Jason Chiappa - February 18

Convective Meteorology (Mesoscale Dynamics) Investigations into the Poor Representation of the 10 August 2020 Midwest Derecho in Operational Numerical Models Jason Chiappa Friday, February 18 03:30 PM Join Google Meet The 10 August 2020 Midwest Derecho was the costliest thunderstorm disaster in modern US history, with damage estimates of ~$11.8 billion.

Start

February 18, 2022 - 3:30 pm

End

February 18, 2022 - 4:30 pm

Convective Meteorology (Mesoscale Dynamics)

Investigations into the Poor Representation of the 10 August 2020 Midwest Derecho in Operational Numerical Models

Jason Chiappa

Friday, February 18

03:30 PM

Join Google Meet

The 10 August 2020 Midwest Derecho was the costliest thunderstorm disaster in modern US history, with damage estimates of ~$11.8 billion. However, most operational model forecasts failed to accurately represent the convection, contributing to a maximum lead time of only about 3 hours before significant wind gusts began to occur. Given that an increase in extreme weather events is likely to occur with climate change, understanding and eliminating the causes for such poor forecasts of extreme convection events are critical societal needs.

Our initial approach was to examine the forecast skill from both convective allowing models (CAMs) and global modeling systems, with a focus on the European Center for Medium Range Forecasts (ECMWF) deterministic and ensemble modeling systems, as it performed the best out of the global models. The hourly High Resolution Rapid Refresh (HRRR) analysis was used as a best guess of environmental conditions and was compared to the ECMWF deterministic model for diagnosing the causes behind its lack of representing the convection. The examination of the ECMWF ensembles included two approaches: i) assessing the ECMWF Extreme Forecast Index (EFI), which employs the deviation of the ECMWF ensembles from the model climatology for parameters likely to be associated with extreme events (e.g., CAPE, vertical shear); ii) comparing the environments between composites of the top 5 “best” and the top 5 “worst” performing members.

Elevated convective initiation (CI) took place about 200-300 km behind a surface cold front around 6-8 UTC in southern South Dakota within a zone of 700 hPa convergence and moisture advection. Most CAMs underrepresented CI and overrepresented earlier surface-based convection, precluding convection the next day. The ECMWF deterministic model underrepresented low-mid level moisture and 700 hPa convergence, likely contributing to the lack of CI in the model. In contrast, the statistical approach of the ECMWF EFI using the CAPE/shear parameters indicated a favorable environment for a significant convective event in this region over the 24-hour forecast period up to 7 days in advance, but the ECMWF ensembles still had poor skill in representing CI in a rapidly evolving, post-frontal environment. The broader implications of this study include that the accurate prediction of CI and especially elevated nocturnal CI remains a major challenge for forecast models, but statistical-based approaches like the EFI that do not use model precipitation output can potentially aid in forecasting extreme convection events with extended lead time.