Evaluation and improvement of TAPM in estimating solar irradiance in Eastern Australia

• TAPM overestimated GHI over 75% of the time when was compared to satellite data.
• Bias and errors do not change significantly for different model resolutions.
• The resolution of 45 km has mostly the lowest RMSE among all sites.
• Fronts, troughs and occasionally high pressures associated with large hourly errors.
• Linear regression approach improved the accuracy of the TAPM-estimated GHI.

Numerical Weather Prediction models provide short and medium-range solar irradiance forecasts. This paper investigates the accuracy of The Air Pollution Model (TAPM) in estimating Global Horizontal Irradiance (GHI). The model uses standard meteorological fields but with no real-time cloud/aerosol information. In order to assess the model, TAPM-estimated GHI is compared to satellite-derived GHI at selected sites in Eastern Australia. The satellite data used in this study was compared to ground-based observations at the Australian Bureau of Meteorology stations and adjusted accordingly. Also, the model-estimated GHI at horizontal spatial resolutions of 45, 15, 5, 1.5 and 0.5 km were compared to adjusted-satellite data. The results revealed that the resolution of 45 km had the lowest Root Mean Square Error (RMSE). This means that variations on scales less than 45 km cannot be captured by the model properly. We also found that during cloudy conditions, the RMSE could increase by around 300% compared to clear-sky conditions for both lower and higher resolutions. Therefore, most of the large errors that occur during cloudy conditions are associated with the misrepresentation of clouds in the model, in particular during deep low-pressure systems, passage of cold fronts, Easterly troughs, cloud bands, and in some conditions even during high pressure systems. Furthermore, correcting the 45 km modelled GHI for clear-sky errors improved the accuracy of the model and reduced the RMSE, Mean Absolute Error (MAE) and Mean Bias Error (MBE) on average by maximum 16%, 26% and 84% respectively.