Attribute Analysis Of Spatial Pattern Of Runoff Changes And Runoff Elasticities For 1003 Catchments Across Contiguous U.S.

It is of great significance for decision-making of water resources management and ecological protection to study the spatial pattern of runoff changes and its elasticities under the impact of climate change and human activities,and attribute the pattern features.This paper taken 1003catchments over the United States as the research area,using the measured runoff data and meteorological data to calculate the runoff changes and its elasticities on the basis of Budyko theory.Fifty-six indexes involved in multi-type,large-scale and high-quality catchment characteristics including climate,hydrology,land use/cover,soil,geology,topography,etc.,and approaches including principal components,K-means clustering,hierarchical clustering,machine learning and interpretable machine learning were used to explore attribution of spatial pattern of runoff changes and its elasticities across the United States.The main conclusions of this study are as follows:1)The spatial distribution characteristics of runoff changes and its elasticities were explored for 1003 catchments across the contiguous U.S.Total changes(-36%/10a-56%/10a),rainfall-(-7%/10a-17%/10a),and catchment characteristics-(-60%/10a-56%/10a)induced changes generally showed an increase in the east and a decrease in the west;the changes caused by potential evaporation showed an increase in the central and southern parts,and a decrease in the central and western parts(-4%/10a-19%/10a);the average temperature-induced changes are unevenly distributed(-240%/10a-430%/10a).The elasticities of rainfall and potential evaporation are large in the west and small in the east(1-2,-1-0,respectively);the elasticity of average temperature fluctuates greatly(-1--16);the distribution of catchment characteristics elasticity in the catchment area is more regular,mainly in-1.2-0.The distribution of nine ecoregions in the contiguous United States can roughly explain the distribution characteristics of runoff changes and its elasticities.2)Based on the cumulative local effect method,the nonlinear relationship between the catchment characteristics and runoff changes and the runoff elasticities was analyzed,also the importance of the impact factors was judged.Except the important role of annual runoff on runoff changes,longitude is also the one catchment attribute that have the greatest impact on total runoff changes and rainfall-induced changes.The annual herbaceous coverage is also a factor that affect potential evaporation-induced changes;rainfall regime had the greatest impact on the average temperature-induced changes.The runoff elasticities in arid areas are larger than that in humid areas,and the annual average runoff is the biggest factor affecting the runoff elasticities.3)Based on the cumulative local effect,the nonlinear relationship between the runoff characteristics and the runoff changes and the runoff elasticities was analyzed,and the importance of the impact factors was judged.The slope of the flow duration curve is the most important factor affecting the total runoff changes and the catchment attribute-induced changes.Runoff coefficient,actual evaporation and annual average runoff are the three runoff characteristics that have the greatest impact on the rainfall elasticities.The annual average runoff is the most influential factor of the catchment characteristics elasticity.The most important runoff characteristics for the parameter value of the catchment attribute are actual evaporation and runoff coefficients.4)The random forest model simulates the runoff changes and its elasticities.When using the catchment attribute to predict the total runoff changes and the catchment attribute-induced changes,the accuracy rate R~2 is the largest(0.6);the accuracy of predicting the elasticities of catchment attribute,rainfall,and potential evapotranspiration is larger(0.96,0.7,0.7,respectively).According to the runoff features,the simulation accuracy of predicting the total runoff changes and the catchment attribute-induced changes was the highest,both of which are 0.6.The accuracy of predicting the parameter value of the catchment attribute and the elasticities of rainfall,potential evaporation,catchment attribute was the highest,which were 0.6,0.8,0.8,and 0.96,respectively.5)According to the cluster results of runoff characteristics and catchment attributes,the runoff changes and its elasticities distribution pattern within each category were analyzed.10 categories can well explain the distribution pattern of runoff changes and its elasticities.The runoff coefficient,annual average runoff and 95%quantile runoff have the greatest impact on the total runoff changes and the catchment attribute.The runoff elasticity coefficient is larger in arid regions.