Predicting winter wheat growth based on integrating remote sensing and crop growth modeling techniques
HUANG Yan,ZHU Yan,WANG Hang,YAO Xinfeng,CAO Weixing,David B．Hannaway,TIAN Yongchao
2011-01-01
Abstract:Historically, remote sensing (RS) and crop growth modeling have been used independently to monitor and predict crop growth. This paper reports a growth prediction technique for winter wheat (Tritium aestivum L.) based on the integration of ground-based and space-borne remote sensing data and a winter wheat growth model (WheatGrow). Leaf area index (LAI) and leaf nitrogen accumulation (LNA) of winter wheat were estimated using ASD field spectrometer, HJ-1 A/ B CCD, and Landsat-5 TM data and statistical remote sensing estimation models. This information was integrated with the WheatGrow model in three different growth stages (jointing, heading, and grain filling). Management parameters included sowing date, sowing rate, and nitrogen rate. Parameterization for regionalization of the integrated model was accomplished using the Shuffled Complex Evolution-University of Arizona (SCE-UA) optimization algorithm. This integrated technique was tested on independent datasets acquired from three winter wheat field tests in different years on different winter wheat varieties and at different treatments of nitrogen rates and sowing densities, and from data obtained from study areas in Haián and Rugao counties in Jiangsu Province (in central eastern China), both of which are main production areas of high-quality, low-gluten wheat in China. The results showed that LNA, one of the most sensitive parameters within WheatGrow, was better than LAI as an integrated parameter for crop model parameter initialization with the best integration period being the heading stage. RMSE values were 5. 32 days, 14.81 kg/hm 2 and 14.11 kg/hm 2 for sowing date, sowing rate, and nitrogen rate based on the ground spectral datasets, and 6. 55 days, 13.94 kg/hm 2 and 84.97 kg/hm 2 based on the space-borne satellite images. The crop model parameterization results were poorest when the grain filling stage was used as the integration period, probably because crop growth at earlier stages would be more influenced by agronomic management measures. In addition, predicted results well described the temporal and spatial distribution of winter wheat growth status and productivity in the study area, with relative error values of 0.13, 0.18 and 0.03 for LAI, LNA and grain yield based on the ground spectral datasets, and 0. 22, 0.23 and 0.06 based on the satellite images. The error may be due to the limited simulation ability of the WheatGrow model, or it may have been generated from the process of remote sensing information extraction and the statistical remote sensing estimating models, all of which need improvement. Nevertheless, the study has provided an important step toward more routine use of using remote sensing and crop modeling techniques together to improve our ability of regional monitoring and yield prediction of winter wheat.