A Preliminary Study On Yield Prediction Of Soybean Based On Active Remote Sensing NDVI/RVI Analysis With "GreenSeeker"

Soybean is one of the most important crops in the world. It plays an important role in the food crops constitute and oil crops structure. Owing to the low level of soybean yield in our country,to increase soybean yield has been the most important goal of all the breeders. Soybean yield performance is affected by genetic factors, the natural ecological conditions and cultural practices and so on. Measurement of growth traits and yield in large-scale breeding of soybeans is urgently needed to be resolved. Remote sensing is the one of the measurement about growth traits and yield of soybean. Therefore, this study conducted a series of experiments of the soybean in2011-2012. Using active remote sensor GreenSeeker at soybean critical growth stages,this study obtained the canopy NDVI and RVI of Huang-Huai region representative breeding materials and RIL(NJRIKY), analysed the variation of the soybean canopy reflectance spectra,and determined the best sensitive growth stage to predict soybean yield. Moreover, the study established the yield prediction model based on NDVI and RVI.The main results are as following:1. In this study, we choosed the Huang-Huai region representative breeding materials and RIL(NJRIKY). There was highly significant difference (P<0.01) in yield of soybean materials in this research. And it was suitable for building yield prediction model.2. This study obtained the canopy NDVI and RVI at seeding,flowering,podding and filling stage.And the soybean canopy NDVI and RVI presented low-high-low trend along with the growth stages. Soybean plants grew well at flowering stage. With the soybean biomass and canopy density increased, the canopy reflectance of soybean increased sharply during near infrared waveband. And the canopy reflectance of soybean decreased during red waveband. Therefore, the value of the soybean canopy NDVI and RVI value was maintained at high level at flowering stage. Soybean population structure was stable during soybean flowering to podding stage. So NDVI and RVI was also maintained at high level and changed little. Soybean canopy NDVI and RVI changed a lot at seedling and filling stage. Owing to seedling soil background, the canopy vegetation index can not accurately reflect the growth of the material.However, the soybean canopy density decreased and plant coverage reached25%-80%at filling stage.So the canopy vegetation index was very sensitive at filling stage.3. This study used NDVI and RVI to predict soybean yield. And remote sensing prediction model had been found based on NDVI and RVI at different growth stages,and most of all were linear models,others were exponential models. Taking into account the different growth characteristics of soybean, the model was closely related to the growth stage. This study took NDVI and RVI as independent variable to predict soybean yield. This research determined that the filling stage can predict soybean yield better.And its estimation accuracy was higher, also SE was smaller. The filling stage was very critical to predict soybean yield. Research shows that the canopy NDVI and RVI at filling stage can be used to predict soybean yield better.This study took NDVI and RVI at different stage to establish multiple regression model.The model based on theer stages(flowering,podding and filling)was better than two stages (podding and filling).And it was better to predict soybean yield using NDVI than RVI.And this study choosed the best soybean prediction model, for soybean material of the Huang-Huai region, and it (y=e6.9-4.1x1+4.3x2+1.4x3,R2=0.66,SE=416.77) can predict the soybean yield749.96-4456.30kg hm-2,and the variation of growth stage range variation was93.0-117.0d. This study validate the best model using NJRIKY. The coefficient of determination between actual and predicted soybean production is0.59.So through model validation, it can be used to predict soybean yield.4. SLW is closely associated with photosynthesis.And it can be used to predict soybean yield. So this study also took NDVI with SLW and RVI with SLW to predict soybean yield. This research discovery that it(y=946.6+199.3RVI+15.4SLW,R2=0.65, SE=208.73) is better to predict soybean yield to use vegetation index with SLW.The main purpose of this study is to select NDVI and RVI to establish the optimal regression model to simplify the identification of soybean yield. And the research discovery that it is feasible to predict soybean yield based on NDVI and RVI.