Study On Influencing Factors And Simulation Of Soil Moisture In The Middle Reaches Of The Yellow River

The middle reaches of the Yellow River span six provinces,including Inner Mongolia,Ningxia,Gansu,Shaanxi,Shanxi and Henan.It is an important base for agriculture,animal husbandry and energy production in China.In this area,evaporation is strong,precipitation is low,topography fluctuates greatly,water resources are scarce,vegetation is sparse,and ecological environment is fragile,which makes soil moisture low and spatial heterogeneity strong.At the same time,the change of soil moisture is influenced by many different factors,and these factors are complex and diverse,which restrict the agricultural and ecological development in this area.However,at present,there is still a lack of clear understanding of the temporal and spatial variation characteristics and main influencing factors of soil moisture in the middle reaches of the Yellow River.Therefore,this paper first analyzes the temporal and spatial changes of soil moisture in the middle reaches of the Yellow River.Secondly,the main influencing factors of soil moisture in the middle reaches of the Yellow River are revealed.Finally,based on the machine learning model,a long-term pixel-scale soil moisture model in the middle reaches of the Yellow River is constructed.The main conclusions of this paper are as follows:(1)The soil moisture in the middle reaches of the Yellow River and its sub-regions showed a sustainable and significant decreasing trend on the annual scale and the sarle of spring and summer,and there was a significant mutation point,and the degree of variation was small on each time scale,and there was a main periodic oscillation of about 24 years and a periodic change of about 22 years.The spatial distribution of soil moisture in each time scale shows the distribution characteristics of low in the north and high in the south;In this area,the areas with small spatial variability of soil moisture are mainly located in Hekou Town to Longmen,the areas with medium variability are mainly located in Longmen to Sanmenxia and Sanmenxia to Huayuankou,and there is no obvious strong variability area.The percentage of the area where the annual scale and seasonal scale soil moisture showed a continuous and significant decreasing trend in the middle reaches of the Yellow River was 81.4%,86.5%,91.1%,3.4%and 23.0%,respectively.(2)The annual NDVI is the main land factor in the whole middle reaches of the Yellow River and the sub-region from Hekou Town to Longmen,and the altitude is the main land factor in Longmen to Sanmenxia and Sanmenxia to Huayuankou,and the synergistic interaction of multiple factors leads to the enhanced influence of land factors on the spatial distribution pattern of soil moisture.Generally speaking,seasonal and annual soil moisture is positively correlated with precipitation,relative humidity and actual evapotranspiration,and negatively correlated with air temperature and net solar radiation,and most of them are significantly correlated.There are obvious differences in the influence of various factors on soil moisture during the period of significant reduction.On the annual scale,the contribution rate of temperature to the reduction of soil moisture in the middle reaches of the Yellow River is the largest,while in spring,the contribution rate of relative humidity to the reduction of soil moisture is the largest,and in summer,the actual evapotranspiration is the largest.(3)Through recursive feature elimination method,the characteristic factors of establishing different monthly machine learning models in each region are precipitation,air temperature,relative humidity,actual evapotranspiration,net solar radiation and NDVI.According to the characteristic factors,machine learning models are established in different regions.The results show that the simulation accuracy of support vector machine model is the best in all regions,and the NSE is greater than 0.8 in the periodic rate and verification period.Secondly,the extreme learning machine model,the NSE is greater than 0.75 in the periodic rate and verification period;However,knn and random forest model are not effective.A pixel-scale machine learning model of soil moisture in the middle reaches of the Yellow River is established by using the support vector machine model.The results show that R2>0.60 and NSE>0.50 of the machine learning model established for each pixel in the regular and verification periods are good,which reveals the changes and influencing factors of soil moisture in the middle reaches of the Yellow River.