Date post: | 21-Feb-2017 |
Category: |
Science |
Upload: | claudia-kamphuis |
View: | 389 times |
Download: | 1 times |
Sensors technologies
replacing or complementing human senses to monitor animal health
What can you expect
Claudia Kamphuis
Sensor technologies in dairy
Theory and Economic potential vs. reality
Performance of sensor technologies
Working with sensor technologies
Current work
3
2004: Graduated, Preventive Animal Heath and Welfare, Wageningen University
2006: PhD, Utrecht University
2010: Defended successfully PhD, Utrecht University
2011: Scientist role at DairyNZ, New Zealand
2013: Post-Doc, Business EconomicsWageningen University
What can you expect
Claudia Kamphuis
Sensor technologies in dairy
Boosted by development of automatic milking systems in 1990s
6 main brands
Boosted by development of automatic milking systems in 1990s
6 main brands
1992 first farm in NL (Bottema, 1992)
>10,000 farms globally 2013 (Rodenburg, 2013)
3,615 (19.5%) Dutch farms (Stichting KOM, 2015)
Forced to replace human senses
Boosted by development of automatic milking systems in 1990s
And further pushed by increased animal welfare concerns
Increasing herds
Government
Society
Cheap technology
Low in maintenance costs
Udder or quarter level
Most used to detect abnormal milk or mastitis
Limited performance for mastitis detection (Rutten et al., 2013)
Electrical Conductivity
handheldIn-line
Other (more sophisticated and expensive) sensor technologies were introduced to monitor cow health and productivity
Udder Health- Electrical Conductivity- Milk yield- Somatic Cell Count- (Milk) Temperature- Colour
Other (more sophisticated and expensive) sensor technologies were introduced to monitor cow health and productivity
Udder Health- Electrical Conductivity- Milk yield- Somatic Cell Count- (Milk) Temperature- Colour
Milk Composition- Milk yield- Fat and protein content- Lactose content- Somatic cell count
Other (more sophisticated and expensive) sensor technologies were introduced to monitor cow health and productivity
Fertility- Progesterone- Activity- Rumination
Cow ‘Composition’- Weight- Body Condition Score
Other (more sophisticated and expensive) sensor technologies were introduced to monitor cow health and productivity
Metabolic disorders- Activity- Rumination- Milk yield- SCC- pH
Cow Mobility- Weight- Activity- Rumination - Milk yield
There are A LOT of sensor technologies
15
With A LOT of benefits
Improve health, welfare
Increase productivity
Increase efficiency
Improve product quality
Objective monitoring
Improve social lifestyle
Use of sensor technologies in the Netherlands(Steeneveld and Hogeveen, 2015)
Survey study1,672 farmers approached via email
512 farmers replied (31%)
202 farmers (41%) replied to have sensor technologies
17
When did CMS farmers invest in sensors (n = 81)(Steeneveld and Hogeveen, 2015)
2004 2005 2006 2007 2008 2009 2010 2011 2012 20130
5
10
15
20
25
30
35
40
Mastitis Rumination Estrus
Year
Farm
ers
(n)
When did AMS farmers invest in sensors (n = 121)(Steeneveld and Hogeveen, 2015)
2004 2005 2006 2007 2008 2009 2010 2011 2012 20130
5
10
15
20
25
30
35
Mastitis Rumination Estrus
Year
Farm
ers
(n)
Use of sensor technologies (%) in the Netherlands(Steeneveld and Hogeveen, 2015)
20
Sensor AMS (n = 121)
CMS (n = 81)
Colour 60 1Electrical Conductivity 93 35Milk temperature 50 6Weighing platform 27 5Fat and protein 20 0Somatic cell count 17 1Activity meters/pedometers dairy cows 41 70Activity meters/pedometers young stock 12 28Temperature 6 14Rumination 9 12Lactate dehydrogenase (LDH) 2 1Progesterone 2 1
What can you expect
Claudia Kamphuis
Sensor technologies in dairy
Theory and Economic potential vs. reality
Reasons why AMS farmers invested in sensors(Steeneveld and Hogeveen, 2015)
22
Investment reason EC(n = 112)
Rumination(n = 11)
Activity (n = 50)
Reduce labor 1 9 6
Improve health / reproduction
14 55 72
Insight in health 14 82 42
Not a conscious decision 97 54 48
Improve farm profitability 13 45 48
Automated mastitis detection: theory
Not a conscious decision (we have to?)
Managing bulk milk SCC levels
Mastitis detection
Dry-cow therapy decisions
23
Automated mastitis detection: economics
24
Automated mastitis detection: reality (Steeneveld et al., 2015)
Farms AMS farms CMS farms No sensors Before sensors After sensors Before sensors After sensors
Number of cows % growth in size Milk production (kg / cow / year)
86
3.5
8,343
82
2.6
8,398
97
4.2
8,558
127
6.0
8,371
159
9.7
8,179
190
195
200
205
210
215
220
225
230
Som
atic
cel
l cou
nt (x
1,00
0 ce
lls/m
l)Automated mastitis detection: reality (Steeneveld et al., 2015)
Automated oestrus detection: theory
Advantages two-fold Improve farm profitability Better detection rates -> improved pregnancy rates
Automated oestrus detection: theory
Advantages twofold Improve farm profitability Better detection rates -> improved pregnancy rates
Clear management (decision support) associated with information
Automated oestrus detection: theory
Advantages twofold Improve farm profitability Better detection rates -> improved pregnancy rates
Clear management (decision support) associated with information
OK performance: 80% SN with 95% SP(Rutten et al., 2013)
Automated oestrus detection: theory
Advantages twofold Improve farm profitability Better detection rates -> improved pregnancy rates
Clear management (decision support) associated with information
OK performance
Investment is economically beneficial(Rutten et al., 2014)
Automated oestrus detection: economics
General culling
Calving
Ovulation
Heat detection
P(1st ovulation)
P(heat)P(heat detected)
P(culling)
P(culling)
P(culling)
Simulated cowParity, production level
Insemination after voluntary waiting period
Culling due to fertility issues- Max 6 inseminations- Not pregnant in wk 35
Replacement heifer
Cow pregnant
P(pregnant)
P(early embryonic death)
Next parity
∆ Milk yield ∆ Number of inseminations∆ Number of calves produced∆ Feed intake∆ Number of culled cows∆ Number of false alerts from PLF
Output cow place /year
Milk priceLabour costsCost for AICosts/revenues of calvesCosts feed Costs for cullingCosts of false alerts PLF (labour or AI)
x €
At farm level
Probabilities are adjusted for each simulated week
Costs of PLF technology: investment, maintenance, depreciation, replacement of faulty sensors
Cow Model
SN 50% SP 100%
SN 80% SP 95%
€108/cow€3600/herd
10yearsChecking each
alert visually
Automated oestrus detection: economics
Cash flow: 2,287 € / yearCost-Benefit ratio: € 1.23Discounted payback period: 8 years
Investment pays off(Rutten et al., 2014)
SN 80%;SP 95%€ 108/cow
€ 3600/herd10years
Checking each alert visually
Automated oestrus detection: reality (Steeneveld et al., 2015)
Farms AMS farms CMS farms No sensors Before
sensors After
sensors Before sensors
After sensors
Number of cows
% growth in size
Milk production (kg/cow/year)
85
3.5
8,342
86
2.8
8,473
102
5.3
8,632
104
4.0
8,245
131
6.1
8,177
Automated oestrus detection: reality (Steeneveld et al., 2015)
70
75
80
85
90
95
100
105
Day
s to
firs
t ser
vice
Investment in sensor technologies: reality (€/100 kg milk) (Steeneveld et al., unpublished)
No sensor AMS CMS Before After Before AfterCapital costs 10.38 9.72a 13.97b 11.08c 11.35c
Labour costs
12.38 11.69a
11.30a 11.30c
10.43c
Variable costs
1945 18.66a 19.80a 18.28c 19.24c
Revenues
46.28 43.93a 46.38b 45.77c 47.18c
Profit 4.07 3.86a 1.31b 5.11c 6.16c
So, just a mid re-cap
1,672 farms approached512 farmers replied202 indicated to have sensors(Steeneveld and Hogeveen, 2015)
Economic theory is not matching reality
12%
What can you expect
Claudia Kamphuis
Sensor technologies in dairy
Theory and Economic potential vs. reality
Performance of sensor technologies
Sensors are state-of-the-art pieces of technology that develop and improve rapidly
Sensors are state-of-the-art pieces of technology that develop and improve rapidly
It’s all about monitoring parameters associated with events of interest, but sensors
May not accurately or precisely monitor these parameters
It’s all about monitoring parameters associated with events of interest, but sensors
May not accurately monitor these parameters
Monitor a proxy for these parameters
viscosity measurements;Whyte et al., 2004
It’s all about monitoring parameters associated with events of interest, but sensors
May not accurately monitor these parameters
Monitor a proxy for these parameters
Monitor parameters that are not unique for the event
It’s all about monitoring parameters associated with events of interest, but sensors
May not accurately monitor these parameters
Monitor a proxy for these parameters
Monitor parameters that are not unique for the event
Monitor one single aspect of a complex event
Always a trade-of between
SensitivityHow many events do you
detect (true positive alerts) and how many do you
miss (false negative alerts)
SpecificityHow many healthy cows do
not receive an alert (true negative alert)
and how many do receive an alert falsely
(false positive alert)
Trade-off dependants
Event being monitored
Dairying system in which sensor is implemented
Economic consequences of decision-making based on inaccurate sensor information
Farmer’s preference (risk attitude)
Example automated mastitis detection
High SN
no additional labour for checking alerts
Checking a few false positives is always better than checking 2,000 cows
High SP
nuisance of fetching cows and checking alerts
Willing to accept mildly infected cows remain undetected(Mollenhorst et al., 2012;
Hogeveen and Steeneveld, 2013)
Example of automated oestrus detection
Field evaluation of SCR systems in New Zealand: 75% SN and 99%SP
Visual observation using tail paint: 91% SN and 99.8% SP
48
Example automated oestrus detection with 75% sensitivity
Year-round calving might be OK
But what about seasonal calving?6wks time to get all cows pregnantEconomic losses in case oestrus events are missed
Farmers’ attitude
Being in control Letting-go
Convenience seekers Business optimisers
Farmers’ attitude
Eager to understand and learn the system Not having the time or skills
Innovators/ambassadorsConvenience seekers/business optimisers
Sensors are not about ‘one size fits all’
Waiting for ‘improved’ systems(Borchers and Bewley, 2015; Steeneveld and Hogeveen, 2015; Russell and Bewley, 2013)
52
What can you expect
Claudia Kamphuis
Sensor technologies in dairy
Theory and Economic potential vs. reality
Performance of sensor technologies
Working with sensor technologies
Reasons why AMS farmers invested in sensors(Steeneveld and Hogeveen, 2015)
54
Investment reason EC(n = 112)
Rumination(n = 11)
Activity (n = 50)
Reduce labor 1 9 6
Improve health / reproduction
14 55 72
Insight in health 14 82 42
Not a conscious decision 97 54 48
Improve farm profitability 13 45 48
Use of sensor information is limited
Sensor AMS (%) CMS (%)
Never/ sometimes
Daily Never/ sometimes
Daily
Colour (n=72 / 1) 49 32 100 0
Fat and protein sensor (n = 24) 63 17
Electrical conductivity (n = 112 / 28) 5 77 25 21
Weighing platform (n = 33 / 4) 39 21 25 50
Activity meters/pedometers dairy cows (n = 50 / 57)
6 74 6 74
Use of sensor information is limited(Hogeveen et al., 2013)
5% of generated mastitis alerts are visually checked
Use of sensor information is limited(Hogeveen et al., 2013)
5% of generated mastitis alert lists are visually checked
Reasons not to check alerts included:
No deviation in yield (19%)No flakes on filter (28%) Repeatedly on list (10%)
Too busy (10%)Malfunctioning (4%) No EC increase (5%)
Use of sensor information is limited(Hogeveen et al., 2013)
5% of generated mastitis alert lists are visually checked
Reasons not to check alerts
Consequence: 75% of detected mastitis is not ‘seen’
190
195
200
205
210
215
220
225
230
Som
atic
cel
l cou
nt (x
1,00
0 ce
lls/m
l)Automated mastitis detection: reality (Steeneveld et al., 2015)
Use of sensor information is limited
22% of farm owners indicated that expectations did not match performance reality
24% of farm owners indicatedthat learning support was not as expected (Eastwood et al., 2015)
Too much information without knowing what to do with it (Russell and Bewley, 2013)
61
What can you expect
Claudia Kamphuis
Sensor technologies in dairy
Theory and Economic potential vs. reality
Performance of sensor technologies
Working with sensor technologies
Current work
63
The cow central
Farmer rules
Real time models of different parties
Sensors of different companies
Other data sources
InfoBroker: Open platform for sensor data
Work instructions
What’s currently being done?
What’s currently being done?
Develop a blueprint for successful PLF technologies
Social impactEconomic viability
What’s currently being done?
Tools to estimate economic and social value
Value Creation Tool potential economic benefits of sensor technology in different dairying situations
Break-even Tool how much change of a parameter is required to break-even with the investment
Adaptive Conjoint Analysis assessing utilities of costumers for economic or social aspects
What can you expect
What I would like you to remember
Sensors are exciting, high-tech and have potential
But we need their information combined with
To complement management decisions on animal health
Thank you for your attention
www.slideshare.net/claudiakamphuis ckamphuis