An Adaptive Approach for the Calculation of Ensemble Gridpoint Probabilities

Abstract

Traditional ensemble probabilities are computed based on the number of members that exceed a threshold at a given point divided by the total number of members. This approach has been employed for many years in coarse-resolution models. However, convection-permitting ensembles of less than ~20 members are generally underdispersive, and spatial displacement at the gridpoint scale is often large. These issues have motivated the development of spatial filtering and neighborhood postprocessing methods, such as fractional coverage and neighborhood maximum value, which address this spatial uncertainty. Two different fractional coverage approaches for the generation of gridpoint probabilities were evaluated. The first method expands the traditional point probability calculation to cover a 100-km radius around a given point. The second method applies the idea that a uniform radius is not appropriate when there is strong agreement between members. In such cases, the traditional fractional coverage approach can reduce the probabilities for these potentially well-handled events. Therefore, a variable radius approach has been developed based upon ensemble agreement scale similarity criteria. In this method, the radius size ranges from 10 km for member forecasts that are in good agreement (e.g., lake-effect snow, orographic precipitation, very short-term forecasts, etc.) to 100 km when the members are more dissimilar. Results from the application of this adaptive technique for the calculation of point probabilities for precipitation forecasts are presented based upon several months of objective verification and subjective feedback from the 2017 Flash Flood and Intense Rainfall Experiment.

Spatial Predictability of Heavy Rainfall Events in East China and the Application of Spatial-Based Methods of Probabilistic Forecasting

“…In this way, the spatial uncertainty in CSEF probabilistic forecasts was considered to address the double penalty problems (as indicated by Gilleland et al [23], the skill at the grid-point in CSEF tends to be relatively poor) and under-dispersion problems in the CSEF. The third method used was the EAS_NEP [31], a modified version of the NEP. In this method, an adaptive radius acquired from the EAS was applied to each grid to deal with the spurious decay in well-handed events for the NEP.…”

“…That is, if the ensembles at a grid point are in very good agreement, a smaller radius is utilized to calculate the ensemble probability, and vice versa. This method was found to be widely applicable to orographic precipitation, lake-effect snow, or very short-term forecasts [31].…”

“…Subsequently, member-member comparisons were calculated to retrieve the mean value for D i,j . In this study, the pairs for S mm (i.e., the mean agreement scale among ensemble members) was N p = N(N−1) 2 [15,31], where N is the number of ensemble members, and there were N pairs for S mo (i.e., the mean agreement scale between ensemble forecasts and observation). Following Blake et al [31], α was set as 0.5 to meet a moderate difficulty to satisfy a smaller radius.…”

“…In this study, the pairs for S mm (i.e., the mean agreement scale among ensemble members) was N p = N(N−1) 2 [15,31], where N is the number of ensemble members, and there were N pairs for S mo (i.e., the mean agreement scale between ensemble forecasts and observation). Following Blake et al [31], α was set as 0.5 to meet a moderate difficulty to satisfy a smaller radius. Moreover, to meet a potential guidance for the subsequent probabilistic forecast, in this study we directly calculated the EAS for certain incidents (i.e., with certain thresholds).…”

“…However, at later lead times, with the consistent growth of initial error characterized by moist convection and large-scale flow [70], the ensemble members gradually evolve in different directions, leading to lower probabilistic values. Under such circumstances, the effect of added uncertainties extracted with spatial-based methods is clearly reduced [31].…”