linear regression model is proposed that relates outputs of weather model forecasts to the variability of solar irradiance at ground level. In combination with numerical weather prediction modelling, this simple model provides up to day ahead forecast of short-term variability in solar irradiance and its performance tends to decrease with forecast horizon time. A measure of intra-hour solar irradiance variability is constructed, and a regression is formed against many candidate predictors from the weather model. The model is refined using a stepwise algorithm. The method is demonstrated using observations over two summers at Melbourne airport, Australia. The hourly clear sky index and the 500-850 hPa geopotential thickness together form useful predictors for the sub-hourly variability in irradiance (R-2 = 0.47 for two hours advance forecasts). The relationship with hourly clear sky index k(t), changes at a threshold near k(t) = 0.79. The variability index was found to be inversely related to the 500-850 hPa geopotential thickness, a relationship that may be due to cloud type variations. Further analysis indicates that improvements in the weather model forecast of hourly clear sky index would substantially increase the ability to infer the intra-hour solar variability, increasing the R-2 value to 0.7 if there was a perfect hourly forecast. Crown Copyright (C) 2016 Published by Elsevier Ltd. All rights reserved.