Parametric Distribution Fitting of Photochemical Oxidants, NOx Titration Regimes, and Leakage-Free Machine Learning Generalization in a Tropical Urban-Industrial Airshed

Ground-level ozone (O3) in tropical urban airsheds exhibits non-linear dependencies on volatile organic compound (VOC) and nitrogen oxide (NOx) precursors. We establish a parametric distributional fitting and machine-learning generalization framework across a 13-station continuous regulatory monitoring network (26 station-years; 16.1 million raw readings) during 2024-2025. Maximum likelihood estimation across 182 spatiotemporal series (seven monitored pollutant variables x 13 stations x two annual cycles) identifies the Generalized Extreme Value (GEV) and log-logistic distributions as optimal models for heavy-tailed ozone exceedances (minimum Akaike and Bayesian Information Criteria across 78.6% of station-years), whereas traditional Gaussian and lognormal approximations underestimate statutory threshold breach probabilities. The 13 stations exhibit severe nitrogen dioxide titration and secondary photochemical transformation gradients, dividing the conurbation into urban-industrial NOx-saturated regimes (VOC-limited) in western Howrah and photochemical accumulation hotspots across eastern Kolkata. Random Forest models trained under rigorous spatial block holdout cross-validation eliminate feature leakage between adjacent microenvironments (test R2 = 0.81-0.89 across stations). Spatial feature importance identifies atmospheric boundary layer height, ambient temperature, and non-methane hydrocarbons as primary non-linear drivers of diurnal ozone production. The parametric tail parameters and spatial machine-learning models provide regulatory authorities with an interpretable, transferable decision framework for photochemical risk mitigation, air pollution forecasting, and precursor reduction policy across dynamic tropical megacity airsheds.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-20
DOI
https://doi.org/10.21203/rs.3.rs-11793475/v1
Primary Topic
Atmospheric chemistry and aerosols
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Parametric Distribution Fitting of Photochemical Oxidants, NOx Titration Regimes, and Leakage-Free Machine Learning Generalization in a Tropical Urban-Industrial Airshed

Saptarshi Sadhu
Research Square
Atmospheric chemistry and aerosols
preprint

Parametric Distribution Fitting of Photochemical Oxidants, NOx Titration Regimes, and Leakage-Free Machine Learning Generalization in a Tropical Urban-Industrial Airshed

Saptarshi Sadhu
preprint en

Abstract

Ground-level ozone (O3) in tropical urban airsheds exhibits non-linear dependencies on volatile organic compound (VOC) and nitrogen oxide (NOx) precursors. We establish a parametric distributional fitting and machine-learning generalization framework across a 13-station continuous regulatory monitoring network (26 station-years; 16.1 million raw readings) during 2024-2025. Maximum likelihood estimation across 182 spatiotemporal series (seven monitored pollutant variables x 13 stations x two annual cycles) identifies the Generalized Extreme Value (GEV) and log-logistic distributions as optimal models for heavy-tailed ozone exceedances (minimum Akaike and Bayesian Information Criteria across 78.6% of station-years), whereas traditional Gaussian and lognormal approximations underestimate statutory threshold breach probabilities. The 13 stations exhibit severe nitrogen dioxide titration and secondary photochemical transformation gradients, dividing the conurbation into urban-industrial NOx-saturated regimes (VOC-limited) in western Howrah and photochemical accumulation hotspots across eastern Kolkata. Random Forest models trained under rigorous spatial block holdout cross-validation eliminate feature leakage between adjacent microenvironments (test R2 = 0.81-0.89 across stations). Spatial feature importance identifies atmospheric boundary layer height, ambient temperature, and non-methane hydrocarbons as primary non-linear drivers of diurnal ozone production. The parametric tail parameters and spatial machine-learning models provide regulatory authorities with an interpretable, transferable decision framework for photochemical risk mitigation, air pollution forecasting, and precursor reduction policy across dynamic tropical megacity airsheds.

Research Square
Adamas University (IN)
Atmospheric chemistry and aerosols
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon