Study On Jujube Yield Estimation Based On DNDC Model And Remote Sensing Data Assimilation

As a digital model which can describe the energy and material state of crop growth process mechanically,crop model has the problems of difficult to obtain input data and to use it regionally.Remote sensing information can obtain continuous data in time and space,but can only reflect the instantaneous canopy surface information of crops.Therefore,the coupling of the two can realize the complementary advantages of the two,and improve the accuracy of crop model production prediction.In this study,Zaoyuan,a part of the area under the jurisdiction of 224 regiment of the 14 th division of production and Construction Corps of Kunyu City,is taken as the research area.According to the jujube trees in 2015-2019,the Lai and DNDC(denitrification decision)retrieved by remote sensing are used to study the jujube trees in 2015-2019,The yield of jujube was simulated by the leaf area index(LAI)simulated by DNDC model.The specific research work and results are as follows:(1)Sentry-2 remote sensing data is used for data preprocessing,and the maximum likelihood method is used to classify the images.The vegetation index of jujube tree is calculated based on the classification results.The relationship between vegetation index and measured leaf area index is determined by statistical regression method,and then leaf area index is retrieved.By comparing the inversion Lai with the measured Lai,the determination coefficient and root mean square error show that the inversion accuracy is higher.(2)The sensitivity characteristics of input parameters and the uncertainty of predicted yield of red date model based on DNDC model are analyzed by EFAST and Monte Carlo method.The results showed that the sensitivity of the indicators such as the proportion of fruit per plant biomass and the maximum crop yield of the whole plant was the highest,the sensitivity of the indexes such as water holding rate and porosity in the soil parameters was the highest,and the sensitivity of irrigation and fertilization in the field management parameters was the highest;according to the simulation in 2018,the fluctuation range of the parameters increased from ±5% to ± 10%,and the predicted yield was predicted The correlation consistency coefficient of normal distribution increases,and the model stability increases.The results of analysis are used to adjust the parameter optimization model and simulate the output in 2015-2019.The relative error of the predicted output results is controlled within ± 8%,which improves the accuracy of the model prediction yield,which shows that the parameters adjustment and optimization of the model tend to be reasonable.(3)Taking leaf area index LAI as assimilation parameter,the data assimilation of Lai and Lai of model simulated leaf area index is carried out by using simulated annealing algorithm.The optimal assimilation result is determined by objective function.The sensitive parameters of the model are adjusted continuously to obtain the new model simulation leaf area index.The assimilation algorithm is used to iterate continuously until the minimum target function is satisfied,and the data is confirmed The optimal Lai and the optimal input parameters are determined.The results of the model production simulation in 2015-2020 were simulated by using the optimized optimal parameters,and the results of the final yield simulation were significantly improved compared with those before assimilation.It is shown that the coupling of remote sensing information with crop model by data assimilation algorithm can improve the accuracy of model simulation and prediction of jujube yield.