Spatial and multivariate analyses of Colorado rangeland grasshopper abundances: pattern and process

Abstract

Grasshoppers are important agricultural pests which inhabit a broad variety of crop and rangeland ecosystems. Despite more than a century of research, little is known of the factors which control the spatio-temporal dynamics of rangeland grasshopper densities. This study (1) reviews the history of grasshopper problems in the western United States, outlines the efforts which have been made to control and understand grasshopper population changes, and discusses the complexities which preclude our best efforts to predict and control rangeland grasshopper outbreaks; (2) identifies and characterizes spatial patterns in Colorado grasshopper abundances; and (3) determines the environmental factors which are most closely linked with changes in grasshopper population densities. Exploratory data analysis and geostatistical techniques were used to identify and characterize spatial dependence among grasshopper counts for 1993 through 1997. Variograms were computed for each year and used with kriging to interpolate grasshopper densities in unsampled areas. Results are compared with previous spatial analyses of grasshoppers in Montana, Idaho, and Alberta. Grasshopper densities showed spatial patterning at scales larger than those reported for other areas of the Great Plains. The results indicate that the current grasshopper survey techniques may be more useful in detecting large-scale (regional) infestations than in identifying localized areas which may require control. To determine the environmental factors which are most closely linked with changes in grasshopper population densities, grasshopper sampling sites were grouped by direction and magnitude of change in density from 1995 to 1996. Multiple regression and stepwise discriminant analysis were used to identify the soil and weather variables which best distinguished among groups. Grasshopper density in 1995 was the most important variable in the discriminant analysis, followed by mean precipitation for April, 1993, and mean temperature in October, 1993. These results indicate that soil characteristics in Colorado are not as important to changes in density as are population level and stochastic weather events and confirm the presence of time lags in the effects of weather conditions on population changes.