I R M Economic Analysis of Dynamic Management Strategies...

INSECTICIDE RESISTANCE AND RESISTANCE MANAGEMENT

Economic Analysis of Dynamic Management Strategies UtilizingTransgenic Corn for Control of Western Corn Rootworm

(Coleoptera: Chrysomelidae)

D. W. CROWDER,1, 2 D. W. ONSTAD,1 M. E. GRAY,3 P. D. MITCHELL,4 J. L. SPENCER,5

AND R. J. BRAZEE1

J. Econ. Entomol. 98(3): 961Ð975 (2005)

ABSTRACT We studied management strategies for western corn rootworm, Diabrotica virgiferavirgifera LeConte, using transgenic corn, Zea mays L., from both a biological and an economicperspective. In areas with and without populations adapted to a 2-yr rotation of corn and soybean(rotation-resistant), the standard management strategy was to plant 80% of a cornÞeld (rotated andcontinuous) to a transgenic cultivar each year. In each area, we also studied dynamic managementstrategies where the proportion of transgenic corn increased over time in a region. We also analyzedmanagement strategies for a single Þeld that is the Þrst to adopt transgenic corn within a largerunmanaged region. In all areas, increasing the expression of the toxin in the plant increased economicreturns. In areas without rotation-resistance, planting 80% transgenic corn in the continuous cornÞeldeach year generated the greatest returns with a medium toxin dose or greater. In areas with allelesfor rotation-resistance at low initial levels, a 2-yr rotation of nontransgenic corn and soybean,Glycinemax (L.) Merr., may be the most economical strategy if resistance to crop rotation is recessive. Ifresistance to crop rotation is additive or dominant, planting transgenic corn in the rotated cornÞeldwas the most effective strategy. In areas where rotation-resistance is already a severe problem, plantingtransgenic corn in the rotated cornÞeld each year was always the most economical strategy. In somecases the strategies that increased the proportion of transgenic corn in the region over time increasedreturns compared with the standard strategies. With these strategies the evolution of resistance to croprotation occurred more rapidly but resistance to transgenic corn was delayed compared with thestandard management strategy. In areas not managed by a regional norm, increasing the proportionof transgenic corn and increasing toxin dose in the managed Þeld generally increased returns. In asensitivity analysis, among the parameters investigated, only density-dependent survival affected theresults.

KEYWORDS Diabrotica virgifera virgifera, modeling, transgenic corn, integrated pest management,insect resistance management

INSECTICIDAL CROPS PRODUCING TOXINS fromBacillus thu-ringiensis(Bt) have the potential to simplify pest prob-lems while meeting integrated pest management(IPM) objectives (Roush 1997). In 2003, the Þrsttransgenic product effective against the western cornrootworm,Diabrotica virgifera virgifera LeConte, wascommercialized and is being incorporated into man-agement strategies for this pest. The western corn

rootworm has historically been managed with a com-bination of crop rotation and insecticides (Zhou et al.2003). However, this pest has a history of developingresistance to these control tactics (Ball and Weekman1962, 1963; Metcalf 1983; Meinke et al. 1998; Miota etal. 1998; Onstad et al. 1999, 2003b; Scharf et al. 1999).Given this history, transgenic corn, Zea mays L., mustbe managed carefully to maximize its effectiveness asa potentially very effective control tactic.

For managing a species such as the western cornrootworm that has historically been controlled with avariety of tactics, replacement of broad-spectrum in-secticides by more speciÞc methods such as transgeniccorn may go a long way to simplifying pest manage-ment problems (Carriere et al. 2004a). Over-relianceon insecticides creates agricultural, environmental,and pest management problems (Walker et al. 1995)such as frequent evolution of insect resistance(Georghiou 1986) and environmental contamination

The ideas expressed in this paper may not represent those of theUSDA.

1 Department of Natural Resources and Environmental Sciences,University of Illinois, Urbana, IL 61801.

2 Current address: 1322 E. Spring St., Tucson, AZ 85704 (e-mail:[email protected]).

3 Department of Crop Sciences, University of Illinois, Urbana, IL61801.

4 Department of Agricultural and Applied Economics, University ofWisconsin, Madison, WI 53706.

5 Center for Economic Entomology, Illinois Natural History Survey,607 E. Peabody Dr., Champaign, IL 61820.

0022-0493/05/0961Ð0975$04.00/0 � 2005 Entomological Society of America

(Pimental et al. 1992; Matson et al. 1997). In Arizona,the introduction of Bt cotton for pink bollworm con-trol led to important declines in the use of insecticides(Carriere et al. 2001, 2004a). The introduction oftransgenic corn may have similar effects on manage-ment practices throughout the Corn Belt. To analyzethe potential impacts of using transgenic crops, bio-logically rich simulation models (Guse et al. 2002,Onstad et al. 2002, Carriere et al. 2003, Storer 2003) inconjunction with spatially explicit empirical studies(Carriere et al. 2003, 2004b) are required to betterunderstand how refuges should be deployed and man-aged (Carriere et al. 2004a), where refuges are areasnot planted to a transgenic cultivar that maintain sus-ceptible individuals that can disperse and mate withresistant individuals surviving in transgenic Þelds.

Simulation models have indicated that transgeniccorn may be an effective management option for farm-ers in areas with rotation-resistant western corn root-worm phenotypes (Onstad et al. 2003a, Crowder andOnstad 2005, Crowder et al. 2005). Crowder and On-stad (2005) and Crowder et al. (2005) showed thattransgenic corn, planted to 80% of rotated cornÞelds,could effectively prevent resistance to crop rotationfrom evolving in areas with 85% of the land in rotationbetween corn and soybean Glycine max (L.) Merr.Results also indicated that resistance to transgeniccorn did not evolve with a management strategy ofplanting 80% transgenic corn in rotated cornÞelds.

Simulation models also have been used to evaluatethe risk of resistance to transgenic corn in areas with-out rotation-resistant populations (Onstad et al. 2001a,Crowder and Onstad 2005, Crowder et al. 2005). Inthese areas, the gene expression of the allele for re-sistance to transgenic corn, R, is the most importantfactor affecting the evolution of resistance to trans-genic corn. If R is recessive, resistance can be pre-vented with any toxin dose and refuges occupying5Ð30% of continuous cornÞelds. If R is dominant, re-sistance may be difÞcult to prevent (Crowder andOnstad 2005, Crowder et al. 2005).

Onstad et al. (2003a) performed an economic anal-ysis of management strategies with transgenic cornplanted in rotated cornÞelds in areas with 85% ofthe landscape in rotation between corn and soybean.They assumed that transgenic corn killed 90% of alllarvae and did not simulate the potential for resistanceto transgenic corn or various doses of the transgenictoxin. In this article, we expand upon the analysis ofOnstad et al. (2003a) and study the biological andeconomic implications of different strategies for usingtransgenic corn with a model that simulates evolutionof resistance tobothcroprotationand transgeniccorn.

Long-term, possibly areawide, strategies are the fo-cus of attention for both IPM and insect resistancemanagement (IRM). Midgarden et al. (1997) showedthat site-speciÞc IPM, a strategy of spraying only areasthat have densities above economic thresholds,slowed the development of insecticide resistance andconserved natural enemies of Colorado potato beetle.This type of IPM management strategy creates in-Þeldrefuges that slow the rate of evolution of resistance

(Midgarden et al. 1997). A similar result was observedby Peck and Ellner (1997), who showed that manage-ment strategies that reduce selection pressure in someareas have a beneÞcial effect on the regional level ofresistance to a pesticide. In this article, we will attemptto determine whether deployment strategies for trans-genic corn that lower the selection pressure for resis-tance to the transgenic crop also were effective atpreventing resistance to crop rotation and maximizingeconomic returns.

Our goal was to determine the most effective strat-egy from both a biological and economic perspectivefor controlling western corn rootworm damage. Weevaluate the impact of using transgenic corn withvarious toxin doses. We focus our analysis on threeareas: areas with 100% continuous corn and no rota-tion-resistant phenotypes, areas where resistance tocrop rotation may develop, and areas where rotation-resistance is already a serious problem and IRM is nolonger feasible or of primary concern.

Materials and Methods

In this section, we describe the creation of themodel and its analysis. First, we describe the functionsused to create the model. All functions except for theeconomic analysis are the same as in the generationaltime-step model of Crowder and Onstad (2005), andthe logic used to derive these functions is not dis-cussed. Second, we describe the economic functionsused in the model. Third, we describe the standardsimulation conditions. Fourth, we describe the varia-tions on the standard management strategies that wereperformed. Finally, we describe the sensitivity anal-yses that were performed.Population Genetics.We assume this is an autoso-

mal, two-locus, two-allele per locus, diploid geneticsystem. The allele for susceptibility to transgenic cornis S; the allele for resistance to transgenic corn is R.With regard to crop rotation, we deÞned the X allelefor no movement out of corn and the Y allele for thetendency to move to all patches (Crowder and Onstad2005). In a 3-yr Þeld study in eastern Illinois, Rondonand Gray (2004) observed increased oviposition innoncorn habitat (soybean, alfalfa, and oat-stubble).This resulted in no signiÞcant differences in the num-bers of western corn rootworm eggs laid in corn,soybean, and oat-stubble on a per liter or per hectarebasis. These results support the assumption that rota-tion-resistant individuals disperse and oviposit into allpatches in the landscape.

Despite a lack of empirical evidence on the popu-lation genetics of D. virgifera virgifera, this geneticsystem was chosen because it is comparable to severalnatural systems described by Onstad et al. (2001b),and has been previously used in models that simulatedthe development of resistance by western corn root-worm to crop rotation (Onstad et al. 2001b, 2003) andtransgenic corn (Onstad et al. 2001a, Storer 2003).Model Landscape. The region consists of 100 ha of

cropland consisting of up to four crops and a maximumof six Þelds. The four crops are corn grown in the same

962 JOURNAL OF ECONOMIC ENTOMOLOGY Vol. 98, no. 3

location each year (continuous corn), Cc; corn thatfollows a soybean Þeld in a 2-yr rotation of corn andsoybean (rotated corn) Cr; soybean, which precedescorn in a 2-yr rotation of corn and soybean, Soy; andthe extra noncorn, Ex. A proportion of both cornpatches can be planted to a transgenic cultivar in ablock conÞguration. The proportion of the continuousand rotated cornÞelds that are planted to a transgeniccultivar is Tc and Tr, respectively. In cases where Tc orTr � 0, we studied a refuge size occupying 20% of theÞeld. In the model equations the abbreviations rc, rr,tc, and tr are used for the continuous corn refuge,rotated corn refuge, transgenic continuous corn, andtransgenic rotated corn, respectively.Movement and Oviposition. We set the fecundity

per individual to 220 viable eggs, which is half of thenumber per female, and ignore gender (Onstad et al.2003a). Normal (ÐXX) individuals move from the na-tal corn patch and distribute themselves (and theireggs) across all corn patches according to their rela-tive proportional areas, where Ð indicates the allelesfor susceptibility to transgenic corn. We assumed thatrotation-resistant (ÐYY) individuals move into allpatches according to their proportional representa-tion in the region. In the additive case, 50% of theheterozygotes (ÐXY) distribute themselves and theireggs only into cornÞelds, whereas the others disperseinto all Þelds.ToxinMortality.Toxin mortality incurred by larvae,

Qtox, is dependent on the dose of the toxin, the ge-notype of the individual at the locus for resistance totransgenic corn and the gene expression at this locus.We assume this mortality is applied at the same timeas overwintering survival. We studied four doses oftoxin (theoretical high, practical high, medium, andlow) based on values used in Crowder and Onstad(2005), Crowder et al. (2005), and Onstad et al.(2001a).

Homozygous-resistant individuals, (RRÐ), alwayshave 100% survival to the transgenic cultivar regard-less of dose (Qtox � 0). With R dominant, heterozy-gous (sRÐ) individuals also always have 100% survivalto the transgenic crop. The survival of homozygoussusceptible individuals (SSÐ), or heterozygotes with Rrecessive (SrÐ), is 0, 0.001, 0.05, and 0.20 with a the-oretical high, practical high, medium, and low toxindose, respectively. With R partially recessive, survivalof the heterozygotes, (srÐ), is 0, 0.01, 0.50, and 0.60with a theoretical high, practical high, medium, andlow toxin dose, respectively. In cases where a propor-tion of both the continuous and rotated cornÞeldsare planted to a transgenic cultivar, we assume that thedose used in each Þeld is the same. The alleles forsusceptibility to crop rotation (denoted by dashes),do not affect survival to the transgenic toxin.Immature Survival. All larvae emerging from eggs

in noncorn patches die. Offspring in continuous androtated cornÞelds incur an overwintering mortality of50% during the egg stage and incur density-dependentmortality during the larval stage (Onstad et al. 2001a,2003a).

Individuals emerging in Þelds planted to a trans-genic cultivar incur density-independent mortality,Qtox, based on the dose of the toxin and the geneexpression. The density-dependent survival of larvaeper stage is 0.21 � exp (�0.058EGG), where EGG isthe density of eggs (in millions per hectare) (Onstadet al. 2003a). The maximum larval survival based onthis function is 21%. We assume that density-depen-dent mortality occurs after mortality due to overwin-tering and toxin exposure.Model Equations. The number of eggs Ei,p (t � 1)

of genotype i in patch p for year t � 1 as a function ofthe number of adults Ai,p in year t is as follows:

Ei,p(t � 1) � 220 � �j�1

9 �Pj,p � �Aj,rc(t)

� �k�1

9

wQk,rc(t) � Aj,rr(t)

� �k�1

9

wQk,rr(t) � Aj,tc(t)

� �k�1

9

wQk,tc(t) � Aj,tr(t)

� �k�1

9

wQk,tr(t)�� [1]

P is the probability of genotype j (nine possible ge-notypes) moving to patch p. We assume that beetlesmate randomly within the Þeld of emergence and thatthe offspring will have an expected frequency distri-bution dependent upon the frequencies of each ge-notypeemerging in thenatalpatch.Therefore,Q is thefrequency of genotype k in natal patches rc, rr, tc, andtr that can reproduce the particular offspring geno-type i when mated with genotype j. Each weight, w,equals the Mendelian proportion of all offspring thatare genotype i when genotypes j and k mate.

To calculate the number of older larvae and adults,we calculated-density dependent survival. First, wecalculated the total density of larvae TL in each kindof corn habitat in the landscape, f (where f � rc, rr, tc,or tr), surviving the winter and toxicity of transgeniccorn roots (1 � Qtox).

TLf � �i�1

9

Ei,f(t) � 0.5 � (1 � Qtox) [2]

The numbers of older larvae or adults in the four cornhabitats f are as follows:

Ai,rc(t) � 0.5 � Ei,rc(t) � 0.21

� exp{�0.058 � [TLrc/(100

� (1 � Tc) � Cc � 106)]} [3]

June 2005 CROWDER ET AL.: DYNAMIC DEPLOYMENT STRATEGIES BY USING TRANSGENIC CORN 963

Ai,tc(t) � 0.5 � Ei,tc(t) � (1 � Qtox) � 0.21

� exp{�0.058 � [TLtc/(100 � Tc � Cc � 106)]}

[4]

Ai,rr(t) � 0.5 � Ei,rr(t) � 0.21

� exp{�0.058 � [TLrr/(100

� (1 � Tr) � Cr � 106)]} [5]

Ai,tr(t) � 0.5 � Ei,tr(t) � (1 � Qtox) � 0.21

� exp{�0.058 � [TLtr/(100 � Tr � Cr � 106)]}

[6]

The number of adults emerging the next year is Ai,s

(t) � Ai,e (t) � 0 in rotated soybean and in extravegetation.Male Dispersal. Previous models simulating resis-

tance to transgenic corn have assumed that 25% ofwestern corn rootworm males disperse out of the natalÞeld before mating (Onstad et al. 2001a, Crowder andOnstad 2005, Crowder et al. 2005). To simulate this, werecalculated the genotype frequencies of males beforemating in each Þeld after dispersal. This was per-formed using our assumption that 50% of emergingadults are male. All females remain in the natal Þeld tomate, whereas a proportion, PM, of the males moveout of the natal Þeld and into the other cornÞeldsaccording to their proportional areas. The standardvalue for PM is 0.25. For example, the adjusted numberof males in the continuous nontransgenic corn is asfollows:

Mi,rc(t) � 0.5 � Mi,rc � (1 � PM) � 0.5

� (Mi,tc � Mi,rr � Mi,tr) � PM

� �P(f)��k�1

3

P(k)� [7]

Mi,f (t) is the adjusted number of males of genotypei in Þeld f (rc, tc, rr, and tr), whereas Mi,f (t) is thenumber of males of genotype i that emerged in Þeldf. P(f) is the proportion of land planted to Þeld f thatmales are dispersing into and P(k) is the proportion ofland planted to each cornÞeld that is not the natal Þeldf for individual Mi,f (t). On the right-hand side ofequation 7, the Þrst product is the number of malesthat remain in Þeld f. The other three products rep-resent the number of males moving into f from theother cornÞelds.

The adjusted number of males after male dispersalis used to recalculate male genotype frequencies ineach Þeld before mating. In the model, these frequen-cies are calculated before equations 1Ð6 and are usedin equation 1 to calculate eggs. However, we did notadjust population densities in any Þeld due to maledispersal. The reason for this is a constraint of themodel. Because the model does not simulate males andfemales separately, if population densities in each Þeldare calculated after male movement there would be

deviation from a 1:1 sex ratio and our assumptionsabout fecundity per individual would not be valid.EconomicAnalysis.Mitchell et al. (2004) compared

two different pest-damage functions for western cornrootworm that predicted proportional yield loss as afunction of measured root damage by corn rootwormlarvae. They indicated that the results of an economicanalysis using a damage function from a composed-error model were similar compared with the results ofa conventional model unless the distribution of lossmatters, such as when the analysis incorporates riskaversion (Mitchell et al. 2004). Because our analysisdid not include risk aversion, we followed the con-ventional approach from Mitchell et al. (2004) andOnstad et al. (2003a) to calculate the proportion ofyield lost due to rootworm damage. The proportion ofyield lost in habitat f is as follows:

LOSSf,t � 0.251[1 � exp(�0.0089Kf,t0.589)], [8]

where Kf,t � TLf,t/(100Cf� 104) is larvae per squaremeter in year t after mortality due to overwinteringand toxin mortality. No loss occurs unless larvae arepresent. As the larval population increases, loss in-creases and asymptotically approaches the maximumof 0.251.

The economic analysis was based on calculation ofreturns for each cornÞeld in the region. In areas withrotation-resistant phenotypes and 85% rotated land-scape, we did not include returns from soybean orextra vegetation in our analysis because the propor-tion of land planted to these patches never changed,and we wanted to focus on the returns for corn. Re-turns (dollars per hectare) for the crop in habitat f inyear t are as follows:

RETURNSf,t � PRICEf � Yf

� (1.0 � LOSSf,t) � COSTf [9]

PRICEf is the price for crop f, Yf is (pest-free) yieldfor crop f, LOSSf,t is the proportional yield loss forcrop f in year t due to corn rootworm, and COSTf isthe variable cost of production for crop f.

For corn, a price of $8.15/quintal (ql) was used, thisis the approximate marketing year average price from2000 to 2003 (Illinois Agricultural Statistics Service2003). Yields and costs were from Illinois crop budgets(Schnitkey 2004). The variable costs for continuousand rotated corn are $531 or $519/ha, respectively.The yield for continuous corn, Ycc, was set equal to theyield for rotated corn, Yrc, 104 ql/ha, because thereported yield implicitly included yield loss due tocorn rootworm and our economic analysis separatelyincorporated this yield loss. Therefore, returns fornontransgenic continuous or rotated corn are approx-imately $317 or $329/ha, respectively. We included acost of $37/ha for transgenic corn ($15/acre). Al-though the price of transgenic corn may vary greatlydepending on the quantity purchased and other fac-tors, we believe this is a reasonable estimate of theprice that may be charged for transgenic seed. The netpresent value per hectare for habitat f is the sum of thediscounted annual returns:


NPVf � �t�1

15

�(t) � RETURNSf,t [10]

where �(t) � 1/(1 � dr)t and dr is the discount rate.A discount rate of 7% is used, because the Illinoisbudgets report this rate as the farmer cost of capital(Schnitkey 2004). The net present value (dollars perhectare)of a strategy is then the sumof thenetpresentvalue for each corn crop in the strategy, weighted bythe proportion of the total landscape planted to eachcrop habitat for the strategy: NPV � Cr � NPVCr �Cc � NPVCc. The annualized net present value(ANPV) of each strategy is the Þxed annual return(dollars per hectare) for 15 yr that generates the samenet present value:

ANPV � NPV/Z, where Z � �t�1

15

�(t). [11]

Simulation Conditions. The model is programmedin Visual Basic in Microsoft Excel 2002 (Microsoft2002). The model has a time-step of 1 yr, a timehorizon of 16 yr, and a spatial unit of 100 ha. The 16-yrtime horizon includes the Þrst year in which the modelis initialized without transgenic corn plus an addi-tional 15 yr. The initial number of adults is 50,000 perha of corn, which is the value used in the model ofOnstad et al. (2001a). The initial adults are distributedproportionally to the areas of continuous and rotatedcorn. The rotation level, RL, is the sum of the pro-portional areas of rotated corn and soybean, which arealways equal in the model (RL � Soy � Cr). In thestandard simulations of areas without rotation-resis-tant phenotypes, we set the initial R-allele frequencyto 10�4 and the Y-allele frequency to 0 in a landscapewith 100% continuous corn (Cc � 1.0). In the standardsimulations of areas with rotation-resistance, the land-scape is deÞned as RL � 0.85, Ex � 0.05, and Cc � 0.10.In these areas, the adults begin at HardyÐWeinbergequilibrium with initial R- and Y-allele frequencies of10�4.

In simulations without rotation-resistance, the stan-dard management strategy was to plant 80% transgeniccorn to the continuous cornÞeld each year. In thesimulations of areas with rotation-resistance the stan-dard management strategy was to plant 80% transgeniccorn to the rotated cornÞeld each year. In these areas,we also performed simulations that represent a stan-dard 2-yr rotation of nontransgenic corn and soybean.All reported R- and Y-allele frequencies are for theentire region simulated.Dynamic Adoption. In addition to our standard

management strategy, we studied a management strat-egy where the proportion of transgenic corn plantedin the region increases over time, while the proportionof corn in the landscape remains constant. In simula-tions of areas without rotation-resistance, the propor-tion of the continuous cornÞeld planted to a trans-genic cultivar increases over time; in simulations ofareas with rotation-resistance, the proportion of therotated cornÞeld planted to a transgenic cultivar in-

creases over time. In these simulations, the proportionof either the continuous or rotated cornÞeld (depend-ing on simulated area) planted to a transgenic cultivarin the Þrst year is 10%. We studied two strategieswhere the proportion of these Þelds planted to trans-genic corn increased by an additional 5 or 10% eachyear until reaching a maximum value of 80%. Thesetwo strategies are referred to as the 5 and 10% dynamicadoption strategies. With the 5 or 10% dynamic adop-tion strategies, the proportion of the Þeld planted to atransgenic cultivar reached 80% in year 15 or 8, re-spectively.Simulations for Severe Problem Areas. We modi-

Þed the simulation of our standard model to determinethe best strategies for managing western corn root-worm in areas where rotation resistance is already asevere problem. For this new case, the landscape hadRL � 0.85 and initial R- and Y-allele frequencies of0.0001 and 0.5, respectively. For this case, we com-pared the standard management strategy of planting80% transgenic corn in the rotated cornÞeld each yearwith the dynamic adoption and 2-yr rotation of non-transgenic corn and soybean management strategies.We did not perform a sensitivity analysis on simula-tions of severe problem areas.Simulations of Areas Not Managed by the RegionalNorm.We also attempted to determine the best strat-egies for managing western corn rootworm in areasnot managed by a regional norm. This was meant toevaluate management options for a single farmerwhen they are the Þrst to adopt transgenic corn withina larger region. To perform this analysis, we Þrst sim-ulated population densities over 15 yr in a 2,000-haregion consisting of 100% continuous corn and notransgenic corn (unmanaged). The simulated numberof adults each year were recorded and used as poten-tial immigrants into a smaller Þeld within the region.We performed an analysis of either a 20- or 100-hamanaged Þeld contained within the larger 2,000-haregion. In these simulations, the number of adults thatimmigrate from the larger region into the smaller Þeldor the number of adults that move out of the smallerÞeld is proportional to the area of the smaller Þeld. Forexample, if the Þeld using transgenic corn is 20 ha, theproportion of adults each year dispersing in from the2,000-ha region is 20/2,000 (1%), whereas the propor-tion of adults moving out of the Þeld into the largerregion is 1,980/2,000 (99%). If the Þeld using trans-genic corn is 100 ha, the proportion of adults dispers-ing in from the 2,000 region is 5%. To simplify thesesimulations, we assumed that adults moving out of thesmaller Þeld into the larger region did not affect pop-ulation densities in the larger region. For each Þeldsize (20 or 100 ha), we simulated transgenic cornoccupying 0, 20, 50, 80, and 95% of the Þeld. We did notperform a sensitivity analysis on simulations of areasnot managed by the regional norm.

This analysis also represents the lower end of thecontinuum of proportional areas planted with trans-genic corn in landscapes without rotation-resistance.In our simulations of areas without rotation-resistance,the standard strategy was to plant 80% of the contin-


uous cornÞeld to a transgenic cultivar each year. Inthese simulations, the maximum proportion of thetotal landscape planted to a transgenic cultivar is(0.95 � 100 ha)/(2,000 ha) or �5%.Sensitivity Analysis. In a sensitivity analysis, we

studied the inßuence of Þve factors on the economicresults of the standard simulations. First, we varied thediscount rate from the standard 7% to values of 1 and15%. Second, we varied the price of corn by 10%.Third, we varied the variable costs for corn by 10%.Fourth, we varied the price of transgenic corn to $24.7and $49.4/ha, or $10 and $20/acre, respectively.

We also simulated population dynamics with dif-ferent versions of the density-dependent survivalfunction. One was based on the model of Onstad et al.(2001a), 1/(1 � 2.42EGG0.7), which has a maximumlarval survival of 100%. One was based on the modelsin Crowder and Onstad (2005) and Crowder et al.(2005), 1/(2.59 � 1.29EGG0.88), which has a maximumlarval survival of 39%. We also tested the model with-out density-dependent mortality but with an addeddensity-independent mortality based on Þeld data col-lected by Hibbard et al. (2004). They revealed thatdensity-independent establishment was between 2.5and 5.7% when plants were sampled on the optimaldate. For this model, we used the higher end of 5.0%for density-independent survival of larvae after over-wintering and toxin mortality similar to Crowder andOnstad (2005) and Crowder et al. (2005).

Results

Biological Analysis of Areas without Rotation-Resistant Phenotypes. In general, in simulations ofareas without rotation-resistant phenotypes, resis-tance to transgenic corn evolved more slowly with thedynamic adoption strategies compared with the stan-dard strategy of planting 80% transgenic corn eachseason. With the dynamic adoption strategies, there isless selection for resistance to transgenic corn com-pared with the standard strategy. This is especially thecase during the early years of the simulations, whenthe proportion of the region planted to a transgeniccultivar is relatively low. As the proportion of trans-genic corn builds in the region, the rate of evolutionof resistance to crop rotation increases.

In simulations of areas with 100% continuous cornand no rotation-resistant phenotypes, the R-allele fre-quency never exceeded 0.001 in 15 yr with R recessiveor partially recessive with a theoretical or practicalhigh dose and a 5 or 10% dynamic adoption strategy.These results were similar when compared with re-sults of the standard simulations with 80% transgeniccorn planted to the continuous cornÞeld each year,where the R-allele frequency did not exceed 0.0001 in15 yr with R recessive or partially recessive and thehigh doses. Similarly, the R-allele frequency did notexceed 0.0001 in 15 yr with R recessive and a mediumor low dose with either a 5 or 10% dynamic adoptionstrategy or the standard strategy.

With R partially recessive and a medium dose, theR-allele frequency exceeded 50% in year 8Ð9 with

either a5or10%dynamicadoption strategy, comparedwith 6 yr with the standard management strategy ofplanting 80% transgenic corn in the continuous corn-Þeld every year. With R partially recessive and a lowdose, the R-allele frequency exceeded 50% in year 13and 11 with a 5 and 10% dynamic adoption strategy,respectively, compared with 10 yr with the standardstrategy.

With R dominant and a 5% dynamic adoption strat-egy, the R-allele frequency exceeded 50% in year 6, 6,7, and 10 with a theoretical high, practical high, me-dium, and low dose, respectively (Fig. 1). With Rdominant and a 10% dynamic adoption strategy, theR-allele frequency exceeded 50% in year 6, 6, 6, and 9with a theoretical high, practical high, medium, andlow dose, respectively. With the standard manage-ment strategy of planting 80% transgenic corn in thecontinuous cornÞeld each year, the R-allele frequencyexceeded 50% in year 5 with a medium or greater doseand year 8 with a low dose and R dominant (Fig. 1).Biological Analysis of AreaswithRotation-ResistantPhenotypes. In areas with rotation-resistant pheno-types, resistance to transgenic corn never evolved in15 yr with any of the simulated management strategies.In general, resistance to crop rotation evolved morequickly with the dynamic adoption strategies com-pared with a strategy of planting transgenic corn eachseason. In years where transgenic corn is not plantedin the region, the selection pressure for rotation-re-sistant phenotypes decreases, causing resistance tocrop rotation to develop more rapidly.

In simulations of areas with 85% rotated landscapeand rotation-resistant phenotypes, with the standardmanagement strategy of planting 80% transgenic cornin the rotated cornÞeld each year, the Y-allele fre-quency never exceeded 0.03 and the R-allele fre-quency never exceeded 0.0001 in 15 yr with any toxindose or allele expression for either trait. Likewise, theR-allele frequency never exceeded 0.0001 in 15 yr withany allele expression for either trait and toxin dosewith a 5 or 10% dynamic adoption strategy. With a Þveor 10% dynamic adoption strategy and Y recessive, theY-allele frequency never exceeded 0.0001 in 15 yr withany combination of R-allele expression or toxin dose.

Resistance to crop rotation did evolve with thedynamic adoption strategies and Y additive or domi-nant. With Y additive and a 5% dynamic adoptionstrategy, the Y-allele frequency exceeded 50% in year14 with a medium or greater dose and year 13 with alow toxin dose with any R-allele expression. With Yadditive and a 10% dynamic adoption strategy, theY-allele frequency never exceeded 50% in 15 yr, reach-ing a maximum value of 0.06. With Y dominant and a5% dynamic adoption strategy, the Y-allele frequencyexceeded 50% in year 9 with any toxin dose and R-allele expression. With Y dominant and a 10% dynamicadoption strategy, the Y-allele frequency exceeded50% in year 15 with a theoretical or practical high doseand in year 14 and 13 with a medium and low dose,respectively, with any R-allele expression.


Economic Analysis of Areas without Rotation-Re-sistant Phenotypes. In areas with 100% continuouscorn and no rotation-resistant phenotypes, returnswith the standard management strategy of planting80% transgenic corn in the continuous cornÞeld eachyear differed depending on toxin dose and R-alleleexpression (Fig. 2). In all cases, returns were greatestwith a theoretical or practical high toxin dose, whereasthe medium and low doses generated lower returns.WithRrecessive, theannualizednetpresentvaluewas$259/ha with a theoretical high dose. Returns de-creasedwith lower toxindoses, byup to40%witha lowdose. With R partially recessive, the annualized net

present value with any dose did not differ from sim-ulations with R recessive. With R dominant, returnsdecreased by 29% with a theoretical or practical highdose and 2Ð4% with a medium or low dose comparedwith the R recessive case.Dynamic Adoption. Toxin dose determined the rel-

ative value of the dynamic adoption strategies. Imple-menting a 5 or 10% dynamic adoption strategy gen-erated lower returns than the standard managementstrategy of planting 80% transgenic corn in the con-tinuous cornÞeld each year with a medium or greaterdose with any R-allele expression. With a theoreticalor practical high toxin dose and a 5 or 10% dynamicadoption strategy, returns were �15 or 8% lower,respectively, compared with the standard strategywith R recessive or partially recessive and 2Ð5% lowerwith R dominant (Fig. 3a). With a medium dose anda 5 or 10% dynamic adoption strategy, returns were6Ð10% less than the standard strategy. With a low toxindose, the 5% dynamic adoption strategy generatedgreater returns than the 10% dynamic adoption strat-egy and the standard management strategy of planting80% transgenic corn in the continuous cornÞeld eachyear with any R-allele expression (Fig. 3b). The great-est change occurred with R dominant, with the 5 or10% dynamic adoption strategies generating returns 6or 3% greater, respectively, than the standard strategy(Fig. 3b).EconomicAnalysis ofAreaswithRotation-ResistantPhenotypes. In areas with rotation-resistant pheno-types and 85% rotated landscape, simulations of a 2-yrcrop rotation without transgenic corn generated anannualized net present value of $300, $258, and

Fig. 1. Years for the allele for resistance to transgenic corn, R, to reach 50% in simulations with 100% continuous corn,four toxin doses, R dominant, and either the standard management strategy of planting 80% transgenic corn to the continuousÞeld each year or the 5 or 10% dynamic adoption strategies.

Fig. 2. Annualized net present value (dollars per hect-are) of simulations with 100% continuous corn and 80%transgenic corn with three types of expression for the allelefor resistance to transgenic corn, R, and four toxin doses.


$240/ha with Y recessive, additive, and dominant,respectively (Fig. 4). The highest returns occurredin the Y recessive case because the Y-allele frequencyremained at 0.0001 over the course of the 15-yr sim-ulation and damage to rotated cornÞelds was min-imal.

In these areas, the standard management strategy ofplanting transgenic corn to 80% of the rotated corn-Þeld resulted in similar returns regardless of geneexpression for either trait. This differed from the 2-yr

rotation of nontransgenic corn and soybean, wherereturns decreased with Y additive or dominant com-pared with Y recessive (Fig. 4). Returns with thestandard strategy were approximately $277/ha withanycombinationof toxindoseandalleleexpression foreither trait (Fig. 4). These returns represented an 8%decrease with Y recessive and a 7 and 13% increasewith Y additive and dominant, respectively, comparedwith the 2-yr rotation of nontransgenic corn and soy-bean (Fig. 4). In each case, the highest returns oc-

Fig. 3. Annualized net present value (dollars per hectare) of simulations with 100% continuous corn and three types ofgene expression for the allele for resistance to transgenic corn, R, with the standard management strategy of planting 80%transgenic corn to the continuous Þeld each year or the 5 or 10% dynamic adoption strategies with a practical high (a) orlow toxin dose (b).


curred with a theoretical high dose and returns de-creased as dose decreased. However, the differencebetween returns with a theoretical high dose and thelower doses was never �0.9%. With any Y-allele ex-pression or toxin dose, the expression of the R-alleledid not affect returns by �0.012%.Dynamic Adoption. As Fig. 5 indicates, the standard

management strategy of planting 80% transgenic cornin the rotated cornÞeld each year generated similarreturns compared with the dynamic adoption strate-gies. Returns for the dynamic adoption strategiesranged from8%less toup to4%more than the standardstrategy with any toxin dose and R-allele expression.With Y recessive, larval densities with the dynamicadoption strategies were 20Ð50% greater in the Þrst2 yr of simulations compared with the standard strat-egy with any dose but �2% different in subsequentyears. Thus, returns were typically greater with thedynamic adoption strategies and Y recessive com-pared with the standard strategy because the gainsfrom planting less transgenic corn outweighed lossesdue to larval damage (Fig. 5a and b). With Y dominant,larval densities were up to 100% greater in soybeanwith the dynamic adoption strategies compared withthe standard strategy. Therefore, the dynamic adop-tion strategies typically generated slightly lower re-turns with Y dominant compared with the standardstrategy (Fig. 5a and b).Simulations of Severe Problem Areas. In simula-

tions of areas with 85% rotation and an initial Y-allelefrequency of 0.5, the Y-allele frequency increasedtoward 1.0 with any management strategy examined.With any management strategy simulated, the R-allelefrequency never exceeded 0.0001 in 15 yr with anycombination of gene expression or toxin dose. In these

areas, the annualized net present value of a 2-yr ro-tation of nontransgenic corn and soybean was approx-imately $198/ha with any gene expression. The stan-dard management strategy of planting 80% transgeniccorn in the rotated cornÞeld each year generatedreturns 15Ð25% greater than a 2-yr rotation of non-transgenic corn and soybean with any toxin dose orgene expression. With any gene expression, the aver-age annualized net present value was approximately$263 with a theoretical high toxin dose. Returns de-creased with the lower doses, up to a maximum of 12%with a low dose, but they were still greater than a 2-yrrotation of nontransgenic corn and soybean. In theseproblem areas the dynamic adoption strategies nevergenerated returns as high as the standard managementstrategy of planting 80% transgenic corn in the rotatedcornÞeld each year. With any allele expression ortoxin dose, the 5 or 10% dynamic adoption strategiesgenerated returns �12 or 7% lower, respectively, thanthe standard strategy. However, the dynamic adoptionstrategies did generate returns 5Ð19% greater com-pared with a 2-yr rotation of nontransgenic corn andsoybean.Simulations of Areas Not Managed by RegionalNorm. We Þrst simulated the 2,000-ha unmanagedregion where the adults were used as potential immi-grants into a smaller, managed Þeld. In the unmanagedregion, populations ßuctuated considerably and neverreached a carrying capacity. For example, in year 3 thepopulation reached a density of �1.18 million adultsper ha but fell to �15,000 adults per ha in year 4, a 99%decrease. Similarly, populations in the unmanaged re-gion reached a maximum density of �1.33 millionadults/ha in year 10 but decreased to a minimum of6,300 adults per ha in year 11. These ßuctuations were

Fig. 4. Annualized net present value (dollars per hectare) of simulations with 85% rotation and three types of geneexpression for the allele for resistance to crop rotation, Y, with a 2-yr rotation of nontransgenic corn and soybean or thestandard management strategy of planting 80% transgenic corn to the rotated cornÞeld each year with a practical high toxindose. Results were similar with the other three doses and are not shown.


a result of density-dependent survival from year toyear. The average number of adults each year in theunmanaged region over 15 yr was �481,000 beetlesper ha.

In simulations of a relatively small Þeld that is theÞrst to use transgenic corn in the larger region, in-creasing the proportion of transgenic corn in the Þeldincreased returns in most cases. In addition, greatertoxin doses increased returns in these Þelds. In sim-ulations of a 20- or 100-ha Þeld of continuous corn with

a proportion of the landscape planted to a transgeniccultivar that is located within a 2,000-ha unmanagedregion of continuous corn, planting 0% transgeniccorn resulted in an annualized net present value of$194/ha.

With a 20-ha managed Þeld, increasing the propor-tion of transgenic corn in the region always increasedreturns over the no management strategy with anytoxin dose (Fig. 6). Expression of the R-allele did notaffect the returns with any toxin dose or refuge size.

Fig. 5. Annualized net present value (dollars per hectare) of simulations with 85% rotation and three types of geneexpression for the allele for resistance to crop rotation, Y, with the standard management strategy of planting 80% transgeniccorn to the rotated Þeld each year or the 5 or 10% dynamic adoption strategies with a practical high (a) or low toxin dose(b).


Returns were greatest with a theoretical high toxinand 95% transgenic corn in the continuous cornÞeld($271/ha). Returns decreased by up to 23% with 20%transgenic corn compared with 95% transgenic corn.Returns with the lower doses also decreased, up to amaximum of 22% with a low dose (Fig. 6). The dif-ferences between returns with each toxin dose in-creased as the refuge size increased (Fig. 6).

In simulations of a 100-ha managed Þeld, increasingthe proportion of transgenic corn from 20 to 95% in theregion increased returns compared with no manage-ment in all cases with one exception. With a low dose,returns with 20% transgenic corn were 0.4% lowerthan returns with no management. The reason for thisdifference is that the savings from not planting trans-genic corn with a low dose exceeded increased returnscaused by decreasing larval damage by planting 20%transgenic corn. Returns were greatest with a theo-retical high dose, with an annualized net present valueof $272/ha when the proportion of transgenic cornplanted to the continuous cornÞeld was 95% with anyR-allele expression. Returns decreased as dose de-creased, up to a maximum of 19% with a low dose.Similar to simulations of a 20-ha managed Þeld, thedifferences between returns with the theoretical highdose and the lower doses increased as the refuge sizeincreased.

In the simulations of a 20- or 100-ha managed Þeld,the R-allele frequency never exceeded 50% in 15 yrwith any toxin dose, refuge size, or allele expression.The fastest evolution occurred with a theoretical orpractical high dose and 95% transgenic corn, wherethe R-allele frequency reached 0.002 after 15 yr withR dominant with either a 20-or 100-ha managed Þeld.With the lower doses the R-allele frequency neverexceeded 0.001 within 15 yr.

Sensitivity Analysis of Areas without Rotation-Re-sistant Phenotypes. In a sensitivity analysis of areaswith 100% continuous corn and no rotation-resistantphenotypes, changing the discount rate to one or 15%had little effect on the results, as the difference be-tween the dynamic adoption strategies and the stan-dard management strategy of planting 80% transgeniccorn in the continuous cornÞeld each year relative toeach other never changed by �3%. Changing the priceof transgenic corn to $24.7 or $49.4/ha also had littleeffect on the results, as the difference between thevarious strategies relative to each other never changedby �3%.

Changes in the density-dependent survival func-tions did affect the results. Under standard conditions,the 5 or 10% dynamic adoption strategies never gen-erated returns as high as the standard managementstrategy of planting 80% transgenic corn in the con-tinuous cornÞeld each year with a medium or greaterdose and any allele expression. This trend did notchange with various density-dependent survivalfunctions, although the differences between the dy-namic adoption strategies and the standard manage-ment strategy increased. For example, with R reces-sive or partially recessive, returns with the 5% dynamicadoption strategy, a practical high dose, and a density-dependent function that allowed 100% maximum sur-vival were 25 or 23% less, respectively, than the stan-dard strategy. With the function that allowed 39%maximum survival or the density-independent func-tion, returns were 23 or 50% less, respectively, thanwith the standard strategy. These results differed fromthe standard conditions where the 5% dynamicadoption strategy generated returns 15% less than thestandard management strategy with R recessive orpartially recessive. The results were similar with a

Fig. 6. Annualized net present value (dollars per hectare) of simulations of a 20-ha adjacent region with 100% continuouscorn, four toxin doses, and varying proportion of the continuous cornÞeld planted to a transgenic cornÞeld. Toxin dose doesnot affect returns in simulations with 0% transgenic corn.


theoretical high or medium toxin dose or a 10% dy-namic adoption strategy. The affect of density-depen-dence was not as great with a low dose, as the differ-ence between the dynamic adoption strategies and thestandard strategy never differed by �8% when com-paring simulations run with different density-depen-dent functions.

With R dominant, a density dependence functionthat allowed 39 or 100% maximum survival did notaffect the results, as the difference between the 5 or10% dynamic adoption strategies compared with thestandard management strategy of planting 80% trans-genic corn in the continuous cornÞeld each year didnot change by �3%. However, with the density-inde-pendent function and R dominant, the difference be-tween the dynamic adoption strategies and the stan-dard management strategy increased by 20Ð27% witha medium or greater dose.

Varying the price or variable costs of corn had thegreatest effect on the returns of each strategy butnever affected the best management strategy in anycase. Because the amount of corn in the landscape wasalways 100%, changes in the price or variable costs ofcorn affected each strategy equally. The ßuctuationsin returns with any strategy were elastic due to ßuc-tuations in price or variable costs. Increasing and de-creasing the price of corn by 10% resulted in annual-ized net present values �30Ð47% greater and 30Ð47%less, respectively, compared with the standard simu-lations. This indicated that returns had an elasticity of3Ð4.7 due to changes in price. Increasing and decreas-ing the variable costs by 10% resulted in returns �20Ð35% lower and 20Ð35% greater, respectively, com-pared with the standard simulations. This representedan elasticity of 2Ð3.5.Sensitivity Analysis of Areas with Rotation-Resis-tantPhenotypes.Changing the discount rate, the priceof transgenic corn, and density-dependent survivalhad little effect on the results, as the difference be-tween the four different management strategies rela-tive to each other never changed by �4%.

Varying the price and variable costs of corn had thegreatest affect on returns generated with each strategybut never changed the best management strategy.Because the proportion of corn in the landscape neverchanged, changing the price or variable costs of cornaffected each strategy equally. Fluctuations in returnswere elastic due to ßuctuations in price or variablecosts. Increasing and decreasing the price of corn by10% resulted in returns �29% greater and 29% less,respectively, than the standard simulations, represent-ing an elasticity of 2.9. Increasing and decreasing thevariable costs of corn by 10% resulted in returns �18%less and 18% greater, respectively, than the standardsimulations, representing an elasticity of 1.8.

Discussion

In this article, we do not want to emphasize theexact annualized net present value of each manage-ment strategy but rather the differences between thevarious management strategies. In areas without ro-

tation-resistant phenotypes, both toxin dose and geneexpression of the R-allele affected returns. In theseareas, greater toxin doses resulted in increased returnsin every case. Returns with the theoretical and prac-tical high doses were similar, but decreased with themedium and low doses. Returns were similar with Rrecessive or partially recessive and the high doses,where resistance to the transgenic corn did not evolve,but decreased with R dominant. With the medium orlow doses, returns were lower with R partially reces-sive or dominant compared with the R recessive caseas resistance to transgenic corn did evolve.

In areas without rotation-resistant phenotypes, themost economical strategy for farmers may be to adopttransgenic corn as quickly as possible, especially if itis highly effective at controlling western corn root-worm (more than a low dose). In these areas, thedynamic adoption strategies never generated greaterreturns than the standard management strategy ofplanting 80% transgenic corn in the continuous corn-Þeld each year with a medium or greater dose. Withthesedoses and thedynamicadoption strategies, larvaldensities increased by up to 80% compared with thestandard strategy. Increased densities, especially inthe early years of the simulations, resulted in �20%yield lost in refuges and transgenic Þelds and de-creased returns compared with the standard strategy,where yield loss never exceeded 20%.

These results are consistent with the theoreticalÞndings of Laxminarayan and Simpson (2002) using ahighly stylized model of Bt corn. They Þnd that thepest populationÕs intrinsic growth rate and the farm-erÕs discount rate determine the optimal refuge andargue that empirically, the conditions can easily bemet such that 0% refuge is economically optimal. Inthe context of our model, this would imply immediateadoption of Bt corn with 80% refuge would be eco-nomically superior to the two dynamic adoption strat-egies we examine when rotation resistant phenotypesare not present. However, Laxminarayan and Simpson(2002) do not examine optimal refuge strategies whenthe pest also can evolve resistance to another controlmethod such as crop rotation.

With a low-dose corn hybrid planted in continuouscornÞelds, there were several cases where the dy-namic adoption strategies increased returns over thestandard management strategy. With a low dose, thepercentage of yield lost in both refuge and transgenicÞelds increased �22% in most years with any R-alleleexpression and the standard strategy as larval densitiesin both Þelds were high, especially with R dominant.Densities were similar with the dynamic adoptionstrategies and returns increased in most cases com-pared with the standard strategy, as the economicgains from planting less transgenic corn exceededlosses due to larval damage.

This different result for a low dose is generallyconsistent with the Þnding of Secchi et al. (2001), whoderive the economically optimal refuge for Bt corn forEuropean corn borer when the farmer can choose adifferent level of refuge each year. They examine thiseconomically optimal refuge time-path under heavy


and light suppression of the pest population by Btcorn, which are respectively comparable to the highand low dose modeled here. With light suppression(low dose), the optimal time path for refuge is cen-tered around 60% refuge, as opposed to around 5%refuge for a high-dose product (Fig. 1 in Secchi et al.2001). With a low dose, we Þnd that a time-pathbeginning with 90% refuge and slowly decreasing to20% superior to a constant time-path of 20%, whereaswith higher doses, we Þnd the opposite.

In areas with rotation-resistant phenotypes, if therotation-resistant alleles are initially rare and reces-sive, planting transgenic corn is uneconomical. In thiscase, population densities remain low in rotated cornregardless of the type of management. Therefore, a2-yr rotation of nontransgenic corn and soybean maybe the most economical strategy if Y is recessive. Inaddition, the dynamic adoption strategies providedgreater returns than the standard management strat-egy of planting 80% transgenic corn to the rotatedcornÞeld each year in simulations of areas with rota-tion-resistant phenotypes with Y recessive. However,the results of Crowder and Onstad (2005) and Crow-der et al. (2005) indicate that rotation-resistance ismost likely not recessive based on comparisons be-tween modeling work and Þeld observations. If Y isadditive or dominant, the most economical strategyfor farmers may be to adopt transgenic corn as quicklyas possible. With Y dominant, the dynamic adoptionstrategies were less effective than the standard man-agement strategy. With Y recessive, the 5% dynamicadoption strategy provided greater returns than the10% dynamic adoption strategy. In contrast, with Ydominant, the 10% dynamic adoption strategy pro-vided greater returns. With Y recessive, larval popu-lations in rotated corn are low and returns from plant-ing less transgenic corn exceed losses from increasedlarval damage. With Y dominant, the opposite is true,as damage from increased larval populations in rotatedcorn exceed the gains from planting less transgeniccorn throughout the region.

In areas where rotation-resistance is already a se-vere problem, planting 80% transgenic corn in therotated cornÞeld each year was always the most eco-nomical management strategy. In these areas, IRMmay no longer be feasible and IPM may be the onlyoption. With any management strategy, the allele forresistance to crop rotation evolved from the initialvalue of 0.5 toward 1.0. In these severe problem areas,transgenic corn effectively reduces populations in ro-tated cornÞelds and was always more cost-effectivethan implementing a dynamic adoption strategy or a2-yr rotation of nontransgenic corn and soybean.

Farmers who are the Þrst to manage with transgeniccorn in a region may beneÞt economically by increas-ing the proportion of their Þeld planted to a transgeniccultivar. In all cases tested in this study, dispersal ofsusceptible insects from the surrounding region that isnotmanagedwith transgeniccorneffectivelypreventsresistance to the transgenic cultivar from developing.If concerns about resistance are minimal, plantingsmaller refuges may be an effective way to increase

returns. Indeed, Hurley et al. (2002) lament that al-though the effect on the evolution of resistance whenÞelds not managed with transgenic crops supply sus-ceptible adults via dispersal has been studied, there isa general lack of economic assessments of optimalrefuge when a farmer is the Þrst to use transgeniccrops in a region.

The biological analysis of the dynamic adoptionstrategies showed that such strategies can be useful ordetrimental to IRM, depending on the resistance al-lele. The dynamic adoption strategies slowed the evo-lution of resistance to transgenic corn, especially insimulations with R dominant, compared with manage-ment strategies with 80% transgenic corn in either thecontinuous or rotated cornÞeld each year. With thedynamic adoption strategies, the selection pressure onsusceptible phenotypes with regards to transgeniccorn is not as intense, especially during the early yearsof the simulations, because the proportion of trans-genic corn in the region is relatively low at the be-ginning of the 15-yr time horizon (10%) and buildsover time. The result is delayed evolution to trans-genic corn. However, the dynamic adoption strategiesresulted in faster evolution of resistance to crop ro-tation compared with a strategy of planting 80% trans-genic corn in the rotated cornÞeld each year. With thedynamic adoption strategies, there is relatively littlecontrol against rotation-resistant phenotypes in theearly years of the simulations as the proportion oftransgenic corn in the region is low. The result isincreased survival of rotation-resistant individuals inrotated cornÞelds compared with the standard strat-egy and the evolution of resistance to crop rotationoccurs more rapidly.

Changes in the density-dependent survival func-tions had the greatest impact on the results in areaswithout rotation-resistant phenotypes and 100% con-tinuous corn, as the differences between the dynamicadoption and standard management strategieschanged by up to 100%. Although changes in thedensity-dependent survival functions did not affectthe most economical management strategy in theseareas, allowing for increased larval survival decreasedreturns in every case.

The application of these results is limited by severalassumptions in the model. First, we assumed that avery simple genetic system is responsible for evolutionof the behavioral changes and rotation resistance. Sec-ond, we assumed that all farms are the same in ahomogeneous region or that areawide pest manage-ment is occurring. Third, we assumed that plants ex-pressing different toxin doses cost the same amount toproduce. In addition, we used a single economic cri-terion to compare strategies. Other economic criteriaexist that take into account more than just farmerearnings, such as the welfare of the agricultural sector,or society as a whole. For example, we did not considera solution based on the application of insecticides inany cornÞeld. It would be possible to develop aneconomic criterion that takes into account the socialcosts and beneÞts of insecticide use. We also did notconsider the use of a composed error-model similar to


Mitchell et al. (2004) to estimate yield loss due to pestinfestations. Changes in these assumptions could haveproduced other outcomes.

We believe that our results emphasize some of thedifferences between IRM and IPM. IPM strategiesfocus on both farmer returns and resistance to man-agement practices, and are affected by the decliningvalue of future proÞts and the rate of increase inresistance over time. However, the focus with IRM isoften on the distant future and the desire to achievesome biological end point, such as allele frequency orpopulation density, regardless of when these end-points occur. However, farmers have little incentive tobear the costs of delaying resistance in the presentwhen the actual costs of resistance will not be paiduntil far in the future.

In general, IPM and IRM have different goals, con-trolling pest damage to maximize returns versus max-imizing years to resistance. However, other goals,such as environmental protection, are also the focus ofIPM. In some instances, the strategies to pursue thesegoals coincide, in others they do not. For example,management strategies involving planting transgeniccorn to rotated cornÞelds were effective at maximiz-ing farmer returns in areas with rotation-resistant pop-ulations. In addition, this type of management pre-vented resistance to both crop rotation and transgeniccorn from developing. In contrast, in areas withoutrotation-resistant phenotypes and with R dominant,the dynamic adoption strategies were effective atdelaying resistance to transgenic corn, suggesting theymay be useful from an IRM standpoint. However, thedynamic adoption strategies failed to delay resistanceto crop rotation. In addition, these strategies gener-ated lower returns than planting 80% transgenic corneach year, showing they were not effective from anIPM standpoint.

The clearest message for the pest management in-dustry is that producing greater toxin doses may bebeneÞcial from both an IRM and IPM standpoint. Inevery case, returns with the theoretical high and prac-tical high doses were greater than with the medium orlow doses. This was especially true in areas withoutrotation-resistant phenotypes. The greater doses werealso the most effective at preventing resistance totransgenic corn with the standard management strat-egies. In addition, returns with the dynamic adoptionstrategies were always similar compared with the stan-dard strategy with a medium or greater dose, but therewas more uncertainty with a low dose. Therefore, ifthe pest management industry can achieve a high doseof toxin, farmers can plant 80% of their cornÞelds to atransgenic cultivar with conÞdence that this strategywill be beneÞcial both biologically and economically.

Too often in the past, evolutionary changes in pestpopulations have caused scientists to emphasize IRMover IPM. In addition, because of contractual agree-ments and regulatory requirements, farmers havebeen more willing to adopt IRM management prac-tices compared with IPM. However, our results showthat in many cases management strategies that focuson IPM are also beneÞcial from an IRM perspective.

Farmers must constantly strive to manage pests byusing strategies that can be beneÞcial from both per-spectives.

Acknowledgments

We thank the editor Michael Brewer and two anonymousreviewers whose comments greatly improved this manu-script. We also thank University of Illinois and the College ofAgricultural, Consumer and Environmental Sciences for sup-porting this research with a Jonathan Baldwin Turner Fel-lowship. This work was supported by a grant to D.W.O. fromthe USDA Biotechnology Risk Assessment program.

References Cited

Ball, H. J., and G. T.Weekman. 1962. Insecticide resistancein the adult western corn rootworm in Nebraska. J. Econ.Entomol. 55: 439Ð441.

Ball,H. J., andG.T.Weekman. 1963. Differential resistanceof corn rootworms to insecticides in Nebraska and ad-joining states. J. Econ. Entomol. 56: 553Ð555.

Carriere, Y., T. J. Dennehy, B. Petersen, S. Haller, C. Ellers-Kirk, L. Antilla, Y.-B. Liu, E.Willot, and B. E. Tabashnik.2001. Large-scale management of insect resistance totransgenic cotton in Arizona: can transgenic insecticidalcrops be sustained? J. Econ. Entomol. 94: 315Ð325.

Carriere, Y., C. Ellers-Kirk,M. Sisterson, L. Antilla,M.With-low, T. J. Dennehy, and B. E. Tabashnik. 2003. Long-term regional suppression of pink bollworm by Bt cotton.Proc. Natl. Acad. Sci. U.S.A. 100: 1519Ð1523.

Carriere, Y., M. S. Sisterson, and B. E. Tabashnik. 2004a.Resistance management for sustainable use of Bacillusthuringiensis crops in integrated pest management,pp. 65Ð95. In A. R. Horowitz and I. Ishaaya [eds], Insectpest management: Þeld and protected crops. Springer,Berlin, Germany.

Carriere, Y., P. Dutilleul, C. Ellers-Kirk, B. Pederson,S. Haller, L. Antilla, T. J. Dennehy, and B. E. Tabashnik.2004b. Sources, sinks, and the zone of inßuence of ref-uges for managing insect resistance to Bt crops. Ecol.Appl. 14: 1615Ð1623.

Crowder, D. W., and D. W. Onstad. 2005. Using a genera-tional time-step model to simulate the dynamics of ad-aptation to transgenic corn and crop rotation by westerncorn rootworm (Coleoptera: Chrysomelidae). J. Econ.Entomol. 98: 518Ð533.

Crowder, D. W., D. W. Onstad, M. E. Gray, C.M.F. Pierce,A.G.Hager, S. T. Ratcliffe, andK. L. Steffey. 2005. Anal-ysis of the dynamics of adaptation to transgenic corn andcrop rotation by western corn rootworm (Coleoptera:Chrysomelidae) using a daily time-step model. J. Econ.Entomol. 98: 534Ð551.

Georghiou, G. P. 1986. The magnitude of the resistanceproblem, pp. 14Ð43. In National Academy of Sciences[ed.], Pesticide resistance: strategies and tactics for man-agement. National Academy Press, Washington, DC.

Guse, C. A., D. W. Onstad, L. L. Bushman, P. Porter,R. A. Higgins, P. E. Sloderbeck, G. B. Cronholm, andF. B. Peairs. 2002. Modeling the development of resis-tance by stalk-boring Lepidoptera (Crambidae) in areaswith irrigated, transgenic corn. Environ. Entomol. 31:676Ð685.

Hibbard, B. E.,M. L.Higdon,D. P.Duran, Y.M. Schweikert,and M. R. Ellersieck. 2004. Role of egg density on es-tablishment and plant-to-plant movement by westerncorn rootworm larvae (Coleoptera: Chrysomelidae).J. Econ. Entomol. 97: 871Ð882.


Hurley, T. M., S. Secchi, B. A. Babcock, and R. L. Hellmich.2002. Managing the risk of European corn borer resis-tance to Bt corn. Environ. Resour. Econ. 22: 537Ð558.

Illinois Agricultural Statistics Service. 2003. Illinois annualsummary: Þeld crops, pp. 21Ð22. Ill. Dept. Agric., Spring-Þeld, IL.

Laxminarayan, R., and R. D. Simpson. 2002. Refuge strat-egies for managing pest resistance in transgenic agricul-ture. Environ. Resour. Econ. 22: 521Ð536.

Matson,P.A.,W. J.Parton,A.G.Power, andM. J. Swift. 1997.Agricultural intensiÞcation and ecosystem properties.Science (Wash. DC) 277: 504Ð509.

Meinke, L. J., B. D. Siegfried, R. J. Wright, and L. D. Chan-dler. 1998. Adult susceptibility of Nebraska westerncorn rootworm (Coleoptera: Chrysomelidae) popula-tions to selected insecticides. J. Econ. Entomol. 91: 594Ð600.

Metcalf, R. L. 1983. Implications and prognosis of resistanceto insecticides, pp. 703Ð733. In G. P. Georghiou andT. Saito [eds.], Pest resistance to pesticides. Plenum, NewYork.

Microsoft. 2002. Microsoft Excel 2002. Microsoft, Seattle,WA.

Midgarden, D., S. J. Fleischer, R. Weisz, and Z. Smilowitz.1997. Site-speciÞc integrated pest management impacton development of esfenvalerate resistance in Coloradopotato beetle (Coleoptera: Chrysomelidae) and on den-sities of natural enemies. J. Econ. Entomol. 90: 855Ð867.

Miota, F., M. E. Scharf, M. Ono, P. Marcon, L. J. Meinke,R. J. Wright, L. D. Chandler, and B. D. Siegfried. 1998.Mechanisms of methyl and ethyl parathion resistance inthe western corn rootworm (Coleoptera: Chrysomeli-dae). Pestic. Biochem. Physiol. 61: 39Ð52.

Mitchell, P. D., M. E. Gray, and K. L. Steffey. 2004. A com-posed-error model for estimating pest-damage functionsand the impact of the western corn rootworm soybeanvariant in Illinois. Am. J. Agric. Econ. 86: 332Ð344.

Onstad, D. W., M. G. Joselyn, S. A. Isard, E. Levine,J. L. Spencer, L. W. Bledsoe, C. R. Edwards, C. D. Di-Fonzo, and H. Willson. 1999. Modeling the spread ofwestern corn rootworm (Coleoptera: Chrysomelidae)populations adapting to soybean-corn rotation. Environ.Entomol. 28: 188Ð194.

Onstad, D. W., C. A. Guse, J. L. Spencer, E. Levine, andM. E. Gray. 2001a. Modeling the dynamics of adapta-tion to transgenic corn by western corn rootworm(Coleoptera: Chrysomelidae). J. Econ. Entomol. 94: 529Ð540.

Onstad, D. W., J. L. Spencer, C. A. Guse, E. Levine, andS. A. Isard. 2001b. Modeling evolution of behavioral re-sistance by an insect to crop rotation. Entomol. Exp. Appl.100: 195Ð201.

Onstad, D. W., C. A. Guse, P. Porter, L. L. Bushman,R. A. Higgins, P. E. Sloderbeck, F. B. Peairs, and G. B.Gronholm. 2002. Modeling the development of resis-tance by stalk-boring lepidopteran insects (Crambidae)in areas with transgenic corn and frequent insecticideuse. J. Econ. Entomol. 95: 1033Ð1043.

Onstad, D. W., D. W. Crowder, P. D. Mitchell, C. A. Guse,J. L. Spencer, E. Levine, and M. E. Gray. 2003a.Economics versus alleles: balancing integrated pest man-agement and insect resistance management for rotation-resistant western corn rootworm (Coleoptera: Chry-somelidae). J. Econ. Entomol. 96: 1872Ð1885.

Onstad, D. W., D. W. Crowder, S. A. Isard, E. Levine,J. L. Spencer, M. O’Neal, S. Ratcliffe, M. E. Gray,L. W. Bledsoe, C. D. Di Fonzo, B. Eisley, and C. R.Edwards. 2003b. Does landscape diversity slow thespread of rotation-resistant western corn rootworm(Coleoptera: Chrysomelidae)? Environ. Entomol. 32:992Ð1001.

Peck, S. L., and S. P. Ellner. 1997. The effect of economicthresholds and life-history parameters on the evolution ofpesticide resistance in a regional setting. Am. Nat. 149:43Ð63.

Pimental, D., H. Acquay, M. Biltonen, P. Rice, M. Silva,J. Nelson, V. Lipner, S. Giordano, A. Horowitz, andM. D’Amore. 1992. Environmental and economic costsof pesticide use. Bioscience 42: 750Ð760.

Rondon, S. I., and M. E. Gray. 2004. Ovarian developmentand ovipositional preference of the western corn root-worm (Coleoptera: Chrysomelidae) variant in east cen-tral Illinois. J. Econ. Entomol. 97: 390Ð396.

Roush, R. T. 1997. Managing resistance to transgenic crops,pp. 271Ð294. InN. Carozzi and M. Koziel [eds.], Advancesin insect control: the role of transgenic plants. Taylor &Francis, London, United Kingdom.

Scharf, M. E., L. J. Meinke, B. D. Siegfried, R. J. Wright,and L. D. Chandler. 1999. Carbaryl susceptibility, diag-nostic concentration determination, and synergism forU.S. populations of western corn rootworm (Coleoptera:Chrysomelidae). J. Econ. Entomol. 92: 33Ð39.

Schnitkey, G. 2004. Estimated cost of crop production inIllinois, 2004. Farm business management handbookFBM-0100. University of Illinois, Urbana-Champaign, IL.

Secchi, S., T. M. Hurley, and R. L. Hellmich. 2001. Manag-ing European corn borer resistance to Bt corn with dy-namic refuges. Center for Agricultural and Rural Devel-opment, Iowa State University, Ames, IA. Working Paper01-WP 287.

Storer, N. P. 2003. A spatially explicit model simulatingwestern corn rootworm (Coleoptera: Chrysomelidae)adaptation in insect-resistant maize. J. Econ. Entomol. 96:1530Ð1547.

Walker, K., J. Liebman, and W. Pease. 1995 Pesticide in-duced disruptions of agricultural ecosystems. CaliforniaPolicy Seminar Report. University of California, Berke-ley.

Zhou, X., M. E. Scharf, L. J. Meinke, L. D. Chandler, andB. D. Siegfried. 2003. Characterization of general ester-ases from methyl parathion-resistant and -susceptiblepopulations of western corn rootworm (Coleoptera:Chrysomelidae). J. Econ. Entomol. 96: 1855Ð1863.

Received 23 July 2004; accepted 27 February 2005.


Date post:	20-Aug-2020
Category:	Documents
Upload:	others
View:	1 times
Download:	0 times

I R M Economic Analysis of Dynamic Management Strategies...

Documents