Bibliographic details:
Lynn, B. H., Rosenzweig, C., Goldberg, R., Rind, D., Hogrefe, C., Druyan, L., Healy, R., Dudhia, J., Rosenthal, J., & Kinney, P. (2010). Testing GISS-MM5 physics configurations for use in regional impacts studies. Climatic Change, 99(3-4), 567-587.
Abstract:
The Mesoscale Modeling System Version 5 (MM5) was one-way nested to the Goddard Institute for Space Studies global climate model (GISS GCM), which provided the boundary conditions for present (1990s) and future (IPCC SRES A2 scenario, 2050s) five-summer “time-slice” simulations over the continental and eastern United States. Five configurations for planetary boundary layer, cumulus parameterization, and radiation scheme were tested, and one set was selected for use in the New York City Climate and Health Project—a multi-disciplinary study investigating the effects of climate change and land-use change on human health in the New York metropolitan region. Although hourly and daily data were used in the health project, in this paper we focus on long-term current and projected mean climate change. The GISS-MM5 was very sensitive to the choice of cumulus parameterization and planetary boundary layer scheme, leading to significantly different temperature and precipitation outcomes for the 1990s. These differences can be linked to precipitation type (convective vs. non-convective), to their effect on solar radiation received at the ground, and ultimately to surface temperature. The projected changes in climate (2050s minus 1990s) were not as sensitive to choice of model physics combination. The range of the projected surface temperature changes at a given grid point among the model versions was much less than the mean change for all five model configurations, indicating relative consensus for simulating surface temperature changes among the different model projections. The MM5 versions, however, offer less consensus regarding 1990s to 2050s changes in precipitation amounts. All of the projected 2050s temperature changes were found to be significant at the 95th percent confidence interval, while the majority of the precipitation changes were not.