Analytical and Cartographic Predictive Modeling of Arable Land Productivity

Abstract

An inventory-based method is suggested to form spatial models of agricultural crop yields for extended regions (size: about 4° × 5°). The method is based on analyzing links between the long-term characteristics of crop yields and environmental variables, such as climate, topography, and soils. The response variable in multiple regressions was the maximal crop yield surplus by fertilization. This is the difference between the maximal crop yield obtained for an optimal dose of fertilizers and the control without fertilizers. This addition to yield appeared closely related to environmental variables, and on the other hand, it was relatively independent of previous crops. The environmental variables were climatic data on long-term annual means of temperature and precipitation for each month and certain periods, a digital elevation model and 18 basic topographic attributes from it, and soil type data. The topographic attributes were transformed nonlinearly to obtain normally distributed residuals. Multiple regression models included validation using an empirically founded criterion, tests for multicollinearity, and determining the statistical significance of each environmental variable. The method was tested based on long-term data on winter wheat yields for the western part of the Oka River basin. Analysis showed that the topography-generated microclimate is a major factor determining the maximum addition to the yield. The microclimate variable is the relative slope insolation, which is one of the topographic attributes that takes into account the slope steepness and exposure, and 2 angles that describe an effective position for the Sun. Relative insolation characterizes the relative power flux of solar radiation on the land surface and allows for statistical comparisons at different azimuths and angles of declination of the Sun above the horizon. On the most heated southwestern slopes of the basin, the largest increase in the maximum addition is observed for winter wheat. The addition also increases during the relatively high precipitation in February, which is the windiest cold winter month in this region. An increase in annual precipitation in the humid climate of the Oka basin leads to a decrease in the addition. Introducing a nonquantitative or indicator variable, soil types, into the multiple regression model results in increasing the strength of the relationships and shows that the maximum addition to the sod-podzolic soils in the northern region depends more strongly on climate than it does for Gray Forest and Chernozem soils in the southern part of the region. The model validated using the criterion introduced explained 74 % of spatial variability in the addition to winter wheat by means of the environmental variables used. Based on the model, a gridded map was constructed for the entire region (4° × 5°). The results of the analysis indicate that the topography, along with environmental factors, may have the largest influence and should also be taken into account when making crop yield prognoses in the conditions of a changing climate.