Crop growth simulation models have extensive application in crop growth monitoring, yield forecasting and utilization of enviromental resources. However, problems arise when scaling crop simulation models from the field to regional scales, especially in acquiring initial conditions and parameters for the model. Fortunately, satellite remote sensing has the potential to improve some of the model parameterization for monitoring crop growth at regional scales. Thus, there is interest in developing an approach and methodology for incorporating remotely-sensed information with crop growth simulation models.
In this paper, the crop model WOFOST (World food study) was adapted for simulating growth of winter wheat by using field experimental data from North China, and the radiative transfer model SAIL-PROSPECT was adapted by adjusting the sensitivity of its parameters. The two models were then coupled using LAI to simulate the vegetation indices SAVI. Finally, WOFOST was re-initialed by minimizing differences between SAVI_s simulated by coupling the model and SAVI_m synthesized from MODIS remote sensed data using an optimization software program (FSEOPT). The results were validated by using field experimental data (including leaf area index, dry weight of leaves, stems and storage organs) in Zhengzhou, Henan Province, Tai'an, Shandong Province, and Gucheng, Hebei Province, and some MODIS data during the growing season of winter wheat from 2001 to 2003.
The main results in this study were as follows: (1) Differences between the simulated mature date, after re-initializing the emergence date using remote sensing data and simulated values, with the actual ermergence date was within 2 days, and simulated LAI and gross above-ground dry matter weight were 3-8 percent of actual values; (2) By re-initializing biomass weight in the turn-green stage, relative errors of the simulated LAI and gross above-ground dry matter weight were within 16% at key development stages, and simulated LAI and storage organ weight were closer to measured values during the entire growing period; (3) Appropriate remote sensing data during the period from the turn-green stage to earing stage was more critical for improving crop modeling when biomass in the turn-green stage was adjusted.
We presented a novel method for validating and adjusting crop models to regional scales. Optimization of the crop simulation model by dynamical adjustment of the initial variables and parameters based on remote sensing data produced highly satisfactory results. This research provides a basis for optimizing crop models by using remotely sensed data at regional scales. However, errors in the simulation results due to uncertainty of remote sensing data and SAIL-PROSPECT parameters still exist and further study is needed.