Impact Of Climate Variability And Vegetation Change On Terrestrial Water Storage In China

Groundwater,soil water,surface water,snow and ice are dynamic components of the terrestrial water cycle.In the absence of strong hydrological and climate changes or intense human activities,the Terrestrial Water Storage(TWS)usually fluctuates within a stable range.However,in recent years the TWS in some basins have undergone dramatic changes caused by climate variability,vegetation changes and human activities.Due to the differences in geology,hydrology,vegetation,climate,topography,and socioeconomic conditions,the responses of TWS to climate variability,vegetation changes,and human activities are highly variable at different spatial and temporal scales.At present,most researches on TWS focus on the analysis of their spatial and temporal changes,but the understanding on the spatial variations of TWS’s response to various driving factors and their associated mechanisms remains limited.This paper investigated the response of TWS to climate variability,vegetation changes and human activities,and the response variations in 214 large watersheds across china,which provided scientific supports for future water resource management and assessment.Firstly,the Mann-Kendall trend test was used to detect the change trend of inter-monthly / inter-annual hydrological variables in the study watersheds from 2003 to 2014.Partial correlation analysis was used to investigate the correlation between monthly/ annual TWS and climate,vegetation or human activities variables,to determine watershed types and the key driving factors for variations in TWS.The artificial neural network models were then established based on the selected key driving factors to quantitatively predict the variations of TWS,and the sensitivity analysis was applied further to estimate the relative contributions of climate variability,vegetation changes and human activities on the variations of TWS.The key findings are as follows.(1)There are obvious spatial variations of TWS trends in the study watersheds across China.The TWS was on a rising trend with a surplus in watersheds from the Northeast,Qaidam Basin,the middle and lower reaches of the Yangtze River Plain,and the Southeastern coastal regions while it generally showed a downward tendency with a loss in watersheds from the Southwest,Xinjiang,the Loess Plateau,and the North China Plain.(2)The seasonality of TWS in the study watersheds was obvious and spatially variable across China.Generally speaking,the TWS was the lowest in winter and spring,and was the highest in summer and autumn.The timing of the maximum or minimum monthly TWS within a year was different among watersheds due to their differences in the timing and duration of rainy season and dry season.The monthly TWS in most watersheds from South China was featured with greater variations within a year than those from North China.(3)The responses of monthly TWS to vegetation in the study watersheds varied with their climate types.According to the response of TWS to vegetation change,the study watersheds were divided into three types: sponge,mixed and pump.Vegetation in sponge watersheds(Dryness index,DI≤0.92)yielded a positive effect on monthly TWS.Vegetation in pump watersheds(DI≥3.3)produced a negative effect on monthly TWS.In mixed watersheds(0.92<DI<3.3)vegetation yielded either a negative or positive impact on monthly TWS.Wavelet neural network and artificial neural network were used to establish the monthly TWS prediction model.The results showed that the monthly TWS prediction model established using the wavelet neural network had the best performance,with an average R2 of 0.9,while the monthly TWS prediction model R2 established using the artificial neural network was only 0.81.The wavelet neural network predicted the best monthly TWS in the sponge type,with R2 as high as 0.92,the pump type was the lowest,with R2 as only 0.87.The relative contribution of climate variability to in monthly TWS variations was greater than that of vegetation.The TWS in the drier watersheds tended to be influenced by climate variability.(4)The responses of annual TWS to climate,vegetation and human activities in the study watershed were spatially variable across China.Bayesian neural network with uncertainty analysis was used to establish the prediction model of annual TWS in the study watersheds classified as seven geographic regions in China.The results showed that the model performance in South China was best,with R2 up to 0.85 while model performance of Northwest China was lowest with R2 being only 0.7.Climate variability(especially precipitation)was the most important driving factor for annual TWS variations in the study watersheds,which was followed by human activities.The anthropogenic impact on annual TWS in the watersheds located in the east of the HeiheTengchong line is greater than that in the watersheds situated in the west of the HeiheTengchong line.Different human activities may yield different impacts on TWS in different regions,and these effects may offset each other,or they may be additive.The relative contribution of vegetation to annual TWS variations was greater than that on monthly TWS.In particular,watersheds with more high-evapotranspiration vegetation tended to have greater relative contribution of vegetation to annual TWS variation.