Observations And Modeling Of Global Potential Evapotranspiration Changes And Their Causes

Potential evapotranspiration(hereinafter referred to as PET)is the ability to describe the evapotranspiration process under certain weather and climate conditions,and is an important part of the water cycle and energy balance in the climate system.In-depth study of the temporal and spatial evolution of PET and its driving factors can help to understand the mechanism of drought in the context of climate warming,and provide a scientific basis for drought warning and prediction research.Based on the CRU-TS v4.02 data and ERA5 reanalysis data,combined with the CMIP5 and CMIP6 multi-models simulation results,this paper analyzes and studies the global spatiotemporal distribution characteristics of PET in the past 40 years,systematically evaluates the simulation ability of the climate model on PET,and quantitatively estimates different Relative contributions of climate drivers and regional differences,and estimated PET response characteristics under different emission scenarios in the future.The main conclusions are as follows:(1)The PET described by CRU and ERA5 has a uniform distribution on the global land for many years.The high-value areas appear in western North America,northern Africa,and central and western Australia(about 6.0 mm/day).Except for low levels in western North America and northern Africa(about 20%),most models can better simulate the spatial distribution characteristics of observation information(CRU and ERA5),and the simulation ability of CMIP6 is improved to a certain extent compared with CMIP5.(2)The CRU results show that since 1979,the PET in most land areas of the world has shown a significant increase(0.2-0.6 mm/day/50year).Among them,the growth rates in southern North America,southern Europe,northern Africa are more significant,while the PET in central and western South America and high latitudes in North America showed a downward trend.Except for high areas,the overall trend of ERA5 is consistent with that of CRU;the CMIP multi-models can basically simulate the spatial pattern of observed trends,but the increase is weak(0.1-0.3 mm/day/50year).Due to the difference in initial conditions of different climate models,there are significant differences in interannual and interdecadal variability between multiple models,resulting in the variability and long-term trend of the results of each model are different from the observation information,especially the multi-models ensemble mean will reduce the internal variability between the models and lead to a weak long-term trend.(3)The results of ERA5 reanalysis data show that there are significant spatial and temporal differences in the impact of different climatic factors on PET.The relative contribution of the saturated water vapor pressure difference is 95.1%,which is the key driving factor of PET;followed by the radiation term(short-wave radiation 40.0%,longwave radiation-20.7%);the contributions of wind speed is small;and surface pressure is basically negligible.The results revealed by the CMIP multi-model results are basically consistent with ERA5,but the contribution of the radiation factor is opposite,mainly due to the large differences in the radiative forcing factors simulated by different models.(4)Under the different emission scenarios in the future,the climatic average of the PET estimated by the CMIP multi-models in the late 21 st century(2079-2099)has increased significantly compared to the climatic average of 1979-1999,especially the CMIP6 multi-model prediction results predict that the increase in PET and the scope of impact will be greater under high-emission scenarios,which means that the probability of future global drought events will continue to increase.The difference in saturated water vapor pressure is still the dominant factor for future PET changes;unlike historical changes,the contribution of long-wave radiation is higher than that of short-wave radiation under high-emission scenarios.To some extent,the difference between the CMIP5 and CMIP6 multi-model prediction results can be attributed to the inconsistency of the external forced fields used.