The impact of weather model resolution on a coupled model of lake circulation

Numerical coupling between weather and water simulation have shown results consistent with observations in the Great Lakes. The transfer of heat and tracers from onshore to offshore have gained interest in the context of ecosystem and hydrodynamics. However, the dynamics close to the shoreline of these lakes may not be well captured because the weather simulation at the boundary between land and water is usually under-resolved (typical numerical weather simulation resolution is around 9km). Increasing the resolution of the weather numerical model, at the scale of the Great Lakes, requires a prohibitive increase of computing time.

Lake George in upstate New York, is considered a medium-sized lake at 55km by 3km. It represents a potential testbed to assess the effects of resolution of a weather model onto a one-way coupled simulation of the lake's temperature and circulation. We perform two sets of numerical simulations of the lake's circulation, one forcing the lake's circulation with a 3km horizontal resolution weather prediction and the other 0.333km resolution for June 2017. We use the WRF-ARW community model for the weather simulation and the SUNTANS community model for the circulation. The latter has a variable horizontal resolution with an average of 40m.

The comparison against water temperature observations suggests both runs were consistent with observed data. However, the case using the coarser resolution weather model underestimated the depth of a downwelling in the middle of June in the south end of the lake. Moreover, in the northern part of Lake George, the depth of the thermocline is better represented in the case using 0.333km resolution weather, suggesting a difference in the total heat stored in the lake water.

Increasing the resolution of the weather model improved the transfer of energy from the atmosphere into the lake, as suggested by the better consistency in the depth of the downwelling and of the thermocline. We conclude the presentation by discussing the causes of the discrepancies between the two cases.