Determining lake surface water temperatures worldwide using a tuned one-dimensional lake model (FLake, v1)
Abstract. A tuning method for FLake, a one-dimensional (1-D) freshwater lake model, is applied for the individual tuning of 244 globally distributed large lakes using observed lake surface water temperatures (LSWTs) derived from along-track scanning radiometers (ATSRs). The model, which was tuned using only three lake properties (lake depth, snow and ice albedo and light extinction coefficient), substantially improves the measured mean differences in various features of the LSWT annual cycle, including the LSWTs of saline and high altitude lakes, when compared to the observed LSWTs. Lakes whose lake-mean LSWT persists below 1 °C for part of the annual cycle are considered to be seasonally ice-covered. For trial seasonally ice-covered lakes (21 lakes), the daily mean and standard deviation (2σ) of absolute differences between the modelled and observed LSWTs are reduced from 3.07 °C ± 2.25 °C to 0.84 °C ± 0.51 °C by tuning the model. For all other trial lakes (14 non-ice-covered lakes), the improvement is from 3.55 °C ± 3.20 °C to 0.96 °C ± 0.63 °C. The post tuning results for the 35 trial lakes (21 seasonally ice-covered lakes and 14 non-ice-covered lakes) are highly representative of the post-tuning results of the 244 lakes.
For the 21 seasonally ice-covered lakes, the modelled response of the summer LSWTs to changes in snow and ice albedo is found to be statistically related to lake depth and latitude, which together explain 0.50 (R2adj, p = 0.001) of the inter-lake variance in summer LSWTs. Lake depth alone explains 0.35 (p = 0.003) of the variance.
Lake characteristic information (snow and ice albedo and light extinction coefficient) is not available for many lakes. The approach taken to tune the model, bypasses the need to acquire detailed lake characteristic values. Furthermore, the tuned values for lake depth, snow and ice albedo and light extinction coefficient for the 244 lakes provide some guidance on improving FLake LSWT modelling.