| Soil moisture is an important variable in climate system and land surface water cycle,and very sensitive to climate change.In addition,changes in climate factors such as precipitation,temperature,and radiation change can directly cause soil moisture change,and land surface geographical factors also have an important impact on the spatial heterogeneity of soil moisture.The change of soil moisture also has an important feedback effect on the occurrence of extreme weather,vegetation condition,and the production of agriculture and animal husbandry.Therefore,systematic study the spatiotemporal changes of soil moisture and its influencing factors has great scientific and practical significane for formulating water and soil resource utilization and ecological protection strategies and arranging agriculture and animal husbandry production.The Tarim River Basin was taken as the research area in this paper.Based on long–term surface observation data,reanalysis product data and other data,semi–variogram,Mann–Kendall nonlinearity statistical test method,ensemble empirical mode decomposition(EEMD)and rescaled range analysis method(R/S analysis method)and other methods were comprehensively applied to analyze the spatial pattern,spatial variability characteristics and temporal change trend of soil moisture in the Tarim River Basin from 1979 to 2018.The Geo Detector model was used to study the influence of land surface geographical environment factors on the spatial pattern of soil moisture changes,the influence of climate factors on soil moisture changes and the multi–time scale correlation between soil moisture and climate factors were analyzed by simple correlation analysis and wavelet correlation analysis,and studied the response of soil moisture changes to climate change.Constructing a random forest model based on the EEMD method(EEMD random forest model)to simulate the soil moisture change process on seasonal,interannual and interdecadal scales to explore the nonlinear relationship between soil moisture changes,and climate factors and normalized difference vegetation index(NDVI)on different time scales.The main conclusions of this study are as follows:(1)In terms of space,the spatial distribution of soil moisture in the Tarim River Basin is highly consistent with altitude and precipitation showing a spatial pattern of“the high value distributes in the north and west,while the low value distributes in the south and east,the value in mountainous areas is higher than that in plain areas”,and it decreases from high altitude mountainou areas to the interior of the basin,showing the characteristics of ring distribution.Structural foctors have a significant effect on the seasonal and annual spatial variability of soil moisture in the entire Tarim River Basin,which is far greater than the effect of random factors on the spatial variability of soil moisture.Among them,the spatial difference and variation range of soil moisture on the south slope of Tianshan Mountain are the smallest,while the largest is on the north slope of Kunlun Mountain.In terms of time,the soil moisture in the Tarim River Basin showed a downward trend in the four seasons of spring,summer,autumn and winter and annual changes,and occurred mutaion in different years.Among them,the decreasing trend of soil moisture in spring and summer on the south slope of Tianshan Mountain,winter in Tarim Basin and spring on the north slope of Kunlun Mountain are more significant.In addition,the EEMD nonlinear variation trend analysis showed that the soil moisture changes in the Tarim River Basin have obvious multi-time scale oscillation periodicity and seasonal,interannual and interdecadal fluctuation characteristics,but the seasonal fluctuation is the most significant.The change cycle of soil moisture on the south slope of Tianshan Mountain,the Tarim Basin and the north slope of Kunlun Mountain are relatively close on the seasonal and interannual scales,the change cycle of soil moisture on the south slope of Tianshan Mountain and the north slope of Kunlun Mountain is basically the same on the interdecadal scale,while the change cycle of soil moisture on the decadal scale is slightly shorter.(2)The land surface geographic environmental factors,including altitude,slope,aspect,landform types,annual average NDVI and land use types have important effects on the spatial difference and temporal changes in soil moisture changes.On the south slope of Tianshan Mountain,the annual average NDVI has the greatest influence on the spatial difference of soil moisture changes,and the other land surface geographical environmental factors also have an important influence on the spatial differences of soil moisture changes.In the Tarim Basin,the annual average NDVI,landform types and land use types are the dominant factors for the spatial differences of soil moisture changes,which are greater than the influence of altitude,slope,and aspect.On the north slope of Kunlun Mountain,altitude has the greatest influence on the spatial difference of soil moisture changes,followed by slope,landform types and annual average NDVI,and aspect and land use types have the least influence.Moreover,the interaction effects of any two different factors have a nonlinear or mutually reinforcing relationship on the spatial pattern of soil moisture changes in the Tarim River Basin.The change trend and amplitude of soil moisture are different on each sub-categorie of different geographic environmental factors.The precipitation,temperature,relative humidity,sunshine duration,and actual evaporation are significantly correlated with changes in soil moisture in the entire Tarim River Basin.In the long term,the soil moisture increased significantly with the increase of precipitation,relative humidity and actual evaporation,while decreased significantly with the increase of temperature and sunshine duration.But there were different resonance periods and phases relationships between soil moisture and climatic factors in different time periods of the study period.Among them,multi-time scale resonance periods occurred frequently,but the duration was short,and the correlation was unstable.Long-term resonance periods lasted long,and the correlation was stable.The influences of precipitation,temperature,relative humidity,sunshine duration,and actual evaporation on soil moisture changes have obvious seasonal differences.Among them,the actual evaporation is a key factor affecting the soil moisture changes in the Tarim River Basin,and the influence on soil moisture change in the Tarim Basin is greater than that on the south slope of Tianshan Mountain and the north slope of Kunlun Mountain.The relative contribution rate of summer precipitation to soil moisture changes is higher than that of temperature,relative humidity,and sunshine duration.And the relative contribution rate of summer precipitation to soil moisture changes on the south slope of Tianshan Mountain is greater that of actual evaporation,indicating that summer precipitation is a crucial factor in soil moisture changes on the south slope of Tianshan Mountain.(3)EEMD random forest model was established to simulate the nonlinearly change process of soil moisture,and the impacts of climate and NDVI chang on soil moisture changes were studied from multiple time scales.The reconstructed seasonal change trend of soil moisture can well describe the changes in the original soil moisture sequence of the Tarim River Basin during the study period.The interannual and interdecadal changes in soil moisture can effectively reveal the alternation of increase and decrease of soil moisture in different periods on the south slope of Tianshan Mountains,Tarim Basin,and north slope of Kunlun Mountains.Moreover,the interdecadal scale is more suitable for exploring the impact of climate fluctuations and NDVI changes on the process of soil moisture changes.The verification results of the credibility and effectiveness of the EEMD random forest model showed that the NSE values on seasonal,interannual and interdecadal scales were above 0.748,0.813 and0.990,respectively,which indicate that the simulation accuracy of EEMD random forest model is relatively high,and EEMD random forest model can be better applied to simulate the process of soil moisture changes. |
PDF Full Text Request