Skip to main content
ARS Home » Southeast Area » Stoneville, Mississippi » Crop Production Systems Research » Research » Publications at this Location » Publication #311219
Title: Improving winter wheat yield estimation by assimilation of the leaf area index from Landsat TM and MODIS data into the WOFOST model

Author
item HUANG, JIANXI - China Agricultural University
item TIAN, LIYAN - China Agricultural University
item LIANG, SHULIN - University Of Maryland
item BECKER-RESHEF, INBAL - University Of Maryland
item Huang, Yanbo
item SU, WEI - China Agricultural University
item FAN, JINLONG - Chinese Academy Of Sciences
item WU, WENBIN - Chinese Academy Of Sciences

Submitted to: Agricultural and Forest Meteorology
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 2/1/2015
Publication Date: 2/17/2015
Citation: Huang, J., Tian, L., Liang, S., Becker-Reshef, I., Huang, Y., Su, W., Fan, J., Wu, W. 2015. Improving winter wheat yield estimation by assimilation of the leaf area index from Landsat TM and MODIS data into the WOFOST model. Agricultural and Forest Meteorology. 204:106-221.

Interpretive Summary: In order to estimate crop yield precisely, a proven scheme is to incorporate remotely sensed data into a crop growth model. Scientists from China Agricultural University, USDA-ARS, Crop Production Research Unit, Stoneville, Mississippi, and Chinese Academy of Agricultural Science developed a framework for winter wheat yield prediction at regional scale. The framework assimilates leaf area index (LAI) values derived from MODIS (Moderate-Resolution Imaging Spectroradiometer) satellite LAI products into the WOFOST crop growth model with field-measured LAI. The WOFOST is particularly suitable for large-scale and regional study. The results indicated that the scale adjustment between the field measurement, crop model and the satellite data substantially improved the accuracy of wheat yield prediction at regional-scale. The framework provided by the study can improve estimation of regional crop yield by integrating spaceborne remotely sensed data, ground-measured data, and a crop growth model.

Technical Abstract: To predict regional-scale winter wheat yield, a framework was developed to assimilate leaf area index (LAI) values derived from MODIS (Moderate-Resolution Imaging Spectroradiometer) LAI remote sensing products into the WOFOST crop growth model. LAIs were measured in field during seven main phenological phases of winter wheat within 53 sample plots in two agricultural districts in Hebei Province, China. To eliminate cloud contamination, the Savitzky-Golay (S-G) filtering algorithm was applied to the MODIS LAI products to obtain filtered LAIs. Regression models were built between field-measured LAI and Landsat TM vegetation indices and derived multi-temporal TM LAIs. A nonlinear method was developed to obtain an adjusted LAI that accounted for the scale mismatch between the observations and crop models. These three LAI datasets were assimilated into the WOFOST model to evaluate the accuracy of crop yield prediction. A cost function was constructed using four-dimensional variational data assimilation (4DVAR) to account for the model errors during the critical phenological stages. The SCE-UA algorithm was used to minimize the cost function between the time series of remotely sensed LAI and the modeled LAI. The importance of LAI in each phenological stage was also evaluated. Finally, winter wheat yield was simulated in a 1-km grid for cells with at least 50% of the area occupied by winter wheat using WOFOST with the two optimized parameters, and aggregated to the county-level results at the regional scale. The scale adjustment substantially improved the accuracy of regional-scale wheat yield prediction. The scale-adjusted LAI achieved better predictions (R2 = 0.47; RMSE = 151.92 kg ha-1) than without assimilation (R2 = 0.23; RMSE = 373.6 kg ha-1), than with assimilated TM LAI (R2 = 0.23; RMSE = 191.6 kg ha-1), and than with S-G filtered MODIS LAI (R2 = 0.29; RMSE = 730.8 kg ha-1). This research provides a scheme to employ remotely sensed data, ground-measured data, and a crop growth model to estimate regional crop yield.