Prediction Technology Of Rainstorm And Extreme Temperature And Spatial Group-Occurring Characteristics Of Their Complex Events

Extreme temperature events have become one of the most widespread and severe natural disasters that affect not only China but also the world,while rainstorm and flood disasters are the natural disasters with the highest and fastest increasing frequency and the largest secondary impact globally.Both pose a huge threat to human society and the balance of natural ecosystems.Furthermore,extreme compound events have stronger disaster-causing properties compared to single extreme climate events such as rainstorm and extreme temperatures.Therefore,exploring the prediction technology of rainstorm and extreme temperature events and their spatial group-occurring characteristics can play a critical role in disaster monitoring and warning.This article extracted four monthly rainstorm indices and six monthly extreme temperature indices（ETIs）from daily meteorological data from 639 stations in China from 1961 to 2021.The study analyzed the spatial and temporal distribution characteristics of rainstorm and extreme temperature events in seven sub-regions of China.Multi-collinearity analysis was used to remove dependent circulation indices,and Pearson correlation analysis was conducted with lag consideration to screen out the key circulation indices and optimal lag time for each rainstorm index and ETI.Then,using the key circulation indices with optimal lag month,Stepwise Multiple Linear Regression（SMLR）and Random Forest（RF）models were established to simulate rainstorm indices and ETIs,respectively.The performance of the two models was compared and evaluated,and the variable importance of circulation indices for each station and sub-region was revealed.The better-performing RF model was used to predict the ETIs and rainstorm indices for May 2022 in China.In addition,this study used the SRTMDEM90 digital elevation model to extract and calculate nine terrain factors for 639stations in China to explore the impact of terrain factors on rainstorm indices.The influence of terrain factors on the R² of rainstorm index models was analyzed.Finally,the response relationship between rainstorm and extreme temperature events was analyzed,and two types of rainstorm and extreme temperature compound events,namely high-temperature rainstorm compound events and low-temperature rainstorm compound events,were defined.The spatial and temporal distribution characteristics and changing trends of rainstorm and extreme temperature compound events were revealed,and the seasonal and spatial group-occurring characteristics of rainstorm and extreme temperature compound events were explored using regional indices.The main results are summarized.（1）Rainstorm was affected by terrain factors and various circulation indices.This study analyzed four rainstorm indices from 1961 to 2021 in seven sub-regions of China.The results showed that these indices fluctuated around their multi-year averages in the different sub-regions.Spatially,rainstorm indices gradually increased from northwest to southeast,and extraordinary rainstorm was concentrated in sub-region VII（humid tropical region of South China）.Terrain factors had varying effects on rainstorm indices across the seven sub-regions,with differences in the number and correlation coefficient r of key terrain factors.Furthermore,different terrain factors had varying impacts on rainstorm indices in the various sub-regions,with elevation being a key terrain factor in six sub-regions.Circulation indices also had a significant impact on extreme rainfall events,with a lag and apparent periodicity.Most circulation indices having the weakest impact on sub-region I（Northwest desert region）and the strongest on sub-region VII.NANRP（North Africa-North Atlantic-North America Ridge Position Index）was found to be the most influential circulation index for the four rainstorm indices.（2）Terrain factors had certain influence on the R² of rainstorm prediction model.This study examines the accuracy of two models,SMLR and RF,in simulating rainstorm indices.During the validation period,the R² values for SMLR and RF models ranged from 0.01 to 0.54and 0.01 to 0.70,respectively.Comparing the two models,the RF model had better R² and LCCC values during the validation period,but the opposite was observed during the calibration period.The MAE and RMSE performances were similar between the two models.Overall,the RF model had better performance in simulating rainstorm indices.Based on the RF model,the range of changes in rainstorm indices for May 2022 in China were as follows:rainstorm days（0-2.1 d）,rainstorm rainfall（0-101.3 mm）,rainstorm intensity(0-44.2 mm·d^-1),and rainstorm contribution rate（0-50%）.Additionally,in sub-region V（humid semi-humid temperate region of North China）,the Pearson correlation coefficient|r|between terrain factors and the rainstorm index model R² was the highest,while it was the lowest in sub-region VI（humid subtropical region of central and southern China）.Overall,the R² of rainstorm index models based on atmospheric circulation in China showed a significant positive correlation with longitude,and a significant negative correlation with latitude,altitude,and the relative height difference within a 15 km buffer radius（RHD15）.（3）The simulation effect of the RF model on ETIs was better than that of the SMLR model.From 1961 to 2021,the frequency and intensity of extreme high-temperature events in China showed a significant overall increase,while extreme low-temperature events showed a downward trend.Spatially,the intensity and frequency of extreme high-temperature events showed a stepwise increasing distribution trend from northwest to southeast,which was basically opposite to that of extreme low-temperature events.Extreme temperature events are influenced by various circulation indices,and there is a lag and obvious periodicity in the effect of circulation indices on ETIs.The effect of the same circulation index on ETIs varies in different sub-regions,with most circulation indices having the strongest impact on FD0 in sub-region II（grassland region of Inner Mongolia）and the weakest impact on sub-region VII.However,the determining circulation indices for the same ETI are basically the same in all sub-regions,and PPVI（Pacific Polar Vortex Intensity Index）is the most important circulation index affecting TXx,TD30,and FD0;NANRP has the greatest impact on TX90p and TN10p;and SSRP（South China Sea Subtropical High Ridge Position Index）is the main circulation index affecting TNn.The SMLR and RF models have good prediction ability for ETIs,with validation period R² ranging from 0.59 to 0.91 and 0.66 to 0.92,respectively.Overall,the simulation effect of the RF model is better than that of the SMLR model,and therefore it is selected to predict ETIs at 639 stations in China.According to the results of the prediction model,the ETIs in May2022 differed in the seven sub-regions,but were similar to the annual average situation.（4）The compound events of rainstorm and extreme temperature were characterized by spatial clustering.The response relationship between rainstorm and extreme high temperature events is significant,with a probability of 0-53%of extreme high temperature events occurring under rainstorm conditions,but the response between rainstorm and extreme low temperature events is weaker for most stations.From 1961 to 2021,the cumulative frequency of high-temperature rainstorm events in China showed a fluctuating increasing trend.The frequency of high-temperature rainstorm events in the Sub-region VII was significantly higher than that in other sub-regions.The cumulative frequency of low-temperature rainstorm events showed a fluctuating decreasing trend.Most stations had never experienced low-temperature rainstorm events,and their frequent occurrence was mainly in the eastern coastal areas.The high-temperature rainstorm events mainly occurred in summer,while the low-temperature rainstorm events mainly occurred in spring and winter.The compound rainstorm and extreme temperature events have spatial group-occurring characteristics.The regional index of high-temperature rainstorm events was highest in Sub-region VI,indicating that more stations experienced high-temperature rainstorm events simultaneously in this sub-region,and its impact was larger.The regional index of low-temperature rainstorm events was generally low,possibly due to the overall low frequency of low-temperature rainstorm events.