<p>Solar radiation estimation is significantly influenced by air pollution, alongside other meteorological factors. However, while global solar radiation models have been developed for various locations in Pakistan, the specific impact of air pollution on solar radiation remains largely unexamined. This study aims to fill this gap by developing empirical and machine learning models to estimate global and diffuse solar radiation in Karachi, focusing on the influence of air pollution. Multiple Linear Regression (MLR) and Artificial Neural Network (ANN) models were constructed using common meteorological parameters, both excluding and including air pollution metrics. Meteorological data were sourced from the weather station at NED University of Engineering and Technology (NEDUET), and air pollution data from the US consulate. After data pre-processing, model accuracy was evaluated using statistical metrics such as normalized root mean square error (nRMSE), normalized mean absolute error (nMAE), correlation coefficient (R), and Global Performance Indicator (GPI). The results indicate that incorporating air pollution parameters enhances model performance. The optimal model for predicting the clearness index (<InlineEquation ID="IEq1"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="704_2025_5499_Article_IEq1.gif" Format="GIF" Height="19" Rendition="HTML" Resolution="72" Type="Linedraw" Width="45" /> </InlineMediaObject> <EquationSource Format="TEX">$H/{H}_{o}$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mi>H</mi> <mo stretchy="false">/</mo> <msub> <mi>H</mi> <mi>o</mi> </msub> </mrow> </math></EquationSource> </InlineEquation>) included wind speed (WS), AQI, and PM<sub>2.5</sub>, achieving R = 0.8204. The diffuse fraction (<InlineEquation ID="IEq2"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="704_2025_5499_Article_IEq2.gif" Format="GIF" Height="19" Rendition="HTML" Resolution="72" Type="Linedraw" Width="48" /> </InlineMediaObject> <EquationSource Format="TEX">${H}_{d}/{H}$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <msub> <mi>H</mi> <mi>d</mi> </msub> <mo stretchy="false">/</mo> <mi>H</mi> </mrow> </math></EquationSource> </InlineEquation>) was best predicted using the clearness index and other meteorological parameters, achieving R = 0.9353. Additionally, the ANN model demonstrated comparable performance for both <InlineEquation ID="IEq3"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="704_2025_5499_Article_IEq1.gif" Format="GIF" Height="19" Rendition="HTML" Resolution="72" Type="Linedraw" Width="45" /> </InlineMediaObject> <EquationSource Format="TEX">$H/{H}_{o}$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mi>H</mi> <mo stretchy="false">/</mo> <msub> <mi>H</mi> <mi>o</mi> </msub> </mrow> </math></EquationSource> </InlineEquation> and <InlineEquation ID="IEq4"> <InlineMediaObject> <ImageObject Color="BlackWhite" FileRef="704_2025_5499_Article_IEq2.gif" Format="GIF" Height="19" Rendition="HTML" Resolution="72" Type="Linedraw" Width="48" /> </InlineMediaObject> <EquationSource Format="TEX">${H}_{d}/{H}$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <msub> <mi>H</mi> <mi>d</mi> </msub> <mo stretchy="false">/</mo> <mi>H</mi> </mrow> </math></EquationSource> </InlineEquation>. Overall, these findings highlight the necessity of integrating both meteorological and air quality parameters to develop robust solar radiation models, particularly in arid environments. The comparative analysis of MLR and ANN models, along with the inclusion of air pollution metrics, represents a novel contribution to improving solar radiation prediction accuracy in arid climates.</p>

Assessing the impact of air pollution on global and diffuse solar radiation in Karachi, Pakistan: a comparative analysis of multiple linear regression and artificial neural network models

期刊论文
发表日期：

2025年4月28日

查看全文

Muhammad Uzair Yousuf,
Abid Ali,
Muhammad Umair

摘要

Solar radiation estimation is significantly influenced by air pollution, alongside other meteorological factors. However, while global solar radiation models have been developed for various locations in Pakistan, the specific impact of air pollution on solar radiation remains largely unexamined. This study aims to fill this gap by developing empirical and machine learning models to estimate global and diffuse solar radiation in Karachi, focusing on the influence of air pollution. Multiple Linear Regression (MLR) and Artificial Neural Network (ANN) models were constructed using common meteorological parameters, both excluding and including air pollution metrics. Meteorological data were sourced from the weather station at NED University of Engineering and Technology (NEDUET), and air pollution data from the US consulate. After data pre-processing, model accuracy was evaluated using statistical metrics such as normalized root mean square error (nRMSE), normalized mean absolute error (nMAE), correlation coefficient (R), and Global Performance Indicator (GPI). The results indicate that incorporating air pollution parameters enhances model performance. The optimal model for predicting the clearness index ( $H/{H}_{o}$ $H / H_{o}$ ) included wind speed (WS), AQI, and PM_2.5, achieving R = 0.8204. The diffuse fraction ( ${H}_{d}/{H}$ $H_{d} / H$ ) was best predicted using the clearness index and other meteorological parameters, achieving R = 0.9353. Additionally, the ANN model demonstrated comparable performance for both $H/{H}_{o}$ $H / H_{o}$ and ${H}_{d}/{H}$ $H_{d} / H$ . Overall, these findings highlight the necessity of integrating both meteorological and air quality parameters to develop robust solar radiation models, particularly in arid environments. The comparative analysis of MLR and ANN models, along with the inclusion of air pollution metrics, represents a novel contribution to improving solar radiation prediction accuracy in arid climates.