Predicting Grain Protein Content of Winter Wheat Based on Landsat TM Images and Leaf Nitrogen Content

摘要

The purpose of this study is to further improve the accuracy of predicting winter wheat grain quality with remote sensing, to enhance the prediction mechanism and to meet the demand for winter wheat production. In order to predict grain protein content (GPC) in winter wheat using landsat TM images, The experiment was carried out in Jiangsu regions during 2007-2009 winter wheat growth season. Based on Landsat TM image, synchronous or quasi-simultaneous ground observations of leaf nitrogen content (LNC) and grain quality indices of winter wheat under different years and different periods. Firstly, this study analysised the relationships between GPC and LNC, and between LNC and satellite remote sensing variables. Secondly, based on remote sensing variable and LNC, the quantitative relationship models were established to predict GPC of winter wheat, and then evaluated with independent samples. Finally, the indirect model of predicting GPC based on remote sensing variable and LNC was compared to the direct moedel based on only NDVI. The results showed that: anthesis stage can be considered as the sensitive period to predict grain protein content, and it was sensitive to predict GPC of winter wheat using NDVI. the indirect and direct moedels were evaluated with independent samples by the determination coefficient(R~2) with 0.412 and 0.379, the root mean square error (RMSE) with 0.367% and 0.418%, respectively. The indirect model based on NDVI and LNC performed better to predict wheat GPC than the direct model based on only NDVI, and obtained the higher accuracy by 8.5% than the direct model. The result of appling the indirect model was correspondent with the actual distribution of wheat GPC. It is concluded that the research can provide an effective way to improve the accuracy of predicting wheat quality based on aerospace remote sensing, and contribute to large-scale application and promotion of the research results.