Prediction And Attribution Analysis Of Future Changes In Key Land Surface Hydrological Variables In China Based On The Budyko Hypothesi

In the context of global warming,changes in climate and land surface characteristics profoundly impacted on the spatio-temporal pattern of the hydrological cycle.Therefore,exploring the response mechanism of key hydrological cycle variables such as water yield(WY)and evapotranspiration(ET)to changes in climate and land surface characteristics has become a current research hotspot in hydrology,especially for the future.However,factors such as insufficient characterization of the water cycle process,inaccurate parameterization,and failure to consider basin-scale processes in the Earth system model affect the credibility and consistency of its prediction of global or regional hydrological variables.To accurately explore the impact of factors such as climate and land surface on key hydrological variables in the future,this study first constructed seven machine learning models that simulate the Budyko model parameters of 259 measured hydrological divisions(HD)in China.Furthermore,based on the optimal machine learning model XGBoost,a Budyko model parameter n map of China’s spatial-temporal continuity(1987-2005 period by period/0.25°grid scale)was generated.Secondly,the study re-simulated the WY and ET under multiple emission scenarios in the future based on the downscaled CMIP6 data and the XGBoost model.The study then attempted to reveal the impact of changes in WY and ET in the future on precipitation and potential evapotranspiration(including its driving factors)and the response degree of land surface characteristics.Finally,the study clarified the temporal and spatial changes of the dominant factors driving WY and ET in the future period in China.The conclusions of this study are as follows:(1)Based on the analysis of Boruta characteristic screening technology,the proportion of cultivated land area,leaf area index,and elevation are important characteristic factors that determine the Budyko parameter n,but other factors cannot be ignored.Among the seven developed machine learning models,the XGBoost model was found to be good at reproducing the spatial pattern and temporal variation of Budyko parameters n and ET.On a long-term time scale(1982–2015),the gridded Budyko parameter n exhibits large spatial variability.However,the gridded ET shows a decreasing trend from southeast to northwest.(2)The Budyko parameter n in China exhibits an increasing trend under the four scenarios relative to the historical period.The increase trend is significant in the North China Plain,the Northeast Plain,and the Sichuan Basin under the four scenarios.ET in almost all regions of China will gradually increase from near-future to far-future.The growth rate of WY is smaller than that of ET.WY will continue to increase under the four scenarios in the future,and the increase in the southern region is more significant.(3)From a Budyko framework perspective analyzing the driving factors(precipitation,potential evapotranspiration,land surface characteristic Budyko parameter n)contributing to WY and ET,precipitation drives a continuous increase in ET.Potential evapotranspiration in China is expected to generally drive an increase in ET from 2021 to 2100,with a spatial pattern showing noticeable north-south differences.Specifically,potential evapotranspiration contributes positively to evapotranspiration in Southwest China,the Huaihe River Basin,and Northeast China,while other regions have negative contributions.The contribution of potential evapotranspiration to ET turns from negative to positive in all scenarios over time.The contribution of Budyko parameter n to ET is similar to that of precipitation.Under the four emission intensities in China,the Budyko parameter n has a positive contribution to ET in most areas,but its contribution decreases over time.For WY,precipitation in China will continue to cause an increase from the short term to the long term.(4)The control of ET in northern and southern China is mainly attributed to precipitation climatology and the Budyko model parameter n,respectively.Notably,the area where temperature dominates ET will expand with the increase in emission intensity.In terms of temporal changes,most of the factors controlling ET exhibit regular variations.Precipitation climatology and Budyko model parameters remain the primary controllers of WY.Regarding temporal changes,except for a slight increase in the area dominated by precipitation under the SSP126 and 370 scenarios from the near-future to the far-future,other primary controlling factors hardly show any changes.