Improvement of spatially and temporally continuous crop leaf area index by integration of CERES-Maize model and MODIS data

The spatially and temporally continuous leaf area index (LAI) mapping is very crucial for many agricultural applications, such as crop yield estimation and growth status monitoring. Data assimilation technology provides an innovational way to improve spatio-temporally continuous crop LAI estimation through integration of remotely sensed observations and crop growth models. In this study, a very fast simulated annealing (VFSA)-based variational assimilation scheme was proposed to integrate the crop growth model (CERES-Maize), MODIS reflectance product (MOD09A1) and a two-layer canopy reflectance model (ACRM) to estimate time-series crop LAI at regional scale. Simultaneously, a new sensitivity analysis method (called “histogram comparison”) was developed to identify sensitive parameters of CERES-Maize and ACRM models. The proposed scheme was applied for continuous crop LAI estimation during the maize growing season in the dominating spring maize planting area of Jilin province, China. Results showed that R2 values between LAI estimations from the proposed assimilation scheme (referred to as assimilated LAI) and fine resolution LAI reference maps were 0.24 and 0.63, with RMSE values of 0.21 and 0.54 for Julian day 176, 2010, and Julian day 196, 2010, respectively. The assimilated results were closer to LAI reference maps than the MODIS LAI product and ACRM-based inversion results (referred to as ACRM LAI). Moreover, by introducing the prior information of LAI dynamics depicted by a crop growth model, the assimilated LAI showed better temporal consistency than the MODIS LAI product, LAI profiles simulated by CERES-Maize model (referred to as CERES-Maize LAI), and ACRM LAI. It was found that the accuracies of LAI estimations could be enhanced by assimilating satellite observations into a crop simulation model in the VFSA framework at a regional scale.