In this study, we evaluate the ability of GCMs participating in the Intergovernmental Panel for Climate Change's (IPCC) Fourth Assessment Report (AR4) to simulate variability in the snow water equivalent (SWE) in New York City Water Supply watersheds located northwest of NYC in the Catskill Mountains. SWE is estimated using an empirical temperature-based degree day model. Inputs to this model are either measured with historical meteorological (1961–2000) data or a GCM model output for the same historical period. The evaluation of the GCMs is based on a skill score developed using probability distribution functions derived from the time series of simulated snowpack. From the skill scores (SS) calculated, the GCMs are ranked based on their ability to simulate the snowpack. These results have implications for selecting a subset of GCM simulations for climate change impact assessments in New York City's water supply.
Results show that the GFDL 2·0 (gf001) model has the highest SS (0·93) and CCSM (ncc09) model has the lowest SS (0·26). On the basis of the SS, the GCM ensemble members are classified into three categories: high, medium and low performance. The probability density functions for the three performance classes show the largest between-model variability for models in low performance class. Differences between the means and medians of observation-based model simulation and GCM-based simulation were greatest in the low-performance class. Copyright © 2011 John Wiley & Sons, Ltd.