Hourly temporal downscaling of CMIP6 data for photovoltaic energy modelling under climate change: A case study for Portugal

• Daily-consistent irradiance reconstructed from CMIP6 using TAG rescaling. • Custom harmonic fit restores the diurnal temperature cycle from 6-hour data. • Method validation shows small bias in reconstructed irradiance and temperature. • PV yield variations under SSPs remain modest (−5% to + 6%) across Portugal. • POA vs temperature factor separation clarifies climatic drivers of PV yield. Climate impact assessments of photovoltaic (PV) performance are often constrained by the low temporal resolution of climate model outputs (typically 3–6 h), limiting the representation of sub-daily variability in irradiance and temperature, the main drivers of PV production. This work presents a temporal downscaling framework that reconstructs hourly PV-relevant meteorological inputs from 6-hourly CMIP6 data, enabling consistent PV energy simulations under future climate scenarios. Hourly global horizontal irradiance is generated using a modified time-dependent autoregressive Gaussian stochastic downscaling of daily clearness indices. Hourly two-meter air temperature is reconstructed using a continuous harmonic (Fourier) adjustment, while wind speed is linearly interpolated. The resulting hourly series is used to simulate three PV configurations: (i) south-facing fixed-tilt monofacial, (ii) east–west facing vertical bifacial, (iii) north–south single-axis tracking bifacial. The framework is applied to Portugal using historical simulations (1995–2014) and future projections for mid-century (2046–2065) and end-century (2081–2100) under the Shared Socioeconomic Pathways (SSPs): SSP2-4.5, SSP3-7.0, and SSP5-8.5. The irradiance reconstruction preserves daily energy with small diurnal biases (≈ +10 W/m 2 mid-morning/afternoon, ≈ − 17 W/m 2 near noon). The temperature reconstruction achieves daytime bias < 1 °C when validated against hourly reanalysed meteorological data. Projections indicate modest national-average changes in annual PV yield, with mid-century increases under intermediate scenarios (up to + 6%) and decreases under high‑emission scenarios (down to − 5%). Differences in relative yield changes across PV configurations are small, suggesting that projected changes are governed more by meteorological shifts than by system geometry. The framework is transferable, providing higher-detail PV assessments under climate change.