PIN: A new metric to evaluate solar irradiance forecast models

Solar irradiance forecasts face challenge by cloud and aerosol variability in the terrestrial atmosphere. Model performance comparison can use known error metric, such as MBE or RMSE, but there is a lack of standardization. This article introduces a new metric called PIN (Peak Irradiance to Noise), designed to access solar irradiance modeling. The PIN metric compares the peak irradiance value to the error between the estimated and observed values (noise), on a logarithmic scale measured in decibels (dB). This approach normalizes errors and smooths outliers effects. PIN was tested using observed datasets from two sites at Brazil’s semi-arid northeast, Casa Nova and Araripina, evaluating ERA5, WRF, and TMY models, and WRF presented the best performance across all cases. Benchmark models from literature were used to calculate PIN metric, and was demonstrated the PIN metric utility in distinguishing competing models and facilitating interpretation of performance differences. A performance rank classification in terms of PIN value is proposed. • Lack of standard metric for irradiance modeling performance. • General used statistic based on smallest error for better modeling. • A way to make error metric values tangible and comparable. • Logarithmic scale, normalized to peak value indicator, distinguishing competing models, and measured in decibels. • An useful easy-interpreted proposed metric for irradiance modeling performance.