The Effects of Lake Representation on the Regional Hydroclimate in the ECMWF Reanalyses

Lakes are an integral part of the geosphere, but they are challenging to represent in Earth system models, which either exclude lakes or prescribe properties without simulating lake dynamics. In the ECMWF interim reanalysis (ERA-Interim), lakes are represented by prescribing lake surface water temperatures (LSWT) from external data sources, while the newer-generation ERAS introduces the Freshwater Lake (FLake) parameterization scheme to the modeling system with different LSWT assimilation data sources. This study assesses the performance of these two reanalyses over three regions with the largest lakes in the world (Laurentian Great Lakes, African Great Lakes, and Lake Baikal) to understand the effects of their simulation differences on hydrometeorological variables. We find that differences in lake representation alter the associated hydrological and atmospheric processes and can affect regional hydroclimatic assessments. There are prominent differences in LSWT between the two datasets that influence the simulation of lake-effect snowstorms in the Laurentian winters and lake-land-breeze circulation patterns in the African region. Generally, ERAS has warmer LSWT in all three regions for most months (by 2-12 K) and its evaporation rates are up to twice the magnitudes in ERA-Interim. In the Laurentian lakes, ERAS has strong biases in LSWT and evaporation magnitudes. Over Lake Baikal and the African Great Lakes, ERAS LSWT magnitudes are closer to satellite-based datasets, albeit with a warm bias (1-4K), while ERA-Interim underestimates the magnitudes. ERAS also simulates intense precipitation hot spots in lake proximity that are not visible in ERA-Interim and other observation-based datasets. Despite these limitations, ERAS improves the simulation of lake-land circulation patterns across the African Great Lakes.

Automated summary

This study evaluates the performance of two reanalysis models over the world's largest lakes, finding that differences in lake representation affect hydrometeorological variables and can lead to biases in regional hydroclimatic assessments.