Research On Identifying Atmospheric Circulation Patterns Associated With Extreme Precipitation And Downscaling Over The Central-Eastern China

Anomalies in atmospheric circulation represent a significant factor contributing to regional extreme precipitation.The objective identification of atmospheric circulation patterns associated with extreme precipitation events and their utilization for regional extreme precipitation estimation,to accurately project the future trends of extreme precipitation are pivotal issues in the scientific fields of climate change and regional response.In this study,based on the analysis of changes in extreme precipitation from1960 to 2015 over the central-eastern China,the self-organizing map(SOM)was used to objectively cluster the daily atmospheric circulation conditions(characterized by geopotential height at 500 h Pa).The dominant atmospheric circulation patterns related to extreme precipitation events were quantified by using the event synchronization method,to elucidate the characteristics of atmospheric conditions and moisture transport that affect extreme precipitation.Subsequently,various statistical indicators,such as coefficient of variation of root mean squared error(CVRMSE),relative bias(RB),and correlation coefficients were selected to systematically evaluate the performance of global climate models(GCMs)from Coupled Model Intercomparison Project Phase 6(CMIP6)in simulating extreme precipitation and the dominant atmospheric circulation patterns.This assessment aimed to ascertain their applicability in regional studies.Then,to accurately project the future trends of extreme precipitation over the central-eastern China in the context of global warming,a new precipitation statistical downscaling framework combining ensemble learning and nonhomogeneous hidden Markov model was developed.Atmospheric circulation characteristics and integrated water vapor transport index were employed as the independent variables for this downscaling approach,which aims to enhance the spatial resolution of GCMs and improve their reliability in simulating extreme precipitation.Finally,the downscaling model was applied to project the future trends of extreme precipitation over the central-eastern China during the early-,medium-and late-21 st centuries,under different warming scenarios.Relevant conclusions can provide a scientific basis for formulating adaptation strategies to extreme weather events over the central-eastern China.The key findings drawn from this study include:(1)There were regional differences in the synchronicity of extreme precipitation events among stations during the periods of 1960–1989 and 1990–2015.Based on modularity information,the rain gauge network was divided into four clusters.Secondly,25 dominant atmospheric circulation patterns were robustly identified from the 500 h Pa geopotential height anomaly field.Among these patterns,those characterized by positive anomalies of 500 h Pa geopotential height over the Eastern Eurasia continent and negative values over the surrounding oceans were highly synchronized with extreme precipitation events.Composite fields of atmospheric circulation and moisture transport displayed that discrepancies in the positions and intensities of circulation systems,including the Western North Pacific Subtropical High,East Asian Trough,and mid-latitude trough,overlaid with the low-level southwestern winds from the Bay of Bengal resulted in diverse intensities and pathways of moisture transport from the northern Indian Ocean to the eastern region of China,thereby affecting the occurrence and spatial distribution of extreme precipitation over the central-eastern China.(2)GCMs exhibited serious underestimation in simulating extreme precipitation amount(with a maximum RB exceeding 85%).These GCMs encountered challenges in accurately replicating extreme precipitation,with the CVRMSE greater than 0.5.When using the 25 dominant atmospheric circulation patterns identified from ERA5 reanalysis data as benchmarks,GCMs reproduced most circulation types for the periods1960–1989 and 1990–2015.The dominant atmospheric circulation patterns identified by GCMs during these two periods showed relatively consistent spatial characteristics compared to ERA5.In terms of the spatial similarity and temporal characteristics of the identified circulation patterns,MPI-ESM1-2-HR showed superior performance and can be used for downscaling studies.(3)Based on extreme gradient boosting(XGBoost)and random forest(RF),an ensemble learning model was constructed to predict the occurrence probabilities of different levels of daily precipitation(no rain,very-light rain,light rain,moderate rain,and heavy rain)in a multi-site ensemble.Subsequently,the daily precipitation probability sequences were utilized as large-scale predictive factors to further establish a non-homogeneous hidden Markov downscaling model,to achieve precipitation statistical downscaling.Validation results demonstrated that downscaled data exhibited lower errors in simulating monthly precipitation amounts,with the CVRMSE decreasing by approximately 43%.Moreover,the downscaled data proved more reliable in simulating the frequency and amount of annual average extreme precipitation.The downscaling model also performed well in historical simulation data of the MPI-ESM1-2-HR model.(4)The downscaling model was applied to three climate scenarios with different shared socioeconomic pathways and representative concentration pathways(SSPs).The downscaled daily precipitation amounts(2016–2100)were obtained to project future changes in extreme precipitation.Results showed that the central-eastern China would receive more precipitation in the future,with significant regional disparities.During the early 21 st century(2023–2048),annual average precipitation under the three warming scenarios(SSP126,SSP245,SSP585)demonstrated a “wetter in the southdrier in the north” trend.In this period,a slight decrease in annual precipitation is anticipated in regions like Shanghai,southern Anhui,and Jiangsu,while areas such as Zhejiang,Fujian,southern Hubei,and Jiangxi are projected to experience increased precipitation.By the mid-21 st century(2049–2074),most regions are projected to face more frequent extreme precipitation events.Compared to the recent 26-year epoch(1990–2015),the frequency and magnitude of extreme precipitation would significantly increase by 21.9–48.1% and 12.3–38.3%,respectively by the end of the 21 st century(2075–2100),under the worst emission scenario(SSP585).In particular,the southern coastal areas are projected to receive more extreme precipitation than the northern regions.