Retrieval Of Canopy Chlorophyll Density And Yield Prediction For Winter Wheat Based On Remote Sensing

Crop yield prediction allows the government to know the crop production and help themto make decisions to import or export food in time. Chlorophyll is the medium forphotosynthesis, the variation in chlorophyll content influences the variability inphotosynthetic rate, as a result, the variation in chlorophyll content influences the crop yield.Monitoring of the chlorophyll content can provide the basis for crop growth and crop yieldprediction. Vegetation Index which is sensitive to the chlorophyll content can be a useful toolfor retrieving the canopy chlorophyll density.In researches on retrieval of canopy chlorophyll content, the unit “mg/g” is popularlyemployed. However, canopy chlorophyll density （g/cm2） has more direct relationship withremote sensed spectral characteristics, and it is also an important monitoring index during theprocess of crop yield formation. But to date, evaluating the universality of remote sensingmodel for retrieving canopy chlorophyll density using different growth period data andbetween different crop types are rarely reported. For crop yield prediction, most of the studiespredicted crop yield based on parameters related to crop growth parameters, such as leaf areaindex and biomass. The model for crop yield prediction based on chlorophyll content is rarelyreported.To develop a universal applicable statistical model for retrieving canopy chlorophylldensity, we analyzed the correlations between hyperspectral vegetation indices and canopychlorophyll density, and established the statistical models for canopy chlorophyll density. Theprecision and universality of the statistical models for canopy chlorophyll density wereexamined by the dataset of winter wheat and summer maize at different growth stages. At last,we found MERIS Terrestrial Chlorophyll Index （MTCI） is the most sensitive index to canopychlorophyll density and got the finest statistical model based on MTCI. Based on this, wecontinued to investigate the potential of MTCI in crop yield prediction, and also to compareits performance to NDVI. The main work and conclusions are listed as follow.（1） We studied the mechanism of radiation transfer in vegetation and analyzed thespectral bands and vegetation indices sensitive to the chlorophyll content. Then, we selectedSRI, RVI I, VIopt, RVI II, SR705, R-M, MSR705and MTCI, which are sensitive to thechlorophyll content, to establish the statistical models for canopy chlorophyll density.（2） The precision and universality of the statistical models for canopy chlorophylldensity based on the8selected vegetation indices were examined by the dataset of winterwheat and summer maize at different growth stages. And the statistical model based on MTCIgives the best validation precision and universality between two crops at different growthstages.（3） Based on the conclusions above and related literatures，we found that MTCI was anvegetation index sensitive to chlorophyll content, while NDVI was an vegetation indexsensitive to LAI. In this study, we investigated the potential of MTCI for crop yield prediction.Firstly, the MTCI and the NDVI products with a temporal resolution of a half month werecalculated from the daily level-2reduced resolution ENVISAT MERIS reflectance product（MER_RR₂P） using a Maximum Value Composite （MVC） algorithm. Then, the accuracyand universality of MTCI for winter wheat yield prediction was examined and compared tothe NDVI in Henan Province, China, from2003to2011. The results show that:1) thecorrelation coefficient between yield and MTCI are significantly higher than that of NDVIexcept for milking stages, and the accumulated vegetation index through the reviving stage tomilking is more significantly correlated to crop yield, with a determination coefficient of0.892for MTCI and0.629for NDVI;2) the optimum phase for crop yield prediction based onMTCI is the heading stage, about30days earlier than that of NDVI （milking stage）, which isanother advantage of MTCI in crop yield prediction;3) the validation error of the crop yieldmodel based on the accumulated MTCI is half of that based on the accumulated NDVI, whichindicates a more universal application of MTCI. The study indicates that MTCI is a potentialvegetation index for crop yield prediction, with higher accuracy and better universalitycompared to NDVI.