Study Of Stochastic Parameterization Perturbation And Probability Matching Methods In Convective-scale Ensemble Forecasts

The scales of severe convections are often small and the physical processes in clould are complex and not fully understood,resulting in the uncertainties in model forecasts.This work concerned the error representations of physical parameterizations and post-process in convective-scale ensemble forecasts(CEFs).The parameterization errors were represented by using the stochastic parameterization perturbation method which perturbs either the parameterization coefficients(SPP)or the tendencies(SPPT).The stochastic approaches use multiplicative factors to represent the model parameterization uncertainties.The perturbations of stochastic approaches vary spatially and temporally and follow several preset spatiotemporal scales,so these perturbations have the capability to represent the spatiotemporal variety of parameterization uncertainties.Considering the features of the uncertainties of physical parameterizations in CEFs,this study proposed several approaches that perturbed the diffusion term of numerical forecast model(SPPTD,D for diffusion)and the microphysics parameterization(SPTTM/SPIPM,TT for temperature tendency and IP for intercept parameter),using the SPP method.Meanwhile,this work designed a spatial localization(horizontally and vertically)scheme to constraint the influence ranges of perturbations for numerical integration stability and constructed a resampling function to transform the Gaussian perturbations to an asymmetric distribution with respect to 1.0.Using the above new approaches,the forecast skills that include the root-mean-square error,ensemble spread,and fraction skill score were evaluated;the relationship between perturbations and the supercell associated mesocyclone intensity and structure were analyzed;the sensitivities of ensemble forecasts to the spatiotemporal decorrelation coefficients of SPP/SPPT approach were discussed.The ensemble forecasts often provide a huge of information that sometimes needs to be concentrated into a deteministtic product for forecast guidance.This study proposed a probability matching(PM)scheme using the ensemble maxima(En Max)as the reference field,which was called PM max.The different performances of PM Max and the scheme using ensemble mean(PM mean)were evaluated and compared with numerous amount of cases.The causes of the above difference were also elucidated.Using idealized and real cases,several conclusions were drawn.(1)The initial condition perturbations cannot represent the error that underestimated the mesocyclone associated with supercell storm caused by the insufficient model resolution,while the SPPTD was able to alleviate the underestimate issue and to intensify the mesocyclone intensity for some members.The ensemble spread and representation of mesocyclone in terms of updraft helicity were also improved.(2)When the SPPTD was enabled,the intensifying of mesocyclone often accompanied by the negative perturbations around the vortex,which corresponds to the short-wave energy intensification and indicates that it is reasonable to moderately intensify the short wave if it is physical.(3)For microphysics scheme,using barotropic temperature tendency perturbations led to synchronous intensifying(positive perturbations)or weakening(negative perturbations)the middle-level heating and low-level cooling;the positive perturbations intensifies the cold pool,which may be conducive to the short-duration strong vortex,while the negative perturbations weaken the convective activity so that substantially reduces the chance for generating strong vortex.(4)As to the intercept parameters,the positive perturbations produced more small precipitation particles that enhanced the low-level evaporation and led to stronger low-level cold pool,which may favor the short-duration vortex;by contrast,the negative perturbations enlarged the mean drop-size,which weakened the evaporation and cold pool and benefited the long-duration strong vortex.(5)In real tornadic supercell cases,the combination of SPPTD and SPIPM improved the forecast skill of mesocyclone and outperformed the CEFs only using SPPTD or SPIPM.This is because in the real cases using only SPPTD cannot address the cold pool bias caused by microphysical parameterization while using only SPIPM cannot allow for the uncertainty due to model resolution.(6)Small size ensemble often suffers from diversity deficiency;one of the causes is the spatial distribution errors which often occur in the tornadic associated vortrics forecast.The PM max was able to obtain a higher forecast skill score than the PM mean for large prognostic variables;the localized PM max outperformed its ensemble mean counterpart at all thresholds of prognostic variables.