Parameterizing Spatial Crop Models with Inverse Modeling: Sources of Error and Unexpected Results

摘要

The literature to date on inverse modeling-based parameterization of spatial ("distributed") crop models using yield data is dominated by one-dimensional uncoupled models, in which the landscape is divided into cells that do not interchange water. We studied three possible sources of error for uncoupled and spatially coupled distributed crop models: (1) error from biases (towards unrepresentatively wet or dry weather) in the sampling of years of yield data used for the parameterization process, (2) errors due to lack of knowledge of initial soil water conditions, and (3) error from lack of spatial coupling and water transport among different landscape locations. We show analytical evidence that the spatiotemporal infiltration behavior of a simple spatially coupled water balance model cannot be reproduced through a modification of the parameters of an uncoupled model. The corresponding yield prediction limitations of the uncoupled model are confirmed, using an example, both at the parameter estimation and the validation stages. In our example, however, parameter error due to weather biases and the error from lack of knowledge of initial conditions greatly impacted the predictive capability of the spatially coupled model, and had less effect on its uncoupled counterpart. This effect of biased weather has not been reported previously in the crop modeling IM literature. We conclude that the use of spatially coupled distributed crop models requires high-quality data. Practical precision agriculture applications are characterized by uncertain initial conditions and the possibility of biased weather. Under these circumstances, the use of a spatially coupled model may not be justified, especially for low landscape positions