We should first recall that this project is not a beauty contest. Each model is developed for a specific purpose. However, a specific model may perform better over a given lake while others do not. The aim is thus to identify the important physical processes for each applications. This section displays by means of table and graphics some of the model simulated outputs.
Sparkling Lake intercomparison experiment
The Sparkling Lake was chosen for the first LakeMIP experiment as a typical representative of relatively shallow lakes located in the temperate latitudes, experiencing long ice cover period. It has been extensively studied experimentally in the framework of North Temperate Lakes Long Term Ecological Research program URL, providing both limnological and meteorological data. This experiment focuses on the simulation of vertical temperature profiles in the lake as well as surface temperature and heat fluxes to the atmosphere. These variables have been selected as being of primary interest in the scope of lake-atmosphere interactions, not only affecting thermodynamic state of water bodies and near water surface air layers, but also as crucial regulators of lake ecological system processes.
Six models participated in the experiment. Among them, there were:
- three models using the k-ε turbulence closure – SIMSTRAT, LAKEoneD, LAKE –, hence expected to rather realistically represent the vertical mixing.
- two one-dimensional models that calculate the heat diffusion coefficient utilizing semi-empirical approaches – Hostetler and MINLAKE96 –.
- the FLake model, one that represents the water column with two layers, a well-mixed layer and the thermocline, the later being described by the self-similarity hypothesis.
The initial conditions (initial vertical profiles), boundary conditions (represented by the meteorological forcing, measured in the near surface layer) and optical parameters of lake water and surface were unified in all models. The experiment covered the period 2002-2005.
The results of the intercomparison (figures 1-5) show that the k-ε models do not demonstrate clear advantages in simulating the thermal regime of this particular lake relative to the models based on simpler approaches. Simstrat and LAKE models met problems in representing lake surface temperature during summertime. From the other hand, the output temperature profiles from FLake model (the “simplest” one) also deviated significantly in some months from measured data, especially during the period of ice cover. However, the correlation of simulated surface temperature with experimental time series is very high for all the models (about 0.99).
The advantages and deficiencies of lake models revealed in Sparkling Lake experiment invite for deeper intercomparison study of turbulence closures and surface flux schemes, utilized in these models. This is planned for modeling studies with other lakes.