Research On Monitoring Method Of Soil Salinization Based On Satellite Remote Sensing Data Assimilation

Soil salinization is a key factor affecting the economic benefits of land resources and crop yield in Hetao Irrigation District in Inner Mongolia.Therefore,timely and accurate prediction of soil salinity is of great significance for the precise management of soil salinization.Because of the spatiotemporal heterogeneity of the soil,it is difficult and complicated to simulate the actual soil salinity.With the improvement of remote sensing technology,a method of data assimilation combining remote sensing data and model simulation results are applied to the process of soil salinity simulation gradually,effectively improving the prediction accuracy of soil salinity.Therefore,in this paper,the soil salinity obtained from the GF-1 satellite data is taken as the remote sensing observed value.The quantitative relationship between the GF-1 satellite images and the measured soil salinity is used as the observation operator.HYDRUS-1D model simulation process is Model operator.At the same time collect meteorological data,soil data and measured salinity data during the experiment period.Through the En KF and the remote sensing observations are assimilated into the HYDRUS-1D model simulation process,thereby optimizing the simulation process of soil salinity,improving the prediction accuracy of soil salinity,and analyzing the assimilation results of the number of collections and observation errors during the assimilation process Sensitivity.The main results obtained are as follows:(1)A time series remote sensing inversion model of soil salinity was constructed and the optimal inversion model in each period was determined.In each period,based on the full subset screening,determine the sensitive spectral index groups of different periods,B1,B2,SI,SI1,EVI;B4,SI3,SR;B4,BI,NDVI;B4,SI2,S2,S3,EVI.With the sensitive spectral index group as the independent variable in each period and the measured salt content as the dependent variable,the soil salinized prediction models were established respectively.In this paper,three models including QR model,ANN model and SVM model can be used to establish the quantitative relationship between the measured soil salinity and Gaofen-1 satellite remote sensing image,modeling set and verification The fitting accuracy of the set is above 0.45,and the RMSE is less than 0.2%,after a comparative analysis of the accuracy of the prediction results,the quantile regression model is better than the other two models after being screened by the entire subset.Therefore,in this plan of assimilation of soil salinity,the observation operator in each period is the quantile regression model after the whole subset screening.(2)The HYDRUS-1D model was used to simulate the soil salinity at different depths under the vegetation cover in the irrigation area.According to the actual situation of sampling points in the irrigation area and previous experience,combined with indoor and outdoor experiments after multiple debugging to determine the relevant parameters of the soil salt content simulation during the experiment,the HYDRUS-1D model simulation output soil profile Simulation results.At the depths of 0-20 cm,20-40 cm and 40-60 cm,the RMSE of the simulated value of the soil salinity is 0.014,0.012 and 0.011,the RE is 0.061,0.080 and 0.087,the overall RMSE is 0.013 and the RE is 0.076.Compared with the measured soil salinity,the error is relatively small,indicating that HYDRUS-1D can simulate the soil salinity more concisely and efficiently,reflecting the salt content of the soil profile in the study area to a certain extent.(3)The assimilation scheme of remote sensing observation soil salinity based on En KF is determined,and the sensitivity of the assimilation process is analyzed.The assimilation algorithm is the En KF,and the assimilation simulation operator is HYDRUS-1D Simulation process of solute transport model.Compared with the simulation results using the HYDRUS-1D model alone,after assimilating remote sensing to observe the soil salinity data,the accuracy of the obtained soil salinity at each depth has been improved significantly.At depths of 0-20 cm,20-40 cm,40-60 cm and 0-60 cm,NSE reached 0.619,0.668,0.533 and 0.749,RMSE decreased by 0.008,0.008,0.001 and 0.006,and RE decreased by 0.024,0.064,0.002 and 0.003.As the number of sets increases,the assimilation results gradually stabilize.When the number of sets is 100,the RMSE of the assimilation values of soil salinity at various depths are 0.006,0.004,0.010,and 0.007,which is the optimal number of sets for this assimilation process.Therefore,the research on the assimilation of soil salinity using GF-1 satellite remote sensing data can provide a certain reference value for soil information forecast at the irrigation area scale.