A study was conducted to measure the photovoltaic (PV) system energy production losses caused by the presence of snow for Colorado and Wisconsin locations, and to develop a model for predicting PV system performance losses from snowfall. The PV system performance and associated meteorological data for six PV systems in Colorado and Wisconsin were measured during the winters of 2010-2011 and 2011-2012. The PV systems included two residential systems with stand-off roof mounts; two small commercial systems - one with stand-off mounts on a tilted roof and the other rack-mounted on a flat roof; a 100-kW rack-mounted PV system on a flat roof; and a 200-kW PV system with ground-mounted racks. The measured monthly PV system energy losses caused by snow were as high as 90%. Losses expressed as a percentage of annual energy production ranged from 1% to 12%. A model was developed that uses daily snow depth to identify the presence of new snow; an hourly plane-of-array irradiance and air temperature relationship to identify when snow slides down the PV array; the PV array tilt angle to determine how far the snow slides; and the extent of snow coverage on the PV array to determine the fractional energy output of the PV array. On average, the model worked well. For the two-year period, the modeled annual energy losses were within 0.5% (absolute) of those measured for the non-residential systems. For the residential systems, the modeled annual energy losses were about 1.5% (absolute) less than measured losses. Larger differences between modeled and measured energy losses should be expected for monthly or shorter time periods. The standard deviation of the differences between modeled and measured monthly losses was 10.5% (absolute). (C) 2013 Published by Elsevier Ltd.