| Increasing drought risk caused by global warming has induced more serious ecological problems,making ecological drought a research hotspot for eco-hydrology.Currently,there is lacking a unified definition of ecological drought,and the corresponding assessment method is difficult to reflect the actual dynamics of water supply and demand for different vegetation types.Additionally,the mechanism behind ecological drought is unclear.Therefore,a research conducted on the assessment of ecological drought and response mechanism of ecological drought to meteorological and groundwater drought may contribute not only to the expansion of the drought classification and drought index,but have certain guidelines for ecological drought warning and ecological protection.This study defined the terrestrial ecological drought by investigating the mechanism of its occurrence and connotation.An ecological drought index(SEWDI)was subsequently constructed based on a dynamic balance between water supply and demand of vegetation.Using the Northwestern China as the study area,the spatio-temporal evolution characteristics of ecological drought were analyzed by using rotation orthogonal decomposition(REOF),Mann Kendall(MK)trend test,and Theil-Sen(TS)median method.Three-dimensional clustering identification method and gradient enhanced regression(GBR)model were combined to determine how water and temperature influences vegetation in different ecological drought events.The traditional downscaling model for groundwater storage anomaly(GWSA)was improved by integrating machine learning and depth learning models.The meteorological drought index and groundwater drought index were constructed based on precipitation and GWSA,respectively.Then,joint and relative effects of the two types of droughts on ecological droughts were quantified by residual analysis.As an extension of the spatio-temporal matching method of two types of drought events,this study proposed a method to match three types of drought events.Using this method,the pairs of meteorologicalgroundwater-ecological drought events with genetic relationship were identified.The response relationship model for duration,intensity and impact area of ecological drought events is constructed on the basis of polynomial and copula functions.In the end,Bayesian network was used to calculate thresholds of meteorological drought and groundwater drought for ecological droughts at different levels.The major conclusions are as follows:(1)The concept of ecological drought for the terrestrial system was defined as a phenomenon that water deficits in vegetation are higher than regional multiyear averages due to supply-demand imbalance,which may affect the structure and function of vegetation and trigger feedback in natural and/or human systems.Based on this concept,a standardized ecological water deficit drought index(SEWDI)was constructed to estimate ecological drought.Compared with four commonly used drought indices,SEWDI has a higher correlation with vegetation index under different time scales,humid conditions and different vegetation water use efficiency.This means SEWDI can monitor ecological droughts more precisely.(2)Ecological droughts are palliating in most parts of Northwestern China,especially in Ningxia,Shaanxi,northern Gansu,and sountern Qinghai,but are intensifying in the Hexi Corridor.Among the 184 ecological drought events,the drought events that occurred before the 21 st century tended to be more serious,and spread northwestward.During most ecological drought events,the effects of water stress on vegetation outweighed those of temperature stress,and both showed a tradeoff effect.(3)Using a random forest model,three deep learning models(Res Net、Res CNN、GRU-FCN),two machine learning models(xgboost and lightgbm)and the optimal output of38 single models were fused to obtain consistent spatial accuracy between GWSA and other datasets.In comparison to the most accurate single model(Res Net),the Kling efficiency coefficient and Nash coefficient of the fusion model increased by 49% and 45%,respectively.(4)The area where ecological drought process dominated by a joint effect of meteorological drought and groundwater drought were mainly concentrated in arid and semiarid regions,accounting for 60% of the study area.In these regions,the joint contribution of meteorological drought and groundwater drought to ecological drought varies between 0.57 and 0.61.In terms of relative contribution,ecological drought dominated by groundwater drought was mainly distributed in southern Shaanxi,Southeastern Gansu and southern Qinghai,accounting for 12.7%～21.8% of the total.In addition,groundwater drought has a larger effect on forestland than grassland,and it is more severe in the spring and winter than during the summer and fall.(5)A total of 31 pairs of meteorological-groundwater-ecological drought events with genetic relationship were identified.The most serious ecological droughts were caused by the joint influence of meteorological drought and groundwater droughts,most of which occurred before the 21 st century.The ecological drought events caused only by meteorological drought were the lightest,mainly distributed in arid and semi-arid areas.(6)The optimal response models for severity,duration and impact area of ecological drought are joint distribution model,bivariate quadratic function,and bivariate quadratic function,respectively.The thresholds of meteorological drought and groundwater drought to trigger mild,moderate,severe,and extreme ecological drought events and the corresponding curves are obtained using the Bayesian network.The results can provide valuable information for early warning of ecological drought. |
PDF Full Text Request