Quantifying the Relationship between Storm Size and Forecast Uncertainty for Enhanced Blending

Short term prediction of convective storms and their macroscale properties (e.g., size, aspect ratio) is critical for aviation planning. The CoSPA forecast system was developed to provide high resolution depiction of the timing and locations of Vertically Integrated Liquid (VIL) and echo tops by blending MIT Lincoln Lab extrapolation with High Resolution Rapid Refresh (HRRR) to provide common situational awareness of aviation hazards. The blending algorithm attempts to take advantage of the relative strengths of these forecast inputs. Generally, extrapolation performs best at the shorter time horizons and high resolution models with advanced data assimilation perform better at longer time horizons. However, relative performance is also a function of storm scale. In this study, we explore the relationship between storm scale and skill. Then, we assess how well forecast uncertainty captures the actual outcomes by assessing key parameters of probabilistic forecasts (e.g., reliability, resolution and sharpness). Storm size is obtained using an object identification algorithm on both forecast data and the observations. Forecast uncertainty is obtained from time-lagged ensembles that are available from rapidly updating forecasts that have multiple realizations available for the same forecast time. Relationships that are developed can be used to both develop improved blending techniques and to better understand uncertainty of high resolution model forecasts in the context of convective storm size.

Disclaimer: This research is in response to requirements and funding by the Federal Aviation Administration (FAA). The views expressed are those of the authors and do not necessarily represent the official policy or position of the FAA.