• Provides the first multi-climate assessment of tilted irradiance biases between pyranometers and silicon reference cells. • Reveals the influence of atmospheric variability as key drivers of sensor discrepancies. • Introduces correction frameworks with cross-site transferability. • Establishes the presence of reproducible pyranometer–reference cell offsets across varied climates worldwide. Accurate plane-of-array (POA) irradiance measurements are critical for assessing photovoltaic (PV) performance, yet systematic discrepancies between thermopile pyranometers and silicon-based instruments remain underexplored across varying climates. This work quantifies the measurement biases of silicon-based sensors relative to thermopile, using data from eight globally distributed sites covering arid, tropical, subtropical, and temperate regimes. While overall correlations were strong, consistent offsets of -12.5 to +3.5 W/m 2 emerged, largely influenced by solar geometry, sky clarity, and aerosol conditions. Reference cells tended to under-respond under clear, beam-dominated skies, but showed slight over-responses in diffuse-rich or cloud-enhanced environments. Three empirical correction schemes were developed and tested using six sites for training (70/30 split). A geometry-only model based on solar zenith angle (SZA) provided limited improvement, whereas incorporating both SZA and the clearness index (K t ) yielded consistent error reductions across climates. At the arid Qatar site, for instance, the Root Mean Square Deviation (RMSD) dropped from 4.2% to 2.8%, with bias effectively eliminated. Independent testing at two temperate settings confirmed the approach’s robustness, with the SZA and K t formulation offering the most balanced and transferable correction. Overall, combining solar geometry with sky-clarity metrics provides a simple yet effective cross-climate framework, reducing RMSD by roughly 0.5-1% and Mean Bias Deviation (MBD) by 1-2%. The method enables reliable harmonization between pyranometer and reference-cell measurements, improving the accuracy of POA inputs for PV modelling and long-term system evaluation.
Musleh et al. (Tue,) studied this question.