Toward the simulation of the effects of the Great Lakes on regional climate

Abstract

This paper describes a set of numerical experiments aimed at evaluating the feasibility of applying a version of the National Center for Atmospheric Research-Pennsylvania State University regional model (MM4) to regional climate simulation over the Great Lakes Basin. The objectives of this initial modeling investigation are 1) to examine whether the MM4 can capture the primary forcing exerted by the Great Lakes on the regional climate and 2) to evaluate what model resolution and configuration are needed to simulate such forcing. Simulations over the Great Lakes region are conducted with and without representation of the lakes at four model gridpoint resolutions ranging from 15 to 90 km. One experiment at 60-km resolution is discussed in which a one-dimensional thermal eddy diffusion model is interactively coupled to the MM4 to represent the lakes. Initial and lateral boundary conditions necessary to drive these simulations are provided by European Centre for Medium-Range Weather Forecasts (ECMWF) analyses of observations. All simulations conducted are 10 days in length, from 22 December 1985 to 1 January 1986. When driven with data from ECMWF analyses of observations, the climate version of the MM4 reproduces the basic characteristics of the distribution of lake-effect precipitation over the Great Lakes Basin. Differences between simulations with and without the lakes represented indicate that the lakes accounted for approximately 25% of the precipitation over the basin during the 10-day period simulated. Over localized areas, identified as the major snowbelts downwind from the lakes, lake effects were responsible for 50%–70% of the precipitation. Basinwide precipitation did not vary greatly among the simulations with resolutions of 60, 30, and 15 km, although biases between model results and station observations did decrease slightly with increasing model resolution. Basinwide maximum and minimum temperature biases decreased more markedly with finer resolution. In the snowbelt regions downwind from the lakes, precipitation was underforecast at all four model resolutions, but precipitation generally increased with finer resolution. Differences between the results from the simulations at the three finest resolutions were greater over snowbelt regions than over the basin as a whole. A simulation was conducted with the MM4 coupled to a lake model in an interactive two-way nested configuration. The implementation of this coupling was accomplished in a straightforward manner, with no model tuning required, and added very little to the computation time needed for the MM4 system. This coupled modeling system was found to produce realistic distributions of lake surface temperatures, evaporation rates, and ice thicknesses across the lakes. In climate simulations where the MM4 is nested in a general circulation model (GCM), we believe that the use of this coupled modeling system is preferable to specifying lake parameters by interpolation from GCM output. The next step in this work is to conduct a simulation of at least one annual cycle over the region to more fully test the coupled MM4-take model system.