Introduction of a Small-Lake Submodel in the HRRR and RAP Model to Improve Near-Surface and Cloud/Storm Predictions

Inaccurate representation of the temperature evolution of small lakes during ice-free periods (i.e., in the warm season and transition seasons) has been identified as a contributor to large 2m temperature forecast errors in the 13km RAP and 3km HRRR, as well as other NCEP modeling systems. The NOAA daily High-Resolution Real-Time Global Sea Surface Temperature (RTG_SST_HR) analysis dataset, used in RAP and HRRR, provides surface temperature estimates for small lakes resolved by its land-sea mask. But errors in these estimates could be substantial due to the lack of information. This can, in turn, result in degradation near these lakes for 2m temperature, 2m dewpoint temperature, sensible and latent heat fluxes, cloud, and even convective storm forecasts in some situations.

Our strategy is to apply the CLM (Community Land Model) lake model for small lakes including full hourly cycling, similar to the RAP and HRRR multi-year land-surface cycling constrained by hourly data assimilation. Part of our strategy is to continue using better-sensed temperatures from satellite ( RTG_SST_HR) for large lakes (e.g., Great Lakes and large lakes in Canada such as Great Slave and Great Bear Lake) while using the CLM lake model for temperatures in small lakes.

We will present results on initial cycling in the 13km RAP and 3km HRRR model at the meeting. We will show hoped-for improvements in 2m temperature/dewpoint forecasts from the summer and fall seasons. If successful, as anticipated, this effort on small-lake modeling will be transferred as an option to NGGPS-related FV3 CAM (convection-allowing modelings) and global applications including subseasonal prediction.