Crop–Climate Modeling Using Spatial Patterns of Yield and Climate. Part 1: Background and an Example from Australia

Abstract

A new technique in statistical crop-climate analysis, the direct linking of spatial patterns of crop yield and spatial patterns of climate, is explored. Yield and climate data from networks of crop reporting districts and meteorological stations are decomposed into orthogonal components using principal components analysis. Each yield component is then expressed as a function of the climate components using multiple regression. These regression equations are then combined to give an equation which relates interannual variations in the spatial patterns of yield to interannual variations in the spatial patterns of selected climate variables. The method is illustrated using wheat yield data from 59 crop reporting districts in southwestern Western Australia covering the period 1929?75. The regression models are calibrated using data for the period 1929?65 and the results are verified using data for the period 1966?75. The climate contribution is shown to be highly significant, with winter precipitation being the most important variable. A single equation relating yield and climate patterns correctly reproduces the differing results obtained for separate parts of the study area by earlier workers. The influence of winter and autumn precipitation is nonlinear and, as a consequence, the study area divides into three zones: a high rainfall area where rainfall is generally more than optimum so that lower rainfall gives higher yields; a low rainfall area where rainfall is, on average, less than optimum so that positive rainfall anomalies are associated with higher yields; and an intermediate zone where average rainfall is close to optimum so that anomalies in either direction tend to suppress yields. Our analysis shows no evidence for any significant change in the sensitivity of wheat yields to climate in spite of a complete change in the variety of wheat cultivated.