15B.4 Impacts of Accounting for Flow-dependent, Interchannel Observation Error Correlations when Assimilating ABI All-sky Radiance Observations for Forecasts of a Severe Convective Event in RRFS

Thursday, 1 February 2024: 2:30 PM
Key 10 (Hilton Baltimore Inner Harbor)
Samuel K. Degelia, University of Oklahoma, Norman, OK; and X. Wang

Assimilating all-sky radiance observations collected by the Advanced Baseline Imager (ABI) aboard GOES-16 is known to significantly improve forecasts of deep convection. Recent works highlight how assimilating multiple water vapor channels from the ABI can further improve results, despite the fact that observation errors from neighboring channels are likely highly correlated. Observation error correlations are typically ignored in most DA systems for computational efficiency and to simplify the implementation of DA routines. Instead, observation error variances are often inflated to account for ignoring the error correlations. This simplification always results in downweighting of observations. However, recent studies reveal that accounting for a fully-correlated observation error covariance (R) during DA can both downweight and upweight observations and has been shown to improve forecast results compared to using only a diagonal R.

In this presentation, we evaluate the impacts of accounting for interchannel observation error correlations when assimilating three ABI water vapor channels for a severe convective event using the RRFS EnVar DA system. This study is the first known work to evaluate the impact of observation error correlations of satellite radiances specifically for convective-scale purposes. We estimate R using the Desroziers et al. (2005) method which has become popular for various applications. Additionally, observation error correlations are likely larger in cloudy conditions compared to clear air given that each band samples similar cloud tops (i.e., anvil clouds). As such, we further compare the impacts of using a correlated but static R to a unique, flow-dependent R based on a function of symmetric cloud affect. Initial results show that this flow-dependent and correlated R yields consistently higher forecast skill for a tornadic supercell event, likely due to improved predictions for the structure of a related cold pool.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner