Global Estimation Of Household Air Pollution Integrating Ambient Air Pollution : A Bayesian Approach
Aththanayake Mudiyanselage, Asangika Sandamini Aththanayake (2025)
2025
Tietojenkäsittelyopin maisteriohjelma - Master's Programme in Computer Science
Informaatioteknologian ja viestinnän tiedekunta - Faculty of Information Technology and Communication Sciences
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2025-12-30
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tuni-2025123012244
https://urn.fi/URN:NBN:fi:tuni-2025123012244
Tiivistelmä
Household Air Pollution (HAP) remains one of the main global health problems as of 2025, mainly in low- and middle-income countries where solid fuels and traditional stoves are widely used. There are multiple large-scale datasets available of indoor particulate matter (PM), however due to differences in measurement protocols and incomplete metadata, make it hinder to obtain accurate global household air pollution estimations. This work addresses these challenges by implementing an improved statistical model incorporating indoor HAP measurements with auxiliary geospatial and socioeconomic data to estimate indoor PM2.5 concentrations.
Through this work, Bayesian models have been explored to acquire improved results. Several models have been fitted after analysing the dataset, and finally, a Bayesian model has been formulated with the identified key predictors. Outdoor PM2.5 was a significant positive predictor, indicating that indoor concentrations are influenced by ambient air pollution. The Human Development Index (HDI) emerged as a strong negative predictor, reflecting that structural development, cleaner energy access and housing quality play a major role in determining the indoor air quality. The key findings of this work demonstrate that integrating geospatial and socioeconomic data substantially improves the explanatory power of indoor PM2.5 models. Other predictors included in the model are fuel type, fuel usage, and urban-rural classification. As expected in the literature, solid fuels consistently resulted in elevated indoor concentrations of PM2.5. The final selected model, a non-hierarchical Bayesian model with priors, achieved strong predictive performance with a Bayesian R2 of 0.52.
This study advances the current approaches to HAP exposure estimations by providing a transparent, comparative evaluation of modelling strategies. The implemented Bayesian model offers a practical tool for estimating indoor air pollution and highlights the combined influence of environmental, socioeconomic, and behavioural factors on household exposures. The findings support the need for improved datasets, sub-national socioeconomic indicators, and expanded monitoring for better estimations of global household air pollution.
Through this work, Bayesian models have been explored to acquire improved results. Several models have been fitted after analysing the dataset, and finally, a Bayesian model has been formulated with the identified key predictors. Outdoor PM2.5 was a significant positive predictor, indicating that indoor concentrations are influenced by ambient air pollution. The Human Development Index (HDI) emerged as a strong negative predictor, reflecting that structural development, cleaner energy access and housing quality play a major role in determining the indoor air quality. The key findings of this work demonstrate that integrating geospatial and socioeconomic data substantially improves the explanatory power of indoor PM2.5 models. Other predictors included in the model are fuel type, fuel usage, and urban-rural classification. As expected in the literature, solid fuels consistently resulted in elevated indoor concentrations of PM2.5. The final selected model, a non-hierarchical Bayesian model with priors, achieved strong predictive performance with a Bayesian R2 of 0.52.
This study advances the current approaches to HAP exposure estimations by providing a transparent, comparative evaluation of modelling strategies. The implemented Bayesian model offers a practical tool for estimating indoor air pollution and highlights the combined influence of environmental, socioeconomic, and behavioural factors on household exposures. The findings support the need for improved datasets, sub-national socioeconomic indicators, and expanded monitoring for better estimations of global household air pollution.