2.1. Simulation Models
To investigate the pest management implications of climate warming, two simulation models were linked. EPIC (Erosion-Productivity Impact Calculator) was originally developed to examine the impact of agricultural practices on soil erosion and productivity [41
]. It simulates hydrology, erosion-sedimentation, nutrient cycling, pesticide fate and plant growth in response to management practices. A single model structure simulates plant growth with species-specific model parameters. State variables are site-specific; they include physical and chemical characteristics of the soil, topography, latitude, longitude and elevation; land management includes crop rotation sequence, planting schedule, fertilization and pest management. EPIC has been widely used to assess the environmental impact of land management practices, water quality and the effect of temperature and CO2
change on crop growth and water use [42
]. GILSYM (Generalized Insect Life-SYstem Model) is a generalized programmable model capable of simulating a wide range of insect life histories. GILSYM simulates daily cohorts through their life-cycle with eggs laid on a given day constituting a cohort that progresses through the nymphal or larval and pupal stages to adult stage. Using stage-specific developmental thresholds and temperature dependent growth rates, the model is driven by degree-days. Food availability and mortality due to natural enemies and other hazards also modify the cohorts daily. Migrant and resident species start the year differently. Resident populations in diapause start development from the overwintering stage when the degree-day accumulation to exit diapause is reached. Migrants arrive in the spring months whenever the winds are in the south and temperatures exceed the threshold for flight. Simulating the source population is unnecessary as the number of migrants arriving is always a small proportion of the overwintering population.
The version of EPIC used (EPIC.0804) adjusts yield loss due to pests at harvest by reducing the yield by a predetermined factor. In the linked model, the abundance and feeding rate of the damaging stage(s) was used to calculate the yield loss factor each year for each pest. If an insect did not feed on the current crop, the loss factor was set to zero, otherwise it ranged from 0 to 1 (no damage to total loss). Plant phenology is not modeled explicitly in EPIC so a tight linkage between pest and crop is not possible. The use of a loss factor at harvest based on pest abundance during the growing season provided a loose but useful linkage between crop and pest subsystems (A later version of EPIC linking crop growth and pest abundance on a daily basis and permitting pest control measures to be implemented [43
] was not available for this project.).
Both subsystems output a variety of variables over a range of time frames: daily, monthly and annual. Simulations were run at the resolution of the county for an eight state region of the Corn Belt: Illinois, Indiana, Iowa, Kansas, Kentucky, Missouri, Nebraska and Ohio. The region comprises 813 counties and occupies an area of approximately 1400 × 400 km (~560,000 km2) of which ~60% is devoted to corn and soybeans agriculture grown both separately and in rotation. More than 60% of U.S. corn and soybeans are produced in this region. The Midwest Corn Belt is diverse both latitudinally and longitudinally in soil, elevation and climate and could reasonably be expected to reveal changes in pest population and crop yield isoclines in response to climate change.
Data of the dominant soil type in each county was used in EPIC’s crop growth model. These data were obtained from the Natural Resource Conservation Service (NRCS) website [44
]. In each county plant-insect simulations were driven by data of maximum and minimum temperatures and precipitation for the closest National Oceanographic and Atmospheric Administration (NOAA) [45
] station for the years 1901–2000. For the years 2001–2100 we used predictions of maximum and minimum temperatures and precipitation from the Geophysical Fluid Dynamics Laboratory (GFDL) climate model CM2.0 [46
] at intervals of 1° of latitude and longitude. A distance-weighted average of the 8 points closest to the centroid of each county was used to estimate weather variables for that county. The scenario used for weather prediction was the SRES-B1 scenario of the Special Report on Emissions Scenarios’ developed for the IPCC 4th Assessment Report (The IPCC 5th Assessment Report was released after the inception of this project.) [47
] that predicts a lower emissions path over the 21st century in which atmospheric CO2
concentrations are assumed to stabilize at double the pre-industrial level of ~280 ppm by the year 2100. Since conducting the study it has become apparent that this scenario is too optimistic. Its very optimism, however, does not take the scenario too far outside our experience and provides a lower bound for likely changes to pest management requirements.
Our study took the single factor modeling approach commonly used in idealized forcing scenarios, such as those developed for the IPCC, in which a single forcing variable is investigated while holding all others constant [48
]. Concentrating on a single forcing agent avoids ambiguities associated with more complicated but realistic scenarios while losing the ability to compare model predictions directly with observations. For this reason, it is usual to express results as proportional change, rather than as absolute numbers.
Agronomic practices vary substantially both spatially and temporally. However, we chose to employ a single agronomic practice in order not to confound climate’s spatial variability with that of agricultural practice. A major variable in the Midwest corn-soybeans rotation is the cultivar used, although even this has become more uniform with the widespread adoption of Roundup Ready® corn and soybeans over the past two decades. Other aspects of crop production have also become less variable over time: with the increased use of GMO crops, minimum and no-till practices have replaced conventional tillage over large areas of the Midwest. To simplify the study and concentrate on climate forcing, we used the agronomic practices most commonly used in Illinois in 2000. This standard was chosen because Illinois is the central state in our study area and 2000 the mid-point in the simulations that started in 1901 and ended in 2100. The following assumptions are implicit in the experiments: agronomic practices are spatially homogeneous; insect temperature-dependent vital rates (growth, survival and reproduction) are spatially homogeneous; insect phenotypic changes are adaptations to the forcing variable only; geographic variation is the only spatially non-homogeneous state variable; and climate is the only forcing variable; there has been no evolution in technology. Although EPIC has a simple rate function to simulate technology change it is not clear how to realistically incorporate technology into the models over such a long period. More importantly, this last assumption is necessitated by the need not to confound technological change with climate change over the same timeframe.
Obviously, absolute values of crop production or insect population variables cannot be interpreted either forwards or backwards, so we concentrated on relative change across four half centuries. Over 29,000 simulations were run and their output entered into a geographic database for analysis and visualization. Both EPIC and GILSYM, when driven by daily weather records, are essentially deterministic, obviating the need for replicate runs and stochastic analysis.
2.2. The Pests
Nine insect pests of corn and/or soybeans were selected for study: bean leaf beetle (Ceratoma trifurcata (Forster, 1771)) and Mexican bean beetle (Epilachna varivestis Mulsant, 1850) (Coleoptera); armyworm (Pseudoletia unipuncta Haw., 1809), black cutworm (Agrotis ipsilon Hufn., 1766), corn earworm (Helicoverpa zea (Boddie, 1850)), European corn borer (Ostrinia nubilalis Hübner, 1796), stalk borer (Papaipema nebris Guenée, 1852) and velvetbean caterpillar (Anticarsia gemmatalis Hübner, 1818) (Lepidoptera); potato leafhopper (Empoasca fabae (Harris, 1841)) (Cicadellidae). Each of these species is currently a sporadic and occasionally serious pest of one or other crop in at least part of our study area. They were chosen to include resident and migrant species and to represent species with single and multiple generations per year that feed primarily on one or both crops.
We present results of four of the nine species to illustrate the most probable effects of climate warming on insect pests in the Corn Belt. The stalk borer is a resident pest of both corn and soybeans that is univoltine throughout its range. Its winter diapause is obligate, effectively resetting the life cycle each spring. Thus, the main impact of climate warming on this species is likely to be increased overwintering survival and more rapid development in spring. The Mexican bean beetle is also a resident pest on soybeans. It overwinters in the egg stage or as diapausing early instar larvae and has two or more generations per year over parts of our study area. Elevated winter temperatures could increase overwintering survival but the fecundity and larval survival of Mexican bean beetles are reduced by high temperatures and low humidity, which could lead to reduced risk of this species in parts of our area. By contrast, the armyworm is a migrant corn pest that overwinters only in the southern states and migrates north every spring. Warmer winters further north could move the overwintering ranges of this moth northwards and/or increase the size of spring migrations. Unlike the armyworm the potato leafhopper is a weak flier that can only fly any distance by taking advantage of the wind. Potato leafhopper behavior is adapted to take advantage of the regular and repeatable wind patterns that form in the central part of the country. These wind patterns enable the leafhopper to migrate north in spring and return to overwintering areas south of the 36th parallel in late summer [49
]. Potato leafhopper is highly polyphagous and is very common in the northern tier of states throughout the summer. At present, it is an occasional pest on soybeans, causing leaf damage known as “hopper burn” that reduces translocation and photosynthetic area. However, owing to its fairly rapid generation time, a lengthening of the growing season could result in an increase in the number of generations and therefore a change in its pest status.
Data and functional relationships governing the vital rates of corn, beans and the insect pests were obtained from the literature (primarily Extension bulletins from the Land Grant Universities in our study area) to parameterize EPIC/GILSYM. The major life history variables used are summarized in Table 1
. The corn/soybean agricultural practices recommended by Illinois Extension were used to program EPIC’s agronomy module. The model, input data and major results of the simulation study are archived at The Ohio State University Libraries (https://hdl.handle.net/1811/45903