A method is presented for estimating the climatological variability of yearly and monthly photovoltaic power production per 1 kWp of installed power. This quantity is computed for a specified portfolio of sources on the basis of historical data. Its climatological variability is derived from a simulation of 33 years of power production with hourly time step. Underlying meteorological variables are taken from the MERRA reanalysis for the years 1979-2011. Since the MERRA reanalysis is not a traditional data source for photovoltaic power modelling, various comparisons to available and more frequently used data sources are included. The method of estimation has the advantage of wide applicability due to the global coverage of the meteorological data. (C) 2013 Elsevier Ltd. All rights reserved.