Inversion And Spatiotemporal Evolution Of Soil Salinity In Coastal Plain

Soil salinization has seriously restricted the sustainable development of agriculture and ecological security.Rapid access to the content,characteristics and temporal and spatial evolution information of soil salinity provides the basis for improving and repairing the effective soil salinization of crops.According to the characteristics of saline soil in the coastal plain,Huanghua city is selected as the research area,and the remote sensing image of OLI/Landsat-8 is used as the data source,the shortwave infrared bands are introduced to the traditional vegetation and salinity indices and the corresponding enhanced indices are proposed respectively.In view of the traditional indices and enhanced indices of vegetation and salinity,the remote sensing inversion model of soil salt content is constructed by means of multiple stepwise regression and support vector machine respectively,and the prediction results of the model are verified and compared,so as to select the best index and the best soil salt inversion model.Finally,based on the best index model,the spatial characteristics and temporal and spatial evolution of soil salinization in the study area from 2014 to 2018 were analyzed.The main conclusions are as follows:(1)After the introduction of shortwave infrared bands by basic combination algorithm,the correlation coefficient between indices and soil salt content increased by 0.01?0.44,and the correlation increased significantly.The variance expansion factor between indexes decreased by 4.06-107.14,and the VIF value between most improved indices was less than 10,which greatly reduced the multicollinearity between indexes.Based on SVM salt inversion model,the improved values of RPD increased 0.49,and the prediction ability of the model is improved from general to better.After enhancement,the predicted and measured values are evenly distributed on the 1:1 line and both sides,which is more linear.It shows that on the basis of traditional indices,the introduction of shortwave infrared bands can effectively improve the inversion accuracy of regional soil salt content.(2)The trend of grey correlation between soil salt content and traditional and enhanced indices are almost the same,among which SI3,SI,SI-T and NDSI have higher correlation with soil salt content,indicating that the change trend of salt indices is more similar to the change trend of soil salt content in this area,which is more suitable for the prediction of soil salt content.Compared with SMLR and SVM models,R2 of SVM is increased by 0.03?0.22,RMSE is decreased by 0.18?1.29 and RPD is increased by 0.85,which shows that the SVM model is better than SMLR model in the prediction of soil salt content based on OLI/Landsat-8 remote sensing image in coastal plain.(3)Based on the ESI3,ESI,ESI-T and ENDSI indices of the remote sensing image of OLI/Landsat-8 as the relevant variables of SVM model,the accuracy of spatial distribution of soil salt is tested and analyzed.The predicted and measured values of soil salt content in the study area are 0.39?36.38 g·kg-1 and 0.35?38.09 g·kg-1,respectively.The difference between them is small and the results are very consistent,which shows that the inversion results are accurate and reliable.The soil salinity in the study area presents an east-west differentiation trend,that is,the salt level in the western plain area is relatively low,mainly non saline soil and light saline soil,and the salt level in the eastern coastal area is relatively high,mainly saline soil,and its spatial distribution is consistent with the field situation.The predicted area is 1512.82 km2,which is the same as the measured area,the accuracy can reach 63.25%and the model inversion ability is high under the condition of accurate prediction.To some extent,this study is of practical significance to improve the accuracy of soil salinity inversion in coastal plain area,quickly and accurately obtain the dynamic change of saline soil and its reasonable planning and utilization.(4)From 2014 to 2018,the soil salinization in the study area showed a significant slowing down trend,among which the non saline soil and medium saline soil kept a stable trend,the light saline soil showed a significant increase trend,and the heavy saline soil and saline soil showed a decrease trend.According to the statistical results in 2018,moderate saline soil and saline soil still account for a large proportion,which seriously inhibit the healthy growth of crops and increase the pressure of high-quality cultivated land and high-quality food production in the region.According to the soil transfer matrix of different salinization degree,it is found that lightly salinized soil,moderately salinized soil,heavily salinized soil and saline soil are mainly transformed into non-saline soil,and a small part of non-saline soil is mainly transformed into lightly salinized soil,which indicates that in recent 5 years,the degree of soil salinization in the study area has been well alleviated,and the environmental ecological situation has shown a trend of improvement.The purpose of this study is to improve the salt prediction ability of coastal saline soil,to achieve the effect of assisting field chemical experiment measurement or even replacing field measurement.It is of practical significance to achieve the rapid acquisition of salt status and guide the rational utilization and management of agricultural land.