Mansingh, Ariana2025-12-192025-12-192025-12-192025-12-10https://hdl.handle.net/10012/22773Lake ice is an important climatic indicator because it responds to and alters surface energy fluxes, which in turn influence weather and climate (e.g., precipitation, air temperature). Therefore, accurately representing lake-atmosphere interactions within climate and weather models has been shown to reduce forecast errors. Despite the recognized importance of lake ice processes within models, few studies have evaluated the quality of the lake ice-related variables available from ERA5-Land, a widely used reanalysis product. To address this gap, this thesis evaluates ERA5-Land’s lake ice estimates against satellite-derived observations from NOAA, CIS, IMS and MODIS. It assesses biases in lake ice fraction, timing and surface temperature across seven Canadian lakes over 20 years (2004-2023). The study lakes are grouped into northern lakes (Great Bear, Great Slave, Athabasca and Winnipeg) and the Laurentian Great Lakes (Superior, Huron and Erie). The biases were quantified using mean bias error (MBE) and mean absolute error (MAE), with ERA5-Land treated as “predicted” values and satellite-derived products as “observed” values. This thesis provides one of the first spatial and temporally extensive evaluations of ERA5-Land’s lake ice estimates. The overall findings show that ERA5-Land consistently overestimated lake ice fraction during freeze-up and break-up. Across all lakes, ERA5-Land generally produced earlier freeze-up timing, earlier break-up start, and later break-up end than observations. While timing bias followed similar patterns across lakes, distinct patterns emerged between the northern lakes and the Laurentian Great Lakes. These patterns indicate broader weaknesses of ERA5-Land, which uses the Freshwater Lake model (FLake) as its lake parameterization scheme, notably the omission of snow cover over lake ice and the tendency of the product to form a full ice cover on the Laurentian Great Lakes, which typically experience 40-80% ice fraction at winter maximum. As a result, high MAE was observed during both freeze-up (max=47%) and break-up (max=62%). Consequently, this overestimation of ice cover in ERA5-Land typically led to a daytime cold bias in surface temperatures during the ice-covered period. Additionally, it contributed to ice timing biases, notably the freeze-up end (~22 days earlier on average) and the break-up start (~25-28 days later on average). The lack of consideration of snow cover on ice in FLake/ERA5-Land prevents accounting for its effects on ice growth, heat absorption, and surface temperature. It is well known that snow slows ice growth by reducing downward heat transfer. Thus, ERA5-Land overestimates lake ice thickness across all lakes, which in turn contributes to the delay in break-up start. The early break-up end timing is likely linked to ERA5-Land neglecting the cooling effect of snow’s high albedo, which would otherwise slow melt by reflecting solar radiation. Additionally, during months of high snowfall, ERA5-Land produced a larger warm surface temperature bias, most notable among the northern lakes. Overall, this thesis quantified the biases arising from two notable weaknesses in ERA5-Land’s parameterization: the omission of snow cover on ice and the inability to account for partial ice coverage in the large lakes examined in this research. The quantification of biases provides insight into how these weaknesses skew ERA5-Land estimates. This thesis provides new insights into the limitations of ERA5-Land regarding lake ice, which should be considered leading to for future product releases.enERA5-Landlake iceFLakereanalysis modelsatellitelake ice phenologyMaster Thesis