Study On Rice Growth Monitoring And Yield Prediction Based On Assimilation Of SAR Data And Crop Growth Model

Rice is a staple food crop in china. With the world’s population growth and economic development, timely, accurate and dynamic rice growth monitoring and yield forecasting is significant to food security and agricultural sustainable development. Rice crop is mainly cultivated in warm tropical climates with plentiful rainfall and dense cloud cover throughout its growing season, and it is difficult to monitor based on optical remote sensing. Hence such research activities entail use of microwave remote sensing, for microwave can penetrate through clouds and have all-weather capabilities. This allows for a more reliable and consistent rice monitoring and yield prediction in terms of radar sensor data, while the crop growth model can reveal the internal mechanism of crop growth and yield formation and completely understand whole information of rice growth. The combination of rice growth model and radar remote sensing helps to improve the capacity of regional applications of the rice growth model, and to enhance the remote sensing monitoring and forecasting mechanistic, respectively.Relied on the National High Technology Research and Development Program, based on the measured data and Radarsat-2data of the study area in Suzhou Dongqiao, this paper using data assimilation to optimize the parameters of crop growth models to improve the accuracy of model simulations, providing a reference for rice growth monitoring and yield estimation.To begin with, the relationship between backscattering coefficient and biomass of rice has been analyzed, and then the rice biomass inversion based on regression model and the MIMICS mechanism model with backscattering coefficient in four polarization. The results show that the relations in quadratic fit equation of backscattering coefficient in the HH polarization are higher than other conditions, and the value of the coefficient of determination is0.77, and the accuracy evaluation results using the points data beyond modeling show that the average minimum residual coefficient of mechanistic model in HH polarization, which indicate that HH polarization backscattering coefficient is sensitive to the biomass, combining the MIMICS mechanism model can better characterize the rice biomass.After adjusting the WOFOST model by using field experiment data, based on assimilation of radar remote sensing and growth model coupled integration point for information in different growth stages of rice biomass, some parameters of growth model which is difficult to accurately obtain is obtained using particle swarm algorithm, and effective prediction of rice yield is achieved on this basis. The results show that the optimized model can simulate the development stage and organ biomass well in the study area, and reflect the change of these values with the rice growing period process. The simulated yield of rice in the study area turns to7982kg/ha, and the error with the measured values fell to22.8%from38.7%before the assimilation. Therefore, the scheme described in this paper is a promising techenique to apply multi-temporal and multi-polarization radar data and rice crop models for regional rice yield estimation, when no accurate information is available or optical data are hampered by heavy clouds during the rice season.