| Soil is the basic condition for human social production and life,and soil degradation affects agricultural development.Soil salinization is one of the forms of soil degradation.It is of great significance for degraded soil management and ecological environment protection to obtain the information of regional soil salinization degree.In this thesis,the Yinchuan Plain is taken as the research area,and the salt impact factor and salt index are used as input parameters to establish three soil salt inversion models:Support Vector Machine(SVM),BP Neural Network(BPNN)and Bayesian Neural Network(BNN).The best model is selected to invert the soil salinization at different depths in the study area,and the spatial distribution and spatial pattern change characteristics of soil salinization in different periods are obtained.The driving factors such as soil physical and chemical properties,elevation and land use intensity were obtained.On this basis,the changes of driving forces in different periods were analyzed by means of geographic detector and grey correlation analysis.In addition,groundwater,soil salt content at different depths,evaporation ratio and GDP were selected as evaluation factors to construct an ecological risk assessment index system for salinization,and the ecological risk of salinization in Yinchuan Plain was evaluated under the support of analysis methods such as weighted average method.The main conclusions are as follows:(1)Through the soil salinity inversion model,it is found that the BNN model is the best inversion model by comparing the different variable modeling and verification effects of the selected algorithm.The introduction of neural network has certain advantages in the training of the model.The BNN model trained by the salinization influencing factors in the 0～20 cm soil salinity inversion has R2=0.797,RMSE=2.751,which is the best soil salinity inversion model.The BNN model R2=0.442,RMSE=1.006,which was trained by the salinity index of 20～40 cm soil salinity,was the best soil salinity inversion model.(2)The spatial distribution characteristics of soil salinity in Yinchuan Plain showed light in the south and heavy in the north,and the interannual variation of soil salinization showed a decreasing trend as a whole.In 2015,the total area of 0～20 cm saline soil was 5176.89km2,accounting for 72.84%of the total area of the region.In 2021,the total area of 0～20 cm salinized soil was 4741.84 km2,accounting for 66.71%of the total area of the region,and the salinized area decreased by 435.05 km2.The total area of salinized soil in 20～40 cm soil was 5078.34km2 in 2015,accounting for 71.44%of the total area.In 2021,the total area of salinized soil is 3515.89 km2,accounting for 49.46%of the total area of the region,and the salinized area is reduced by 1562.45 km2.(3)The analysis of the driving factors of salinization shows that groundwater salinity,groundwater depth,land use intensity,potential evapotranspiration and soil salinization have a strong correlation.The driving force of human activity factors is increasing,and the increase of land use intensity is the most significant.The driving force of groundwater salinity on soil salinization is the main driving force.(4)The ecological risk of soil salinization in Yinchuan Plain is mainly mild risk area,accounting for 45.36%of the total area,followed by risk-free area,accounting for 25.90%of the total area,and the area of maximum risk area is the smallest,accounting for 2.87%of the total area.The ecological risk of salinization is mainly affected by factors such as 0～20 cm soil salinity,surface temperature,groundwater salinity,elevation,soil organic matter,and evaporation ratio,and is least affected by soil pH. |
PDF Full Text Request