Study On Monitoring And Predicting Winter Wheat Quality Using Hyper-spectral Remote Sensing Data And Crop Growth Model

In order to solve the problem of lacking high quality wheat in China, it is the only way to develop special wheat with stable grain protein content. The advent of crop growth model a nd remote sensing technology offers a new way to efficiently monitor and predict crop qualit y. Crop growth model has the properties of mechanism and time expansibility, while remote sensing data have the properties of spatiality and real-time. The coupling of them provides a new method to improve crop quality monitoring and predicting accuracy.The study site is located in Yangling district and its surrounding areas of Shanxi, China. The study object is winter wheat grown during 2012-2014. The DSSAT as selected crop gro wth model is combined with ground collected ASD data and PSO algorithm to study on mon itoring and predicting winter wheat GPC. First, the localization of DSSAT crop model param eters was conducted with the field experimental data collected during 2013-2014, and the fie ld experimental data collected during 2012-2013 were used for validation. Then, the ground hyper-spectral data collected during the whole growth period were used to retrieve biomass t o improve the retrieval accuracy with only one growth stage. Canopy nitrogen accumulation content was obtained after multiplying biomass by plant nitrogen content. It is an assimilatio n variable to optimize the crop simulation value and remote sensing retrieval value. The expe rimental results showed that the simulation error values of DSSAT during 2012-2013 and 2013-2014 with the whole growth period were less than 1 day. The error value of each growth stage was less than 2 days. The RMSE values between the simulated and measured biomass during 2013-2014 and 2012-2013 were 0.996 kg ha-1 and 1.678 kg ha-1, respectively. The a bsolute error values between the simulated and measured yield during 2013-2014 and 2012-2013 were-59 kg ha-1 and-26 kg ha-1. The absolute error values between the simulated and measured GPC during 2013-2014 and 2012-2013 were-1.2% and-1.55%. It demonstrates th at the DSSAT model has certain applicability in Yangling district after localization. Finally, t he optimization of the six single points in field environment was conducted by assimilation a lgorithm. The RMSE values of 6 points between the simulated and measured GPC decreasedfrom 2.39% before assimilation to 2.26% after assimilation. It shows that the coupling of re mote sensing data and crop growth model by assimilation algorithm can improve the retrieva l accuracy of winter wheat GPC.