are indicated below:
Beef – within group 0.010084Dairy – within group 0.005007Beef and Dairy – between group 5% of within-groupBeef and Dairy to badgers 0.0000504Cattle have no differences in rates between infectious and superinfectiousSuper-infectious Badger – within group 0.0063Super-infectious Badger – between group 0.0005Super-infectious Badger to cow 0.00005Rates for infectious badgers are half that of super-infectious badgers, except for Badger to Cow, which is zero
1. Badger dispersalDispersal probabilities are sex-dependent. The dispersal routine occurs every time step, and badgers disperse only as far as their neighbouring group, and tend to move to a smaller group if available. Badgers were randomly selected for dispersal, and an individual was not selected more than twice in a single time step. Thus a badger that had moved into a territory could be selected to move from that new territory, but could not move again. The furthest that a badger could move in six months was therefore to the next neighbour of the neighbouring territory of its point of origin.
2. Badger social perturbationThis process moves badgers to fill vacancies in social groups. Sexes are checked independently. Groups that already have two members of a sex cannot receive a third, and the donor social group must also have at least three more badgers of that sex than the recipient group. Badgers can be sourced from the neighbour and next-nearest neighbour of the vacant social group, but badgers are moved shorter distances in preference. If a move is appropriate, then a badger of the correct sex is chosen at random from the donor group and assigned to the recipient group. A badger is not allowed to make more than one move in each time step.
3. Cattle movementEach time step the herd size was corrected back to the ideal herd size by moving extra cattle on or off the farm. All spare cattle for all farms are initially moved to a holding stock, with males moving off a dairy farm as priority, and females off a beef farm with priority, but otherwise cattle are moved at random. Extra animals are then chosen to send to market to ensure that about 40% of cattle move each year. Cattle are then moved from market to herds that are short of cattle, females moving to dairy herds and males to beef herds as first priority, then proximity to donor farm as second priority. Following all movements, if more cattle are needed, they are added to the market (and each has a pre-defined probability of being infected).
4. Cattle testingA counter updated in each time step triggers the annual test for each cattle herd, and a separate counter triggers the repeat test for an isolated herd. Each cow within a herd is tested using probabilities that the animal will be assigned a conclusive or inconclusive reactor status based on its individual health status and the test used
25
(standard or severe). Inconclusive reactors are isolated; but if the herd has received this status four times running (herd-based, not individual-based) then the verdict is conclusive reactor instead. Conclusive reactors are slaughtered and all infected cows are confirmed, flagging this herd as a confirmed cattle herd breakdown (CHB), triggering movement restrictions and repeat tests. If tests are negative, isolated cattle are returned to the herd.
5. Cattle herd managementIn each time step, each herd is re-assigned to a different field, randomly chosen from all those fields associated with that herd.
3.2.8 Perturbation Processes3.2.8.1 Extra movement of badgers.
a) This is in addition to the dispersal of badgers, and like dispersal the procedure is run every 2-month time step. It provides an opportunity for territories with surplus badgers to “donate” badgers to territories with few or no badgers. The movement aspect of perturbation runs throughout the model, whether or not any control happens to be taking place, but obviously movements are much more frequent after culling has occurred and small/empty social groups exist. This routine thus counters the stochastic extinction of social groups by chance, as well as producing immigration into culled areas. [This extra-movements aspect is unrelated to the classification of perturbed groups in 3.2.8.3 below, which solely relates to increased transmission rates].
b) Any individual badger movement is only allowed between social groups that are neighbours or neighbours-but-one, and this rule holds whether or not culling has occurred. It is possible that a badger may move further over time (stepping a maximum of two social groups per time step) if the above conditions are met for each move. If there is a choice of donor, preference is given to the shorter-distance move, i.e. neighbours rather than neighbours-but-one
c) Every group is checked as a potential donor or recipient, regardless of location relative to any cull area. The process runs for the badger sexes separately. The donor social group must have 3 or more badgers (of the sex being processed) more than the recipient group, for a movement to be allowed. Also the recipient must have less than two of that sex. If there is more than one potential donor, one is chosen at random. A badger is chosen to move at random from those available of that sex (i.e. independent of age and health status) and is placed in a temporary holding-array so that it will not be moved twice in the same time step.
d) Group sizes are dynamically adjusted so that no group loses or gains too many, and all badgers in the holding-array are transferred back to the main badger array after all social groups have been processed.
3.2.8.2 Increase in badger transmission ratese) For a fixed period of the time the between-group badger transmission rates are
increased to simulate extra contacts between badgers, during and after culling. This is assumed to occur as a result of the extra roaming movements that have been seen in the field after culling.
f) This effect is simulated from and including the time step when culling is first implemented (the disease transmission procedure always runs after the culling procedure within a given time step) and carries on for eight years – three years longer than the culling period.
26
g) Within-group transmission rates remain unaffected, and disease progression rates are also unchanged.
h) If a badger group is classed as perturbed (see 3 below), then between-group transmission rates are increased up to the same level as the within-group rates.
i) Between-group transmission still only occurs between neighbouring territories. It is assumed that transmission does not occur beyond the neighbouring group in any one step.
j) Badger-to-cattle, cattle-to-badger and cattle-to-cattle transmission rates are unaffected
3.2.8.3 Definition of a perturbed groupk) A badger group is classified as perturbed if either it has itself been subject to culling
within the previous three years, or it is close to a badger group that has been subject to culling within the previous three years. The definition of close has been tested in two forms – firstly as being a neighbour (one-layer type), secondly as being a neighbour or a neighbour-but-one (two-layers type).
l) In terms of the transmission effects (3.2.8.2 above) there is no distinction between groups that are classed as perturbed because they have been culled, or those classed as perturbed because they are close (either layer-1 or layer-2), since no evidence exists to show a different level of response.
3.2.8.4 Outputs to examine temporal perturbation effectsFor each control method, and for each year, the following metrics were output from the model as measured at the end of each year (aggregated across all simulations and the whole grid: culled and unculled areas):
Mean badger group size.
Mean number of infected badgers per group
Mean badger prevalence
Mean Cattle Herd Breakdown (CHB) rate per farm
3.2.8.5 Outputs to examine spatial perturbation effectsThe output metrics above were also produced, separated into different zones of the grid:
Core area – the area within a boundary drawn 2km inside the edge of the control area.
Inner layer – the area between the core area and the outer edge of the control area.
Outer layer – the area between the outer edge of the control area and a boundary drawn 2km outside that.
m) The core area and the inner layer together comprise the whole control area.
n) Where a badger territory or a cattle grazing area overlaps more than one zone, the metric value is divided proportionally, based on the areas within each zone.
27
4. Verification
4.1 Badger-to-Cattle Transmission RatesIt is expected that if the badger-to-cattle transmission rates were set to zero, the CHB rate should decline to zero. This is equivalent to removing infection in the badger but continuing the cattle testing. As part of the model verification process, this was tested with the CSL model, and the CHB rate did indeed rapidly decline, reaching zero within about 5 years.
4.2 Cattle Mortality RatesReal life slaughter rates of cattle, obtained from the Cattle Tracing Scheme database, were examined to determine the most appropriate mortality rates to apply in the models (Appendix C). The 6-monthly mortality rates used are listed in Appendix E.1.4.
4.3 Cattle Herd Breakdown RateAlthough the actual number of CHBs has been rising, when the increased area and number of farms is taken into account (Appendix D) it can clearly be seen that the CHB rate (CHBs per 100 farms per year) in all of the parish test-interval types has been remarkably constant between 2003 and 2005. The annual CHB rate for T1 areas is about 8%, so TB transmission rates were set in the models to give approximately this value once disease and population dynamics had stabilised.
4.4 Sensitivity AnalysesThere is no single standard procedure for testing the sensitivity of a stochastic model.
The CSL used a “One-at a time” (OAAT) sensitivity analysis using logical parameter bounds (see Appendix F). Thus each parameter was changed from its default value to the minimum, or maximum, while the remaining parameters were fixed at the default value.
NU used a Latin Hypercube Sampling (LHS) approach for the NU model sensitivity analysis using the restricted pairing technique of Iman and Conover (1982) to eliminate correlation between input variables. In addition, the calculation of partial correlation coefficients for each input variable takes into account the variance in model results caused by other input variables and calculates the proportion of the variance in the output which is uniquely accounted for by each input variable. A LHS strategy following the methods of Vose (1996, chapter 4) was used to select input parameters for the model from the known or estimated ranges of the different variables in the model. The aim was to provide a range of input values for each variable that could potentially occur under field conditions. In other words the model would be run a sufficiently large number of times to encompass the potential range of conditions that occur naturally rather than simply worst and best case scenarios (sensu Bart, 1995). A uniform distribution was assumed for each variable with upper and lower limits derived from the literature. Variables were also assumed to be independent of each other. This approach will lead to an overestimate of the size of the likely universe of possible values that each life history parameter could take, since firstly, it is likely to lead to the selection of values for parameters that are near the extremes of their distributions more frequently than would be expected in reality. Secondly, the assumption of non-independence between the life history variables will lead to variable pairs being selected in the model that are unlikely to occur in the field (e.g. high mortality and high fecundity). On the other hand it also ensures that all potential values (within the known range of observed behaviours for each variable) are sampled.
28
Pre-movement testing was switched on prior to badger control (in year 100), so its effects could be distinguished from badger control, which started in year 120. The key output parameters were the badger population, the badger TB prevalence, the cattle CHB rate, and the NPV benefits and the percentage of simulations that gave a net financial cost (benefit had negative value).
The ranked results for the sensitivity analyses are tabulated in Appendix G.
Both models were in agreement in terms of the most important parameters, and the sensitivities came out very much as expected from previous analyses. The parameters that the badger population was the most sensitive to were the badger mortality rates, the badger female breeding probabilities, and the within-group TB transmission probabilities. The badger prevalence was most sensitive to badger mortality rates, badger TB transmission rates, and TB disease progression rates. Also, in both models, the cattle CHB rate was most sensitive to badger mortality rates, and cattle test sensitivity for both models. The third in the CHB rate sensitivity ranking for the NU model was cattle TB disease progression, followed by within-herd transmission rates and cattle test-frequency, whereas for the CSL model the third in the ranking was badger within-group transmission rates followed by badger TB disease progression.
4.5 Cross-Model ValidationIn this section the results from the two models are examined against each other. If the two models produce similar (although not identical) results, them more faith can be given to the model output.
Both the CSL and the NU models were cross-validated by setting parameter values to match wherever possible, and comparing the outputs closely with each other, across at least 100 simulations per model during years 119 to 135. Not all TB transmission rates could be exactly matched, as their values had been fine-tuned in each model to give the required mean CHB rate of about 8%. The main outputs compared were badger prevalence and cattle CHB rate, first with pre-movement testing switched off, then with it switched on. Analysis of variance was performed on the output data.
29
Figure 4.5.1 Cross-validation of the models. (a) Mean badger prevalence with pre-movement testing switched off.
Badger Prevalence (No Control) PMToff
0.160
0.165
0.170
0.175
0.180
0.185
0.190
0.195
0.200
119 121 123 125 127 129 131 133 135
Year
Bad
ger P
reva
lenc
e
CSL modelNewcastle model
(b) Mean badger prevalence with pre-movement testing switched on.Badger Prevalence (No Control) PMTon
0.160
0.165
0.170
0.175
0.180
0.185
0.190
0.195
0.200
119 121 123 125 127 129 131 133 135
Year
Bad
ger P
reva
lenc
e
CSL modelNewcastle model
(c) Mean cattle-herd breakdown rate with pre-movement testing switched off.Cattle Herd Breakdowns (No Control) PMToff
0.000
0.020
0.040
0.060
0.080
0.100
0.120
0.140
119 121 123 125 127 129 131 133 135
Year
Mea
n C
attle
Her
d B
reak
dow
ns p
er fa
rm
CSL modelNewcastle model
30
(d) Mean cattle-herd breakdown rate with pre-movement testing switched on.Cattle Herd Breakdowns (No Control) PMTon
0.000
0.020
0.040
0.060
0.080
0.100
0.120
0.140
119 121 123 125 127 129 131 133 135
Year
Mea
n C
attle
Her
d B
reak
dow
ns p
er fa
rm
CSL modelNewcastle model
(e) Mean (± 1 standard deviation) cattle-herd breakdown rate with pre-movement testing switched on.
Cattle Herd Breakdowns (No Control) PMTon
0.000
0.020
0.040
0.060
0.080
0.100
0.120
0.140
119 121 123 125 127 129 131 133 135
Year
Mea
n C
attle
Her
d B
reak
dow
ns p
er fa
rm
CSL meanNewcastle meanCSL SD+CSL SD-NCU SD+NCU SD-
These results clearly show that although the NU model has more variation in badger prevalence through time, there is otherwise little difference between the two models (Figures 4.5.1a & b). There is, however, a statistical difference between the two models when pre-movement testing is switched off, but not when pre-movement testing is switched on.
Figures 4.5.1(c) & (d) show the CHB rates are very close and the two lines of the first cross over several times. However, again, there is a statistical difference between the two models in both cases. Figure 4.5.1(e) shows the overlap of the plots of Figure 4.5.1(d) when the standard deviations are shown. It is necessary to consider what is actually an important difference, rather than what might be a very small but statistically significant difference. It was considered that in all the graphs of Figure 4.5.1, none of the differences were sufficiently large to be important in the context of their application, and the project could proceed without further model modification.
31
4.6 Verification OverviewVerification is a process used to check that the model(s) do what is expected of them. The verification processes described above (4.1 – 4.5) together indicate that the model processes represent well the processes that are happening in the field. The overall similarity between the output of the two models indicates that no significant errors have been incorporated into the programming, thus increasing the likelihood that the models are a useful representation of reality.
5. ValidationIn order for a model to be considered ‘useful’, it has to be validated against field data. There are no generally accepted protocols for model validation, however, the usual approach is similar to hypothesis testing: i.e. the model must not be invalidated (be statistically significantly different) from a variety of field data. The more such test that have been performed, then the more ‘valid’ the model may be considered. The model(s) can be validated against any output data that are an emergent property. For example, badger social group size is not defined or limited directly in the model, but emerges from the combination of fecundity, mortality and emigration. Therefore, social group size can be validated against field data. Similarly, economic costs of herd breakdowns are not input into the model, but emerge from the addition of the individual costs for each animal and the time spent under restriction.
5.1 Badger ValidationVarious aspects of the badger models have been subjected to previous validation (Shirley et al., 2003; Smith, Cheeseman & Clifton-Hadley, 1997; Smith et al., 1995). Thus, we did not perform additional simulations to validate badger population and epidemiological output. The previous work has demonstrated the badger population dynamics simulation is realistic when compared against field data on social group size and population trend over time. Discussion of the validity of the mechanisms of social perturbation in the badger is raised in Section 7.1.5.
5.2 Cattle Validation
5.2.1 Tests and reactorsAs part of the model validation in respect of the cattle component, a distribution of points was plotted, each from a single year of the CSL model simulation, to show the number of reactors found, against the number of cattle TB tests performed (Figure 5.2.1.1). This was compared with the data for 2003, 2004, and 2005. The latter three points fall within the distribution of simulation points.
Figure 5.2.1.1. Number of reactors plotted against number of annual cattle TB tests.Results from the model (diamonds), and the field (large Xs, left to right: 2003, 2004, and 2005).
32
XX XX XX
5.2.2 Reactors per Cattle Herd BreakdownTo further validate the CSL model in respect of the numbers of reactor cattle identified per CHB, a frequency distribution of reactor numbers per breakdown was plotted for the real data supplied by the VLA from the Vetnet database, and compared with a simulated dataset from the model (Figure 5.2.2.1). The simulated distribution was compiled from sufficient model runs to give a smooth distribution without gaps.
The mean number of reactors per CHB calculated from the field dataset was 6.4 and the comparable mean from the model was 5.6. The shapes of the distributions were very similar (Figure 5.2.2.1). These were considered close enough to justify proceeding with the model without the need to alter parameter settings.
33
Figure 5.2.2.1. Frequency distributions of number of reactors per CHB. A comparison of real and simulated data.
(a) Distribution from field data (13,233 confirmed CHB cases) that occurred between January 1996 and June 2006, as recorded on the Vetnet database. Mean number of reactors per CHB = 6.4.
0
500
1000
1500
2000
2500
3000
3500
0 3 6 9 12 15 18 21 24 27 30 33 36 39 42 45 48
Number of confirmed reactors per CHB
Freq
uenc
y
(b) Simulated data from the CSL model. Mean number of reactors per CHB = 5.6.
0
5000
10000
15000
20000
25000
30000
35000
40000
0 3 6 9 12 15 18 21 24 27 30 33 36 39 42 45 48
Number of confirmed reactors per simulated CHB
Freq
uenc
y
5.3 Economic Validation
5.3.1 Costs of TB breakdownsAs part of the CSL model validation in respect of the cost component, a distribution of costs per CHB was plotted against the number of confirmed new incidents (CNIs) (Figure 5.3.1.1). The real value for 2005 was an estimated £23,000, which lies well within the distribution.
34
Figure 5.3.1.1. Scatter distribution of cost of each individual CHB in 2005, plotted against the number of confirmed new incidents (CNIs).
5.3.2 Annual costs of TB breakdownsA distribution of total annual costs of the CHBs simulated by the CSL model was plotted (Figure 5.3.2.1). The median value from the model was about £100M. The real value for 2005 was approximately £90M.
Figure 5.3.2.1. Distribution of the simulated annual cost (£) of all CHBs, multiplied up from the modelled grid area to the whole T1 and T2 areas.
0
45
0 100,000,000 200,000,000 300,000,000 400,000,000
5.4 RBCT ValidationThe NU model was validated against the proactive areas of the RBCT by using a GIS-based map of the proactive RBCT triplets as a substrate for the model. Badger territories were randomly located within each triplet centred around points placed in woodland habitat. The pastureland was divided up into randomly generated farms at the appropriate density; and compliance was determined by the compliant and non-compliant areas indicated in the data for each triplet.
35
The length of the proactive period varied between triplets; from 36 months (triplets I and J) to 83 months (triplet B). The average length of a proactive trial across all triplets was 57 months; consequently the model simulated a five year span of culling, and the number of simulated and observed cattle herd breakdowns was divided by the number of years of the trial to produce comparable results (Figure 5.4.1).
Figure 5.4.1. RBCT validation of the NU model.The post-cull cattle herd breakdowns per herd per year (y-axis) plotted against the observed rate for the proactive triplets. The straight line represents exact parity between the model and the field data. The upper and lower quartiles of the modelled scenarios are indicated.
Rate of CHB per herd per year
AB
CD
E
F
H
IJ
G
0
0.05
0.1
0.15
0.2
0 0.05 0.1 0.15 0.2
field
med
ian
sim
ulat
ed
Each simulated trial consisted of 50 replicates. The number of herds simulated in each trial was randomly determined according to the distribution described above; although using the same number of baseline herds for each simulation would have provided smaller estimates of error; the results would have had less general applicability in interpretation.
For each simulation and the observed field data, a post-cull cattle herd breakdown rate was calculated as the number of breakdowns per baseline herd per year in each triplet. It was not sensible to compare the field aspects of the survey-only areas, since these will have stochastic elements, which will not be exactly simulated within the model: i.e. some survey-only areas will have a higher than ‘expected’ herd breakdown rate, thus reducing the comparability between the RBCT and the simulation model. It is therefore, more suitable to compare pre-and post-cull herd breakdown rates for each of the proactive areas.
In general the range of simulated RBCT proactive triplets fell close to the line that indicates a perfect match between the field trial and the simulated data. The model showed no consistent tendency to over-predict or under-predict the rate of CHB, suggesting that the variations are due to unknown temporal and/or spatial heterogeneity that are not
36
represented in the model. Table 5.4.1 indicates a goodness-of-fit for the model based on the rank of the validation data set within the simulated data. Four of the triplets (A, B, F, H) fall within the interquartile range of the simulated data (five would be expected to do so), and none display results that were outside the simulated maxima (rank 51) or minima (rank 1). Trials F & G falls outside the range that encompasses 95% of the data, which would indicate the presence of a significant result in a significance test. Triplets D, I and J were most strongly under-predicted by the model and were also the triplets which joined the trial immediately after the 2001 FMD outbreak and the associated lack of cattle testing. Thus, these trial areas had a greater than usual herd breakdown rate in 2002, their first year in the trial. Since the model did not simulate the effect of the 2001 FMD epidemic on herd testing (i.e. cessation of testing), we would expect these areas to be under-predicted since cattle herds would have been carrying TB undetected during this period. Had FMD not occurred in 2001, these triplets would have had a lower field breakdown rate and thus the points would move to the left, bringing them closer to the line.
Even without such a correction, the distribution of the triplets is not statistically different from the simulated results, thus the model cannot be invalidated against the RBCT herd breakdown data.
Table 5.4.1 A comparison of the RBCT CHB rate with simulated data.The rank of the CHB rate from the field data and the 50 simulated CHB rates. A rank of 26 is by definition the median of the simulated data; a rank less than the median is under-predicted by the model and a rank greater than the median is over-predicted by the model. The interquartile range encapsulates 50% of the simulated data.
Triplet Rank field data
median simulation
interquartile rangeof simulations
95% boundsof simulations
A 28 0.112 0.117 0.073 - 0.189 0.036 - 0.329B 34 0.101 0.118 0.082 - 0.145 0.067 - 0.201C 38 0.058 0.075 0.059 - 0.105 0.030 - 0.177D 12 0.153 0.090 0.063 - 0.138 0.021 - 0.240E 45 0.086 0.114 0.096 - 0.130 0.069 - 0.152F 50 0.026 0.059 0.047 - 0.084 0.028 - 0.152G 48 0.076 0.120 0.100 - 0.153 0.078 - 0.261H 24 0.128 0.116 0.093 - 0.185 0.070 - 0.385I 12 0.122 0.076 0.062 - 0.110 0.036 - 0.371J 8 0.111 0.056 0.039 - 0.080 0.021 - 0.221
6. Methods
6.1 Control and Management Options TestedFollowing model verification and validation, a number of control and management options were identified with discussion with Defra, for generating the main model output:
6.1.1 Default settings and assumptionsA full list of assumptions is detailed in Appendix K.Key model settings and assumptions are outlined below:
37
a) Pre-Movement Testing switched on (2 levels: >16 months for one year, and thereafter > 2 months [to match as closely as possible the 15 months and 6 weeks, since the model had a 2 month time-step])
b) Compliance set to 70% (with non-compliant farms randomly distributed)
c) Compliance for vaccination was 100%
d) Social perturbation was switched on (i.e. extra perturbation from the start of culling to three years after culling ends). This is in contrast to the Phase I simulations where social perturbation was added to a single example, and was simulated more simply
e) All control was carried out by farmers for five consecutive years and costs estimated appropriately
f) All healthy badgers that eat vaccine bait were fully protected against TB
g) 100 model simulations were used for each scenario
h) Outputs included badger population, badger TB prevalence, cattle-herd-breakdown rate, economic NPV and cost comparisons. All output means were calculated over the whole grid area, not just for the control area.
6.1.2 Main ScenariosFollowing discussions, a variety of different control areas, control rates and compliance were investigated (Table 6.1.2). These were combined into seven scenarios. One hundred simulations of each scenario were run, and the mean values calculated across these simulations. For each scenario listed below, a “no-control” option was run in each simulation, before any control options.
38
Table 6.1.2. Details of the 7 badger control scenarios
ScenarioControl
Area (km2)Control Rates
(%)Compliance
(%)
1 100 Shooting 50Trapping 70Snaring 80Gassing 80Vaccine 60
70
2 300 Shooting 50Trapping 70Snaring 80Gassing 80Vaccine 60
70
3 400 Shooting 50Trapping 70Snaring 80Gassing 80Vaccine 60
70
4 300 Shooting 50Trapping 70Snaring 80Gassing 80Vaccine 60
60
5 300 Shooting 50Trapping 70Snaring 80Gassing 80Vaccine 60
80
6 300 Shooting 30Trapping 60Snaring 70Gassing 70Vaccine 40
70
7 300 Shooting 70Trapping 80Snaring 90Gassing 90Vaccine 80
70
39
7. Results
7.1 Sensitivity Analysis
7.1.1 Cattle skin test sensitivityThe sensitivity of the CSL model to the cattle TB skin test was examined in detail, since this was by far the most important parameter that could be easily manipulated. It was also clear that the lower bound was ranked much higher in sensitivity of CHB rate than the upper bound (Appendix G, table G.3). When the CHB rate was plotted across time it is obvious that the upper bound is virtually identical to the default value used (Figure 7.1.1.1). In other words, the CHB rate is insensitive to increasing the sensitivity of the TB skin test, but very sensitive to decreasing it. Halving the test sensitivity approximately doubled the CHB rate.
Figure 7.1.1.1 Cattle skin test sensitivity.Mean CHB rate per farm with three different levels of the cattle TB skin test: lower, default, and upper. The lower bounds of the test sensitivity were set to half the default values (see Appendix F).
Cattle Herd Breakdowns and Skin-Test Sensitivity
0.00
0.05
0.10
0.15
0 2 4 6 8 10Years
Mea
n C
attle
Her
d B
reak
dow
ns p
er fa
rm
Default skin test sensitivityUpper-bounds of skin test sensitivityLower-bounds of skin test sensitivity
7.1.2 Cattle test-intervalFor the first 49 years all the simulated parishes used annual cattle testing. From year 50 onwards each parish was assessed annually, and parishes with a low enough CHB rate could be reclassified to a longer test interval (T1 to T2, T2 to T3, T3 to T4), and vice versa. The algorithm used to reclassify parishes is detailed in Appendix I.
It should be noted that the algorithm used to change a parish from one type to another, which is the same as the one used in the field, increased the CHB rate in the model (Figure 7.1.2.1 – change at year 50). The size of that increase was dependent on the particular disease transmission rates being modelled.
40
Figure 7.1.2.1 Effect of changing parish test-interval dependent on CHB rate.The algorithm was instigated in year 50. Two different sets of transmission rates were tested – one that gave rise to a high badger contribution to CHBs (~60%), and one that gave rise to a low badger contribution (~40%). Pre-movement testing was switched off.
Cattle Herd Breakdowns and Badger Contribution
0.00
0.05
0.10
0.15
20 40 60 80 100Years
Mea
n C
attle
Her
d B
reak
dow
ns p
er fa
rm
High Badger Contribution (~60%)Low Badger Contribution (~40%)
In the main sensitivity analysis, the CSL model was surprisingly insensitive to the cattle-testing interval, but it was thought that since pre-movement testing was then switched on by default, that might be an important factor. Further analysis with the CSL model revealed this was the case. Figures 7.1.2.2(a) & (b) show that, as expected, the badger parameters were completely insensitive to the cattle test-interval. Figure 7.1.2.2(c) shows that the CHB rate was very insensitive to the test-interval when pre-movement testing is in effect, but if pre-movement testing is switched off, the CHB rate is much more sensitive. It would appear that the pre-movement testing could compensate for less frequent whole-herd regular testing. The intersection of the two lines at about 2 months suggests that the pre-movement testing results in disease control equivalent to a very short test interval. Figure 7.1.2.2(d) shows the true TB prevalence level in the model, which does show an increase as the test interval is increased. This indicates that some of the insensitivity of the CHB rate to test-interval with pre-movement testing switched off is due to more TB cases being missed by pre-movement testing, rather than there being no room for further improvement.
41
Figure 7.1.2.2 Effect of changing the routine cattle-test-interval.Pre-movement testing is switched off (triangles), or on (circles). Values are presented for the final 10 years of the simulations to ensure that the population and disease dynamics had stabilised.(a) badger social group size
7.0
7.2
7.4
7.6
7.8
8.0
0 5 10 15 20 25 30 35Cattle Test Interval (months)
Mea
n ba
dger
gro
up s
ize
PMToffPMTon
(b) badger TB prevalence
0.10
0.12
0.14
0.16
0.18
0.20
0 5 10 15 20 25 30 35Cattle Test Interval (months)
Mea
n ba
dger
TB
pre
vale
nce
PMToffPMTon
(c) cattle herd breakdown rate per farm
0.00
0.02
0.04
0.06
0.08
0.10
0.12
0 5 10 15 20 25 30 35
Cattle Test Interval (months)
Mea
n C
attle
Her
d B
reak
dow
ns p
er fa
rm
PMToffPMTon
42
(d) cattle TB prevalence
0.000
0.004
0.008
0.012
0.016
0 5 10 15 20 25 30 35
Cattle Test Interval (months)
Mea
n ca
ttle
TB p
reva
lenc
e
PMToffPMTon
7.1.3 Disease transmission ratesThe badger-to-cattle and the cattle-to-cattle transmission rates of TB can be adjusted in the model to give similar CHB rates as seen in the field. However, this can be done either by setting badger-to-cattle rates higher and cattle-to-cattle rates lower, giving a higher badger contribution to CHBs (e.g. about 60%) or vice versa, giving a lower badger contribution to CHBs (e.g. about 40%) (see Figure 7.1.3.1.a). An important effect seen when further analysed with the CSL model is that the effectiveness of the pre-movement testing (PrMT) regime is dependent on the level of badger contribution – the higher the badger contribution the less effective is the PrMT regime (Figure 7.1.3.1.b). We will assume that the badger contribution is around 40% for all further simulations. It should be noted that these badger contributions are those measured before PrMT is applied. Once PrMT is applied the badger contribution increases from around 40% to around 60%, because a significant proportion of the cattle-to-cattle transmission is prevented once PrMT starts.
Figure 7.1.3.1 Badger contribution to the CHB rate.(a) The effect of changing disease transmission rates on the CHB rate and the level of badger contribution to those rates. Pre-movement testing is applied from year 100.
Cattle Herd Breakdowns and transmission rates
0.00
0.05
0.10
0.15
80 90 100 110 120 130Years
Mea
n C
attle
Her
d Br
eakd
owns
per
farm
High Badger Contribution (~60%)Low Badger Contribution (~40%)
43
(b) The effect of different badger contributions to CHBs, on the effectiveness of pre-movement testing (PrMT). PrMT is first applied in year 120. Approximate badger contributions quoted are those before PrMT is applied.
Cattle Herd Breakdowns and Badger Contributions
0.00
0.05
0.10
0.15
100 105 110 115 120 125 130 135Years
Mea
n C
attle
Her
d B
reak
dow
ns p
er F
arm
No Control, High badger contribution (~60%)PrMT, High badger contribution (~60%)No Control, Low badger contribution (~40%)PrMT, Low badger contribution (~40%)
7.1.4 Pre-movement testing – sensitivity to scaleThe effects of PrMT on two different scales were compared using the CSL model. The first was as used in the field – namely being applied to parishes of types T1 and T2 (annual and two-yearly routine cattle TB testing), and the second was simulated for all parish types (T1-4) simulated in the grid. It is important to note, however, that the T2, T3, and T4 areas simulated in the model (see section 7.1.2 and Appendix I) were those that switched from T1 parishes, and would not represent the actual proportions seen in all of England/Wales. A range of different transmission settings was tested for each of these two scenarios (Figure 7.1.4.1).
Figure 7.1.4.1 Effects of extending the area over which PrMT is applied.PrMT is first applied in year 100. Each individual line is from using a particular set of transmission rates, and with PrMT either just for T1 and T2, or for all areas.
Cattle Herd Breakdowns and pre-movement testing
0.00
0.05
0.10
0.15
80 90 100 110 120 130Years
Mea
n C
attle
Her
d B
reak
dow
ns p
er fa
rm
PMT for T3 & T4 areas as well
PMT for T1 & T2 areas only
44
These results show that applying PrMT to all areas could significantly reduce the CHB rate further. The variation from the different transmission rates is relatively very small compared with the effect of extending the PrMT. The cost effectiveness of extending the PrMT to all areas was not studied in this project.
7.1.5 PerturbationAs part of model validation, in particular related to the spatial effects of perturbation, the models were modified to give extra output, matching spatial metrics analysed from the RBCT. Different zones of the RBCT proactive trapping areas and their surroundings were defined as follows:
Core area – the area within a boundary drawn 2km inside the edge of the control area
Inner layer – the area between the core area and the outer edge of the control area
Outer layer – the area between the outer edge of the control area and a boundary drawn 2km outside that.
The core area and the inner layer together comprise the whole control area. Where a badger territory or a cattle grazing area overlaps more than one zone, the metric value is divided proportionally, based on the areas within each zone.
The CSL model with initial settings showed a reduction in CHB rates following trapping (compared with no-control) for all three layers (Table 7.1.5.1). The level of reduction increased as one moved from outer to inner to the core. The trend was similar for all the methods of culling, and the level of reduction was also dependent on the cull rate. This indicated that any effect of social perturbation, as simulated in Phase I, was not sufficient to produce an increase in CHB in the outer layer.
Table 7.1.5.1. Change in simulated CHB rate following control. The percentage change in CHB rate following control over 100km2 (compared with no-control) for the three zones (as defined in 7.1.5). The original default perturbation processes were used, as in Phase I. (The trap option is highlighted as this one should be compared with the RBCT data shown in table 7.1.5.2)
Outer(2km ring
no-control
Inner(2km ring
of control)
Core (control)
Shoot -3 -11 -21Trap -9 -18 -37Snare -13 -27 -45Gas -13 -22 -45Vaccinate -0 -4 -9
Several processes were adjusted, one-at-a-time, to ascertain which would give similar spatial effects of perturbation as seen in the RBCT. These included (a) increasing the transmission rates even further when perturbation was simulated, plus raising the within-group badger-to-badger transmission rates in addition to the between-group rates, (b)
45
preventing or limiting the number of moves from larger to smaller groups, (c) allowing badgers that roam further due to perturbation to infect cattle further afield than normal, as well as other badgers, and (d) extending the perturbation effect spatially to two rings of social groups neighbouring the cull area, rather than just one, i.e. the perturbation continues for approximately 2km outside the culled area.
The first three (increasing transmission rates, limiting movement, and increasing distance of badger-to-cattle transmission) all resulted in an increase in the CHB rates of the outer and inner zones in all culling scenarios. Increasing the transmission rates even resulted in an increase in the core CHB rates for culling by shooting or trapping, and option (c) increased the CHB rates in the core zone for all types of culling.
Option (d), extending the perturbation effect to two rings of badger groups (approximately 2km beyond the edge of the cull area), was the only option that gave spatial trends similar to the RBCT, specifically with an increase in the outer ring and the trend of increasing effectiveness of culling closer to the centre (Table 7.1.5.2). Due to the extended ring of badger perturbation, we ensured that the grid size was not affecting the disease dynamics by repeating the simulation on a larger grid.
Table 7.1.5.2. Change in simulated CHB rate following control. The percentage change in the simulated CHB rate following control (compared with no-control) for the three zones (as defined in 7.1.5). The extra ring of perturbation was implemented, but in all other respects the default processes were used. The last two rows of RBCT field data are taken from Donnelly et al (2007), and should be compared with the “Trap” option of the model, and also show (in brackets) the 95% confidence interval values.
Outer
(2km ring no-control)
Inner(2km ring of control)
Core (control)
Shoot 28 2 -9Trap 20 -9 -33Snare 13 -18 -38Gas 14 -15 -37Vaccinate 1 -10 1RBCT (actual field) based on VetNet location data
24.5(-0.6, 56.0)
-23.2(-32.7, -12.4)
RBCT (actual field) based on RBCT location data
35.3(5.8, 73.0)
-17.4(-27.2, -6.2)
The double ring of badger social groups (approximately 2km beyond the edge of the cull area) with extended movement caused by perturbation in the CSL model was able to reproduce the spatial trends seen in the RBCT. All further model simulations were therefore done with this double ring of perturbation being implemented (option d).
46
For the NU model, badger social groups were assigned to one of four layers (see Figure 7.1.5.1) based on the location of their centroid, and all ten RBCT proactive areas were simulated. The average prevalence of TB in badger social groups in each of the four layers was as follows (pooled over all 10 triplets):
Zone Prevalence of TB in badgers
Percentage difference to 'No control' zone
No control 0.170 n/aOuter Layer 0.200 +17.5%Inner Layer 0.162 –5.0%Core Area 0.126 –26.3%
The 2km buffer around the border of the RBCT areas is approximately equal to the width of two badger social groups. The implication is that control is less effective at reducing TB prevalence in the clans around the edge of the control zone due to the effects of social perturbation; as individuals emigrate from the social groups neighbouring the control zones into the territory left vacant by the cull. This increased movement results in a decreased effect of the cull in the inner layer, and an increased TB prevalence in badgers in the outer layer since the latter do not also suffer the effects of the cull itself.
Figure 7.1.5.1 Illustration of the spatial effect of culling in the NU model.The black line indicates the boundary of a proactive square (in this case, from triplet E), and the blue lines indicate the inner and outer layers generated by a 2 km buffer. The underlying map presents the landscape prevalence of TB in badgers; with red indicating the highest prevalence (which in this simulation is 0.60) and pale yellow, zero prevalence.
47
The method used in the RBCT was to compare the mean CHB rate in each of the above areas, with the equivalent area in the no-control option, calculating a percentage difference from the no-control equivalent. This was done for the period of proactive control. Over that period, the RBCT showed a reduction of CHBs in the cull area of around 20%, but a significant increase of about 30% in the 2km ring immediately outside the cull area (Donnelly et al., 2007); though the variation and confidence intervals were large (see Table 7.1.5.2). The closer to the centre of the cull area, the greater appeared to be the effectiveness of the culling in reducing CHB rates (Donnelly et al., 2007). The same zones and period were compared using the models.
7.2 Population Management
7.2.1 Badger populationBadger control and compliance (land access)It was demonstrated in the RBCT (Donnelly et al., 2007) that if only a proportion of a badgers’ territory can be accessed for control, the whole of that badger group can be successfully controlled (“drawing-out”). The CSL model was used to check what proportion of badger groups that had been marked for requiring control could actually be controlled. The assumption made was that if at least 10 % of a badger territory lay on land for which permission was given, the badgers of that territory could be caught as if all the territory was accessible. This would assume that badgers frequently move round the edge of their territory, and is thus probably not valid for gassing, since access to the main sett is required. Nonetheless, we have not changed this routine by control method to retain consistency of results. Any territory with less than 10% accessible area was assumed to be completely inaccessible for control. Based on this assumption a range of different compliances was tested, using the default spatial characteristics of the CSL model (tessellated badger territories, farms and parishes). Figure 7.2.1.1 indicates that with a farm compliance of 70% (the default value for the models) 94% of the badger groups marked for control could be culled.
Figure 7.2.1.1. The percentage of badger groups subjected to culling for different rates of farm compliance.This assumes that if 10% or more of a badger territory is accessible, it is equivalent to full access.
0
20
40
60
80
100
0 20 40 60 80 100
Land Access (Farm Compliance)
Perc
enta
ge o
f tar
get b
adge
r gro
ups
culle
d by
"D
raw
ing-
Out
"
70% compliance
94% of badger groups culled
Assuming a group can be culled if at least 10% lies on accessable land
48
Badger population response to cullingThe CSL model output for badger population, represented by mean badger group size, kept stable at 7.5 adult badgers per social group when no badger culling was being applied (Figure 7.2.1.2). The no-control and vaccination options gave virtually identical results for badger group size. The badger population decreased in the culling scenarios during the five control years, and then recovered during the following 10 years. The amount of population depression was proportionate to the effectiveness of control: shooting at a rate of 50% reduced mean group size down to a minimum of 5.3 adult badgers, trapping at 70% down to about 4.8 badgers, and the snaring and gassing options (both at 80% control rate) down to about 4.6 adult badgers.
In the CSL model the other main scenarios (changing compliance, control area, and control rates) gave very similar results to scenario 1 in terms of badger population. The main differences were that when the shooting rate was reduced to 30% (scenario 6) or increased to 70% (scenario 7), the minimum badger population after five years of control shifted to 5.8 or 4.7 adult badgers respectively.
Figure 7.2.1.2 Scenario 1: badger social group size.The effect of different badger control options on the mean badger group size (adults). Badger control starts in year 120 and finishes in year 124. The modelled control area was 100km2 and the land compliance was 70% (CSL model).
Badger Population
0
2
4
6
8
10
115 120 125 130 135Years
Mea
n B
adge
r Gro
up S
ize
No ControlCull 50%Cull 70%Cull 80%Vaccinate 60%
7.2.2 Prevalence of TB in badgersThe models’ output for badger TB prevalence kept stable at about 18% when no badger culling was being applied (Figure 7.2.2.1). The badger control options showed similar trends to each other, with an initial rise in prevalence in year 120, due to the perturbation effect of the culling. This perturbation effect over-rode any suppression that might have occurred due to culling, and the badger prevalence continued to rise to a peak in year 127, and then declined once the perturbation effect had been switched off in the model. The degree of the effects were in proportion to the cull rates, the lower cull rate of 50% for the shooting giving the highest rise in prevalence throughout the whole period. The NU model
49
showed very similar trends but with a slightly lower rise in prevalence and a slightly quicker recovery to background levels once culling ceased. Vaccination gave a sustained reduction to about 13% prevalence in the badgers, as there was no cull-induced perturbation effect with the vaccine option.
Figure 7.2.2.1 Scenario 1: badger TB prevalence.Badger control starts in year 120 and finishes in year 124. The perturbation effect starts in year 120 and finishes in year 127. The modelled control area was 100km2 and the land compliance was 70%.
(a) CSL Model:
0.0
0.2
0.4
0.6
0.8
1.0
115 120 125 130 135Years
Badg
er T
B pr
eval
ence
No ControlCull 50%Cull 70%Cull 80%Vaccinate 60%
(b) NU Model:
0
0.2
0.4
0.6
0.8
1
115 120 125 130 135Years
Badg
er T
B pr
eval
ence
No ControlCull 50%Cull 70%Cull 80%
An example of the output produced by the NU model is shown in Figure 7.2.2.2. This landscape shows the field boundaries and badger prevalence. The infected badger social groups can be clearly identified, and the portion of the landscape 'contaminated' by their presence can cross several field boundaries.
50
Figure 7.2.2.2 Example output from the NU model.This is a 20km x 20km landscape; the black lines delineate simulated field boundaries overlaid on a map of TB prevalence amongst badgers in the landscape. Red indicates the highest prevalence of TB, fading down to the pale yellow indicating zero TB.
7.2.3 Number of TB-infected badgersThe actual number of infected badgers remaining after a cull is a better measure of risk to cattle than prevalence. The model output for the mean number of infected badgers per social group kept stable at about 1.3 when no badger culling was being applied (Figure 7.2.3.1). The patterns for infected badgers followed a similar trend to badger prevalence. Again, vaccination gave a sustained reduction of the number of infected badgers per social group.
Figure 7.2.3.1 Scenario 1: Number of infected badgers.Effects of badger control options on the number of infected badgers. Badger control starts in year 120 and finishes in year 124. The perturbation effect starts in year 120 and finishes in year 127. The simulated control area was 100km2 and the land compliance was 70% (CSL model).
0.0
1.0
2.0
3.0
4.0
5.0
115 120 125 130 135Years
Mea
n TB
-infe
cted
bad
gers
per
gro
up
No ControlCull 50%Cull 70%Cull 80%Vaccinate 60%
51
7.2.4 Parish typesThe mean proportions of the different parish types varied with time once the switching algorithm was turned on (Figure 7.2.4.1), before stabilising. Frequencies changed again after pre-movement testing (PrMT) was implemented in year 100. PrMT had the overall effect of decreasing the numbers of T1 and increasing the numbers of T4. It can also been seen that there was some degree of variation after testing changed, before the proportion of parishes stabilised.
Figure 7.2.4.1 Scenario 1: Parish test frequency.Relative proportions of the four parish types in the no-control option. All parishes were annually tested (T1), until the switching algorithm was implemented in year 50. Pre-movement testing was implemented in year 100 (CSL model).
Test Interval Profile
0
5
10
15
20
25
30
40 60 80 100 120Years
Num
bers
of p
aris
hes
in th
e Te
st In
terv
al C
ateg
ory
T1T2T3T4
7.2.5 Cattle herd prevalenceThe model output for cattle-herd prevalence (Figure 7.2.5.1) exhibited more background variation than badger prevalence. Without control, the prevalence was about 7% of cattle herds. All the badger-cull methods resulted in an increase in herd prevalence. This increase was due to the perturbation effect in the badgers completely over-riding any culling advantage, and the increased weight of badger infection transmitting over to cattle herds. Once the perturbation effect was switched off in year 127, the cattle prevalence dropped, but had not reached the pre-cull levels by the end of the simulation (year 135). With the vaccine option the herd prevalence slowly reduced to about 6.5%.
52
Figure 7.2.5.1 Scenario 1: Cattle herd prevalence.Effect of badger control options on cattle-herd prevalence. Badger control starts in year 120 and finishes in year 124. The perturbation effect starts in year 120 and finishes in year 127. The modelled control area was 100km2 and the land compliance was 70% (CSL model).
0.00
0.05
0.10
0.15
115 120 125 130 135Years
Cattl
e He
rd T
B Pr
eval
ence
No ControlCull 50%Cull 70%Cull 80%Vaccinate 60%
7.2.6 Cattle herd breakdown rateThe model output for cattle herd breakdowns shows a very similar trend to herd prevalence (Figure 7.2.6.1). Without badger control, the mean CHB rate was about 6.3% of cattle herds, pre-movement testing having already caused some reduction. Badger culling caused an increase in the CHB rate when measured across the whole simulation grid, with the greatest increase being realised by the lowest cull rate. For the CHB rate, as for the herd prevalence, the perturbation effect over-rode the culling effect, and the rates had not returned to pre-cull rates eight years after the perturbation effect was switched off in the model. Very similar results were seen in the NU model.
53
Figure 7.2.6.1 Scenario 1: CHB rate.Effect of different badger control options on the cattle herd breakdown rate. Badger control starts in year 120 and finishes in year 124. The perturbation effect starts in year 120 and finishes in year 127. The modelled control area was 100km2 and the land compliance was 70%.(a) CSL Model:
0.00
0.05
0.10
0.15
115 120 125 130 135Years
Mea
n C
attle
Her
d B
reak
dow
ns p
er fa
rm
No Control
Cull 50%Cull 70%
Cull 80%Vaccinate 60%
(b) NU Model:
0
0.05
0.1
0.15
115 120 125 130 135Years
Mea
n C
attle
Her
d B
reak
dow
ns p
er fa
rm No ControlCull 50%Cull 70%Cull 80%
Vaccination resulted in a decrease in the CHB rate (Figure 7.2.6.1a), but the effect is relatively small. However, because the vaccine option does not cause the perturbation effect, the reduction to around 5.5% is maintained throughout the remainder of the simulation period, even though vaccination is stopped after year 124.
7.2.7 Varying control areaThe model outputs for different areas of culling were compared. The grid size was scaled up appropriately to accommodate the larger control areas. As expected, neither the grid size nor the control area affected the mean badger group size (Table 7.2.7.1). The reduction in mean group size was dependent on the cull rate, and the percentage of the grid being subject to badger culling.
54
Table 7.2.7.1 Varying control area: Badger social group size.Effect of control area on the mean badger group size. The means were calculated over ten years starting from the first year of culling (CSL model).
Mean badger group size
ScenarioControl
Area (km2)Grid Size
(km2)No
ControlShoot (50%)
Trap (70%)
Snare (80%)
Gas (80%)
Vaccinate (60%)
1 100 400 7.5 5.8 5.5 5.3 5.3 7.5
2 300 1024 7.5 5.7 5.3 5.1 5.1 7.5
3 400 1600 7.5 6.0 5.6 5.4 5.4 7.5
The greater the size of the control area, the lesser was the increase in the mean number of infected badgers per group (Table 7.2.7.2). All the culling scenarios increased the numbers of infected badgers due to the perturbation effect. This increase is less for larger cull areas because the ratio of perturbed edge area to cull area was smaller for the larger cull areas. The cull rates also affected the numbers of infected badgers, the high cull rates (such as snaring or gassing) counteracting the perturbation effect better than the lower cull-rate options. Vaccination, due to lack of perturbation, showed very similar reductions for all the control areas.
Table 7.2.7.2 Varying control area: Number of infected badgers.Effect of control area on the mean number of TB infected badgers per group. The means were calculated over ten years starting from the first year of culling (CSL model).
Mean number of infected badgers per group
ScenarioControl
Area (km2)Grid Size
(km2)No
ControlShoot (50%)
Trap (70%)
Snare (80%)
Gas (80%)
Vaccinate (60%)
1 100 400 1.3 3.2 2.8 2.6 2.6 1.1
2 300 1024 1.3 2.5 2.1 2.0 2.0 1.0
3 400 1600 1.3 2.2 1.9 1.8 1.8 1.1
Control area size had a very similar effect on badger prevalence (Table 7.2.7.3). The greater the control area, the smaller the increase in prevalence. Again, there was no such effect for vaccination, which reduced the prevalence rather than increased it.
55
Table 7.2.7.3 Varying control area: Badger prevalence.Effect of control area on the badger prevalence, calculated as means over ten years starting from the first year of culling (CSL model).
Mean badger prevalence
ScenarioControl
Area (km2)
Grid Size (km2)
No Control
Shoot (50%)
Trap (70%)
Snare (80%)
Gas (80%)
Vaccinate (60%)
1 100 400 0.18 0.55 0.51 0.49 0.50 0.14
2 300 1024 0.17 0.44 0.40 0.39 0.39 0.14
3 400 1600 0.17 0.38 0.34 0.33 0.33 0.14
Applying badger control over a wider area had the effect of increasing the core area within each simulated badger control area. Since the core area has the biggest change in badger prevalence due to control, increasing the size of this area causes this greater effect of control on more badger social groups. Further, the relative size of the inner zone is smaller, so the effects of social perturbation are felt less strongly. This is clearly seen in the graphical results from the NU model (Figure 7.2.7.1).
Figure 7.2.7.1 Varying control area: Badger prevalence.Results for different rates of control and different areas (NU model).
(a) 100km2
0
0.2
0.4
0.6
0.8
1
115 120 125 130 135Years
Badg
er T
B pr
eval
ence
No ControlCull 50%Cull 70%Cull 80%
56
(b) 300km2
0
0.2
0.4
0.6
0.8
1
115 120 125 130 135
Years
Bad
ger T
B pr
eval
ence
No ControlCull 50%Cull 70%Cull 80%
(c) 400km2
0
0.2
0.4
0.6
0.8
1
115 120 125 130 135Years
Badg
er T
B p
reva
lenc
e
No ControlCull 50%Cull 70%Cull 80%
The size of the control area also affected the CHB rate in the same way as it affected the prevalence and numbers of infected badgers (Table 7.2.7.4 and Figure 7.2.7.2). Vaccination was not affected by control area, but note that the decrease in CHBs due to vaccination is proportionally much smaller than the changes seen for badger prevalence or number of infected badgers.
57
Table 7.2.7.4 Varying control area: Mean CHB rate.Effect of control area on mean CHB rate per farm. Means were calculated over ten years starting from the first year of culling (CSL model).
Mean number of CHBs per farm
ScenarioControl
Area (km2)Grid Size
(km2)No
ControlShoot (50%)
Trap (70%)
Snare (80%)
Gas (80%)
Vaccinate (60%)
1 100 400 0.063 0.098 0.092 0.087 0.089 0.058
2 300 1024 0.063 0.086 0.078 0.074 0.075 0.056
3 400 1600 0.063 0.080 0.073 0.071 0.071 0.058
Figure 7.2.7.2 Varying control area: Mean CHB rate.Results for different rates of control and different areas (NU model).
(a) 100km2
0
0.05
0.1
0.15
115 120 125 130 135Years
Mea
n C
attle
Her
d B
reak
dow
ns p
er fa
rm
No ControlCull 50%Cull 70%Cull 80%
(b) 300km2
0
0.05
0.1
0.15
115 120 125 130 135Years
Mea
n C
attle
Her
d B
reak
dow
ns p
er y
ear
No ControlCull 50%Cull 70%Cull 80%
58
(c) 400km2
0
0.05
0.1
0.15
115 120 125 130 135Years
Mea
n C
attle
Her
d B
reak
dow
ns p
er fa
rm No ControlCull 50%Cull 70%Cull 80%
7.2.8 Varying farm complianceThe model outputs for different levels of farm compliance were compared. The default control area of 300km2 was used for all these three scenarios. Compliance did not have any notable effects on the badger mean group size (Table 7.2.8.1). As expected, no differences were seen between the three scenarios for the no-control and vaccination option. All cull options gave a significant reduction in mean group size, the cull rate determining the size of that reduction. Over the ten years, farm compliance did not have an effect on the number of infected badgers (Table 7.2.8.2) or prevalence (Table 7.2.8.3). It was the cull rate and social perturbation that influenced these results. Vaccination, with its absence of perturbation, gave a decrease in both measures.
Table 7.2.8.1 Varying compliance: Badger social group size.Effect of farm compliance on mean badger group size. Means were calculated over ten years starting from the first year of culling (CSL model).
Mean badger group size
ScenarioFarm
ComplianceNo
ControlShoot (50%)
Trap (70%)
Snare (80%)
Gas (80%)
Vaccinate (60%)
4 60% 7.5 5.8 5.4 5.2 5.2 7.6
2 70% 7.5 5.7 5.3 5.1 5.1 7.5
5 80% 7.5 5.7 5.2 5.0 5.1 7.5
59
Table 7.2.8.2 Varying compliance: Number of infected badgers.Effect of farm compliance on mean number of TB infected badgers per group. Means were calculated over ten years starting from the first year of culling (CSL model).
Mean number of infected badgers per group
ScenarioFarm
ComplianceNo
ControlShoot (50%)
Trap (70%)
Snare (80%)
Gas (80%)
Vaccinate (60%)
4 60% 1.3 2.6 2.2 2.0 2.0 1.0
2 70% 1.3 2.5 2.1 2.0 2.0 1.0
5 80% 1.3 2.5 2.1 1.9 1.9 1.0
Table 7.2.8.3 Varying compliance: Badger prevalence.Effect of farm compliance on badger prevalence. Means were calculated over ten years starting from the first year of culling (CSL model).
Mean badger prevalence
ScenarioFarm
ComplianceNo
ControlShoot (50%)
Trap (70%)
Snare (80%)
Gas (80%)
Vaccinate (60%)
4 60% 0.17 0.45 0.41 0.40 0.39 0.13
2 70% 0.17 0.44 0.40 0.39 0.39 0.14
5 80% 0.17 0.44 0.40 0.38 0.38 0.14
Increasing compliance reduced the peak in TB prevalence in badgers caused by social perturbation for all three simulated control methods (Figure 7.2.8.1), with higher levels of compliance resulting in less difference between the methods simulated.
Figure 7.2.8.1 Varying compliance: Badger prevalence.Results for different levels of farmer compliance and rate of control (NU model).
(a) 60% farmer compliance
0
0.2
0.4
0.6
0.8
1
115 120 125 130 135Years
Bad
ger T
B p
reva
lenc
e
No ControlCull 50%Cull 70%Cull 80%
60
(b) 70% farmer compliance
0
0.2
0.4
0.6
0.8
1
115 120 125 130 135
Years
Badg
er T
B p
reva
lenc
e No ControlCull 50%Cull 70%Cull 80%
(c) 80% farmer compliance
0
0.2
0.4
0.6
0.8
1
115 120 125 130 135Years
Badg
er T
B p
reva
lenc
e
No ControlCull 50%Cull 70%Cull 80%
Within the simulated range, compliance did not have any notable effects on the CHB rate (Table 7.2.8.4). Again, the control rate and the perturbation effect were the factors that influenced the CHB rates. Compliance did not significantly affect the peak number of cattle herd breakdowns in the NU model (data not presented).
Table 7.2.8.4 Varying compliance: CHB rate.Effect of farm compliance on the mean cattle herd breakdown rate per farm. Means were calculated over ten years starting from the first year of culling (CSL model).
Mean number of CHBs per farm
ScenarioFarm
ComplianceNo
ControlShoot (50%)
Trap (70%)
Snare (80%)
Gas (80%)
Vaccinate (60%)
4 60% 0.063 0.086 0.078 0.076 0.075 0.056
2 70% 0.063 0.086 0.078 0.074 0.075 0.056
5 80% 0.064 0.086 0.077 0.075 0.075 0.057
61
7.2.9 Varying control ratesThe model outputs for different control rates were compared (see section 6.1.2 for adjusted values of control rates for each scenario). The default control area of 300km2 and farm compliance of 70% was used for all these three scenarios. All the cull options reduced the mean badger group size significantly, with vaccination having no effect (Table 7.2.9.1). The higher the cull rate, the larger the reduction in mean group size.
All the culling options resulted in an increase in the mean number of infected badgers per group, but the increase was least for the higher cull rates (Table 7.2.9.2). It can be seen in Figure 7.2.9.1 that the perturbation effect is over-riding any reduction that one might expect from the culling. The lower the control rates, the more the perturbation effect dominates. Conversely, the vaccination option, which does not cause any perturbation effect, results in a significant reduction in infected badgers, but varying the vaccination rates had no noticeable effect.
Table 7.2.9.1 Varying control rates: Badger social group size.Effects of control rates on mean badger group size as calculated over ten years starting from the first year of culling.
Mean badger group size
ScenarioControl Rates No
Control Shoot Trap Snare Gas Vaccinate
6 Low 7.5 6.3 5.5 5.3 5.3 7.5
2 Medium 7.5 5.7 5.3 5.1 5.1 7.5
7 High 7.5 5.3 5.1 5.0 5.0 7.5
Table 7.2.9.2 Varying control rates: Number of infected badgers.Effects of control rates on mean number of TB infected badgers per group. The means were calculated over ten years starting from the first year of culling (CSL model).
Mean number of infected badgers per group
ScenarioControl Rates No
Control Shoot Trap Snare Gas Vaccinate
6 Low 1.3 3.1 2.3 2.1 2.1 1.1
2 Medium 1.3 2.5 2.1 2.0 2.0 1.0
7 High 1.3 2.1 2.0 1.9 1.9 1.0
62
Figure 7.2.9.1 Varying control rates: Number of infected badgers.Effect of control rates on mean number of TB infected badgers per social group (CSL model). (a) Number of infected badgers at low control rates
0.0
1.0
2.0
3.0
4.0
5.0
115 120 125 130 135Years
Mea
n TB
-infe
cted
bad
gers
per
gro
up
No ControlShoot 30%Trap 60%Snare 70%Gas 70%Vaccinate 40%
(b) Number of infected badgers at medium control rates
0.0
1.0
2.0
3.0
4.0
5.0
115 120 125 130 135Years
Mea
n TB
-infe
cted
bad
gers
per
gro
up
No ControlShoot 50%Trap 70%Snare 80%Gas 80%Vaccinate 60%
(c) Number of infected badgers at high control rates
0.0
1.0
2.0
3.0
4.0
5.0
115 120 125 130 135Years
Mea
n TB
-infe
cted
bad
gers
per
gro
up
No ControlShoot 70%Trap 80%Snare 90%Gas 90%Vaccinate 80%
63
Most of the control options showed small differences in badger TB prevalence when comparing their low and high rates (Table 7.2.9.3). Low cull rates led to a higher prevalence. However, even the highest cull rates increased prevalence above the ‘no control’ option. Again, perturbation dominated the effects. Vaccination showed a reduction in prevalence, but as with the numbers of infected badgers, that reduction did not appear very sensitive to the vaccine rate.
Table 7.2.9.3 Varying control rates: Badger prevalence.Effect of control rates on badger prevalence, calculated as means over ten years starting from the first year of culling (CSL model).
Mean badger prevalence
ScenarioControl Rates No
Control Shoot Trap Snare Gas Vaccinate
6 Low 0.18 0.50 0.42 0.40 0.40 0.14
2 Medium 0.17 0.44 0.40 0.39 0.39 0.14
7 High 0.17 0.40 0.39 0.38 0.38 0.13
The effect of control rates on the CHB rate is proportional and consistent across the different methods (Table 7.2.9.4). The lower the control rate, the higher the CHB rate. The consistency of this effect, and the over-riding importance of perturbation, can clearly be seen in Figure 7.2.9.2, with substantial rises after year 122, except for vaccination (where there was no perturbation effect).
Table 7.2.9.4 Varying control rates: CHB rateEffect of control rates on mean CHB rate per farm. Means were calculated over ten years starting from the first year of culling (CSL model).
Mean number of CHBs per farm
ScenarioControl Rates No
Control Shoot Trap Snare Gas Vaccinate
6 Low 0.063 0.097 0.081 0.077 0.078 0.059
2 Medium 0.063 0.086 0.078 0.074 0.075 0.056
7 High 0.063 0.078 0.074 0.073 0.073 0.055
64
Figure 7.2.9.2 Varying control rates: CHB rate.Effects of control rates on the mean cattle herd breakdown (CHB) rate per farm (CSL model). (a) CHB rate at low control rates
0.00
0.05
0.10
0.15
115 120 125 130 135Years
Mea
n C
attle
Her
d B
reak
dow
ns p
er fa
rm
No ControlShoot 30%Trap 60%Snare 70%Gas 70%Vaccinate 40%
(b) CHB rate at medium control rates
0.00
0.05
0.10
0.15
115 120 125 130 135Years
Mea
n C
attle
Her
d B
reak
dow
ns p
er fa
rm
No ControlShoot 50%Trap 70%Snare 80%Gas 80%Vaccinate 60%
(c) CHB rate at high control rates
0.00
0.05
0.10
0.15
115 120 125 130 135Years
Mea
n C
attle
Her
d B
reak
dow
ns p
er fa
rm
No ControlShoot 70%Trap 80%Snare 90%Gas 90%Vaccinate 80%
65
7.2.10 No-ImmigrationA modification to the application of the control methods was tested with the CSL model, where badgers were prevented from crossing the cull-area boundary – i.e. post-cull immigration (and emigration) of badgers was prevented. This was to simulate a situation where land area(s) might be identified for control which had borders across which badgers (infected or otherwise) would be very unlikely to pass (e.g. very wide rivers, estuary, sea, areas of very low badger density). This would remove the immigration aspect of the perturbation effect. However, the badgers remaining within the culled area that escaped a cull were still subjected to perturbation – i.e. the transmission rates were raised for the default perturbation period (cull period plus 3 years).
Preventing immigration into the culled area had very little effect on the size of reduction of the badger population (Figure 7.2.10.1 compared with Figure 7.2.1.2). However, since the results are reported for the whole simulated grid (to compare with previous results), the true level of reduction within the culled area is much higher. As expected, after culling stops in year 124 the badger population recovery was slower without badger immigration (Figure 7.2.10.1) than with it (Figure 7.2.1.2). Vaccination had no significant effect on the modelled badger population.
Figure 7.2.10.1 No immigration: Badger social group size.Effect of preventing badger immigration into the culled area on the mean badger group size (adults per social group). Control was applied to 100km2
of the 400km2 grid, and as for other outputs the means reported were for the whole grid. Farm compliance was 70% (CSL model).
Badger Population: No immigration
0
2
4
6
8
10
115 120 125 130 135Years
Mea
n ba
dger
gro
up s
ize
No ControlCull 50%Cull 70%Cull 80%Vaccinate 60%
For culling without immigration, the CHB rate per farm was reduced in all the control options (Figure 7.2.10.2), in complete contrast to the scenarios where badger immigration occurred (Figure 7.2.6.1). This difference is attributable to the fact that when immigration (and emigration) is prevented, firstly no perturbation is occurring beyond the cull boundary, and secondly any effects of perturbation within the cull area are greatly limited by the smaller numbers of badgers remaining from the cull. Again, note that within the control area the effect of CHB would be much greater than shown, since 75% of the simulated area was not subjected to control.
66
Figure 7.2.10.2 No immigration: CHB rate.Effect of preventing badger immigration into the culled area on the mean CHB rate per farm. Control was applied to 100km2 of the 400km2 grid, and as for other outputs the means reported were for the whole grid. Farm compliance was 70% (CSL model).
0.00
0.05
0.10
0.15
0 5 10 15 20Years
Mea
n C
attle
Her
d B
reak
dow
ns p
er fa
rm No ControlCull 50%Cull 70%Cull 80%Vaccinate 60%
7.2.11 Gamma-InterferonThe CSL model was modified to briefly study the inclusion of gamma-interferon (gIFN) testing of cattle for TB. The test was simulated to be used in conjunction with the normal skin test, and was used in the model either with every routine skin test that was taking place, or with every Pre-Movement Test (PrMT) skin test that was being done, or with every test, both routine and PrMT. The results show that, in the absence of badger control, gIFN testing reduced the CHB rate when used in conjunction with routine tests, but had no effect when used with PrMT (Figure 7.2.11.1).
Figure 7.2.11.1 Gamma-interferon: CHB rate.Effect of combining gamma-interferon testing of cattle with the skin tests on the mean CHB rate per farm (CSL model). The lines represent routine cattle testing with pre-movement tests (No Control), gIFN combined with the routine tests (gIFN ROU), the pre-movement tests (gIFN PMT) or both (gIFN ROU+PMT).
0.00
0.05
0.10
0.15
110 115 120 125 130 135Years
Mea
n C
attle
Her
d B
reak
dow
ns p
er fa
rm No ControlgIFN ROUgIFN PMTgIFN ROU+PMT
67
7.3 Economic ModellingThe following results are all based on the CSL model, as the Newcastle model did not include the costs. All 38 parameters used in the estimation of economic outcomes are listed in Appendix A. We have estimated cash flows and overall outcomes for farmers and for Defra separately, based on our assumptions that:
a. all control actions will be at farmers’ expense, including farm survey and purchase of equipmentb. farms will require a licence to cull, involving some administration costs for Defrac. Defra will wish to pay for an assessment of outcomes at the end of the control period.
Whatever strategy is chosen, the payoff (benefit) would be achieved by a reduction in the incidence of TB in cattle. The cost-benefit calculations can be multiplied up to represent all parishes in England and Wales that are tested annually (T1) or every two years (T2) since these are the TB-affected areas of interest; however it should be noted that multiplying-up in such a way will tend to overestimate the overall losses or benefits, so results are presented for the simulated areas (not just the area subjected to control). Four badger culling methods were costed in some detail:
Shooting of visible badgers, by farmers or their agents (50% reduction in population),
Trapping, with subsequent shooting, using protocol of former WLU (70% reduction),
Restraints, with subsequent shooting, as costed for Northern Ireland (80% reduction),
Gassing, with hypothetical protocols (80% effective).
Vaccination of badgers against TB using an oral vaccine was also modelled (60% of healthy badgers in the control area take vaccine bait and become fully protected against TB), and costs were estimated for two types of proposed oral vaccine. The model predicts the rate of occurrence of TB breakdowns/year in the equilibrium state (no control) and also the rates during the five-year implementation of each control method, and for a further ten years after.
Economic benefits of control methods are assumed to arise during the 15-year period solely from the reduction in CHBs achieved as a consequence of the control method. The stochastic variation in the model is relatively large. When the output from sets of 100 simulations are aggregated, the between-set standard deviation of CHBs from year to year with no control is 3.6 % of the mean.
The overall year-by-year cash flow (Figure 7.3.1) shows the costs associated with each of the strategies and the ongoing costs associated with herd breakdowns. The peaks in year 0 and 4 for the badger management strategies are related to the set-up costs and the ecological monitoring costs at the end of control. For all scenarios presented here the double ring of social perturbation was used. For an assessment of reduced perturbation see Section 7.4
68
Figure 7.3.1. Cash flow by method of badger control.The overall year-by-year cash flow (expenditure £million) at present cost for the baseline no control strategy (NC) and for each of the badger management strategies. These results are from 200 simulations of 100km2
control, and assume the policy is implemented over all annual and biennial tested parishes.
Cash flow(100 km2; double perturbation)
0
50
100
150
0 5 10 15
year
£M/y
r
No Controlshoot50trap70snare80gas80vaccine60
It can bee seen that the costs associated with shooting, trapping, snaring and gassing are greater than those for no additional control over the entire period simulated. In contrast, after initial higher costs, vaccination exhibits lower cost cash flows than for ‘no control’ after year 4.
The net present cost of ‘no control’ and of each control method, discounted over the 15-year period, is shown in Table 7.3.1, together with the difference between each control method and the ‘no control’ equilibrium state.
Table 7.3.1. NPV for badger control.The 15-year Net Present Value associated with the 400km2 simulation, which includes 100km2 control area for each badger management strategy.
Control method Efficacy NPV£M
relative toNo Control
£M(No Control) -9.38Vaccination 60 -9.43 -0.06Gassing 80 -11.73 -2.35Restraint 80 -12.11 -2.73Shoot 50 -12.57 -3.19Trap 70 -13.43 -4.05
None of the five control methods were predicted to show a positive net benefit, with losses of between £0.06million and £4million over 15 years. For the scenario with vaccination, the deficit is very small and not statistically significant. In the case of trapping, the relatively large net present cost is due largely to the high cost of the trapping method, which involves the purchase of traps and substantial labour input. In the case of shooting,
69
the net present cost is largely due to the assumed relative poor efficacy of the method in culling badgers.
The allocation of costs to Defra and farmers (Figure 7.3.1) was made using the assumptions stated above. However, these costs can be broken down. The Defra portion of the cash flow is shown in Figure 7.3.2 and that for farmers in Figure 7.3.3.
Figure 7.3.2. Cash flow to Defra for badger control.The year-to-year cash flow (expenditure in £million) to Defra, at present cost, for the baseline no control strategy (NC) and for each of the badger management strategies (100km2) and assuming the policy is implemented over all annual and biennial tested parishes.
DEFRA cash flow(100 km2; double perturbation)
0
25
50
75
0 5 10 15
year
£M/y
r
No Control
shoot
trap
snare
gas
vaccine
Again, the increase in year zero arises from the Defra cost of licensing, and in year four from the assumption that Defra will pay for an assessment at the end of the control period. All of the control methods, with the notable exception of vaccination, show higher levels of expenditure over time compared to no control.
Figure 7.3.3. Cash flow to industry for badger control.The year-to-year cash flow (expenditure £million) to Farmers, at present cost, for the baseline no control strategy (NC) and for each of the badger management strategies (100km2) and assuming the policy is implemented over all annual and biennial tested parishes.
Farm cash flow(100 km2; double perturbation)
0
25
50
75
100
0 5 10 15
year
No Control
shoot
trap
snare
gas
vaccine
70
Figure 7.3.3 shows that none of the control methods, except vaccination, give any positive cash flows over time for the farming industry. These results confirm that Defra costs for control of TB are less than half the total costs, with farmers/landowners bearing the greater part of the costs whether or not a control method is applied.
7.3.1 Pre-movement testing – economic analysisA basic economic analysis of PrMT was performed. The PrMT was applied only to T1 and T2 areas, and was started in two phases based on cattle age, as in the field (see Section 3.2.7). In each simulation the model was run without PrMT, then all model values were reset to the state they were at the start of year 120, and rerun from then onwards with PrMT switched on. This allowed a truer comparison, with both options having the same start points.
The (discounted) net benefit distribution for Defra and industry combined (Figure 7.3.1.1) indicates that PrMT is slightly more likely to give an economic benefit than a loss, with 57 of the 100 simulations giving a benefit. The badger contribution (percentage of CHBs caused initial by badger transmission as opposed to cattle transmission) was about 40%, measured before PrMT started. Simulations with different transmission rates (giving lower badger contribution and higher cattle contribution to CHBs) showed that the economic viability of the PrMT was sensitive to this badger/cattle contribution ratio. As expected, the higher the cattle contribution to CHBs, the more benefit was likely from applying a PrMT regime.
Figure 7.3.1.1 Discounted Net Benefit: PrMT.The effects of PrMT applied to T1 and T2 parishes on the Discounted net benefit distribution. Each bar is the overall economic benefit (whole 400km2
grid area, Defra and industry combined) of one simulation, taking into account the costs of applying PrMT, and cattle disease management. This example assumes that badgers were responsible for 40% of all herd breakdowns prior to PrMT.
Pre-movement Testing
-4,000,000
-3,000,000
-2,000,000
-1,000,000
-
1,000,000
2,000,000
3,000,000
4,000,000
1 7 13 19 25 31 37 43 49 55 61 67 73 79 85 91 97
Simulation
Net
Pre
sent
Ben
efit
(£)
71
7.3.2 Economic benefit distribution – scenario 1The economic benefits for each of the control options are presented as distributions of net benefit for Defra and industry combined (Figure 7.3.2.1). This allows us to compare the likelihood for each of the options to give an overall economic gain or loss. The percentage of the bars that show a negative benefit is a measure of the risk of a control option actually causing more loss than gain – i.e. being a cost rather than a benefit. On the assumption that the modelled perturbation process is sufficiently close enough to what happens in the field, trapping badgers (70% removal) is clearly the worst option economically, with not only 100% of the simulations showing a net economic loss, but the magnitude of the losses being greater than the other culling options. The magnitude of the losses was mainly due to the high set-up cost of trapping badgers. However, the economic benefit distributions of the shooting, snaring and gassing options are similar, with 100%, 99%, and 98% of simulations respectively giving a net economic loss. Vaccination of badgers (60% of badgers protected) shows a better net benefit distribution, with only 46 of the 100 simulations showing a net economic loss.
Figure 7.3.2.1 Net Benefit: Scenario 1.Effect of different badger control options on the discounted net benefit distribution. Each bar is the overall economic benefit (Defra and industry combined) of one simulation, taking into account the economics of the badger control, and cattle disease management. (a) Net benefit distribution –shooting (50% removal)
-9,000,000
-7,000,000
-5,000,000
-3,000,000
-1,000,000
1,000,000
3,000,000
5,000,000
7,000,000
9,000,000
1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91 96
Simulation
Net
Pre
sent
Ben
efit
(£)
(b) Net benefit distribution –trapping (70% removal)
-9,000,000
-7,000,000
-5,000,000
-3,000,000
-1,000,000
1,000,000
3,000,000
5,000,000
7,000,000
9,000,000
1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91 96
Simulation
Net
Pre
sent
Ben
efit
(£)
72
(c) Net benefit distribution –snaring (80% removal)
-9,000,000
-7,000,000
-5,000,000
-3,000,000
-1,000,000
1,000,000
3,000,000
5,000,000
7,000,000
9,000,000
1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91 96
Simulation
Net P
rese
nt B
enef
it (£
)
(d) Net benefit distribution –gassing (80% removal)
-9,000,000
-7,000,000
-5,000,000
-3,000,000
-1,000,000
1,000,000
3,000,000
5,000,000
7,000,000
9,000,000
1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91 96
Simulation
Net P
rese
nt B
enef
it (£
)
(e) Net benefit distribution – badger vaccination (60% protection)
-9,000,000
-7,000,000
-5,000,000
-3,000,000
-1,000,000
1,000,000
3,000,000
5,000,000
7,000,000
9,000,000
1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91 96
Simulation
Net P
rese
nt B
enef
it (£
)
73
7.3.3 Economic comparison of different scenariosThe economic results for the different control scenarios are presented in Table 7.3.3.1. The values represent the risk of economic loss: the percentage value being equivalent to the cross over of the x-axis in the net benefit distributions such as those in Figure 7.3.2.1. For each simulation, the NPV of each control method in turn was compared with the no-control option: the difference comprising the value of loss or benefit. In all scenarios, the culling options were economically unviable (Table 7.3.3.1). Consistently the vaccination option was the most viable, economically, but even with vaccination the model shows that between 45% and 60% of the time we might expect the costs to be greater than the benefits.
Table 7.3.3.1 Percentage chance of economic loss.The risk of economic loss represented by the percentage of model simulations that gave an economic loss rather than a benefit. 100 simulations were run for each scenario. (The most economically viable of the options for each scenario is highlighted in bold).
Scenario DescriptionShoot% loss
Trap% loss
Snare% loss
Gas% loss
Vaccine% loss
1 100km2, 70% compliance, medium control rates 100 100 99 98 46
2 300km2, 70% compliance, medium control rates 100 100 98 97 53
3 400km2, 70% compliance, medium control rates 99 100 96 97 54
4 300km2, 60% compliance, medium control rates 100 100 99 99 54
5 300km2, 80% compliance, medium control rates 100 100 98 96 48
6 300km2, 70% compliance, low control rates 100 100 100 98 59
7 300km2, 70% compliance, high control rates 100 100 98 98 58
As an illustration of the two extremes of the economic distribution output, Figure 7.3.3.1 shows the distribution for shooting option with low control rate (30%), and that of the vaccine option at high control rate (80%).
74
Figure 7.3.3.1 Discounted Net Benefit: shooting and vaccination.Discounted net benefit distributions of two control options. Each bar is the overall economic benefit of one simulation, taking into account the economics of the badger control, and cattle disease management. The control area was 300km2, and farm compliance 70%.
(a) badger control by shooting (30% control rate: scenario 6 )
-900,000,000
-700,000,000
-500,000,000
-300,000,000
-100,000,000
100,000,000
300,000,000
500,000,000
700,000,000
900,000,000
1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91 96
Simulation
Net P
rese
nt B
enef
it (£
)
(b) badger control by vaccination (80% control rate: scenario 7)
-900000000
-700000000
-500000000
-300000000
-100000000
100000000
300000000
500000000
700000000
900000000
1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91 96
Simulation
Net P
rese
nt B
enef
it (£
)
Discounted net benefit distributions were also separated in terms of costs/benefits that would be borne by Government (Defra) and those by Industry (farmers). Figures 7.3.3.2 shows that Industry bears a greater risk from some of the control options than does Defra.
75
Figure 7.3.3.2 Discounted Net Benefits: Defra and Industry.Discounted net benefit distributions of two control options separated into Defra + Industry benefits (a & d), Defra benefits (b & e), and Industry benefits (c & f). Each bar is the overall economic benefit of one simulation. The control area was 100km2, and farm compliance 70%.
(a) Defra & Industry – snare (80%)
-9,000,000
-7,000,000
-5,000,000
-3,000,000
-1,000,000
1,000,000
3,000,000
5,000,000
7,000,000
9,000,000
1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91 96
Simulation
Net P
rese
nt B
enef
it (£
)
(b) Defra-only – snare (80%)
-9,000,000
-7,000,000
-5,000,000
-3,000,000
-1,000,000
1,000,000
3,000,000
5,000,000
7,000,000
9,000,000
1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91 96
Simulation
Net P
rese
nt B
enef
it (£
)
(c) Industry-only – snare (80%)
-9,000,000
-7,000,000
-5,000,000
-3,000,000
-1,000,000
1,000,000
3,000,000
5,000,000
7,000,000
9,000,000
1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91 96
Simulation
Net P
rese
nt B
enef
it (£
)
76
(d) Defra & Industry – trap (70%)
-9,000,000
-7,000,000
-5,000,000
-3,000,000
-1,000,000
1,000,000
3,000,000
5,000,000
7,000,000
9,000,000
1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91 96
Simulation
Net P
rese
nt B
enef
it (£
)
(e) Defra-only – trap (70%)
-9,000,000
-7,000,000
-5,000,000
-3,000,000
-1,000,000
1,000,000
3,000,000
5,000,000
7,000,000
9,000,000
1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91 96
Simulation
Net P
rese
nt B
enef
it (£
)
(f) Industry-only – trap (70%)
-9,000,000
-7,000,000
-5,000,000
-3,000,000
-1,000,000
1,000,000
3,000,000
5,000,000
7,000,000
9,000,000
1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91 96
Simulation
Net P
rese
nt B
enef
it (£
)
77
7.3.4 No Immigration – economic analysisThe no-immigration control, described in Section 7.2.10, where the ‘edge-effects’ of culling due to badger perturbation were minimised (e.g. due to physical boundaries preventing badger immigration) showed improved economic benefits for the shooting, trapping, gassing and snaring control options. The gassing control gives a clear example of the greater benefit (Figure 7.3.4.1) where 67% of the simulations showed an economic benefit, compared with Figure 7.3.2.1(d) where only about 2% showed an economic benefit. With no-immigration control the trapping option was almost as economically unviable as with the open area with only 21% of the simulations showing an economic benefit compared to 0%. This was probably because of the high set-up costs of trapping, and controlling. The other cull methods were more successful because the costs were relatively much lower. This clearly demonstrates the importance of badger perturbation, particularly immigration (and of assumptions concerning perturbation within the modelling exercise) to the economic viability of badger control methods.
Figure 7.3.4.1 Discounted Net Benefit: No immigration.Discounted net benefit distribution of gassing control. Each bar is the overall economic benefit of one simulation. The control area was 100km2, and farm compliance 70%.
-9,000,000
-7,000,000
-5,000,000
-3,000,000
-1,000,000
1,000,000
3,000,000
5,000,000
7,000,000
9,000,000
1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91 96
Simulation
Net P
rese
nt B
enef
it (£
)
7.3.5 Gamma-Interferon testing – economic analysis The simulation of gamma-interferon (gIFN) testing applied in combination with cattle skin tests is described in section 7.2.13, and was applied in the model either in every case where a routine skin test was used, or in every case where a PrMT skin test was used, or in every case regardless. The cost of such widespread testing on top of the skin test costs was high and analysis showed a large economic loss in every simulation, as is clearly seen in the discounted net benefit distribution for gIFN testing in conjunction with routine skin tests (Figure 7.3.5.1).
78
Figure 7.3.5.1 Discounted Net Benefit: gamma-interferon.Discounted net benefit distribution for simulations of gIFN testing combined with routine cattle skin-testing. Each bar is the overall economic benefit of one simulation. The control area was 100km2, and farm compliance 70%.
-60,000,000
-50,000,000
-40,000,000
-30,000,000
-20,000,000
-10,000,000
-
1 6 11 16 21 26 31 36 41 46 51 56 61 66 71 76 81 86 91 96
Simulation
Net P
rese
nt B
enef
it (£
)
7.4 Sensitivity Analysis of Economic OutputThe dominant feature of the farm costs of TB is the loss of slaughtered animals. Defra pay compensation for the market value of compulsorily slaughtered animals according to a scale that is adjusted from time to time (Defra, 2006a), but this does not take account of other farm costs. Data from the 2002/3 Reading Farm Survey included distributions of CHB cost per animal for beef and dairy farms. Lognormal curves provide a close fit to the distributions of cost per animal, and random draws are made from these statistical distributions (see Figure 3.1.1) in each simulation. This provides a stochastic input to the simulation of economic outcomes. The sensitivity to assumptions about this key cost component has been tested by offsetting the mean of the slaughter value distribution to higher and lower values, the offset being equivalent to one standard deviation of the lognormal distribution (Appendix B).
Components of farm costs such as farmers’ costs of testing and isolation of inconclusive reactors (IRs), collected in the Reading Farm Survey, are included in the lognormal distribution. Movement restriction costs are included separately.
Away from hotspot areas, there were 3.8 reactors per CHB in 2005: Defra (2006) [‘Detailed TB Statistics: 1 January to 31 December 2005, by SVS Region]. The difference in reactors/CHB between hotspot regions and other areas has only a small effect on overall predicted costs, because the number of reactors per CHB outside hotspot areas is only used to estimate the savings from any reduction in CHBs attributable to the export of infected animals; the export of infected animals has already been reduced to a small number by pre-movement testing, and any further reduction will be of little overall economic benefit. A stochastic value for the farm cost of CHBs arising from export of infected animals is obtained by drawing at random from the lognormal distribution of farm CHB costs reported in the Reading Farm Survey, adjusted for price changes between 2002 and 2006. The stochastic value is scaled down in proportion to the number of reactors per CHB, because the Reading data was collected entirely from farms within hotspot areas.
79
The outcomes of further simulations performed to test the sensitivity confirms that a high farm cost per slaughtered animal tends to lead to greater benefit for badger control, whereas a high number of reactors per CHB leads to reduced benefit, because the number of CHBs predicted by the model is unchanged.
There is no direct precedent for any of the five badger control methods being assessed, and the costs of preparing for and implementing a cull of badgers can only be estimated by making a number of assumptions. Preparation costs include costs of equipment and training. The main component of culling costs is farm labour. Labour rates are known, but the hours per cull are less certain. In the vaccination scenario, costs of vaccination are also significant, and not known with any certainty at the time of writing.
When further simulations are performed to test sensitivity to setup and culling costs, there is no clear pattern in the values (Table 7.4.1). This is because the ranges of values used to test for sensitivity lead to outcomes that are close in value relative to the extent of stochastic variability in the model. In most instances, higher set up and operating costs lead to reduced benefit, as would be expected.
Table 7.4.1. Economic sensitivity to culling and vaccination.Results from 100 simulations.
Control method
Relative best case£M
Relative worst case£M
Vaccination 0.12 -0.17Gassing -2.16 -2.50Restraint -2.25 -3.40Shoot -2.97 -4.14Trap -3.11 -4.43
Sensitivity to set-up cost assumptions show similar ranges. Results were found to be less sensitive to other assumptions. Overall, sensitivity to economic variables was not found to alter the overall findings. Each of the control methods had substantial net costs associated with them compared to no control, apart from vaccination, which was near the borderline of being a net cost or a net benefit.
7.5 Perturbation – Alternative OptionsFollowing the validation work on perturbation (section 7.1.5), the models were run with a double-ring of perturbation around the cull area to match the spatial profile of the RBCT results. However, very little detail is known about either the spatial or the temporal rules of perturbation, so all conclusions made from the models should consider these uncertainties in the simulation of perturbation. To contrast with the results presented above (default of double-ring perturbation), the following graphs and tables show some of the CSL model results when only a single-ring of perturbation was simulated around the cull area.
80
Figure 7.5.1. Badger prevalence.Relative level of disease (cattle herd breakdowns) for each type of badger management: shooting, cage trapping, snaring, gassing and vaccination (single ring perturbation).
Disease prevalence100km2 control (200 sims)
0
50
100
150
0 5 10 15
year
% re
lativ
e to
yea
r 0
NCShoot50Trap70Restraint80Gas80Vaccine60
It can be seen that each of the badger culling strategies are more effective in reducing the prevalence of TB in badgers, and thus the incidence of TB in cattle than under the double ring perturbation assumption, although the shooting method still causes the incidence to increase over time. Table 7.5.1 shows the discounted net benefits associated with these different methods.
Table 7.5.1 Discounted net benefits of control with reduced perturbation.Results shown for a single ring of perturbation (simulated area 400km2: control area 100km2).
Control method Efficacy NPV£M
Relative toNo Control
£M(No Control) -7.61Gassing 80 -7.28 0.33Snaring 80 -7.41 0.20Vaccination 60 -7.54 0.06Shoot 50 -8.57 -0.96Trap 70 -9.11 -1.50
It can be seen that gassing, snaring and badger vaccination have relatively small net benefits associated with them compared to no control, whilst both shooting and trapping have relatively large net costs associated with them. However, the net benefits associated with gassing, snaring and vaccination are not significantly different from no control.
Table 7.5.2 shows the risk of economic loss associated with the model simulations for each control scenario. By using this reduced assumption of social perturbation gassing and
81
snaring may sometimes be more economically viable than vaccination. However, even with a large control area, high compliance and high control rates, an economic loss is still likely in at least one third of all simulations.
Although, with more spatially-limited perturbation effects, the model generally gave lower percentages of simulations showing controls with net economic loss, still none of the control methods were by any means certain to give a benefit. Trapping remained the worst option and nearly always gave a net economic cost rather than a benefit. Gassing, even at the high control rates, still resulted in an economic loss in one third to one half of all simulations.
Table 7.5.2 Percentage economic loss.The risk of economic loss represented by the percentage of model simulations that gave an economic loss rather than a benefit. 100 simulations were run for each scenario, and a single-ring of perturbation was used. (The most economically viable of the options for each scenario is highlighted in bold).
Scenario DescriptionShoot% loss
Trap% loss
Snare% loss
Gas% loss
Vaccine% loss
1 100km2, 70% compliance, medium control rates 76 93 64 52 63
2 300km2, 70% compliance, medium control rates 79 99 58 54 54
3 400km2, 70% compliance, medium control rates 82 99 46 44 60
4 300km2, 60% compliance, medium control rates 79 99 51 48 48
5 300km2, 80% compliance, medium control rates 87 96 50 49 65
6 300km2, 80% compliance, low control rates 97 100 56 58 60
7 300km2, 80% compliance, high control rates 48 96 37 32 45
8 300km2, 60% compliance, low control rates 98 99 64 59 56
9 300km2, 60% compliance, high control rates 63 98 48 35 46
8. DiscussionThis project built on Phase I of the Cost-Benefit Analysis, which examined the costs and benefits of badger management using a pure badger model. The report of Phase I was submitted previously. Phase II involved putting the economic costs within the simulation model, and extending the simulation model to include cattle farming and more realistic badger behaviour. This involved a cattle-rearing aspect, which included births, sale and
82
slaughter of cattle for dairy, beef and mixed herds. Data on parameter values (e.g. sale frequency and distance of farm-to-farm movement) were obtained from the relevant VLA-held databases. A cattle epidemiological model was also parameterised, which included disease transmission, disease progression, test frequency and test performance. Parameter values were obtained from a variety of sources. In addition, a GIS version of the model was produced by Newcastle University but without economic parameters. This permitted improved model verification and allowed further simulations in specific landscapes.
8.1 Verification and ValidationThe models were verified by a number of different approaches. Model verification ensures that model output changes in an expected manner if parameter values are changed. This process included switching off badger-to-cattle TB transmission. This resulted in a decline and elimination of cattle TB, which indicates that the current cattle testing regime is capable of removing TB from cattle herds, as reported historically, in the absence of infection in the badger. Sensitivity analysis by One-At-A-Time and Latin Hypercube Sampling gave very similar results between the CSL and NU models for badger social group size, prevalence and cattle herd breakdowns rates. Output from the two models was compared statistically and although significantly different, the mean results were very similar. Statistical significance occurred because of the large number of iterations; the difference in all cases was too small to be measurable in the field. Output from the CSL model was more variable than the NU version. This is because of the randomisation procedure to initialise the locations of badger setts and farms performed in each iteration in the CSL model, i.e. each iteration is a randomised landscape. The NU version uses a GI landscape, which by definition has some degree of similarity in each iteration. Thus for any particular geographical configuration, the NU model output should be less variable, and may be slightly higher or lower than the CSL mean output.
Model output validation was performed by comparing simulation output with field data. Badger social group size, population dynamics and historical population growth has previously been validated for both models (Shirley et al., 2003; Smith, Cheeseman & Clifton-Hadley, 1997; Smith et al., 1995). The number of reactors per herd breakdown and their frequency distribution, the economic costs of breakdowns and annual TB costs were not very different from historical field data. In addition, the CSL model replicated the general spatial findings of the proactive culling in the RBCT. The NU model replicated the RBCT trial by simulating 50 iterations of each proactive area. In all cases the range of CHBs in the simulated trials encompassed the field data. For two out of ten triplets the field data lay outside the simulated 95%ile distribution and for four of the triplets the field data lay outside the simulated 50%ile distribution. For the latter results no consistent bias occurred in the data. These results are very encouraging, since with as few as 50 simulations, the range of the simulated data encompasses the field results. However, it also indicates that further heterogeneities are likely to exist. One such was the effect of the 2001 FMD outbreak, since all three triplets that entered the proactive culling of the RBCT after FMD had a higher CHB rate than predicted by the model.
The above verification and validation procedures together fail to invalidate the models. Thus, we have some evidence on the reliability of the model results. This model has thus been verified and validated to the available data at the time of construction. It is capable of simulating most badger management strategies, many cattle management strategies and combined management strategies. These could include, but are not limited to, badger culling by any means, badger vaccination, changes to cattle testing (frequency), and changes to the rules to change cattle test frequency. However, without further data and
83
programming, it is not yet able to simulate specific farm types (e.g. finishing farms) or simulate changes to on-farm biosecurity. Whilst the model is validated to the RBCT research results at the end of the trial, further analysis is required to take account of changes in herd breakdown rate following the cessation of culling. It is noted that, from analysis of the RBCT results after the completion of this model (Jenkins, Woodroffe & Donnelly, 2008), the assumptions used in the model on the temporal aspects of social perturbation no longer produce herd breakdown rates consistent with the available data, and the model will need to be further updated to take account of this. The recent data report on changes in herd breakdown rate and have not attempted to ascribe this to particular mechanisms, particularly since no new data has been produced on badger territoriality and density. However, it is possible to examine the potential role of decreasing social perturbation to see if temporal parameter changes can produce similar changes in geographical herd breakdown rates, or whether such a mechanism is not a potential causative function. Thus the model is validated to the end of the RBCT, and further work is ongoing to ensure validation is obtained for the time period following the cessation of culling.
8.2 Cattle TestingOne important result from the sensitivity analysis was that the cattle herd breakdown rate was sensitive to a decrease in the sensitivity of the cattle skin test (see section 7.1.1). Although the degree of this response would depend on the contribution of badger to the overall rate, in our simulations the CHB rate doubled if the skin test sensitivity was halved. Thus it appears to be important to ensure that the skin testing is performed correctly, since systematic poor performance would allow TB to spread within the cattle population and result in a noticeable increase in herd breakdown rate.
It also appears that the algorithm used to switch parishes to and from annual/biennial testing has scope for improvement (see section 7.1.2) and this should be investigated with and without the effects of Pre-Movement Testing. Indeed the whole area of cattle test frequency and the geographical size of the areas to which these are applied should be investigated further. Some of the simulation results suggested that the rules used to switch test frequency could result in a subsequent increase in herd breakdown rate.
We investigated the effect of Pre-Movement Testing with different proportionate contributions of badgers to the overall herd breakdowns rate. With the default parameter setting, switching on Pre-Movement Testing resulted in badgers causing 60% of all CHBs in subsequent years. Thus, if badgers contributed less than 60% of all CHBs then a larger decrease in CHBs is seen when PrMT is simulated. This is an important result. Whilst we cannot be sure what the level of contribution from badgers actually is prior to Pre-Movement Testing (although the results from the RBCT suggest it is at least 20% (Donnelly et al., 2006), and earlier analysis suggested it must be less than 80%: Phase I report), the level of contribution appears to harmonise to close to 60% after the introduction of Pre-Movement Testing. This means that the economic outputs of badger and cattle management will be more reliable, since they are less dependent on the (unknown) level of badger contribution.
We also note that a further reduction in CHB rate may be obtained from instigating Pre-Movement Testing in all areas (i.e. including areas subjected to three- and four-yearly testing), although we did not examine the economic benefit of this. Preliminary investigations suggested that the use of IFN-gamma testing could result in a reduction in herd breakdowns, but no economic benefit if this was applied in a blanket manner in
84
conjunction with Pre-Movement Testing. Thus, we suggest that there is potential to use the IFN-gamma test in certain circumstances, but these need a thorough investigation, including the economics of use.
8.3 Badger CullingIf badger culling were performed in such a way as to be able to kill badgers without access to their setts (e.g. snaring or trapping), it is expected that access to the whole badger territory is not necessary to effectively remove individuals. We have no real evidence of how much territorial access is required in order to catch badgers, but since many badgers may patrol the edge of their territory at some point every few nights, we assumed that access to 10% of the territory would be sufficient. This indicated that compliance to at least 60% of farms would result in access to over 90% of badger territories. However, it is important to note that this result will vary geographically, since it depends not only on average farm size, but also how farmland is distributed (i.e. in a contiguous manner or not) and the relative size of badger territories.
Badger culling tended to result in an increase in TB prevalence in the badger. The level of this increase was dependent on the level of badger control, but no method or realistic level of control was simulated which was capable of reducing the burden of TB in badgers. However, this was dependent on the introduction of a single level of perturbation; i.e. any badger culling would result in the same level of perturbation. Whilst this may appear unrealistic, we have no evidence on which to base a graduated response. However, it is the number of infected (or infectious) badgers that is a better measure of risk to cattle. This measure increased in the simulation grid as a whole, until the effects of perturbation ceased.
Since disease transmission is a relatively uncommon stochastic process, the TB prevalence in cattle always appeared more variable than the prevalence in the badger, or number of infected badgers. However, the same qualitative results were seen. By using the most valid data on social perturbation, the models suggested that the overall CHB rate could increase above its original level, even though the CHB rate decreased within the centre of the culled area. This level of increase depended to a small extent on the efficacy of control. However, it will also depend strongly on the actual level of perturbation in the badger. Thus, following the CHBs in the proactive areas of the RBCT over the next few years will help to determine the quantitative effects of badger social perturbation on the disease in cattle.
Changing the simulated area of culling from 100km2 up to 400km2 had relatively little effect on the overall level of disease (in badgers or cattle) within the whole simulated area. Changing the level of land compliance between 60-80% also had little effect on the disease in badgers or cattle. This result is expected if badgers can be successfully culled with limited access to the territory (see Figure 7.2.1.1).
The largest effect of badger culling on TB in the badger, or CHBs, was seen if badger immigration was prohibited. This resulted in a significant decrease in cattle herd breakdown rate. This strongly suggests that culling isolated areas (e.g. as occurred to some extent in the Irish Four Areas Trial) would give the best overall outcome in terms of reducing CHBs. This result provides an explanation for the different outcomes reported in the RBCT (Donnelly et al., 2006) and the Irish Four Areas Trial (Griffin et al., 2005).
85
8.4 Badger VaccinationBadger vaccination (applied to 100% of territories) resulted in a decrease in TB prevalence in the badger while vaccination was being performed (i.e. for five years). After vaccination ceased, prevalence stabilised at its new level during the remainder of the simulation. This result was expected, since vaccination would reduce the spread of disease locally (i.e. the local R0 would reduce to less than unity in some areas) and focal areas would become disease free. Since TB in the badger spread slowly geographically, it may take some years, or even decades for the disease to return to pre-control levels.
8.5 EconomicsThe economic output from the results will of course depend on a variety of factors. This includes the proportionate contribution of badgers to TB cattle infection, the assumption that the available cattle movement data captures the essence of cattle movement in real life, the epidemiological parameters used are suitable, and many other assumptions. One of the most critical is that the data used to parameterise the model will also describe the future dynamics. If farmers change their cattle management behaviour, e.g. frequency and timing of cattle sales, this may bias the results in some, as yet, unknown way. Further we did not assume there was any spatial or temporal heterogeneity in badger culling, which may occur in real life. However, given such caveats the economic results are at least consistent with each other, and thus are likely to remain relatively correct with respect to each badger management method.
We performed a single evaluation of Pre-Movement Testing, which showed a greater probability of making an economic benefit, rather than a loss, and so this appears to be a reasonable policy option for Defra. We have not attempted to simulate the change in cattle sale frequency that has been recorded in the last year, since this is unlikely to continue for the next few years. The preference to sell cattle that are not subjected to PrMT will probably reduce the real benefit (in terms of CHB) seen. This benefit will likely be delayed until the second phase of PrMT, and thus the economic benefit realised will have been delayed as well.
Regardless of the percentage efficacy of control simulated, all forms of badger management simulated make an economic loss. This is greatest for trapping (100% of iterations make a loss), then shooting (99-100% make a loss), and then snaring and gassing (over 95% make a loss). Vaccination is the least costly option (with 46-59% of iterations making a loss). It is not surprising that trapping is shown to be the most uneconomical. The high cost of purchasing cages, and the necessity to trap for a number of continuous nights incurs a high cost. The high probability of an economic loss occurring as a result of shooting (despite very low costs associated with it) is due to the low removal rate. Therefore, a large number of badgers are left alive and perturbed. The simulations of shooting, trapping and snaring never gave economic results better than gassing or vaccination (with a single exception) across all eight scenarios where compliance, area and efficacy were varied. It must also be recalled that gassing (badger setts) was also assumed to be able to ‘draw out’ badgers (i.e. culling efficacy was not reduced with access to only part of the territory). Since this assumption is less accurate for gassing compared to all the other methods of culling, these results suggest that a low-cost high efficacy method is a necessity for badger culling to be economically viable. Even in the best scenarios for gassing, an economic loss occurred in 96% of iterations.
86
However, the best way of obtaining the greatest probability of an economic benefit was to remove badger immigration, and thus avoid the ‘edge effect’ of culling. For gassing, this gave a 67% chance of an economic benefit (compared to 2% for the same scenario with badger immigration). This highlights the importance of badger perturbation (and assumptions regarding perturbation in the modelling exercise) in determining the economic viability of the various badger control methods. Indeed, if badger perturbation was adjusted to be less extreme (and thus unable to replicate the RBCT findings) this same scenario resulted in a 52% chance of an economic loss.
The routine use of the IFN-gamma gave an overall economic loss in all cases. However, this was only a simple preliminary investigation, and does not mean that there are no strategies that could be employed, where an economic gain would be achieved.
9. ConclusionsThis project utilised a previously validated badger/TB model and added a full cattle/TB simulation layer to produce two models. In one the economics of badger management and cattle TB epidemiology was included. In the other a full GIS version was constructed to permit simulation of specific geographical areas. The combined model(s) were subjected to verification, validation and sensitivity analysis. The completed model(s) are therefore capable of simulating a variety of different badger and cattle management strategies, and predicting their outcomes in terms of badger epidemiology, cattle herd breakdowns and economics. Both models successfully simulated the spatial effect of the RBCT in terms of cattle herd breakdowns by simulating realistic levels of increased ranging behaviour in badgers following culling.
Together, both models predicted that badger culling could reduce herd breakdowns in the centre of the culled area, but the increase in herd breakdowns at the periphery resulted in no real overall benefit, despite simulating culling areas of up to 400km2. In these scenarios no method of badger culling gave a certainty, or even a high probability, of a net economic benefit over 15 years.
By reducing the level of social perturbation in the badger, the models predicted that some methods of badger culling reduced the herd breakdown rate overall. However, these methods only had up to a 50% chance of achieving an economic benefit.
The most influential component of the system that led to the economic loss was the immigration of badgers from outside the culled area. Simulations with no immigration resulted in a dramatic decrease in herd breakdown rate (as seen in the Irish Four Areas Trial), and a much higher probability of an economic gain. This holds true for different realisations of social perturbation, and thus must be considered the main influence on the success of any badger culling policy.
By contrast badger vaccination (albeit with large uncertainty in the costings) always had about a 50:50 chance of achieving an economic benefit. However, the simulation of vaccination in this model was necessarily simplistic and more realistic approaches are necessary. The outcome of the BCG badger vaccine trial will produce much improved information on which to base future predictions of vaccine economics.
Analysis of cattle management indicated that herd breakdown rates could easily increase if the skin test was performed with a reduced sensitivity and that pre-movement testing
87
should reduce herd breakdowns (and also make the models more reliable as it reduced the variation in the contribution of badgers to herd breakdowns).
Further work should be done on badger culling methods where immigration is reduced/absent, measuring badger immigration rates, simulating badger vaccination and cattle testing (e.g. routine test interval, IFN-gamma).
10. References
Bart, J. (1995) Acceptance criteria for using individual-based model to make management decisions. Ecological Applications, 5, 411-20.
Bennett, R.M. & Cooke, R.J. (2006) Costs to farmers of a tuberculosis breakdown. Veterinary Record, 158, 429-32.
de Lisle, G.W., Buddle, B.M. & Collins, D.M. (2007) Bovine tuberculosis – Another view from down under. The Veterinary Journal, 173, 233-34.
Defra. (2006a). Compensation for Bovine TB, BSE, Brucellosis, and Enzootic Bovine Leukosis. Available at http://www.defra.gov.uk/news/2007/070301a.htm. Accessed on 21/05/2007.
Defra. (2006b). SE3112 Assessment of the economic impacts of TB and alternative control policies. In Review of the Bovine TB Research Programme [available online http://www.defra.gov.uk/science/documents/publications/BTB.pdf], pp. 71-74.
Defra. (2007). Pre-movement testing in England: updated assessment of costs and benefits, January 2007. Available at http://www.defra.gov.uk/animalh/tb/pdf/costs_benefits0107.pdf. Accessed on 21/05/2007.
Delahay, R.J., Cheeseman, C.L., Mallinson, P.J., Rogers, L.M. & Smith, G.C. (1998) Badgers and bovine tuberculosis: a review of studies in the ecology of a wildlife disease reservoir. Cattle Practice, 6, 83-87.
Donnelly, C.A., Wei, G., Johnston, W.T., Cox, D.R., Woodroffe, R., Bourne, F.J., Cheeseman, C.L., Clifton-Hadley, R.S., Gettinby, G., Gilks, P., Jenkins, H.E., Le Fevre, A.M., McInerney, J.P. & Morrison, W.I. (2007) Impacts of widespread badger culling on cattle tuberculosis: concluding analyses from a large-scale field trial. International Journal of Infectious Diseases, in press.
Donnelly, C.A., Woodroffe, R., Cox, D.R., Bourne, F.J., Cheeseman, C.L., Clifton-Hadley, R.S., Wei, G., Gettinby, G., Gilks, P., Jenkins, H., Johnston, W.T., Le Fevre, A.M., McInerney, J.P. & Morrison, W.I. (2006) Positive and negative effects of widespread badger culling on tuberculosis in cattle. Nature, 439, 843-46.
Donnelly, C.A., Woodroffe, R., Cox, D.R., Bourne, J., Gettinby, G., Le Fevre, A.M., McInerney, J.P. & Morrison, W.I. (2003) Impact of localized badger culling on tuberculosis incidence in British cattle. Nature, 426, 834-37.
Fischer, E.A.J., van Roermund, H.J.W., Hemerik, L., van Asseldonk, M.A.P.M. & de Jong, M.C.M. (2005) Evaluation of surveillance strategies for bovine tuberculosis (Mycobacterium bovis) using an individual based epidemiological model. Preventive Veterinary Medicine, 67, 283-301.
88
Griffin, J.M., Williams, D.H., Kelly, G.E., Clegg, T.A., O'Boyle, I., Collins, J.D. & More, S.J. (2005) The impact of badger removal on the control of tuberculosis in cattle herds in Ireland. Preventive Veterinary Medicine, 67, 237-66.
Grimm, V., Berger, U., Bastiansen, F., Eliassen, S., Ginot, V., Giske, J., Goss-Custard, J., Grand, T., Heinz, S., Huse, G., Huth, A., Jepsen, J.U., Jørgensen, C., Mooij, W.M., Müller, B., Pe’er, G., Piou, C., Railsback, S.F., Robbins, A.M., Robbins, M.M., Rossmanith, E., Rüger, N., Strand, E., Souissi, S., Stillman, R.A., Vabø, R., Visser, U. & DeAngelis, D.L. (2006) A standard protocol for describing individual-based and agent-based models. Ecological Modelling, 198, 115-26.
Iman, R.L. & Conover, W.J. (1982) A distribution-free approach to inducing rank correlation among input variables. Comm. Stat, B11, 311-34.
Independent Scientific Group. (2007). Bovine TB: The scientific evidence. Defra Publications, London, UK.
Jenkins H.E., Woodroffe R., Donnelly C.A. (2008) The effects of annual widespread badger culls on cattle tuberculosis following the cessation of culling. International Journal of Infectious Diseases, in press
Krebs, J.R., Anderson, R.M., Clutton-Brock, T., Donnelly, C.A., Frost, S., Morrison, W.I., Woodroffe, R. & Young, D. (1998) Badgers and bovine TB: conflicts between conservation and health. Science, 279, 817-18.
Munroe, F.A. & Dohoo, I.R. (1999). Estimating the incidence rate of within-herd spread of M. bovis. In Proceedings of the Society for Veterinary Epidemiology and Preventive Medicine Conference.
Neal, E. & Cheeseman, C. (1996) Badgers T & AD Poyser, London, UK.
Shirley, M.D.F., Rushton, S.P., Smith, G.C., South, A.B. & Lurz, P.W.W. (2003) Investigating the spatial dynamics of bovine tuberculosis in badger populations: evaluating an individual-based simulation model. Ecological Modelling, 167, 139-57.
Smith, G.C., Bennet, R., Wilkinson, D. & Cooke, R. (2007) A cost-benefit analysis of culling badgers to control bovine tuberculosis. The Veterinary Journal, 173, 302-10.
Smith, G.C., Cheeseman, C.L. & Clifton-Hadley, R.S. (1997) Modelling the control of bovine tuberculosis in badgers in England: culling and the release of lactating females. Journal of Applied Ecology, 34, 1375-86.
Smith, G.C., Cheeseman, C.L., Clifton-Hadley, R.S. & Wilkinson, D. (2001a) A model of bovine tuberculosis in the badger Meles meles: an evaluation of control strategies. Journal of Applied Ecology, 38, 509-19.
Smith, G.C., Cheeseman, C.L., Wilkinson, D. & Clifton-Hadley, R.S. (2001b) A model of bovine tuberculosis in the badger Meles meles: the inclusion of cattle and the use of a live test. Journal of Applied Ecology, 38, 520-35.
Smith, G.C., Richards, M.S., Clifton-Hadley, R.S. & Cheeseman, C.L. (1995) Modelling bovine tuberculosis in badgers in England: preliminary results. Mammalia, 59, 639-50.
Smith, R.M.M., Drobniewski, F., Gibson, A., Montague, J.D.E., Logan, M.N., Hunt, D., Hewinson, G., Salmon, R.L. & O'Neil, B.D. (2004) Mycobacterium bovis infection, United Kingdom. Emerging Infectious Diseases, 10, [online http://www.cdc.gov/ncidod/EID/vol10no3/02-0819.htm].
Swinton, J., Tuyttens, F., Macdonald, D., Nokes, D.J., Cheeseman, C.L. & Clifton-Hadley, R. (1997) A comparison of fertility control and lethal control of bovine tuberculosis in
89
badgers: the impact of perturbation induced transmission. Philosophical Transactions of the Royal Society of London B, 352, 619-31.
VLA. (2006). Surveillance project SB4008: laboratory testing and epidemiology support for the National Gamma Interferon Field Trial. Available at www.defra.gov.uk/animalh/tb/pdf/gifn_trialfinalreport.pdf. Accessed on 21/05/2007.
Vose, D. (1996) Quantitative risk analysis: a guide to Monte Carlo simulation modelling John Wiley & Sons, Chichester.
Wilkinson, D., Smith, G.C., Delahay, R., Rogers, L.M., Cheeseman, C.L. & Clifton-Hadley, R.S. (2000) The effects of bovine tuberculosis (Mycobacterium bovis) on mortality in a badger (Meles meles) population in England. Journal of Zoology, London, 250, 389-95.
Wilkinson, D., Smith, G.C., Delahay, R.J. & Cheeseman, C.L. (2004) A model of bovine tuberculosis in the badger Meles meles: an evaluation of different vaccination strategies. Journal of Applied Ecology, 41, 492-501.
Woodroffe, R., Donnelly, C.A., Cox, D.R., Bourne, F.J., Cheeseman, C.L., Delahay, R.J., Gettinby, G., McInerney, J.P. & Morrison, W.I. (2006) Effects of culling on badger Meles meles spatial organization: implications for the control of bovine tuberculosis. Journal of Applied Ecology, 43, 1-10.
90
Appendix A – Economic ParametersEconomic parameters and their values, as used in the CSL model.Parameter GENERAL Value unit Notes
1 Disc_rate 3.5 %2 SVStoDefra 1.46 -3 public_share 0.4 -
BADGER4 SetUpCost(1) 400 £/km2 one ‘hunting’ firearm5 SetUpCost(2) 8675 £/km2 costs as for RBCT6 SetUpCost(3) 1175 £/km2 costs as for Northern Ireland
trials7 SetUpCost(4) 1000 £/km2 estimate for training and
equipment8 SetUpCost(5) 1000 £/km2 estimate for training9 HotSpotArea 45177 km210 EcoMonitoringCost 1000 £/km211 IndustrySurveyCost 131.92 £/km2 two man-days farm labour (Nix,
2007)12 CullingCost(1) 350 £/km2 estimate based on hourly wage
rate13 CullingCost(2) 1387.53 £/km2 costs as for RBCT14 CullingCost(3) 462.77 £/km2 costs as for Northern Ireland
trials15 CullingCost(4) 306.38 £/km2 estimate based on hourly wage
plus supplies16 CullingCost(5) 297.72 £/km2 estimate based on hourly wage
plus supplies17 DisposalCost 20 £/animal
CATTLE18 beefSVmu 7.18 -19 beefSVsd 0.43 -20 dairySVmu 7.28 -21 dairySVsd 0.36 -22 farmCHBmu 8.27 -23 farmCHBsd 0.99 -24 InflationBeef 9 % annual Defra values, 2002-200625 InflationDairy 20 % annual Defra values, 2002-200626 CHBperiCattle 1.2 - CHB per exported outbreak27 RCHBhotspot 9.7 -28 RCHBreg 3.78 -29 SlaughterPerReactor 1.18 - Defra TB statistics30 SVScostPerTest(1,1) 44.72 £31 SVScostPerTest(1,2) 170.32 £32 SVScostPerTest(2,1) 3.14 £/animal33 SVScostPerTest(2,2) 2.016 £/animal34 VLAsampleCulturing 400.53 £/CHB35 MovemtRestr(1) 32.7 £/farm/day36 MovemtRestr(2) 6 £/farm/day37 Cval(1) 500 £38 Cval(2) 600 £
91
Appendix B – Farm Costs
Lognormal distributions were fitted to outcome of Reading farm survey for beef cattle, dairy cattle and additional on-farm costs.
beefSVmu = 7.18 £1313beefSVsd = 0.43 (95% prob range £555 - £3103)dairySVmu = 7.28 £1451dairySVsd = 0.36 (95% prob range £706 - £2980)farmCHBmu = 8.27 £3905farmCHBsd = 0.99 (95% prob range £539 - £28,283)
Sensitivity tests:SA107 (one s.d. high)beefSVmu = 7.61 £2018beefSVsd = 0.43 (95% prob range £854 - £4770)dairySVmu = 7.64 £2080dairySVsd = 0.36 (95% prob range £1012 - £4273)
SA108 (one s.d. low)beefSVmu = 6.75 £854beefSVsd = 0.43 (95% prob range £361 - £2018)dairySVmu = 6.92 £1012dairySVsd = 0.36 (95% prob range £493 - £2080)
92
Curve fit - beef
0
0.01
0.02
0.03
0.04
0.05
0 1000 2000 3000 4000 5000
farm cost per slaughtered animal £
Curve fit - dairy
0
0.01
0.02
0.03
0.04
0.05
0 1000 2000 3000 4000 5000
farm cost per slaughtered animal £
Curve fit - CHB
0
0.1
0.2
0.3
0.4
0.5
0 10,000 20,000 30,000 40,000 50,000
farm cost £
93
Appendix C – Cattle Mortality
Cattle mortality (at slaughter) data, obtained from the Cattle Tracing Scheme, was plotted for beef and dairy, male and female. These are presented as monthly rates and cumulative distribution functions.
Beef Mortality Rates
0.00
0.20
0.40
0.60
0.80
1.00
0 10 20 30 40 50 60 70 80 90 100
Age (months)
MalesFemales
Beef Mortality Rates
0.00
0.20
0.40
0.60
0.80
1.00
0 10 20 30 40 50 60 70 80 90 100
Age (months)
MalesFemales
94
Dairy Mortality Rates
0.00
0.20
0.40
0.60
0.80
1.00
0 10 20 30 40 50 60 70 80 90 100
Age (months)
MalesFemales
CDF
0.00
0.20
0.40
0.60
0.80
1.00
0 10 20 30 40 50 60 70 80 90 100
Age (months)
Males
Females
On the basis of these results it was decided to use a 6-month resolution for the mortality rates (see Appendix E for actual rates used). More coarse resolution (e.g. annual) did not give realistic cattle/age distributions in the model.
95
Appendix D – Cattle Herd Breakdown Rates (historic)
Cattle Herd Breakdown rates compiled from veterinary records. These graphs show the relative stability in herd breakdown risk per farm, despite the increasing number of breakdowns, as this increase is associated with the increasing area under annual or biennial testing.
CHBs
0
200
400
600
800
1000
1200
1400
1600
1800
2000
2002 2003 2004 2005 2006
T1T2T3T4
Areas
0
5000
10000
15000
20000
25000
30000
35000
40000
2002 2003 2004 2005 2006
T1
T2
T3
T4
CHBs per 100 herds
0
2
4
6
8
10
12
2002 2003 2004 2005 2006
T1T2T3T4
96
Appendix E – Parameter Values (CSL model)
E.1 Default Settings and Parameter Values used in the CSL modelNote: probabilities or rates are adjusted for a two-month time step where appropriate
E.1.1 Temporal Settings:Year Badgers added to grid 1Year cattle added to grid 20First Year that routine Test Interval switching is introduced (where parishes can change status from one test-interval type to another if CHB rate low or high).The algorithm to determine switching is as follows:
(1) Take the confirmed breaks in a parish over the previous 2 years and divide by the total herds in the parish, if the % is over 1 then the testing interval is yearly (T1)
(2) If not (1) then do as above but consider breaks over 4 years, if the % is greater than 0.2 then the interval is 2 yearly (T2)
(3) If not (2) then do as above but consider breaks over 6 years, if the % is greater than 0.1 then the interval is 3 yearly (T3)
(4) If none of the above then the interval is 4 yearly (T4)This algorithm was supplied by the veterinary agency, and is the one advised by the EU, and used in the field.
50
First year that Pre-Movement Testing (PrMT) is introduced 100Years of Badger Control 120-124Years that badger perturbation is applied (cull options only) 120-127Last Year of each simulation 135
E.1.2 Spatial Settings:Grid: A B C
Target Control Area (km2) 100 300 400Grid size (squares per side) 100 160 200Grid-Cell size (km) 0.2 0.2 0.2Grid Area (km2) 400 1024 1600Total Parishes 30 77 120Parishes subject to Control 7 21 28Mean Parish Size (km2) 13.3 13.3 13.3Mean Control Area (km2) 93 279 373Control Area relative to Grid A 1.0 3.0 4.0Control Area as proportion of simulation grid 0.23 0.27 0.23Badger Groups 300 768 1200Mean Badger Territory Size (km2) 1.33 1.33 1.33Beef farms 70 179 280Dairy farms 58 148 232Mixed farms 18 46 72X4 mixed other sp (mainly cattle) 37 95 148X3 farms 37 95 148X2 37 95 148X1 mixed other sp (mainly others) 37 95 148Total Farms 312 799 1248
97
E.1.3 Badger Settings:
Initial Badgers Added per social groupFigures obtained by iterative process to give ratio of badger ages and sexes that the model stabilises at.Juvenile male 0 or 1
(mean 0.8)Yearling male 0 or 1
(mean 0.6)Adult male 1 or 2
(mean 1.2)Juvenile female 0 or 1
(mean 0.9)Yearling female 0 or 1
(mean 0.7)Adult female 2 or 3
(mean 2.1)Mortality Rates Taken from fusion life tables of always-healthy badgers (Wilkinson et a.l 2000) at Woodchester Park. Mortality of first 2-months from pre-capture mortality estimates. Mortality of 2-month periods (including first 2-months immediately after pre-capture) from life table of annual mortality: male = 0.304, female = 0.236. Annual mortality of super-excretors: male = 0.667, female = 0.480.male 1st 2m pre-emergence 0.2400female 1st 2m pre-emergence 0.2400male not-super-excretor 0.0586female not-super-excretor 0.0439male super-excretor 0.1675female super-excretor 0.1033Breeding ProbabilitiesFirst female 0.852nd female [adjustable(1)] 0.40 +/- 3rd female [adjustable(1)] 0.40 +/- 4th female [adjustable(1)] 0.40 +/- (1) Note: the probabilities of 2nd/3rd/4th female breeding are adjusted to be linearly inversely proportional to group size – so smaller groups may breed back up to size faster. The adjustment is based on the equation: “0.40 + (n - 6.7) * -0.079”, but limited between the values 0.00 and 0.85.Litter Size ProbabilitiesTaken from Neal & Cheeseman (Neal & Cheeseman, 1996), p1601 cub 0.082 cubs 0.183 cubs 0.514 cubs 0.185 cubs 0.05Dispersal Probabilitiesmale 0.009390female 0.000834Health-Status Transfer Probabilitiesinfected to infectious 0.0309
98
infected to super- 0.0274male infectious to infected 0.1660male infectious to super- 0.2511female infectious to infected 0.1660female infectious to super- 0.2511Infection Transmission Probabilitiesi=infectious, si=super-infectiousRates set to give a badger prevalence of ~18%, and a CHB rate of ~ 8% i-badger-badger within-group 0.021000i-badger-badger between-group 0.001050i-badger-cow 0si-badger-badger within-group 0.042000si-badger-badger between-group 0.002100si-badger-cow 0.0005
E.1.4 Cattle Settings:Farm NumbersCalculated from June Census 2004. Farms mixed with other species (pigs, sheep etc) divided equally between X1-X4. The values given below are for the 400km2 grid. Numbers are multiplied proportionally for larger grid-sizes to maintain the right farm densities (see E.1.2).Total Farms 312Beef farms 70Dairy farms 58Mixed farms 18X4 (mixed other sp (mainly cattle)) 37X3 farms 37X2 37X1 (mixed other sp (mainly others)) 37Grazing ProportionsCalculated from June Census 2004Beef farms 0.26Dairy farms 0.46X4 0.20X3 0.15X2 0.10X1 0.05Stocking Density (Beef) ProbabilitiesCalculated from June Census 20040.5 cattle per hectare 0.0931.0 0.1441.5 0.1902.0 0.1872.5 0.1303.0 0.0893.5 0.0544.0 0.0344.5 0.0225.0 0.0165.5 0.0116.0 0.007
99
6.5 0.0057.0 0.0047.5 0.0048.0 0.0028.5 0.0029.0 0.0029.5 0.00210.0 0.002Stocking Density (Dairy) ProbabilitiesCalculated from June Census 20040.5 cattle per hectare 0.0331.0 0.0581.5 0.1152.0 0.2172.5 0.2043.0 0.1513.5 0.0854.0 0.0514.5 0.0285.0 0.0195.5 0.0116.0 0.0106.5 0.0047.0 0.0037.5 0.0018.0 0.0038.5 0.0039.0 0.0009.5 0.00310.0 0.001Over-Stock Limit(the herd-size multiplier which determines the threshold when herd-size must be corrected by "selling-movements"). A value of 1.0 means zero-tolerance to herd-size changes, and cattle are sold/bought the same time-step to correct.
1.0
Under-Stock Limit(the herd-size multiplier which determines the threshold when herd-size must be corrected by "buying-movements". A value of 1.0 means zero-tolerance to herd-size changes, and cattle are sold/bought the same time-step to correct).
1.0
Beef Age/Sex ProfileCalculated from June Census 2004male 1-yr-old 0.16male 2-yr-old 0.13male 3-yr-old 0.04male 4-yr-old 0.00male 5-yr-old 0.00female 1-yr-old 0.14female 2-yr-old 0.18female 3-yr-old 0.20female 4-yr-old 0.10female 5-yr-old 0.05
100
Dairy Age/Sex ProfileCalculated from June Census 2004male 1-yr-old 0.07male 2-yr-old 0.05male 3-yr-old 0.01male 4-yr-old 0.00male 5-yr-old 0.00female 1-yr-old 0.13female 2-yr-old 0.22female 3-yr-old 0.30female 4-yr-old 0.15female 5-yr-old 0.07Cattle BirthRate (per 2m time step)Taken from Defra Stats Report (Economics of Milk Production - England and Wales 2002/03, chapter 5: autumn calvers = 0.93 calves per year, spring calvers = 0.96 calves per year)
0.159
Mortality RatesCalculated from CTS slaughter data 2002-2004beef male, 6months x 1 0.0186beef male, 6months x 2 0.0113beef male, 6months x 3 0.0821beef male, 6months x 4 0.0698beef male, 6months x 5 0.3958beef male, 6months x 6 0.5479beef male, 6months x 7 0.1642beef male, 6months x 8 0.1796beef male, 6months x 9 0.1573beef male, 6months x 10 0.2028beef male, 6months x 11+ 0.1565beef female, 6months x 1 0.0182beef female, 6months x 2 0.0072beef female, 6months x 3 0.0145beef female, 6months x 4 0.1127beef female, 6months x 5 0.3354beef female, 6months x 6 0.2353beef female, 6months x 7 0.1025beef female, 6months x 8 0.1229beef female, 6months x 9 0.1238beef female, 6months x 10 0.1347beef female, 6months x 11+ 0.1807dairy male, 6months x 1 0.1394dairy male, 6months x 2 0.0180dairy male, 6months x 3 0.0974dairy male, 6months x 4 0.0594dairy male, 6months x 5 0.3738dairy male, 6months x 6 0.5367dairy male, 6months x 7 0.1691dairy male, 6months x 8 0.1684dairy male, 6months x 9 0.1258dairy male, 6months x 10 0.1616dairy male, 6months x 11+ 0.1484dairy female, 6months x 1 0.0565
101
dairy female, 6months x 2 0.0116dairy female, 6months x 3 0.0096dairy female, 6months x 4 0.0303dairy female, 6months x 5 0.0500dairy female, 6months x 6 0.0801dairy female, 6months x 7 0.0804dairy female, 6months x 8 0.0910dairy female, 6months x 9 0.1209dairy female, 6months x 10 0.1328dairy female, 6months x 11+ 0.1921TB-Test probabilities based on data analysis from Tony Goodchild giving
sensitivity of 70% for standard interpretation test and 90% for severe interpretation test and ratio of Inconclusive Reactors (IRs) to Conclusive Reactors (CRs) of 1.0 for infected cattle and 0.011 for infectious cattle.
based on data analysis from Tony Goodchild giving specificity of 99.7% for IRs (standard and severe test), and 99.935% (standard test) and 99.8 (severe test) for CRs.
and based on finding that "super-excretor" cattle do not respond to the TB test
Standard TB-Test probabilities of CRhealth category 1 0.0007health category 2 0.3500health category 3 0.6900health category 4 0.0007health category 5 0.3500health category 6 0.3500Standard TB-Test probabilities of IRhealth category 1 0.0030health category 2 0.3500health category 3 0.0100health category 4 0.0030health category 5 0.3500health category 6 0.3500Severe TB-Test probabilities of CRhealth category 1 0.0020health category 2 0.4500health category 3 0.8900health category 4 0.0020health category 5 0.4500health category 6 0.4500Severe TB-Test probabilities of IRhealth category 1 0.0030health category 2 0.4500health category 3 0.0100health category 4 0.0030health category 5 0.4500health category 6 0.4500TB-Detect Probability at Slaughter (of an infected animal)Calculated from CTS data
0.217
Infection Transmission Probabilities
102
Infectious & super-infectious transmission rates not differentiated for cattle. Dairy and Beef are differentiated (Munroe & Dohoo, 1999)dairy-cow to cow within-group 0.007100dairy-cow to cow between-group 0.000355Super-infectious dairy-cow to badger 0.000050Beef-cow to cow within-group 0.014300Beef-cow to cow between-group 0.000715Super-infectious Beef-cow to badger 0.000050Health-Status Transfer Probs (Disease progression)from (Fischer et al., 2005)male infected to infectious 0.42male infected to super- 0.001female infected to infectious 0.42female infected to super- 0.001male infectious to infected 0male infectious to super- 0.001female infectious to infected 0female infectious to super- 0.001
103
Appendix F – Parameter Bounds (CSL model Sensitivity Analysis)
Upper and lower parameter bounds used in the CSL model sensitivity analysis. 50 simulations were run for each parameter setting, and each run was simulated for 135 years. Output parameters were calculated for the final 10-year period. The discounted net benefits of four control options (shooting, trapping, snaring and gassing) were also included in the CBA output of the CSL model sensitivity analysis.
SA#, min
SA#, max Badger Settings: Default Min Max
1 2 Badger Groups 300 150 4003 4 Average Carrying Capacity 3.00 2.5 3.5
First Control Year 120Last Control Year 124Mortality Rates
5 6 male 1st 2m pre-emergence 0.2400 0.14 0.34female 1st 2m pre-emergence 0.2400 0.14 0.34
7 8 male not-super-excretor 0.0586 0.04 0.08female not-super-excretor 0.0439 0.025 0.06
9 10 male super-excretor 0.1675 0.1 0.25female super-excretor 0.1033 0.05 0.15Breeding Probabilities
11 12
a female 0.85 0.70 0.90a 2nd female 0.40 0.30 0.45a 3rd female 0.40 0.30 0.45a 4th female 0.40 0.30 0.45Dispersal Probabilities
13 14 male 0.009390 0.00751 0.011female 0.000834 0.000667 0.001Health-Status Transfer Probs
15 16 bt23 infected to infectious 0.0309 0.012 0.0517 18 bt24 infected to super-infectious 0.0274 0.01 0.04
19 20 bt32 male infectious to infected 0.1660 0.07 0.27bt32 female infectious to infected 0.1660 0.07 0.27
21 22 bt34 male infectious to super-infectious 0.2511 0.1 0.4bt34 female infectious to super-infectious 0.2511 0.1 0.4Infection Transmission Probs
23 24 i-badger-badger within-group 0.021200 0.01 0.03si-badger-badger within-group 0.042400 0.02 0.06
25 26 i-badger-badger between-group 0.001060 0.0005 0.0015si-badger-badger between-group 0.002120 0.001 0.003
27 28 i-badger-cow 0 0 0.000275si-badger-cow 0.000393 0.0002360.000549
104
SA#, min
SA#, max Cattle Settings: Default Min Max
Farm Numbers
29 30
Total Farms 312 154 416Beef farms 70 35 94Dairy farms 58 29 78Mixed farms 18 9 24X4 (mixed other sp (mainly cattle) 37 18 49X3 farms 37 18 49X2 37 18 49X1 (mixed other sp (mainly others) 37 18 49
31 32 Beef farms 162 312 0Dairy farms 150 0 312Grazing Proportions
33 34 Beef farms 0.26 0.13 0.3935 36 Dairy farms 0.46 0.23 0.69
37 38
X4 0.20 0.1 0.3X3 0.15 0.075 0.225X2 0.10 0.05 0.15X1 0.05 0.025 0.075Stocking Density (Beef) probs
39 40
0.5 cattle per hectare 0.093 0.18 0.0061.0 0.144 0.2 0.0211.5 0.190 0.18 0.0452.0 0.187 0.146 0.0872.5 0.130 0.101 0.143.0 0.089 0.071 0.183.5 0.054 0.04 0.144.0 0.034 0.026 0.0874.5 0.022 0.016 0.0525.0 0.016 0.01 0.03025.5 0.011 0.008 0.01516.0 0.007 0.005 0.01116.5 0.005 0.003 0.0087.0 0.004 0.002 0.00527.5 0.004 0.002 0.0048.0 0.002 0.002 0.0028.5 0.002 0.002 0.0029.0 0.002 0.002 0.0029.5 0.002 0.002 0.002Stocking Density (Dairy) probs
41 42 0.5 cattle per hectare 0.033 0.156 0.01151.0 0.058 0.194 0.0271.5 0.115 0.235 0.04432.0 0.217 0.187 0.08772.5 0.204 0.138 0.11233.0 0.151 0.09 0.1885
105
SA#, min
SA#, max Cattle Settings: Default Min Max
3.5 0.085 0.045 0.19184.0 0.051 0.02 0.12624.5 0.028 0.012 0.085.0 0.019 0.009 0.02545.5 0.011 0.007 0.01646.0 0.010 0.005 0.016.5 0.004 0.004 0.0087.0 0.003 0.003 0.0057.5 0.001 0.002 0.0038.0 0.003 0.001 0.0028.5 0.003 0 0.0019.0 0.000 0 09.5 0.003 0 0
43 44 Over-Stock Limit 1.2 1.05 1.4Under-Stock Limit 0.8 0.95 0.6Beef Age/Sex Profilemale 1-yr-old 0.16male 2-yr-old 0.13male 3-yr-old 0.04male 4-yr-old 0.00male 5-yr-old 0.00female 1-yr-old 0.14female 2-yr-old 0.18female 3-yr-old 0.20female 4-yr-old 0.10female 5-yr-old 0.05Dairy Age/Sex Profilemale 1-yr-old 0.07male 2-yr-old 0.05male 3-yr-old 0.01male 4-yr-old 0.00male 5-yr-old 0.00female 1-yr-old 0.13female 2-yr-old 0.22female 3-yr-old 0.30female 4-yr-old 0.15female 5-yr-old 0.07Cattle BirthRate (per 2m time step) 0.159Mortality Rates (2 monthly)
45 46 beef male, 6months x 1 0.0186 0.0167 0.0205beef male, 6months x 2 0.0113 0.0102 0.0124beef male, 6months x 3 0.0821 0.0739 0.0903beef male, 6months x 4 0.0698 0.0628 0.0768beef male, 6months x 5 0.3958 0.3562 0.4354beef male, 6months x 6 0.5479 0.4931 0.6027beef male, 6months x 7 0.1642 0.1478 0.1806
106
SA#, min
SA#, max Cattle Settings: Default Min Max
beef male, 6months x 8 0.1796 0.1616 0.1976beef male, 6months x 9 0.1573 0.1416 0.1730beef male, 6months x 10 0.2028 0.1825 0.2231beef male, 6months x 11+ 0.1565 0.1409 0.1722beef female, 6months x 1 0.0182 0.0164 0.0200beef female, 6months x 2 0.0072 0.0065 0.0079beef female, 6months x 3 0.0145 0.0131 0.0160beef female, 6months x 4 0.1127 0.1014 0.1240beef female, 6months x 5 0.3354 0.3019 0.3689beef female, 6months x 6 0.2353 0.2118 0.2588beef female, 6months x 7 0.1025 0.0923 0.1128beef female, 6months x 8 0.1229 0.1106 0.1352beef female, 6months x 9 0.1238 0.1114 0.1362beef female, 6months x 10 0.1347 0.1212 0.1482beef female, 6months x 11+ 0.1807 0.1626 0.1988
47 48
dairy male, 6months x 1 0.1394 0.1255 0.1533dairy male, 6months x 2 0.0180 0.0162 0.0198dairy male, 6months x 3 0.0974 0.0877 0.1071dairy male, 6months x 4 0.0594 0.0535 0.0653dairy male, 6months x 5 0.3738 0.3364 0.4112dairy male, 6months x 6 0.5367 0.4830 0.5904dairy male, 6months x 7 0.1691 0.1522 0.1860dairy male, 6months x 8 0.1684 0.1516 0.1852dairy male, 6months x 9 0.1258 0.1132 0.1384dairy male, 6months x 10 0.1616 0.1454 0.1778dairy male, 6months x 11+ 0.1484 0.1336 0.1632dairy female, 6months x 1 0.0565 0.0509 0.0622dairy female, 6months x 2 0.0116 0.0104 0.0128dairy female, 6months x 3 0.0096 0.0086 0.0106dairy female, 6months x 4 0.0303 0.0273 0.0333dairy female, 6months x 5 0.0500 0.0450 0.0550dairy female, 6months x 6 0.0801 0.0721 0.0881dairy female, 6months x 7 0.0804 0.0724 0.0884dairy female, 6months x 8 0.0910 0.0819 0.1001dairy female, 6months x 9 0.1209 0.1088 0.1330dairy female, 6months x 10 0.1328 0.1195 0.1461dairy female, 6months x 11+ 0.1921 0.1729 0.2113Standard TB-Test probabilities of CR
49 50 health category 1 0.0007 0.00035 0.0014
51 52 health category 2 0.3500 0.175 0.700health category 3 0.6900 0.35 1.0
49 50 health category 4 0.0007 0.00035 0.0014
51 52 health category 5 0.3500 0.175 0.700health category 6 0.3500 0.175 0.700Standard TB-Test probabilities of IR
49 50 health category 1 0.0030 0.0015 0.006
107
SA#, min
SA#, max Cattle Settings: Default Min Max
51 52 health category 2 0.3500 0.175 0.7health category 3 0.0100 0.005 0.02
49 50 health category 4 0.0030 0.0015 0.006
51 52 health category 5 0.3500 0.175 0.7health category 6 0.3500 0.175 0.7Severe TB-Test probabilities of CR
49 50 health category 1 0.0020 0.001 0.004
51 52 health category 2 0.4500 0.225 0.9health category 3 0.8900 0.445 1.78
49 50 health category 4 0.0020 0.001 0.004
51 52 health category 5 0.4500 0.225 0.9health category 6 0.4500 0.225 0.9Severe TB-Test probabilities of IR
49 50 health category 1 0.0030 0.0015 0.006
51 52 health category 2 0.4500 0.225 0.9health category 3 0.0100 0.005 0.02
49 50 health category 4 0.0030 0.0015 0.006
51 52 health category 5 0.4500 0.225 0.9health category 6 0.4500 0.225 0.9
53 54 TB-Detect Probability at Slaughter 0.217 0.1085 0.434
Infection Transmission Probs
55 56
i-beefcow-cow within-group 0.0204320.010216 0.040864si-beefcow-cow within-group 0.0204320.010216 0.040864i-dairycow-cow within-group 0.0100840.005042 0.020168si-dairycow-cow within-group 0.0100840.005042 0.020168
57 58
i-beefcow-cow between-group 0.0010220.000511 0.002044si-beefcow-cow between-group 0.0010220.000511 0.002044i-dairycow-cow between-group 0.0005040.000252 0.001008si-dairycow-cow between-group 0.0005040.000252 0.001008
59 60
i-beefcow-badger 0.000000 0 0.000112si-beefcow-badger 0.0000560.000028 0.000112i-dairycow-badger 0.000000 0 0.000112si-dairycow-badger 0.0000560.000028 0.000112Health-Status Transfer Probs
61 62 ct23 infected to infectious 0.420 0.21 0.8463 64 ct24 infected to super-infectious 0.000 0 0.00265 66 ct34 infectious to super-infectious 0.001 0.0005 0.002
Pre-Movement TestingPMT switched on? TRUEPMTAge1 (months) for yr 120 16PMTAge2 (months) for yr 121+ 2
108
Appendix G – Sensitivity Analysis Output
Sensitivity Analysis Outputs:
L = Lower bound, U = Upper boundYellow = default, blue = badger parameters, green = cattle parameters, and brown = costing parameters.
Table G.1. CSL model, ranked parameters – sensitivity of badger populationFile Parameter Bound bPop %Change
SA000 All Default Values Used - 7.465 -SA008 Badger Mortality rates - not Super-infectious U 5.709 -23.53SA007 Badger Mortality rates - not Super-infectious L 8.953 19.93SA011 Breeding probabilities for females L 6.659 -10.80SA006 Badger Mortality rates - pre-emerge U 6.990 -6.36SA005 Badger Mortality rates - pre-emerge L 7.928 6.20SA010 Badger Mortality rates - Super-infectious U 7.917 6.05SA023 Infection Transmision probabilities - within group L 7.914 6.01SA017 Health Status Transfer probabilities, badger 2 to 4 L 7.883 5.60SA024 Infection Transmision probabilities - within group U 7.060 -5.43SA015 Health Status Transfer probabilities, badger 2 to 3 L 7.867 5.38SA025 Infection Transmision probabilities - between group L 7.848 5.13SA003 Carrying Capacity L 7.101 -4.88SA012 Breeding probabilities for females U 7.818 4.72SA016 Health Status Transfer probabilities, badger 2 to 3 U 7.123 -4.59SA018 Health Status Transfer probabilities, badger 2 to 4 U 7.132 -4.47SA026 Infection Transmision probabilities - between group U 7.227 -3.19SA004 Carrying Capacity U 7.682 2.91SA019 Health Status Transfer probabilities, badger 3 to 2 L 7.308 -2.11SA020 Health Status Transfer probabilities, badger 3 to 2 U 7.622 2.09SA021 Health Status Transfer probabilities, badger 3 to 4 L 7.612 1.96SA022 Health Status Transfer probabilities, badger 3 to 4 U 7.417 -0.65SA032 All Dairy Farms - 7.513 0.63SA109 Reactors per CHB + MovemtRestrCosts L 7.509 0.59SA045 Mortality Rates - beef L 7.509 0.58SA108 Reactors per CHB + MovemtRestrCosts U 7.506 0.54SA063 Health Status Transfer probabilities, cattle 2 to 4 L 7.506 0.54SA103 Culling Costs L 7.504 0.52SA053 TB-Detect probability at Slaughter L 7.501 0.48SA033 Grazing Proportions - beef farms L 7.500 0.47SA041 Stocking Density distribution - dairy farms L 7.500 0.46SA064 Health Status Transfer probabilities, cattle 2 to 4 U 7.500 0.46SA059 Infection Transmision probabilities - cow to badger L 7.500 0.46SA043 Over/Under Stock Limits L 7.499 0.45SA107 beefSVmu + dairySVmu + farmCHBmu L 7.499 0.45SA046 Mortality Rates - beef U 7.499 0.45SA002 Number of Badger Groups in grid (group density) U 7.499 0.45SA042 Stocking Density distribution - dairy farms U 7.499 0.44SA069 Test Interval L 7.496 0.41SA058 Infection Transmision probabilities - between herd U 7.495 0.39SA102 Culling Costs U 7.495 0.39SA048 Mortality Rates - dairy U 7.494 0.39SA067 PreMovementTesting switched off - 7.494 0.38SA028 Infection Transmision probabilities - badger to cow U 7.494 0.38
109
SA056 Infection Transmision probabilities - within herd U 7.493 0.37SA052 TB Test probabilities - Sensitivity U 7.493 0.37SA030 Farm Numbers U 7.492 0.36SA031 All Beef Farms - 7.492 0.35SA054 TB-Detect probability at Slaughter U 7.491 0.35SA055 Infection Transmision probabilities - within herd L 7.490 0.34SA027 Infection Transmision probabilities - badger to cow L 7.490 0.33SA106 beefSVmu + dairySVmu + farmCHBmu U 7.488 0.30SA047 Mortality Rates - dairy L 7.486 0.28SA060 Infection Transmision probabilities - cow to badger U 7.446 -0.26SA104 Setup Costs U 7.483 0.23SA013 Dispersal probabilities L 7.482 0.23SA009 Badger Mortality rates - Super-infectious L 7.482 0.22SA035 Grazing Proportions - dairy farms L 7.481 0.21SA050 TB Test probabilities - Specificity U 7.480 0.19SA061 Health Status Transfer probabilities, cattle 2 to 3 L 7.479 0.18SA036 Grazing Proportions - dairy farms U 7.478 0.17SA065 Health Status Transfer probabilities, cattle 3 to 4 L 7.478 0.17SA057 Infection Transmision probabilities - between herd L 7.478 0.16SA001 Number of Badger Groups in grid (group density) L 7.454 -0.16SA029 Farm Numbers L 7.477 0.16SA044 Over/Under Stock Limits U 7.477 0.15SA039 Stocking Density distribution - beef farms L 7.477 0.15SA014 Dispersal probabilities U 7.477 0.15SA105 Setup Costs L 7.477 0.15SA051 TB Test probabilities - Sensitivity L 7.476 0.14SA037 Grazing Proportions - mixed-species farms L 7.475 0.13SA049 TB Test probabilities - Specificity L 7.474 0.11SA034 Grazing Proportions - beef farms U 7.474 0.11SA070 Test Interval U 7.474 0.11SA038 Grazing Proportions - mixed-species farms U 7.472 0.09SA040 Stocking Density distribution - beef farms U 7.471 0.07SA066 Health Status Transfer probabilities, cattle 3 to 4 U 7.461 -0.06SA062 Health Status Transfer probabilities, cattle 2 to 3 U 7.464 -0.03
Table G.2 CSL model, ranked parameters – sensitivity of badger prevalenceFile Parameter Bound bPrev %Change
SA000 All Default Values Used - 0.1889 -SA009 Badger Mortality rates - Super-infectious L 0.7264 284.46SA007 Badger Mortality rates - not Super-infectious L 0.5650 199.06SA008 Badger Mortality rates - not Super-infectious U 0.0002 -99.87SA010 Badger Mortality rates - Super-infectious U 0.0003 -99.86SA023 Infection Transmision probabilities - within group L 0.0007 -99.65SA017 Health Status Transfer probabilities, badger 2 to 4 L 0.0088 -95.35SA025 Infection Transmision probabilities - between group L 0.0120 -93.62SA024 Infection Transmision probabilities - within group U 0.3646 92.99SA015 Health Status Transfer probabilities, badger 2 to 3 L 0.0236 -87.49SA011 Breeding probabilities for females L 0.0291 -84.58SA026 Infection Transmision probabilities - between group U 0.3188 68.71SA016 Health Status Transfer probabilities, badger 2 to 3 U 0.3069 62.41SA006 Badger Mortality rates - pre-emerge U 0.0716 -62.13SA018 Health Status Transfer probabilities, badger 2 to 4 U 0.2865 51.62SA005 Badger Mortality rates - pre-emerge L 0.2738 44.94SA003 Carrying Capacity L 0.1080 -42.85SA020 Health Status Transfer probabilities, badger 3 to 2 U 0.1220 -35.44SA012 Breeding probabilities for females U 0.2426 28.38
110
SA019 Health Status Transfer probabilities, badger 3 to 2 L 0.2401 27.08SA021 Health Status Transfer probabilities, badger 3 to 4 L 0.1402 -25.81SA004 Carrying Capacity U 0.2327 23.17SA041 Stocking Density distribution - dairy farms L 0.1702 -9.93SA053 TB-Detect probability at Slaughter L 0.1703 -9.84SA067 PreMovementTesting switched off - 0.172 -8.83SA002 Number of Badger Groups in grid (group density) U 0.1723 -8.80SA032 All Dairy Farms - 0.1744 -7.69SA033 Grazing Proportions - beef farms L 0.1747 -7.54SA028 Infection Transmision probabilities - badger to cow U 0.1755 -7.11SA035 Grazing Proportions - dairy farms L 0.1758 -6.94SA044 Over/Under Stock Limits U 0.1760 -6.84SA045 Mortality Rates - beef L 0.1762 -6.74SA054 TB-Detect probability at Slaughter U 0.1763 -6.68SA055 Infection Transmision probabilities - within herd L 0.1765 -6.58SA013 Dispersal probabilities L 0.1767 -6.46SA031 All Beef Farms - 0.1769 -6.35SA059 Infection Transmision probabilities - cow to badger L 0.1773 -6.15SA069 Test Interval L 0.177 -6.08SA049 TB Test probabilities - Specificity L 0.1775 -6.05SA103 Culling Costs L 0.178 -6.00SA104 Setup Costs U 0.178 -5.88SA034 Grazing Proportions - beef farms U 0.1779 -5.86SA046 Mortality Rates - beef U 0.1779 -5.84SA109 Reactors per CHB + MovemtRestrCosts L 0.178 -5.79SA065 Health Status Transfer probabilities, cattle 3 to 4 L 0.1784 -5.56SA043 Over/Under Stock Limits L 0.1785 -5.51SA056 Infection Transmision probabilities - within herd U 0.1787 -5.43SA042 Stocking Density distribution - dairy farms U 0.1790 -5.27SA047 Mortality Rates - dairy L 0.1792 -5.14SA038 Grazing Proportions - mixed-species farms U 0.1800 -4.74SA106 beefSVmu + dairySVmu + farmCHBmu U 0.180 -4.73SA070 Test Interval U 0.180 -4.53SA102 Culling Costs U 0.181 -4.20SA058 Infection Transmision probabilities - between herd U 0.1812 -4.11SA036 Grazing Proportions - dairy farms U 0.1814 -3.98SA107 beefSVmu + dairySVmu + farmCHBmu L 0.182 -3.89SA037 Grazing Proportions - mixed-species farms L 0.1816 -3.87SA064 Health Status Transfer probabilities, cattle 2 to 4 U 0.1817 -3.81SA051 TB Test probabilities - Sensitivity L 0.1819 -3.75SA063 Health Status Transfer probabilities, cattle 2 to 4 L 0.1821 -3.61SA061 Health Status Transfer probabilities, cattle 2 to 3 L 0.1822 -3.57SA105 Setup Costs L 0.182 -3.55SA108 Reactors per CHB + MovemtRestrCosts U 0.182 -3.45SA027 Infection Transmision probabilities - badger to cow L 0.1825 -3.39SA052 TB Test probabilities - Sensitivity U 0.1827 -3.32SA001 Number of Badger Groups in grid (group density) L 0.1827 -3.30SA057 Infection Transmision probabilities - between herd L 0.1830 -3.15SA030 Farm Numbers U 0.1833 -2.97SA048 Mortality Rates - dairy U 0.1834 -2.93SA040 Stocking Density distribution - beef farms U 0.1839 -2.67SA039 Stocking Density distribution - beef farms L 0.1840 -2.60SA062 Health Status Transfer probabilities, cattle 2 to 3 U 0.1856 -1.76SA022 Health Status Transfer probabilities, badger 3 to 4 U 0.1921 1.65SA066 Health Status Transfer probabilities, cattle 3 to 4 U 0.1859 -1.60SA029 Farm Numbers L 0.1868 -1.13
111
SA060 Infection Transmision probabilities - cow to badger U 0.1879 -0.56SA050 TB Test probabilities - Specificity U 0.1896 0.35SA014 Dispersal probabilities U 0.1891 0.10
Table G.3 CSL model, ranked parameters – sensitivity of cattle herd breakdown rateFile Parameter Bound cCHBs %Change
SA000 All Default Values Used - 0.0462 -SA009 Badger Mortality rates - Super-infectious L 0.1581 242.13SA007 Badger Mortality rates - not Super-infectious L 0.1200 159.60SA051 TB Test probabilities - Sensitivity L 0.0934 102.11SA008 Badger Mortality rates - not Super-infectious U 0.0000 -99.93SA010 Badger Mortality rates - Super-infectious U 0.0001 -99.75SA023 Infection Transmision probabilities - within group L 0.0002 -99.58SA067 PreMovementTesting switched off - 0.092 98.25SA017 Health Status Transfer probabilities, badger 2 to 4 L 0.0021 -95.40SA025 Infection Transmision probabilities - between group L 0.0038 -91.72SA015 Health Status Transfer probabilities, badger 2 to 3 L 0.0068 -85.23SA011 Breeding probabilities for females L 0.0080 -82.60SA029 Farm Numbers L 0.0797 72.34SA024 Infection Transmision probabilities - within group U 0.0745 61.10SA006 Badger Mortality rates - pre-emerge U 0.0191 -58.77SA018 Health Status Transfer probabilities, badger 2 to 4 U 0.0718 55.24SA016 Health Status Transfer probabilities, badger 2 to 3 U 0.0715 54.65SA026 Infection Transmision probabilities - between group U 0.0687 48.68SA003 Carrying Capacity L 0.0274 -40.80SA005 Badger Mortality rates - pre-emerge L 0.0649 40.38SA020 Health Status Transfer probabilities, badger 3 to 2 U 0.0317 -31.41SA027 Infection Transmision probabilities - badger to cow L 0.0323 -30.07SA040 Stocking Density distribution - beef farms U 0.0599 29.55SA012 Breeding probabilities for females U 0.0591 27.96SA019 Health Status Transfer probabilities, badger 3 to 2 L 0.0587 26.92SA021 Health Status Transfer probabilities, badger 3 to 4 L 0.0339 -26.70SA028 Infection Transmision probabilities - badger to cow U 0.0582 25.85SA004 Carrying Capacity U 0.0574 24.24SA032 All Dairy Farms - 0.0573 24.05SA033 Grazing Proportions - beef farms L 0.0355 -23.30SA030 Farm Numbers U 0.0358 -22.56SA031 All Beef Farms - 0.0565 22.30SA035 Grazing Proportions - dairy farms L 0.0363 -21.58SA041 Stocking Density distribution - dairy farms L 0.0388 -16.13SA037 Grazing Proportions - mixed-species farms L 0.0390 -15.57SA050 TB Test probabilities - Specificity U 0.0390 -15.53SA034 Grazing Proportions - beef farms U 0.0533 15.38SA001 Number of Badger Groups in grid (group density) L 0.0392 -15.09SA061 Health Status Transfer probabilities, cattle 2 to 3 L 0.0408 -11.72SA038 Grazing Proportions - mixed-species farms U 0.0513 10.97SA055 Infection Transmision probabilities - within herd L 0.0412 -10.83SA053 TB-Detect probability at Slaughter L 0.0418 -9.55SA044 Over/Under Stock Limits U 0.0418 -9.53SA022 Health Status Transfer probabilities, badger 3 to 4 U 0.0502 8.65SA036 Grazing Proportions - dairy farms U 0.0502 8.50SA054 TB-Detect probability at Slaughter U 0.0426 -7.85SA062 Health Status Transfer probabilities, cattle 2 to 3 U 0.0492 6.45SA065 Health Status Transfer probabilities, cattle 3 to 4 L 0.0433 -6.30SA039 Stocking Density distribution - beef farms L 0.0434 -6.18SA109 Reactors per CHB + MovemtRestrCosts L 0.043 -6.10
112
SA045 Mortality Rates - beef L 0.0436 -5.64SA047 Mortality Rates - dairy L 0.0436 -5.60SA103 Culling Costs L 0.044 -4.69SA059 Infection Transmision probabilities - cow to badger L 0.0442 -4.40SA107 beefSVmu + dairySVmu + farmCHBmu L 0.044 -4.22SA049 TB Test probabilities - Specificity L 0.0481 3.95SA046 Mortality Rates - beef U 0.0446 -3.48SA104 Setup Costs U 0.045 -3.40SA052 TB Test probabilities - Sensitivity U 0.0447 -3.23SA048 Mortality Rates - dairy U 0.0475 2.82SA063 Health Status Transfer probabilities, cattle 2 to 4 L 0.0450 -2.70SA069 Test Interval L 0.045 -2.23SA013 Dispersal probabilities L 0.0453 -2.04SA105 Setup Costs L 0.045 -1.86SA108 Reactors per CHB + MovemtRestrCosts U 0.045 -1.80SA002 Number of Badger Groups in grid (group density) U 0.0455 -1.47SA057 Infection Transmision probabilities - between herd L 0.0456 -1.46SA060 Infection Transmision probabilities - cow to badger U 0.0469 1.41SA042 Stocking Density distribution - dairy farms U 0.0468 1.16SA064 Health Status Transfer probabilities, cattle 2 to 4 U 0.0457 -1.04SA066 Health Status Transfer probabilities, cattle 3 to 4 U 0.0466 0.78SA014 Dispersal probabilities U 0.0465 0.69SA058 Infection Transmision probabilities - between herd U 0.0459 -0.68SA106 beefSVmu + dairySVmu + farmCHBmu U 0.046 -0.67SA056 Infection Transmision probabilities - within herd U 0.0460 -0.51SA043 Over/Under Stock Limits L 0.0464 0.37SA102 Culling Costs U 0.046 -0.31SA070 Test Interval U 0.046 0.19
113
Table G.4 CSL model, ranked parameters – sensitivity of benefit break-even (% of simulations that give a net cost) with shooting.Only the top 20 of the ranked parameters are shown.
NPV 'BE'File Parameter Bound Shooting %Change
SA000 All Default Values Used - 87 -SA009 Badger Mortality rates - Super-infectious L 21 66SA007 Badger Mortality rates - not Super-infectious L 41 46SA051 TB Test probabilities - Sensitivity L 57 30SA036 Grazing Proportions - dairy farms U 57 30SA026 Infection Transmision probabilities - between group U 57 30SA016 Health Status Transfer probabilities, badger 2 to 3 U 59 28SA029 Farm Numbers L 59 28SA056 Infection Transmision probabilities - within herd U 65 22SA024 Infection Transmision probabilities - within group U 67 20SA004 Carrying Capacity U 67 20SA018 Health Status Transfer probabilities, badger 2 to 4 U 69 18SA042 Stocking Density distribution - dairy farms U 69 18SA028 Infection Transmision probabilities - badger to cow U 69 18SA062 Health Status Transfer probabilities, cattle 2 to 3 U 69 18SA057 Infection Transmision probabilities - between herd L 69 18SA032 All Dairy Farms - 71 16SA012 Breeding probabilities for females U 71 16SA103 Culling Costs L 71 16SA106 beefSVmu + dairySVmu + farmCHBmu U 71 16SA102 Culling Costs U 71 16
Table G.5 CSL model, ranked parameters – sensitivity of benefit break-even (% of simulations that give a net cost) with trappingOnly the top 20 of the ranked parameters are shown.
NPV 'BE'File Parameter Bound Trapping %Change
SA000 All Default Values Used - 101 -SA051 TB Test probabilities - Sensitivity L 91 10SA056 Infection Transmision probabilities - within herd U 91 10SA105 Setup Costs L 91 10SA036 Grazing Proportions - dairy farms U 97 4SA047 Mortality Rates - dairy L 97 4SA062 Health Status Transfer probabilities, cattle 2 to 3 U 97 4SA004 Carrying Capacity U 99 2SA029 Farm Numbers L 99 2SA102 Culling Costs U 99 2SA103 Culling Costs L 99 2SA001 Number of Badger Groups in grid (group density) L 101 0SA002 Number of Badger Groups in grid (group density) U 101 0SA003 Carrying Capacity L 101 0SA005 Badger Mortality rates - pre-emerge L 101 0SA006 Badger Mortality rates - pre-emerge U 101 0SA007 Badger Mortality rates - not Super-infectious L 101 0SA008 Badger Mortality rates - not Super-infectious U 101 0SA009 Badger Mortality rates - Super-infectious L 101 0SA010 Badger Mortality rates - Super-infectious U 101 0SA011 Breeding probabilities for females L 101 0
114
Table G.6 CSL model, ranked parameters – sensitivity of benefit break-even (% of simulations that give a net cost) with snaringOnly the top 20 of the ranked parameters are shown.
NPV 'BE'File Parameter Bound Snaring %Change
SA000 All Default Values Used - 49 -SA008 Badger Mortality rates - not Super-infectious U 101 -52SA010 Badger Mortality rates - Super-infectious U 101 -52SA023 Infection Transmision probabilities - within group L 101 -52SA017 Health Status Transfer probabilities, badger 2 to 4 L 99 -50SA025 Infection Transmision probabilities - between group L 97 -48SA009 Badger Mortality rates - Super-infectious L 7 42SA015 Health Status Transfer probabilities, badger 2 to 3 L 91 -42SA035 Grazing Proportions - dairy farms L 91 -42SA011 Breeding probabilities for females L 87 -38SA007 Badger Mortality rates - not Super-infectious L 13 36SA031 All Beef Farms - 83 -34SA032 All Dairy Farms - 15 34SA018 Health Status Transfer probabilities, badger 2 to 4 U 17 32SA016 Health Status Transfer probabilities, badger 2 to 3 U 19 30SA006 Badger Mortality rates - pre-emerge U 77 -28SA028 Infection Transmision probabilities - badger to cow U 21 28SA041 Stocking Density distribution - dairy farms L 77 -28SA105 Setup Costs L 21 28SA004 Carrying Capacity U 25 24SA012 Breeding probabilities for females U 25 24
Table G.7 CSL model, ranked parameters – sensitivity of benefit break-even (% of simulations that give a net cost) with gassingOnly the top 20 of the ranked parameters are shown.
NPV 'BE'File Parameter Bound Gassing %Change
SA000 All Default Values Used - 29 -SA008 Badger Mortality rates - not Super-infectious U 101 -72SA010 Badger Mortality rates - Super-infectious U 101 -72SA023 Infection Transmision probabilities - within group L 101 -72SA017 Health Status Transfer probabilities, badger 2 to 4 L 99 -70SA025 Infection Transmision probabilities - between group L 97 -68SA015 Health Status Transfer probabilities, badger 2 to 3 L 91 -62SA011 Breeding probabilities for females L 83 -54SA031 All Beef Farms - 67 -38SA035 Grazing Proportions - dairy farms L 67 -38SA006 Badger Mortality rates - pre-emerge U 59 -30SA007 Badger Mortality rates - not Super-infectious L 1 28SA009 Badger Mortality rates - Super-infectious L 1 28SA018 Health Status Transfer probabilities, badger 2 to 4 U 3 26SA041 Stocking Density distribution - dairy farms L 55 -26SA104 Setup Costs U 55 -26SA024 Infection Transmision probabilities - within group U 7 22SA067 PreMovementTesting switched off - 51 -22SA027 Infection Transmision probabilities - badger to cow L 49 -20SA051 TB Test probabilities - Sensitivity L 49 -20SA003 Carrying Capacity L 47 -18
115
Table G.8 NU model, ranked parameters – sensitivity of badger population
Variable
Partial regression coefficient F-value significance CSL Rank
Mortality - female adults season 1 -0.397 13.441 0.007472 1,2Excretor infection rate (within group) 0.288 7.219 0.008774 7,9Prob. 1st female breeding 0.295 6.871 0.009643 3,13Mortality - female cubs season 2 -0.268 5.582 0.01448 1,2
Table G.9 NU model, ranked parameters – sensitivity of badger prevalence
Variable
Partial regression coefficient F-value significance CSL Rank
Excretor infection rate (between group) 0.576791 39.884 0 7,11Prob. infected to excretor 0.376442 13.208 0.000491 9,12Super-excretor infection rate (between group) 0.284311 7.035 0.009634 7,11Mortality - female adults season 1 -0.26915 6.248 0.01448 2,3Excretor infection rate (within group) 0.251329 5.394 0.022749 5,8Mortality - female cubs season 1 -0.24169 4.963 0.028701 2,3 (13,15?)Mortality - male juveniles season 1 -0.20866 3.770 0.07641 2,3Mortality - female cubs season 2 -0.20461 3.496 0.065193 2,3
Table G.10 NU model, ranked parameters – sensitivity of cattle CHB rate
Variable
Partial regression coefficient F-value significance CSL Rank
Mortality - female adults season 1 -0.87021 249.5846 0 1,2,4,5Prob. infected to excretor 0.741865 97.9214 0 38,46Mortality - female cubs season 1 -0.60242 45.5716 0 2,4 (14,19?)Within farm infection 0.501986 26.95039 0.000002 40,74Standard TB-test sensitivity -0.27865 6.73441 0.011247 3,58Test frequency -0.25531 5.578093 0.020618 61,77
116
Appendix H – Cattle Testing Flowchart
Flowchart of cattle-testing procedures simulated in the model.
Cattle TestingAnnual
(6/12/24mo)Test
RT severe
RT Std
AbattoirSuspect
NR: Wait interval and retest
IR: MR on IR
IR: remove IR.+ve? No goto RT std after 60 days
Yes goto RT severe
NR = no reactorsIR = Inconclusive reactorsCR = Conclusive reactors+ve = Visib le Lesions or Culture PositiveMR = Movement Restrictions
IR: repeat test after 60 days
IR: repeat test after 60 days
CR: MR, remove CR+ve ?
Previous TB within 3 years ?
No: After 60 days RT Std
Yes: MR on whole herd, after 60 days RT Std
NR: MR lifted, after 6 months go to Annual Test
CR: +ve?No: goto RT std after 60 days
Yes: after 60 days goto RT severe, contiguous herd goto immediate annual test
Yes: after 60 days go to RT severe, contiguous herds go to immediate Annual Test, (within 6-
month?)
No:after 60 days go to RT std
CR:after 60 days goto RT severe
NR: after 60 daysRetest RT
severe
CR:+ve? noyes
CR:+ve? noyes
MR on whole herd go immediately to RT severe
117
Appendix I – Test Interval Algorithm
The Algorithm to determine whether the routine test-interval of a parish should be regraded, either upwards or downwards, was supplied by the veterinary agency, and is the one advised by the EU, and used in the field.
The algorithm to determine switching is as follows:(1) Take the confirmed breaks in a parish over the previous 2 years and divide by the
total herds in the parish, if the % is over 1 then the testing interval is yearly (T1)(2) If not (1) then do as above but consider breaks over 4 years, if the % is greater than
0.2 then the interval is 2 yearly (T2)(3) If not (2) then do as above but consider breaks over 6 years, if the % is greater than
0.1 then the interval is 3 yearly (T3)(4) If none of the above then the interval is 4 yearly (T4)
118
Appendix J – Badger Movement Flowchart
Flow-chart of simulated Badger movements from populous badger groups to smaller ones. More movements occur following culling, as has been seen in real life.
119
For each badger group
For each sex
Count number of badgers [A] of that sex
[A] is less than 2
Yes
Next group
Next sex
For each close group (both immediate neighbours and
“neighbours-but-one”)
Count number of badgers [B] (of sex being checked) in each close group
[B]-[A] is more than 2
Next group
No
Yes No Shortlist close group
All close groups checked ?
No
Select one short-listed “donor” group at random, then randomly
select one badger of required sex, and move to “holding bay”
for “recipient” group”
Both sexes checked? No
All groups checked? No
Yes
START
FINISH
Transfer all badgers in holding-bays to destination groups
Yes
Yes
Appendix K – Model Assumptions
K.1 Model Assumptions
K.1.1 Temporal Assumptions1. Badger perturbation, in terms of increased transmission rates, starts when culling
starts, continues at the same rate (per perturbed badger) and finishes exactly 3 years after the culling finishes.
2. There was no temporal heterogeneity in the rates of badger culling.
K.1.2 Spatial Assumptions1. Farms compliant to badger culling are distributed at random across the space
modelled.2. All land is always compliant to badger vaccination (100% compliance).3. If at least 10 % of a badger territory lies on land that is compliant to culling, the
badgers of that territory could be caught as if all the territory was accessible. Any territory with less than 10% accessible area was assumed to be completely inaccessible for control. This model assumption is likely to result in a cull of more badgers with the gassing option than might occur in the field.
4. There was no spatial heterogeneity in the rates of badger culling.5. The distribution of dairy and beef herds, and their grazing areas in the model are
sufficiently representative of the real landscape, and there is no clustering in real life that affects the disease dynamics.
6. The algorithms for moving cattle in the model include any processes or spatial elements that occur in the field and which are important in the dynamics of the TB disease.
K.1.3 Badger Assumptions1. The transmission rate of TB between badger groups is a twentieth of the within-
group rate, and only occurs where badger territories are contiguous.2. 100% of healthy badgers that eat vaccine bait are consequently fully protected
against TB infection, and this state continues for the rest of its life.3. Before PrMT is instigated, ~40% of CHBs are caused by an initial transmission of
TB infection from a badger, as opposed to cattle to cattle across the farm boundary, or the moving-in of infected cattle onto the farm.
K.1.4 Cattle Assumptions1. The modelled cattle population dynamics, including birth, mortality and cattle
movements, sufficiently represent farming management in the field, and there are no processes in the field that are important drivers in the disease dynamics that aren’t included in the model.
2. The cattle TB transmission rates reported in the Canadian study (Munroe and Dohoo, 1999) are valid parameter values to be used for the UK.
3. The transmission rate of TB between cattle herds is a twentieth of the within-group rate, and only occurs where herd-grazing areas are contiguous.
4. Only TB infectious or super-infectious cattle are given a probability of detection at routine slaughter, as opposed to infected cattle. However, cattle that are tested after having been identified as a reactor to a TB test are detected at post mortem if they are infected, infectious, or super-infectious, under the assumption that cultures would be done.
120
5. The algorithm in the model determining the parish type for testing interval gives a spatial distribution and parish type proportions that represents the high risk areas that would be considered for control.
K.1.5 Economic Assumptions1. Farmers/landowners pay for all the set-up costs for badger culling/vaccination.2. Defra pay for an initial admin cost (licensing and ecological survey) and a later cost
(follow-up ecological survey and analysis of results).3. Farmers/landowners bear the cost of surveying for badger setts at 2 days farm
labour per km2.4. Farmers/landowners pay for all the badger culling/vaccination costs including
equipment, labour, despatch, and disposal costs. These costs are the same as those met during the RBCT.
5. Gamma Interferon tests are done at the same time as skin tests, and so doesn’t require any additional farm labour.
6. The lognormal distribution used for stochastically determining the value of cattle on individual farms gives a close representation of cost distribution in the real landscape.
K.1.6 Badger Control Assumptions1. Control lasts for five years.2. The control area is contiguous, with a focus on and around the parish with the
highest historical record for cattle herd breakdowns (as measured during the three years leading up to culling.
3. Non-compliant farms are spatially distributed at random.4. A minimum threshold of 10% land access to a badger territory is required to be able
to control the badgers, and that being the case, badgers are controlled at the same rate as they would be with 100% access – i.e. efficient “drawing-out” is assumed with access of 10% or more, and zero “drawing-out is assumed with less than 10% access.
5. Control is applied in the third time step of each year that control is being simulated, which is equivalent to May/June.
6. The badger control is applied at the specified rate to badgers chosen at random from the population, i.e. independent of sex, age, or health status.
7. Trap shyness is not simulated8. The higher control rates for snaring/gassing, and the alternative control rates tested
in the model, have no economic consequence. In other words, it is assumed that the higher rates can somehow be achieved without any extra costs.
9. The vaccine control rate represents the proportion of badgers that eat vaccine bait AND receive from the vaccine full protection against TB.
10. Vaccine protection lasts throughout the life of a vaccinated badger.
121
Appendix L – Glossary
Algorithm: a section of computer code which models or calculates a specific aspect, eg cattle testing or badger mortality.
Anergic: Giving a negative TB test result despite being infected/infectious. Believed to be the case for cattle in the late “super-infectious” stage of TB infection.
Badger immigration: Movement of badgers from higher to lower density areas, usually following badger removal operations.
Badger territory: the area used by a badger social group, which they normally patrol and defend against incursions from badgers of other groups.
Bovine tuberculosis (TB): the disease caused by the mycobacterium M. bovis.
Cattle Herd Breakdown (CHB): an incident of bovine tuberculosis identified in a cattle herd. Reactors to the bovine tuberculosis skin and gamma interferon tests constitute an “unconfirmed” CHB until it is “confirmed” in the laboratory with a positive bacterial culture test.
Cattle Tracing Scheme (CTS): – the scheme whereby cattle are registered and cattle movements are recorded.
CBA: Cost-Benefit Analysis, a method for evaluating the economic merits of a course of action.
CHB: Cattle Herd Breakdown
CSL: Central Science Laboratory
Compliance: (as in land compliance for badger control): the proportion of the land that is accessible for badger control purposes. Non-compliant land is that land for which access has been refused by the land-owner or farmer.
CTS: Cattle Tracing Scheme.
Culling: the killing of badgers usually over a targeted area of land.
Ecological monitoring: the monitoring of areas both before and after a badger control operation to try and measure any effects the control might have on the badger population.
Defra: Department for Environment, Food and Rural Affairs.
Gamma-interferon test: (see IFN-gamma test).
GIS: Geographical Information System; a computer technique for analysing data plotted on maps.
IFN-gamma test: the gamma interferon test – a blood test of cattle which measures a specific product of immunological reaction to bovine tuberculosis.
ISG: Independent Scientific Group on Cattle TB – a group of independent scientists who were gathered to advise the Secretary of State for Environment, Food and Rural Affairs on how best to tackle the problem of cattle TB.
LHS: Latin Hypercube Sampling – a sensitivity analysis technique which helps prevent confounding effects of correlation between input variables.
122
M. bovis: the bacterium Mycobacterium bovis, the causative agent of bovine tuberculosis.
Movement restriction: enforced restrictions on cattle movements, following a cattle herd breakdown (confirmed or unconfirmed), to prevent between-farm transfers and spread of TB. Non-reactors can still be sent to slaughter.
NPV: Net Present Value
NU: Newcastle University
Perturbation: (see Social perturbation)
Prevalence (of TB): the proportion of the population (badger or cattle) that is infected with bovine tuberculosis.
Proactive badger culling: killing of badgers as a preventative method and not in response to any current cattle herd breakdown (CHB), though the decision on which area to cull proactively may be determined by the historic CHB rates. (Also see reactive badger culling)
RBCT: Randomised Badger Culling Trial
Reactive badger culling: killing of badgers in response to a confirmed cattle herd breakdown. (Also see proactive badger culling).
Reactor: an animal that gives a positive result to a disease test, eg a cow that gives a positive reaction to the skin test for TB. An Inconclusive Reactor is an animal that gives a borderline test result – i.e. an inconclusive test result indicates that the animal may be or have been infected, but further tests are required to confirm, one way or the other.
Sensitivity (of diagnostic test): the proportion of infected animals correctly identified as such.
Sett: a burrow system that badgers use for shelter and breeding. There is usually a main sett (primary burrow), and secondary setts (used less frequently but may house a second breeding female).
Skin test: cattle test on cervical skin to quantify the skin reaction to bovine tuberculosis antigen.
Social group: a group of badgers that live together and occupy one or more setts usually within a well-defined and defended territory.
Social perturbation: Disruption of the badgers’ social group structure following culling. There is mounting evidence that this includes extra roaming movements of badgers, which lead to greater contact rates and disease transmission.
Specificity (of diagnostic test): the proportion of uninfected animals correctly identified as such.Stochastic model: a model that includes random elements to represent the uncertainty (within pre-defined limits) of some parameter values or process decisions. The outputs of stochastic models can usually be represented as probability distributions rather than single values.
Super-infectious: term used to describe a badger or cow that is infectious with TB and is shedding substantial quantities of M. bovis. A specific definition has been adopted for the badger – namely one that has two or more successive culture positive TB tests, or two or more culture-positive tests from different body sites.
SVS: State Veterinary Service
123
TB: bovine tuberculosis
Tesselate: create area shapes (e.g. badger territories or parishes) that abut with each other, filling the area with no gaps and no overlapping.
Test frequency: (see Test interval)
Test interval: the interval between successive routine tests such as the cattle skin test which is every year in areas where bovine TB is most prevalent, to once every four years in areas where it is least prevalent. A short test interval means a high test frequency.
Tuberculin test: the standard skin-test for cattle to measure the immune reaction to antigens as evidence of M.bovis infection.
VETNET: the database system for records of bovine TB tests and cattle herd breakdowns.
VLA: Veterinary Laboratories Agency
Woodchester: Woodchester Park – Central Science Laboratory Badger Research Unit based in Gloucestershire.
Zoonotic disease (zoonosis): a disease of animals that can be transmitted to humans.
124
