Unraveling Heterogeneous Solar Irradiance Variability in Subtropical Regions Through a Multi-source Interpretable Stacking Ensemble Under Climate Change Scenarios
Journal
Earth Systems and Environment
ISSN
25099426
Date Issued
2026
Author(s)
Abstract
Global horizontal irradiance (GHI) is a key determinant of solar energy potential and long-term renewable energy planning. However, assessments based directly on global climate model (GCM) radiative outputs are often constrained by coarse spatial resolution and structural biases, limiting their applicability at regional and local scales. To address these challenges, this study develops a scalable and interpretable framework to assess climate change impacts on solar resources by directly predicting GHI from multi-source geospatial and meteorological information. A reconstructed satellite-derived Typical Meteorological Year (TMY) is employed as a climate-consistent baseline to constrain historical variability and anchor future projections. Building on this observation-constrained framework, future GHI trajectories are estimated by integrating projected meteorological variables derived from CMIP6 emission pathways over a climate-relevant two-decade timescale (2026–2045). Scenario-based analysis reveals divergent responses, with modest but systematic increases in GHI (+ 0.93%) under low-emission pathways and slight declines (− 0.34%) under high-emission scenarios. Model interpretation using SHAP analysis identifies air temperature and precipitation as the dominant atmospheric drivers governing GHI variability, highlighting the roles of thermodynamic and hydrological processes in regulating surface solar irradiance. The influence of meteorological drivers is amplified in regions characterized by high climatic variability, whereas calendar-related factors play a more prominent role under relatively stable climate conditions. By explicitly bridging the scale mismatch between global climate projections and local solar resource assessment, this study provides a transferable and interpretable tool for climate-aware solar energy planning, supporting long-term investment decisions and low-carbon transition strategies under future climate change. These results underscore the importance of incorporating observation-constrained and interpretable modeling frameworks into climate impact assessments of renewable energy systems.
Subjects
Climate change impacts
Climate-aware energy planning
CMIP6
Global horizontal irradiance
Interpretable ensemble learning
Solar resource assessment
Publisher
Springer Science and Business Media Deutschland GmbH
Type
journal article