Sensitivity of lake‒atmosphere interaction on water clarity

Lakes impact regional weather in lake-abundant areas and consequently lake subroutines have recently been implemented in numerical weather prediction (NWP) models, leading to an improved forecast skill. Global data on lake coverage and depth are essential for lake subroutines and these data have recently been created. Another essential external parameter is water clarity, for which global data are not yet available and a global constant value is used. Clarity is described in lake models with a diffuse light extinction coefficient (K_d), which determines how the received radiative energy is distributed within the lake water body as a function of depth. Its value influences directly the lake‒atmosphere interaction through changes in the surface temperature and heat storage in lakes: a low-clarity lake (high K_d) acts partly as a shallow lake since solar heat is immediately stored only into the top layer of thickness ~ K_d^–1.

The aim of this work is to test the sensitivity of two 1-dimensional lake models on K_d. The chosen models are a multilayer model LAKE and the FLake model which has a parameterized temperature profile. Measurements from Lake Kuivajärvi, Finland, are used. The suite of instrumentation provides data on the full surface energy budget (including turbulent fluxes of sensible and latent heat, using the eddy-covariance technique, and radiation budget), water temperature profile, K_d and basic meteorology. The models are able to simulate the water temperature profile in addition to surface fluxes. The models are run for a full year with a constant time-independent K_d which is varied in a realistic range from run to run. In addition, model runs are made using direct measurements of K_d providing a time-dependent evolution throughout the year. The turbulent fluxes and the water temperature profile are studied, and recommendations based on this sensitivity analysis are given: Is water clarity an important variable in lake‒atmosphere interaction modeling and should a global database of it be included in NWP? Is the seasonal variation of K_d substantial enough that a constant value in time cannot be used?