Real-Time Monitoring and Forecasting of Crop Production by Combining Remotely Sensed Data and Mechanistic Crop Models.

Real-time crop assessments over large land areas are desired by many public and private entities. While there have been a number of attempts to use remotely sensed spectral signatures to evaluate crop status, these data are not able to gauge physiological activity that ultimately determines crop yield. An alternative is to use mechanistic crop models that track crop physiological activity, but even the simpler models require critical input data to initiate the simulations. Dr. Paul Doraiswamy pioneered an effort to combine the power of each approach to do assessments of crop production over large land areas. A relatively simple crop model was selected that could simulate crop development, growth, and yield with a comparatively small amount of initialization input. The remote sensing data was used to identify the specific crop species and the emergence date in each individual field across the landscape. In addition, the estimates of crop leaf area from the remotely sensed data allowed real-time confirmation of the capacity of the models to simulate this key aspect of crop development. Adjustments in model parameters such as assumed plant population density or soil depth may be required if the discrepancies become large. Combining yield predictions on a field based with the land area of each crop resulted in a powerful tool to assess crop production over relatively large land areas.