Improving Crop Model Simulations by Assimilating Leaf Area Index Estimated From Canopy Spectral Reflectance.

Great efforts have been taken to programmatically synthesize current agronomic knowledge into computer simulation models.  However, the applicability of these models has been limited, because many input parameters must be specified and model calibration against field measurements is often necessary to insure adequate model performance.  Remote sensing has been proposed as a relatively quick, easy, and inexpensive source of information that relates well with key model state variables, such as green leaf area index (GLAI).  Our objectives were to 1) develop a strategy for assimilating remote sensing estimates of GLAI into the CERES-Wheat crop model and to 2) evaluate the ability of the assimilation strategy to improve simulations of crop yield and evapotranspiration using data from a field study in central Arizona.  Preliminary results suggest that the data assimilation strategy can improve wheat yield simulations by 100-200 kg ha^-1 and evapotranspiration by 10-20 mm.  The data assimilation strategies were shown to be effective under moderate model parameter uncertainty.