Using convection-allowing ensembles to understand the predictability of extreme rainfall
The meteorological community has well established the usefulness of ensemble-based numerical weather prediction for precipitation guidance, since trusting one possible atmospheric solution can lead to, in some cases, particularly bad forecasts for precipitation guidance, owing to inherent uncertainties in precipitation processes that make deterministic prediction impractical. However, continued predictive challenges associated with intense convective rainfall has led to an increasing need to determine the most effective use of these ensemble systems in high impact, extreme precipitation events. Further, it cannot be assumed that ensembles will evolve similarly in both extreme precipitation and more benign events, due to the importance and error growth associated with convective-scale motions. This error growth associated with the chaotic nature of moist convective dynamics can also serve to limit the predictability of an extreme rainfall event (known as intrinsic predictability), in addition to predictability limits imposed by deficiencies in observing systems and numerical models (known as practical predictability). This research will focus on using a recently developed, operationally based ensemble dataset, specifically the National Oceanic and Atmospheric Administration's (NOAA) Second Generation Global Medium-Range Ensemble Reforecast Dataset (Reforecast-2), to create downscaled ensemble reforecasts of the extreme precipitation events. Some events examined during the course of this research are the inland movement of tropical storm Erin in 2007 and flooding associated with mesoscale convective vortices in Arkansas in 2010 and San Antonio, Texas in 2013. The global reforecasts are used to force an ensemble of convection-allowing WRF-ARW numerical simulations for the purpose of evaluating ensemble-based precipitation forecasts associated with specific extreme rainfall events. Using these ensemble forecasts, we address several questions related to the practical versus the intrinsic predictability of the extreme rainfall events examined. Experiments that vary the magnitude of the perturbations to the initial and lateral boundary conditions (ICs and LBCs) reveal a seemingly proportional scaling of ensemble spread early in the simulations associated with the magnitude of the perturbation, but this scaling is not maintained throughout the simulations. Additionally, a diurnal cycle in ensemble spread growth is observed with large growth associated with afternoon convection, but the growth rate then reduced after convection dissipates the next morning rather than continuing to grow. The specific characteristics of the diurnal cycle, however, vary based upon region and flow regime. Lastly, the ensemble spread was found to be influenced by the size of the IC perturbations out to at least 48 hours. These spread evolution characteristics speak to the viability of running convection-allowing ensembles for prediction on multi-day timescales, since no saturation of the ensemble spread is seen despite extreme precipitation within the modeled time period. In addition to the overall ensemble characteristics, terrain-induced precipitation variability associated with the terrain feature known as the Balcones Escarpment, located in central Texas, is analyzed in multiple instances of heavy rainfall in San Antonio and the surrounding area. Simulations in which the Balcones Escarpment is removed reveal that when the synoptic to mesoscale forcing for heavy rainfall are in place over the Balcones Escarpment, the terrain does not directly affect the occurrence or magnitude of precipitation. It does affect the spatial distribution of the precipitation in a small but consistent way. This shift in precipitation associated with removing the Balcones Escarpment, when compared to a WRF-ARW ensemble for the event, is smaller than shifts associated with typical atmospheric variability. The combined results of these experiments demonstrate that downscaled ensemble NWP systems using readily available global ensemble forecasts can faithfully represent previously unresolved mesoscale features, precipitation totals, and depict ensemble-spread characteristics associated with moist convection.
Includes bibliographical references.
Includes bibliographical references.
numerical weather prediction