Enhancing predictive performance for spectroscopic studies in wildlife science through a multi-model approach: A case study for species classification of live amphibians.

Near infrared spectroscopy coupled with predictive modeling is a growing field of study for addressing questions in wildlife science aimed at improving management strategies and conservation outcomes for managed and threatened fauna. To date, the majority of spectroscopic studies in wildlife and fisheries applied chemometrics and predictive modeling with a single-algorithm approach. By contrast, multi-model approaches are used routinely for analyzing spectroscopic datasets across many major industries (e.g., medicine, agriculture) to maximize predictive outcomes for real-world applications. In this study, we conducted a benchmark modeling exercise to compare the performance of several machine learning algorithms in a multi-class problem utilizing a multivariate spectroscopic dataset obtained from live animals. Spectra obtained from live individuals representing eleven amphibian species were classified according to taxonomic designation. Seven modeling techniques were applied to generate prediction models, which varied significantly (p < 0.05) with regard to mean classification accuracy (e.g., support vector machine: 95.8 ± 0.8% vs. K-nearest neighbors: 89.3 ± 1.0%). Through the use of a multi-algorithm approach, candidate algorithms can be identified and applied to more effectively model complex spectroscopic data collected for wildlife sciences. Other key considerations in the predictive modeling workflow that serve to optimize spectroscopic model performance (e.g., variable selection and cross-validation procedures) are also discussed.

Models of Diabrotica Populations: Demography, Population Genetics, Geographic Spread, and Management.

Both Diabrotica virgifera virgifera LeConte and D. barberi Smith and Lawrence are among the most damaging insects impacting corn in North America. D. virgifera virgifera has also invaded Europe and has become an important pest in that region. Computer models have become an important tool for understanding the impact and spread of these important pests. Over the past 30 years, over 40 models have been published related to these pests. The focus of these models range from occupancy models (particularly for Europe), impact of climate change, range expansion, economics of pest management, phenology, to the evolution of resistance to toxins and crop rotation. All of these models share characteristics. We elaborate on the methods in which modelers have incorporated the biology of these pests, including density-dependence, movement, fecundity and overwintering mortality. We discuss the utility of both spatially-explicit, complex models and spatially-implicit, generational models and where each might be appropriate. We review resistance models that either explain past evolution to crop rotation, insecticides or insecticidal traits or attempt to predict the consequences of resistance management strategies.

The Corn-Cotton Agroecosystem in the Mid-Southern United States: What Insecticidal Event Pyramids Should be Used in Each Crop to Extend Vip3A Durability.

Recent studies suggest that resistance in Helicoverpa zea (Boddie) (Lepidoptera, Noctuidae) to Cry1A(b/c) and Cry2Ab2 toxins from the bacterium Bacillus thuringiensis Berliner (Bacillales: Bacillaceae) has increased and field efficacy is impacted in transgenic corn and cotton expressing these toxins. A third toxin, Vip3A, is available in pyramids expressing two or more Bt toxins in corn hybrids and cotton varieties, but uncertainty exists regarding deployment strategies. During a growing season, H. zea infests corn and cotton, and debate arises over use of Vip3A toxin in corn where H. zea is not an economic pest. We used a three-locus, spatially explicit simulation model to evaluate when using Vip3A in corn might hasten evolution of resistance to Vip3A, with implications in cotton where H. zea is a key pest. When using a conventional refuge in corn and initial resistance allele frequencies of Cry1A and Cry2A were 10%, transforming corn with Vip3A slowed resistance to these toxins and delayed resistance evolution to the three-toxin pyramid as a whole. When Cry resistance allele frequencies exceeded 30%, transforming corn with Vip3A hastened the evolution of resistance to the three-toxin pyramid in cotton. When using a seed blend refuge strategy, resistance was delayed longest when Vip3A was not incorporated into corn and used only in cotton. Simulations of conventional refuges were generally more durable than seed blends, even when 75% of the required refuge was not planted. Extended durability of conventional refuges compared to other models of resistance evolution are discussed as well as causes for unusual survivorship in seed blends.

Density Dependence and Growth Rate: Evolutionary Effects on Resistance Development to Bt (Bacillus thuringiensis).

It has long been recognized that pest population dynamics can affect the durability of a pesticide, but dose remains the primary component of insect resistance management (IRM). For transgenic pesticidal traits such as Bt (Bacillus thuringiensis Berliner (Bacillales: Bacillaceae)), dose (measured as the mortality of susceptibles caused by a toxin) is a relatively fixed characteristic and often falls below the standard definition of high dose. Hence, it is important to understand how pest population dynamics modify durability and what targets they present for IRM. We used a deterministic model of a generic arthropod pest to examine how timing and strength of density dependence interacted with population growth rate and Bt mortality to affect time to resistance. As in previous studies, durability typically reached a minimum at intermediate doses. However, high population growth rates could eliminate benefits of high dose. The timing of density dependence had a more subtle effect. If density dependence operated simultaneously with Bt mortality, durability was insensitive to its strengths. However, if density dependence was driven by postselection densities, decreasing its strength could increase durability. The strength of density dependence could affect durability of both single traits and pyramids, but its influence depended on the timing of density dependence and size of the refuge. Our findings suggest the utility of a broader definition of high dose, one that incorporates population-dynamic context. That maximum growth rates and timing and strength of interactions causing density dependent mortality can all affect durability, also highlights the need for ecologically integrated approaches to IRM research.

The Impact of Inter-Kernel Movement in the Evolution of Resistance to Dual-Toxin Bt-Corn Varieties in Helicoverpa zea (Lepidoptera: Noctuidae).

Seeds or kernels on hybrid plants are primarily F(2) tissue and will segregate for heterozygous alleles present in the parental F(1) hybrids. In the case of plants expressing Bt-toxins, the F(2) tissue in the kernels will express toxins as they would segregate in any F(2) tissue. In the case of plants expressing two unlinked toxins, the kernels on a Bt plant fertilized by another Bt plant would express anywhere from 0 to 2 toxins. Larvae of corn earworm [Helicoverpa zea (Boddie)] feed on a number of kernels during development and would therefore be exposed to local habitats (kernels) that varied in their toxin expression. Three models were developed for plants expressing two Bt-toxins, one where the traits are unlinked, a second where the traits were linked and a third model assuming that maternal traits were expressed in all kernels as well as paternally inherited traits. Results suggest that increasing larval movement rates off of expressing kernels tended to increase durability while increasing movement rates off of nonexpressing kernels always decreased durability. An ideal block refuge (no pollen flow between blocks and refuges) was more durable than a seed blend because the refuge expressed no toxins, while pollen contamination from plants expressing toxins in a seed blend reduced durability. A linked-trait model in an ideal refuge model predicted the longest durability. The results suggest that using a seed-blend strategy for a kernel feeding insect on a hybrid crop could dramatically reduce durability through the loss of refuge due to extensive cross-pollination.

Risk assessment for Helicoverpa zea (Lepidoptera: Noctuidae) resistance on dual-gene versus single-gene corn.

Recent Environmental Protection Agency (EPA) decisions regarding resistance management in Bt-cropping systems have prompted concern in some experts that dual-gene Bt-corn (CrylA.105 and Cry2Ab2 toxins) may result in more rapid selection for resistance in Helicoverpa zea (Boddie) than single-gene Bacillus thuringiensis (Bt)-corn (CrylAb toxin). The concern is that Bt-toxin longevity could be significantly reduced with recent adoption of a natural refuge for dual-gene Bt-cotton (CrylAc and Cry2Ab2 toxins) and concurrent reduction in dual-gene corn refuge from 50 to 20%. A population genetics framework that simulates complex landscapes was applied to risk assessment. Expert opinions on effectiveness of several transgenic corn and cotton varieties were captured and used to assign probabilities to different scenarios in the assessment. At least 350 replicate simulations with randomly drawn parameters were completed for each of four risk assessments. Resistance evolved within 30 yr in 22.5% of simulations with single-gene corn and cotton with no volunteer corn. When volunteer corn was added to this assessment, risk of resistance evolving within 30 yr declined to 13.8%. When dual-gene Bt-cotton planted with a natural refuge and single-gene corn planted with a 50% structured refuge was simulated, simultaneous resistance to both toxins never occurred within 30 yr, but in 38.5% of simulations, resistance evolved to toxin present in single-gene Bt-corn (CrylAb). When both corn and cotton were simulated as dual-gene products, cotton with a natural refuge and corn with a 20% refuge, 3% of simulations evolved resistance to both toxins simultaneously within 30 yr, while 10.4% of simulations evolved resistance to CrylAb/c toxin.

Future fitness of female insect pests in temporally stable and unstable habitats and its impact on habitat utility as refugees for insect resistance management.

The long-term fitness of individuals is examined in complex and temporally dynamic ecosystems. We call this multigeneration fitness measure "future fitness". Helicoverpa zea (Boddie) (Lepidoptera: Noctuidae) is a polyphagous insect that feeds on many wild and cultivated hosts. While four generations of H. zea occur during the cropping season in the U.S. Mid Southern agroecosysem, the latter two generations were of most interest, as corn (which has been largely nontransgenic in the Mid-South) dominates the first two generations in the cropping system. In simulations of the evolution of resistance to Bt-transgenic crops, cotton refuge areas were found to be significantly more effective than similar soybean acreages at delaying the evolution of resistance. Cotton is a suitable host for H. zea during two late summer generations, while a soybean field is suitable for only one of these generations, therefore soybean fields of other maturity groups were simulated as being attractive during the alternative generation. A hypothetical soybean variety was tested in which a single field would be attractive over both generations and it was found to be significantly more effective at delaying resistance than simulated conventional soybean varieties. Finally, the placement of individuals emerging at the start of the 3rd (first without corn) generation was simulated in either refuge cotton, conventional soybean and the hypothetical long attractive soybean and the mean number of offspring produced was measured at the end of the season. Although females in conventional and long soybean crops had the same expected fecundity, because of differences in temporal stability of the two crops, the long soybean simulations had significantly more H. zea individuals at the end of the season than the conventional soybean simulations. These simulations demonstrate that the long-term fecundity associated with an individual is dependent not only on the fecundity of that individual in its current habitat, but also the temporal stability of habitats, the ecosystem at large and the likelihood that the individual's offspring will move into different habitats.

Characteristics of 11 polymorphic microsatellite markers in the red imported fire ant, Solenopsis invicta Buren.

We have characterized 11 polymorphic microsatellite loci in the invasive ant Solenopsis invicta. Primer pairs were evaluated on fire ants collected from monogyne mounds in Lauderdale County, Mississippi. The observed and effective number of alleles ranged from two to six and from 1.31 to 2.64, respectively. The observed and expected heterozygosity values ranged from 0.1613 to 0.7826 and from 0.1491 to 0.6242, respectively. The polymorphism information content of the microsatellites ranged from 0.1482 to 0.6208. Probability tests indicated significant deviations from the Hardy-Weinberg equilibrium at three loci. Pairwise tests did not detect linkage disequilibrium between any pair of loci.

Assessing risk of resistance to aerial applications of methyl-parathion in western corn rootworm (Coleoptera: Chrysomelidae).

We validated a stochastic model of the evolution of resistance to adulticidal sprays of methyl-parathion in western corn rootworm, Diabrotica virgifera virgifera LeConte, populations in Nebraska. The population dynamics predicted by the model resembled that reported for field populations, and time until control failures occurred closely matched reports by commercial crop consultants. We incorporated uncertainty about the values used for 18 model parameters by replacing default values with random draws taken from a normal distribution. One parameter, the initial resistance allele frequency, was no longer measurable because of the evolution of resistance. We therefore proposed five candidate initial allele frequencies and developed probability distributions for the time to resistance for each by running 1000 simulations with parameters randomly varied. These distributions included variation because of stochastic effects as well as parameter uncertainty. We used Bayesian inference to estimate the candidate frequency most likely, given reported times to field control failures. The initial allele frequency of 10(-4) was most likely (29%), 10(-3) was less likely (28%), whereas 10(-6) was relatively unlikely (5%). Results from sensitivity analysis depended upon how evolution of resistance was measured. When resistance was examined as a genetic phenomenon, the rate of increase of the resistance allele depended almost entirely on genetic factors (LC50 values), the characteristics of the pesticide (residual activity), and the variance associated with emergence of adults. When resistance was measured as failure of methyl-parathion to reduce populations below threshold levels (0.5 gravid females per plant), parameters that contributed to population growth rate (mortality and fecundity) were also important. These data suggest two important phases in resistance evolution in corn rootworms: a genetic phase associated with negative growth rates and rapid changes in resistance allele frequencies and a rebound phase associated with positive growth rates and near fixation of the resistance allele.

Modeling the impact of alternative hosts on Helicoverpa zea adaptation to bollgard cotton.

For highly polyphagous cotton, Gossypium hirsutum L., pests such as Helicoverpa zea (Boddie), a substantial portion of the larval population develops on noncotton alternative hosts. These noncotton hosts potentially provide a natural refuge for H. zea, thereby slowing the evolution of resistance to the Bacillus thuringiensis Berliner (Bt)-derived Cry1Ac protein present in Bollgard cotton. Here, we demonstrate how the measured contribution of such alternative hosts can be included in estimating the "effective refuge" present for H. zea and in modeling resistance evolution in this species. A single-gene, two-compartment model was used in which one compartment represented corn, Zea mays L., and cotton that express the Cry1Ac protein or similar proteins, and the other compartment was the effective refuge, made up of a weighted average of non-Bt cotton and noncotton hosts. The effective refuge was estimated for each of six generations of H. zea based upon available data on larval population densities on different hosts and cropping patterns. Model runs were performed for regions centered on three states: Georgia, Mississippi, and North Carolina. Three sets of fitness cost assumptions for the putative resistance gene were used: none, low, and moderate, with either recessive or additive inheritance for resistance and fitness costs. For Georgia and North Carolina, resistance was predicted to take >30 yr to evolve except in the absence of fitness costs. For Mississippi, results were sensitive to fitness costs: >30 yr with moderate costs, 7-14 yr with low costs, and 6-10 yr without such costs.

Evaluating the impacts of refuge width on source-sink dynamics between transgenic and non-transgenic cotton.

Resistance management for Bt-transgenic crops relies in part on the production of sufficient numbers of susceptible insects in non-toxic refuges. Simulation models suggested that source-sink dynamics could interact with the structure of refuges to impact the production of insects in these areas. We tested the hypothesis that altering isolation between refuges and transgenic cotton by manipulating the width of refuges embedded within cotton fields would alter the density of Heliothis virescens and Helicoverpa zea eggs oviposited in refuges. Three categories of refuge widths were tested over two years: they included narrow (16-24 m wide), medium (32-48 m wide) and wide (80-96 m wide) refuges. Isolation between the two habitats increased as refuge width increased. In 1996, eggs of H. virescens from H. zea were not distinguished, but a significant increase in the density of eggs and a significant decrease in relative yield (refuge yield compared to the yield from immediately surrounding Bt-cotton) was found as refuge width increased. In 1997, eggs from H. virescens were analyzed separately from H. zea using an ELISA test. The density of H. virescens eggs increased with increasing refuge width, and there was a significant decline in density of H. virescens eggs with increasing distance from the refuge. In contrast, there was no impact of refuge width on the density of H. zea eggs, nor was the slope of a regression of egg density and distance from the refuge significantly different from zero. We suggest that these differences reflect differences in the biology of the two insects.

Genetics and fitness costs of cyromazine resistance in the house fly (Diptera: Muscidae).

The genetic basis of cyromazine resistance was investigated in the house fly, Musca domestica L. The ED-R strain, which was collected in Mississippi and selected further in the laboratory, was 116.5-fold resistant compared with the laboratory susceptible strain, OR-S. The SEL strain, which was created by crossing ED-R with OR-S followed by three cycles of reselection and backcrossing to OR-S, was 84.7-fold resistant relative to the susceptible strain. Mortality data from reciprocal crosses of resistant and susceptible flies indicated that resistance was autosomal and not influenced by maternal effects. The relative position of probit lines from the parental strains and reciprocal crosses showed that resistance was expressed as an incompletely dominant trait with D = 0.30 and 0.32 for ED-R and SEL, respectively. To determine the number of genes involved, models of one, two, three, four, and five loci were used to compare observed and expected mortality of F1ED-R x susceptible backcross. Resistance was best described by a polygenic model of three loci when equal and additive effects of loci were assumed. Another approach, which was based on phenotypic variances, showed that nE, or the minimum number of freely segregating genetic factors for ED-R, equaled 3.07. ED-R showed greater reductions in fitness compared with SEL independent of the presence or absence of sublethal concentrations of cyromazine. These data suggested that reduced fitness was not due to deleterious pleiotropic effects of the resistance genes themselves but arose from other loci in the ED-R genotype.

Heritability of wing-beat frequency in Anopheles quadrimaculatus.

The repeatability of male wing-beat frequency measurements of Anopheles quadrimaculatus was determined by using mosquitoes allowed free flight in a confined space. Heritability of the wing-beat frequency trait was estimated for a laboratory and a wild-strain population of An. quadrimaculatus by using free-flight measurement with a parent-offspring regression of offspring on dams. Repeatability was 0.75 for free flight. Wing-beat frequency rose for the 1st day after adult emergence and then became steady. Female heritability of wing-beat frequency was 21.6% for colony and 24.0% for wild-strain mosquitoes. Male heritability was 57.2% for colony and 53.7% for wild-strain mosquitoes. Male heritability was significantly different from 0 when probabilities were combined across both populations.