Modeling Large- and Small-Lake Temperature and Ice Evolution in the RAP/HRRR Models

An accurate representation of the evolution of near-surface water temperature and surface ice coverage is an important component of short-term numerical weather prediction near lakes. In this presentation, we describe recent progress towards modeling the evolution of both large, deep lakes (such as the Great Lakes), as well as smaller, shallower lakes (for example, Lake Champlain in the northeastern United States), in a computationally efficient way within the framework of the experimental Rapid Refresh (RAPX) and High-Resolution Rapid Refresh (HRRRX) forecast models used within NOAA.

For large lakes, the RAPX and HRRRX take somewhat different approaches to specifying the lower boundary conditions. The RAPX uses lake temperature conditions provided by the global RTG dataset, and ice coverage information provided by the IMS snow dataset. The HRRRX, on the other hand, uses real-time model predictions of lake surface water temperature and fractional lake ice coverage from the real-time experimental FVCOM lake forecast model developed and run by the NOAA Great Lakes Environmental Research Laboratory (GLERL). These lake model forecasts are not subject to the same cloudiness limitations of real-time satellite-derived lake datasets, and allow for much more accurate weather predictions in regions near the Great Lakes. The HRRRX has been using FVCOM lake temperature and ice predictions since early 2019.

For small lakes, both the RAPX and HRRRX now use WRF’s Community Land Model (CLM) lake model to predict the evolution of temperature and ice coverage. The RAPX and HRRRX model/assimilation cycles have been using the CLM lake model since late 2018, resulting in improvements to local weather forecasts in the vicinity of small lakes. Most important, initialization of the CLM lake model for HRRRX and RAPX has been accomplished quite accurately by cycling the lake model’s 10-layer temperature and ice fields through the hourly HRRRX and RAPX data assimilation cycle over this nearly 1-year period. Atmospheric forcing (temperature, wind, radiation) of the CLM lake model is closely constrained by the HRRRX and RAPX hourly assimilation of near-surface, radar, satellite, and other observations.

The use of the WRF CLM lake model and the FVCOM datasets is planned for implementation in the next operational versions of the RAP and HRRR, slated for 2020.