الوصف: |
Extremely intensive hourly precipitation is one of the major cause of floods. Hence, it is necessary to investigate hourly precipitation extremes to be proactive at flood risk management. This paper applys Metastatistical Extreme Value (MEV) for estimation of current and future 1 in 10 to 100 year return values of hourly precipitation from short records with the duration around 10 years. Extreme Value Theory (EVT), including Generalized Extreme Value (GEV) and Peak-Over-Threshold (POT), has been prevalent in hydrology frequency analysis since last century. However, the asymptotic assumption of EVT and very often not-large-enough data records in practical cases limit the application and development of EVT. MEV is a novel method in extreme value analysis by considering ordinary events as well as the extremes. Compared with traditional EVT, MEV is a non-asymptotic approach and it estimates high quantiles by considering all or most of independent ordinary events in the process of estimation return values. In this paper, MEV is applied to three different datasets to investigate the mechanism of MEV from various aspects and to assess the performance under different thresholds. The first is KNMI dataset. With up to 95 years of historical hourly rainfall observations in the Netherlands, it is used to analyze the influence of sample data and thresholds on the MEV estimates, as well as for error analysis. Subsequently, a DWD dataset, with 1035 station observations covering Germany, is used for the analysis of the spatial distribution of the hourly rainfall extremes. Finally, an EUCP dataset of kilometre-scale gridded data covering Europe generated by the UKMO Connective Permitting Regional Climate Model (CPRCM) is used. It includes an evaluation run forced by ERA interim reanalysis, a historical run and two future runs downscaled from CMIP5 GCM. The evaluation run is also used for spatial analysis to support the applicability of MEV to climate model simulation data. The historical run and two future runs were used to investigate the mid-century and end-of-century variability of 10-year, 50-year and 100-year hourly precipitation. MEV, with 75 percentile as threshold, is more advantageous than traditional extreme value analysis methods in the case of small sample data. This is reflected in the fact that MEV has smaller errors, less uncertainty and better spatial expressiveness. Applying MEV to the EUCP dataset provides a glimpse of the variation of hourly rainfall extremes in the future: by mid-century, the increase of hourly rainfall extremes is not significant, while the second half of the century faces a significant increase. This increase is also related to the topography, with more pronounced in the Alpine and Mediterranean coastal regions. Izjemno intenzivne urne padavine so eden glavnih vzrokov za poplave. Zato je treba raziskati urne ekstremne količine padavin, da bi bili proaktivni pri obvladovanju poplavne ogroženosti. Ta članek uporablja metastatistično ekstremno vrednost (MEV) za ocenjevanje sedanjih in prihodnjih urnih padavin s povratno vrednostjo 1 na 10 do 100 let iz kratkih zapisov, ki trajajo približno 10 let. Teorija ekstremnih vrednosti (EVT), vključno s posplošeno ekstremno vrednostjo (GEV) in teorijo vršnega praga (POT), je v hidrološki frekvenčni analizi razširjena že od prejšnjega stoletja. Vendar asimptotska predpostavka EVT in zelo pogosto premajhna količina podatkov v praktičnih primerih omejujeta uporabo in razvoj EVT. MEV je nova metoda v analizi ekstremnih vrednosti, saj upošteva tako običajne dogodke kot tudi ekstreme. V primerjavi s tradicionalno EVT je MEV neasimptotski pristop in ocenjuje visoke kvantile z upoštevanjem vseh ali večine neodvisnih običajnih dogodkov v procesu ocenjevanja vrednosti donosa. V tem članku je MEV uporabljen za tri različne nabore podatkov, da bi raziskali mehanizem MEV z različnih vidikov in ocenili učinkovitost pri različnih pragovih. Prvi je nabor podatkov KNMI. Ta vsebuje 95 let zgodovinskih urnih opazovanj padavin na Nizozemskem in se uporablja za analizo vpliva vzorčnih podatkov in pragov na ocene MEV ter za analizo napak. Nato se za analizo prostorske porazdelitve ekstremnih urnih padavin uporabi podatkovni niz DWD z 1035 opazovanji postaj, ki pokriva Nemčijo. Nazadnje je uporabljen nabor podatkov EUCP z mrežnimi podatki v kilometrskem merilu, ki pokrivajo Evropo in jih je ustvaril regionalni podnebni model UKMO Connective Permitting Regional Climate Model (CPRCM). Vključuje ocenjevalno izvedbo, ki jo je izsilila vmesna reanaliza ERA, zgodovinsko izvedbo in dve prihodnji izvedbi, znižani na podlagi modela CMIP5 GCM. Ocenjevalni potek se uporablja tudi za prostorsko analizo, ki podpira uporabnost MEV za simulacijske podatke podnebnih modelov. Zgodovinski potek in dva prihodnja poteka so bili uporabljeni za preučevanje spremenljivosti 10-letnih, 50-letnih in 100-letnih urnih padavin sredi stoletja in ob koncu stoletja. MEV s 75 percentilom kot pragom je v primeru majhnih vzorčnih podatkov bolj ugodna od tradicionalnih metod analize ekstremnih vrednosti. To se kaže v dejstvu, da ima MEV manjše napake, manjšo negotovost in boljšo prostorsko izraznost. Uporaba MEV za podatkovno zbirko EUCP omogoča vpogled v spreminjanje urnih ekstremnih vrednosti padavin v prihodnosti: do sredine stoletja povečanje urnih ekstremnih vrednosti padavin ni znatno, medtem ko se v drugi polovici stoletja znatno poveča. To povečanje je povezano tudi s topografijo in je izrazitejše v alpskih in sredozemskih obalnih regijah. |