In areas of intensive maize–soybean rotation, rootworm larvae from eggs that are oviposited into and overwinter in soybean emerge in maize the following spring. Alternatively, eggs laid in maize fields hatch the fol- lowing year in a non-host field (i.e. soybean). Given that WCR larvae do not feed on soybean and eggs do not exhibit an extended diapause (Levine and Oloumi-Sadeghi, 1996; Levine et al., 2002), there is selection pressure in a landscape predominantly rotated between maize and soybean that favours WCR that lay eggs in soybeans. Both issues are important to managing this pest.
We created a set of simple meteorological and behavioural models to predict the spread of the beetle infesting soybeans throughout the north- central USA. Data collected from 1987 to 2001 in Illinois, Indiana, Michigan and Ohio identified geographical areas where WCR in soybean fields exceeded a detection threshold of 20 beetles per 100 sweeps and two beetles per yellow sticky trap per day. Counts above a detection threshold represent populations that lack fidelity to maize and are adapted to circumvent maize–soybean rotation. Maps of these observa- tions were used for evaluation of the model.
Figure 8.1 presents the observed infestations of soybean by WCR over time in Illinois, Indiana, Michigan and Ohio. The counties are shaded according to the year during which WCR adults were first observed in excess of the detection threshold in soybean (20 beetles per 100 sweeps or two beetles per trap per day). Since 1986, the rootworm has expanded its range in Illinois to the west, north and south. The southern, eastern and northern fronts in Indiana, Ohio and Michigan, respectively, did not change much after 1997. Analysis of the counties exceeding the higher detection threshold indicated that the rate of spread from 1986 to 1997 was approximately 27 km/year to the east and 8.5 km/year to the west.
From 1998 to 2001, the rate of spread slowed to approximately 16 km/year to the east and 7.75 km/year to the west. This indicates that some factor has limited the spread of the rootworm in these directions.
The models are based on wind speed and direction, the direction of storms with rainfall over 2.54 cm, knowledge of beetle flight speeds and the probability of a beetle flying long distances (Onstad et al., 1999, 2003a). We assumed that beetles can be carried a maximum of 33 km by storm fronts (Onstad et al., 1999, 2003a). Onstad et al. (1999) described the first dozen years of the geographical spread of rotation-resistant WCR.
156 D.W. Onstad et al.
Model results supported the hypothesis that the population of WCR infesting soybean originated in Ford County, Illinois. The predictions of the simple model fitted an independent set of observations well on three of four fronts or directions up to 1997. Some of the newer models invoked a landscape-diversity function that included the proportion of non-maize, non-rotated soybean vegetation on farmland in each county (Onstad et al., 2003a). An example of how the values of extra vegetation were used in the model is shown in Fig. 8.2. East-central Illinois and western Indiana have the lowest levels in the region. The proportions increase in Ohio, Michigan and north-eastern and southern Indiana. We assume that land- scape diversity increases as the proportion of this extra vegetation increases.
The best model for the period 1997 to 2001 reduces the distance that beetles spread each year by the proportion of extra vegetation in a county.
This version is superior to the model published in 1999 and to two new
Heterogeneous Landscapes and Variable Behaviour 157
Fig. 8.1. Counties in the north-central USA with soybeans infested by western corn root- worm at 20 beetles per 100 sweeps or 2.0 beetles per trap per day with initial year of obser- vation indicated by colour and shading.
1986 1996–1997
1992–1993 1998–1999
1994–1995 2000–2001
chap08.qxd 15/11/04 2:51 PM Page 157
models that do not consider landscape diversity but reduced the spread each year by a constant factor (0.85 or 0.80). Most of the models predicted spread at too high a rate between 1997 and 2001 compared to observa- tions, but a few new models with rates of spread reduced by a landscape- diversity function matched the observations relatively well.
Figure 8.3 compares the results of two models and the model pub- lished in 1999 with the observations. The dark contour lines represent the 12th (inner line for 1997) and 16th (outer line for 2001) years of the model simulations. A model that reduced the distance that beetles spread each year by the mean extra vegetation (MEV) per county performed relatively well by 2001 but failed to predict the earlier infestations in Ohio (Fig.
8.3a). The use of a threshold level of extra vegetation (30%), where there is no reduction in the distance a beetle can travel from counties with levels of extra vegetation below 30%, improved the 1997 predictions, especially in Ohio, but by 2001 the predicted wave front extended too far south-east, south-west and north into Wisconsin (Fig. 8.3b). All three of
158 D.W. Onstad et al.
Fig. 8.2. Percentage of extra vegetation on farmland in each county in the north-central USA.
< 30 50–70
30–50 > 70
these models outperformed the previously published 1999 model, which overpredicted the spread on the northern, western and eastern fronts by 2001 (Fig. 8.3c). The results of these models suggest that landscape diver- sity affects the spread of rotation-resistant individuals primarily to the north and east, while wind is a limiting factor to dispersal on the western and southern fronts.
Results suggest that the conclusions based on a linear model using proportion of extra vegetation as the key parameter are likely to be robust.
Thus, we hypothesize that, as the landscape diversity represented by the proportion of non-maize and non-rotated soybean vegetation in a geo- graphical region increases, the rate of regional spread of the rotation- resistant WCR decreases over several years.