Crop, semi-natural, and water features of the cotton agroecosystem as indicators of risk of infestation of two plant bug (Hemiptera: Miridae) pests

Introduction: This study considers concepts and tools of landscape ecology and geographic information systems (GIS) to prioritize insect monitoring in large-scale crops, using the cotton agroecosystem of the Texas Gulf Coast and two plant bug species (Creontiades signatus Distant and Pseudatomoscelis seriatus (Reuter) [Hemiptera: Miridae]) as a case study. The two species differed in host plants and time span as cotton pests.

Methods: C. signatus and P. seriatus abundance in early growth of cotton were regressed on landscape metrics. Comparisons of three approaches to select landscape variables in stepwise multiple regressions were made across spatial scales and two weeks of insect data extracted from monitoring of 21 cotton fields, years 2010 through 2013.

Results and discussion: The spatial variation of plant bug abundance and the landscape features were substantial, aiding the regression approach. For full stepwise regression models using 18 landscape variables, regression model fit using C. signatus data was modestly better in week one of sampling when C. signatus adults and young nymphs were detected (R ² range of 0.56 to 0.82), as compared with model fit at week two (R ² range of 0.49 to 0.77). The smallest scale (2.5 km radius) models had the greatest number of variables selected and highest R ², while two broader scales (5 and 10 km) and truncating the models to three variables produced a narrower range of R ²s (0.49 to 0.62) and more consistent entry of variables. Wetland composition had a consistent positive association with C. signatus abundance, supporting its association with seepweeds which are common in coastal wetlands. When selected, the composition of cotton and grassland/shrubland/pasture also had a positive association with C. signatus abundance. Aggregation metrics were also relevant, but composition metrics in the models were arguably more easily utilized in prioritizing insect monitoring. In contrast, there were few significant regressions using P. seriatus data, possibly due to the widespread distribution of its weedy host plants and lower abundance. Overall, selected landscape features served as indicators of C. signatus infestation potential in cotton particularly grown near coastal wetlands, but landscape features were not useful for P. seriatus infestation potential in cotton.