How well do RCMs and ESDs reproduce the occurrence of extreme precipitation events over Southeastern South America?

SOLMAN, S.; BETTOLLI MARIA LAURA; DOYLE, M; BLAZQUEZ, J; FEIJOO, M; OLMO, MATIAS; DIAZ, G; BALMACEDA HUARTE, R; POGGI, M MERCEDES

Beijing

Conferencia; ICRC-CORDEX 2019; 2019

CORDEX

Individual extreme precipitation eventsover Southeastern SouthAmerica(SESA)duringthespringand Summertimeareresponsibleofmorethan40%of thetotalaccumulatedseasonalprecipitation.Theseextremeeventsareassociatedwiththeoccurrenceoforganizedconvectionintheregion.Giventhemesoscale features involved in their development, modelling their main features and lifecycles ischallenging.Inthisworkwegatherdifferentmodellingstrategies,includingseveralEmpiricalStatistical Downscaling(ESD) models,severalCORDEXRegionalClimateModels(RCMs)fortheSouthAmericandomain at various horizontal resolutions and several convective permitting simulations performedwiththeWRFmodelforselectedcasestudiestoassessthecapabilityofdifferentmethodologiesincapturingthespatialdistributionofrainfallduringtheoccurrenceofanextremeevent.Theevaluationofdifferentmethodologiesalsoallowsidentifyingtheircapabilityincapturing(ornot)theassociatedphysicalforcingstriggeringextremeevents.Tenindividualeventswereselectedbasedondatafromthe TRMM dataset and the CPC-Unified gridded dataset for the period 1979-2015 satisfying thefollowingcriteria:dailyprecipitationexceedingthe95thpercentileandwithacoverageofmorethan10%ofgridpointswithinSESA.Duetothelargeobservationaluncertainty,wealsoincludedseveralobservational datasets to characterize the main features of the individual cases evaluated, whichincludestationdata,griddedproducts(CPC-Unifieddata)andseveralprecipitationestimatesbasedonsatellitedata (CHIRPS;MSWEP;TRMM;PERSIANN;CMOPRH).Foreachindividualevent,evaluationRCMsimulations fromtheCORDEXdatabaseat50kmand25kmresolutions;72-hourssimulationsperformed withtheWRFmodeldrivenbyERA-Interimreanalysisatroughly20km,12kmand4km,and several ESD models based on different techniques and different predictor variables werecompared.