May 22 - July 28
Analyzing the Performance of Different Parameter Settings with the Ensemble Nowcasting of Tropical Cyclone Precipitation
Steven Rearden, Steven Martinaitis, Jackson Anthony, and Dean Meyer
What is already known:
- Flash floods are a common, yet deadly hazard associated with tropical cyclones.
- Efforts to increase lead times in flash flood warnings have slowed considerably over the past decade.
- A new, parameter-based method of nowcasting is being tested using MRMS data in the STEPS framework with the goal of increasing flash flood warning lead times.
- Hurricanes Ian and Henri had substantially different precipitation structures and characteristics that allow for comprehensive case study analyses.
What this study adds:
- Differences in advection tracking thresholds lead to more significant changes in performance.
- Tropical cyclones with varying precipitation structures/intensities require different advection tracking thresholds for the best performance.
- Differences in number of ensemble members and seed values lead to insignificant changes in performance.
Abstract:
Flash flooding is a dangerous weather hazard that can result in substantial property damage and loss of life if the public is not prepared. Because of this, continuous advancements dedicated to extending the lead times of flash flood warnings is of great importance. To allow for new progress in this, a new precipitation nowcasting scheme which uses data from the MRMS system was tested in the STEPS framework. This study focused on nowcasting precipitation with tropical cyclones. Hurricanes Ian (2022) and Henri (2021) were chosen as case studies due to their differing precipitation structures. Varying parameter value combinations were used to analyze nowcast performance. These parameters include the advection tracking threshold, number of ensemble members, and seed value for ensemble perturbations. Results indicate that differences in advection tracking threshold caused the most significant changes in performance, while differences in number of ensemble members and seed value cause generally insignificant performance differences. Performance when the seed value was randomized was also found to have a low variability, leading to the conclusion that the particular seed value is generally insignificant in nowcast performance. Analysis of both cases revealed that the best performance for each tropical cyclone utilized different advection tracking thresholds. This underscores the challenge of finding the right parameters to use within the MRMS system.