DERIVING REGRESSION EQUATIONS (META-MODELS) FROM DETERMINISTIC SIMULATION MODELING FOR CROP FERTILIZATION

摘要

Operating complex mathematical models, describing the plant-water-soil environment, is difficult and generally only done by experts. Complex data handling and parameterization of those models hamper the use by less skilled people and are a serious drawback on using information from deterministic simulation modeling in decision support systems for precision agriculture. In this study regression or meta-models for nitrogen fertilizer application in Winter Wheat on the Van Bergeijk farm in the Netherlands were derived from long-term simulation modeling. These meta-models only need simple, easily obtainable, weather data and basic soil information, such as soil organic matter content, to predict accurate timing and amount of spatial variable nitrogen applications. In a validation study, timing and amount of the various nitrogen applications in 1999 on a 15 ha field on the Van Bergeijk farm was always within 10 days and in almost all cases within 7 days compared to the recommendation of the decision support system. Predicted amounts of an individual application were accurate within 5-10 kg ha-1 compared to the decision support system. This study illustrates that results of complex deterministic simulation models can be used to derive simple regressionbased meta-models that can be used by farmers in simple straightforward calculators such as spreadsheets or pocket calculators.