Portal de Periódicos da CAPES

Data assimilation of MODIS and TM observations into CERES-Maize model to estimate regional maize yield

2010; SPIE; Volume: 7809; Linguagem: Inglês

10.1117/12.860315

1996-756X

Huaan Jin, Jindi Wang, Yanchen Bo, Guifen Chen, Huazhu Xue,

Land Use and Ecosystem Services

Accurate and real-time estimation of crop yield over large areas is critical for many applications such as crop management, and agricultural management decision-making. This study presents a scheme to assimilate multi-temporal MODIS and Landsat TM reflectance data into the CERES-Maize crop growth model which is coupled with the radiative transfer model SAIL for maize yield estimation. We extract the directional reflectance data of MODIS subpixels corresponding to pure maize conditions with the objective to increase time series observations at the TM scale. The variables to be assimilated were chosen by conducting the sensitivity analysis on the coupled model. The SCE-UA algorithm was applied to determine the optimal set of these sensitive variables. Finally the maize yields maps were produced at TM scale with the coupled assimilation model. The proposed scheme was applied over Yushu County located in Jilin province of Northeast China and validated by using field yield measurement dataset during the maize growing season in 2007. The measurement data include the species of planting maize, soil type and fertility, field observed leaf, canopy and soil reflectance data etc. Furthermore, yield data were gained in specially designed experimental campaigns. The validation results indicate that the yield estimation scheme using multiple remote sensing data assimilation is very promising. The accuracy of TM yield map produced by adding time series MODIS subpixel information was improved comparing with that only using TM data.