Thursday, November 16, 2006 - 9:00 AM

Use of Aggregated Environmental Data to Predict Spatial Variation of Crop Yields Across a Landscape.

Carol Williams1, Matt Liebman1, Jode Edwards2, David E. James3, Daryl Herzmann4, Jeremy Singer3, and Ray Arritt5. (1) Agronomy Department, Iowa State Univ, 1123 Roosevelt Ave., Ames, IA 50010, (2) USDA, 1505 Agronomy Hall, Ames, IA 50011, (3) USDA-ARS, National Soil Tilth Laboratory, 2150 Pammel Dr, Ames, IA 50011, (4) Univ.of Iowa, 1123 Roosevelt Ave., Ames, IA 50010, (5) Iowa State Univ., 1123 Roosevelt Ave., Ames, IA 50010

Crop yield variability is effected by environmental heterogeneity at various scales. “Scaling-up” of locally-derived process-based models has not been universally successful in accurately modeling patterns of yield association with environmental characteristics at broader spatial scales. Use of remotely-sensed spatially-referenced environmental data and geographic information systems (GIS), combined with autoregression statistical techniques offers an opportunity to meet the challenge of predicting broader-scale patterns of yield variability in relation to crop-relevant environmental parameters. Such information could aid in improved strategic agro-ecological and economic decision-making beyond the farm enterprise. The aim of this study was to model spatial distribution of crop yields and their interannual variability among counties in Iowa, U.S.A, based on the hypothesis that mean county-level environmental characteristics were predictors of mean county-level, long-term crop yields. We used a raster GIS to derive values for a limited set of environmental predictors which were then used to predict yield using autoregression. Predictors included county-level means and standard deviations of climatic, edaphic and topographic environmental attributes. Both predictor types were significant in predicting yields of corn (Zea mays), soybean (Glycine max), alfalfa (Medicago sativa), and oat (Avena sativa).