Correlation Analysis of Canopy Reflectance and Growth of Winter Wheat in a Semi-Wet Rainfed Agriculture Area
半湿润雨养农业区冬小麦冠层反射光谱与长势的相关性分析

Crop growth characteristics such as leaf area index and coverage are critical to yield formation throughout the entire period of growth. Thus, crop growth monitoring is important for prediction of crop yield and evaluation of agricultural production. At present, ground-based remote sensing technology, with many advantages as it is timely, dynamic, economical and readily available, has become a main method for crop growth monitoring. By studying the relationship of spectral features and growth indices and constructing models between them, the technology can better be used to monitor crop growth conditions. However, due to differences in crop type and variety, environmental conditions, and available sensors, existing monitoring models can not be used across all regions, and regionally relevant models are required. This research studies the relationship between growth indices and canopy reflectance to construct relevant monitoring models and provide information on winter wheat production in semi-wet rainfed agricultural areas. The focus is the relationship between growth condition indices (coverage and leaf area index) and canopy reflectance of winter wheat during the main growth stages (regreening-heading) in Xifeng district of Qingyang City of Gansu, which is a typical semi-wet region. The result showed that winter wheat growth had a negative correlation with the spectral reflectance bands of 450nm, 550nm, 650nm, and 1650nm, and positive correlation with the 850nm band. The correlation was not significant before the erecting stage, and was constrained by crop growing stages and spectral characteristics. After the erecting stage, it was significant because the wheat was growing. To improve monitoring of crop growth, the correlations of eight normal vegetation indexes including NDVI, EVI, RVI, DVI, SAVI, MSAVI, RDVI, and NIR/G with coverage and leaf area index were analyzed. Just as with the single band reflectance, the correlation was not significant before the erecting stage but highly significant after erecting. The correlations were higher than with single band reflectance, which suggests that vegetation indices are better for monitoring winter wheat growth. Finally, linear and exponential monitoring models of winter wheat growth were constructed using the eight vegetation indexes. The correlation coefficients of models all exceeded 0.6. For coverage, all linear regression models performed better than the exponential models, but the opposite was true for leaf area index. Among all linear and exponential models, the fitting degree of the model using NDVI was best. Overall, for semi-wet rainfed agricultural regions, a linear regression model of NDVI is better for monitoring winter wheat coverage, while an exponential model of NDVI is better for monitoring winter wheat leaf area index.