Entropy-based evaluation of empirical daily global solar radiation models for Mogadishu, Somalia
Year: 2026
Type: article
Abstract: Accurate estimation of global solar radiation (H) is critical for solar energy system design, building energy simulation, and renewable energy planning, sparticularly in data-scarce regions. This study evaluates four empirical models namely the Angström–Prescott (A–P) model, Local Polynomial Regression (LPR), quadratic spline regression, and cubic spline regression for estimating daily H in Mogadishu, Somalia. Models were assessed using conventional statistical metrics ( \(R^2\) , RMSE, MAE, MAPE) and entropy-based indicators (Shannon entropy of residuals, Kullback–Leibler divergence). Results demonstrate that the LPR model achieved superior performance ( \(R^2 = 0.907\) , RMSE = 0.156 kWhm \(^{-2}\) d \(^{-1}\) , 95% CI [0.132, 0.178]; Shannon entropy = 1.679, 95% CI [1.577, 1.917]), compared to the A–P formulation ( \(R^2 = 0.797\) , RMSE = 0.231 kWhm \(^{-2}\) d \(^{-1}\) , 95% CI [0.215, 0.262]). Entropy-based metrics indicated that non-parametric models preserved the statistical structure of observed radiation series, thereby minimizing information loss. The validated LPR estimator offers a deployable baseline for solar resource assessment in regions lacking ground-based meteorological infrastructure, while the locally calibrated A–P coefficients provide a parsimonious, site-adapted formula when sunshine duration data are available.
Source: Discover Applied Sciences
Authors: Shuaib Mursal Ibrahim, Aşır Genç
Institutions: SIMAD University, Necmettin Erbakan University
Language: English
Field: Energy
Domain: Physical Sciences
SDG: Climate action
Open Access status: diamond
OpenAlex ID: W7229700227
