Submitted to: Agricultural Systems
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 5/8/2000
Publication Date: N/A
Citation: Interpretive Summary: Recent expansion of precision agriculture technology means that many farmers, extension agents, service providers, and researchers will have to re-think recommendations for fertilizer and other management techniques. The problem is that field experiments to provide this information would take many years. Consequently, researchers and others hoped that computer models of crop growth and yield could answer the questions now arising. However, models were developed to predict average yields for clearly different cultivars, soils, and weather. Now, they are being used to simulate point yields for one cultivar with soils and weather that are only slightly different within a field. It is not clear that current models can perform so far in excess of their original design. In this paper, we report certain aspects of the CERES-Maize model that have potential to be adapted to site-specific farming conditions and report other aspects that lack the sensitivity to perform well for this purpose. Some enhancements that could improve the suitability of the model are accounting for summer storm runoff, for within-field redistribution or rain that includes runoff, and for water stress effects on canopy temperature. Other enhancements that improve spatial data handling in the model would further improve its application to site-specific farming.
Technical Abstract: When site-specific farming became technologically feasible, existing crop models made computer simulation a natural choice for predicting yield under various combinations of soil, weather, and management. However, modeling for site-specific farming may require both greater accuracy and sensitivity to more parameters than current models allow. The objective for research reported in this paper was to evaluate the DSSAT V3.1 and V3.5 corn models for sensitivity to parameters important to site-specific farming. Compared with V3.1, V3.5 simulated a consistent 1000+ kg/ha reduction in yield for these conditions. The models were unexpectedly insensitive to inputs for soil type, depth to clay, nitrogen, and plant population, suggesting areas for attention. Although they were appropriately sensitive to rainfall, indicating sensitivity to soil water content is generally correct, there are known problems with the curve number procedure that calculates runoff. The runoff routine needs improvement, and a separate routine may be needed to accommodate within-field redistribution of runoff. The models also responded to maximum air temperature, but since crop temperature varies more than air temperature, perhaps crop temperature should be calculated from air temperature and water stress. Model accuracy issues aside, accommodating spatial inputs and model runs requires enhanced interfaces. These and other suggested enhancements to the current version of this model would improve its applicability for site-specific agriculture.