May 23 - July 29
Anti-Goldilocks Forecasting: Modeling Forecast Probability Going From Small to Large Scales
Joseph Dale (University of Northern Colorado), Dr. Burkely Twiest Gallo (OU/CIWRO & NOAA/SPC), and Dr. Harold Brooks (NOAA/NSSL)
What is already known:
- Gaps currently exist between convective forecast scales (i.e., watch to warning scales)
- Attempts are being made to bridge these scales with consistent probabilities
- There are specific issues in areal coverage probabilities in the watch to warning space
What this study adds:
- A method to understand areal coverage along time and space scales
- This method may provide a diagnostic tool for looking at forecast model behavior
- Forecast interpretation differs between long and short time and space scales
- A reasoning for why there are issues in probability in the watch to warning scale: Understanding storm organization is crucial between these scales
Abstract:
The Forecasting a Continuum of Environmental Threats (FACETs) project aims to span the range of convective forecast scales with consistent probabilistic forecasts. To investigate how areal coverage probabilities behave across a continuum of space and time, this study analyzes simulations from an idealized model in time and space. Two grids are created. On the first, “events” are randomly placed throughout the grid. On the second, reports are placed with the same overall coverage but organized in time and space (e.g., lines). Aggregation is done over time and space scales on the grids to calculate the coverage probability as the size of the aggregation changes. Dividing the “organized” coverage probability by the “random” yields a u-shaped curve as a function of aggregation size. The depth and location of the minimum of the u-curve is related to the organization of the threat and its underlying coverage on the finest grid. Experiments with this framework - plotting synthetic data and numerical model proxies as organized events - indicate that the location where the u-curve reaches max depth is between the watch and warning time and space scales. These results show that forecast interpretation is different between long and short scales, and that organization of storms strongly influences forecasts in the watch-to-warning space. These results characterize a challenge in creating consistent probabilities across the spectrum of scales, a goal of FACETs.