Analysis of the Predictability of Warm-Season Derechos Using a Convection-Allowing MPAS Ensemble

This study investigates the predictability of warm-season (May-August) progressive derechos using convection-allowing ensemble (CAE) forecasts from the Model for Prediction Across Scales (MPAS). The main goal is to verify if CAE forecasts can increase forecaster awareness of the potential for a derecho in the medium (3-5 days before) and short (1-2 days before) range, and what synoptic and mesoscale processes are associated with varying levels of predictability.

Warm-season progressive derechos in the 2005-2020 period are identified using severe storm reports and radar imagery. The Storm Prediction Center Convective Outlooks from five days before (D5) to the day of the event (D1) are used to assess the operational predictability of severe weather in general and derechos, stratified into categories of low, moderate, and high predictability.

MPAS CAE forecasts were run for 24 events (12 low and 12 high/moderate predictability) with initial conditions from the Global Ensemble Forecast System. The MPAS mesh has 60 km horizontal spacing over the globe decreasing to 3 km in an area centered in the derecho. Ten members were initialized at 0000 UTC at 24h intervals from D5 to D1 for each event. The convective evolution in each MPAS member was subjectively classified into six categories ranging from one dominant, progressive mesoscale convective system (MCS; more successful forecast) to the lack of organized storms in the area of derecho occurrence (less successful).

For high/moderate predictability cases, D1 and D2 forecasts by the MPAS CAE have 80-90% of the members indicating a dominant bow echo MCS (Fig. 1). D3-D5 forecasts have at least 20% of the members indicating a dominant bow echo MCS. For low predictability derechos, less than 60% of the members are able to produce a bow echo MCS on D1 forecasts and this percentage reduces significantly on D2-D5. More predictable derechos are typically associated with better-defined, higher amplitude troughs that are more accurately forecast by the CAE, whereas low predictability derechos tend to be associated with poorly resolved, transient short-wave troughs. Storm-scale processes appear to be larger factors in low predictability derechos.

Figure 1: Percentage of members in each convective class for initializations from D1 to D5 for low and high/moderate predictability cases.