Regional Potato Production Estimation Based On Data Assimilation Of Remote Sensing Information And DSSAT-SUBSTOR Model

Potatoes are the third largest food crop in the world.Since 2015,the Ministry of Agriculture of China proposed the"Staple Food"strategy for potatoes,potatoes will beome the fourth most important food crop in China.So it is of great practical significance to estimate the production of potatoes.Improving the estimation accuracy of crop production by combining remote sensing data with crop growth model is a research hotspot in recent years.This paper selected Xinglong Town and Jijia Town of Jiutai District,Changchun City,Jilin Province as the research area,and the potato as the research object.On the basis of collecting the canopy reflectance spectrum and leaf area index(LAI)of the potato,the potato planting area was extracted by GF-1 image,and the remote sensing retrieval of LAI was completed by statistical regression method.The EFAST global sensitivity analysis method was used to analyze the sensitivity of the input parameters of the DSSAT-SUBSTOR crop growth model,and parameters sensitive to the potato production and LAI would be used as parameters to be optimized in the data assimilation process.Then,remote sensing data and crop growth model were combined by constructing the minimum cost function of model-simulated LAI and remote sensing retrieved LAI.The data assimilation of remote sensing data and crop growth model was completed by adjusting the parameters to be optimized with the SCE-UA optimization algorithm.The potato production in the research area was estimated based on the data assimilation results.The main conclusions of the paper were as follows:(1)Based on the pretreatment of the GF-1 remote sensing image,the maximum likelihood method was used to classify the land use status in the research area and extract the potato planting area.The classification results showed that the overall classification accuracy of different land use types was95.22%,kappa coefficient was 0.93,and the classification accuracy was relatively high.By calculating the vegetation index and leaf area index of the potato,the remote sensing retrieval of LAI was completed by statistical regression method.The retrieval results showed that the fitting degree of normalized difference vegetation index NDVI and LAI of the potato in the research area was the highest,and the determination coefficient(R~2)was 0.73.The accuracy of remote sensing retrieved LAI obtained by NDVI-LAI statistical regression model was higher.The determination coefficient(R~2)of retrieved LAI and measured LAI was 0.88,and the root mean square error(RMSE)was 0.72.(2)The parameter sensitivity of input parameters of DSSAT-SUBSTOR crop growth model was analyzed by EFAST global sensitivity analysis method.The results showed that the upper limit of the critical temperature at which the tuber begins to grow,photoperiod coefficient,potential tuber growth rate,root growth coefficient,soil drainage upper limit,drainage rate,organic carbon content,soil total nitrogen content,soil solution pH,sowing date,irrigation date,irrigation amount,and nitrogen application amount were all parameters that were sensitive to the potato production and LAI.In the calibration process of DSSAT-SUBSTOR model,the applicability of the model was evaluated by four indicators:emergence date,tuber start date,maximum LAI value and production.The calibration results showed that the model-simulated emergence date and tuber start date were the same as the measured values.The mean relative error MRE of model-simulated maximum LAI value and measured maximum LAI value was 1.10%.The mean relative error MRE of the model-simulated production value and the measured production value was 5.72%.The results showed that DSSAT-SUBSTOR model could simulate the potato production with high accuracy on a single point scale.(3)First,the input parameters of DSSAT-SUBSTOR model were regionalized to obtain potato production estimation results without the unassimilated of remote sensing data.Then,the minimum cost function of remote sensing retrieved LAI and model-simulated LAI was constructed,and the SCE-UA optimization algorithm was used to adjust the parameters to be optimized to achieve data assimilation of remote sensing data and crop growth model.The results showed that the production estimation accuracy of the model without assimilated remote sensing data was low,and the production estimation result of assimilated remote sensing data was closer to the measured value,and the estimation accuracy was higher.The mean relative error MRE of assimilated remote sensing data was 6.17%,which was 9.45%lower than that unassimilated remote sensing data.It showed that the combination of remote sensing LAI data and DSSAT-SUBSTOR model can effectively improve the estimation accuracy of regional potato production.