Chronic obstructive pulmonary disease (COPD) is strongly associated with the inhalation of harmful particulate matter in ambient air. This study examined 786,290 COPD-related hospital admissions among patients aged 45–64 in Istanbul from 2013 to 2015, using a Generalized Linear Model (GLM) with meteorological variables included as covariates and air pollutant effects evaluated across lag days 0–9. Daily mean concentrations of PM₁₀, PM_2.5, and NO₂ were used as air pollution indicators, while average temperature and relative humidity were considered as meteorological variables. Relative risk (RR) and excess relative risk (ERR) estimates were calculated for a 10 μg/m³ increase in pollutant concentrations across the lag period. Significant associations were found between air pollution and COPD-related hospital admissions in overall analyses as well as seasonal assessments, especially for temperature-related effects. A 10 μg/m³ increase in PM_2.5 was associated with an ERR of 1.26% in females and 1.07% in males at lag 1, while NO₂ exposure showed ERRs of 1.31% in males and 1.30% in females. The effects of PM₁₀ were comparatively smaller, peaking at about 1.13% ERR at lag 5. Stronger associations were observed in both summer and winter seasons. PM_2.5 demonstrated the highest overall impact, particularly among females, with an excess risk of 1.7%. Pollutant effects were more pronounced at ambient temperatures around 0 °C and 25 °C.

Urban air pollution results from complex interactions between vehicle emissions, meteorological conditions, and atmospheric chemistry. While machine learning models achieve high accuracy in air quality prediction, their limited transparency hinders policy adoption. We present an integrated (M-ETAQI) framework combining multiple XAI techniques, temporal decomposition, and causal inference to quantify traffic and meteorological contributions to PM₁₀, PM_2.5, NO_X, and NO₂ concentrations in the Istanbul FSM Bridge corridor (2022–2023 hourly data). Five machine learning models, including XGBoost, LightGBM, CatBoost, Random Forest, and CNN–LSTM–Attention, were trained with temporal cross-validation. SHAP, LIME, PDP, and ALE were applied for interpretability; STL decomposition isolated temporal components, and CCM tested causal links. Tree-based models achieved R² > 0.80 for all pollutants, with CatBoost reaching PM_2.5 R² = 0.876. SHAP confirmed Lag1 as the dominant feature. Wind speed had a significant negative effect on NO_X, while traffic contributed ~20% to NO_X, twice that of other pollutants. STL showed the trend component dominated total variance; NO₂ trend variance = 56.3%. CCM revealed wind speed as the strongest causal driver of NO_X (ρ = 0.37) and confirmed direct traffic–NO_X links. Knowledge distillation from CatBoost improved CNN–LSTM–Attention performance. The four XAI methods yielded consistent attributions, providing robust, cross-validated evidence for traffic management and air-quality policy.

The impact of air pollution on human health has been widely studied in recent decades. Recent findings show that even low levels of air pollution can be harmful to our health, causing disease and early death. However, these studies are very limited in the central region of Türkiye. Therefore, this study focused on the association between the daily variations in air pollutants (PM₁₀, PM_2.5, SO₂, and NO₂) and hospital admissions due to respiratory, cardiovascular, and total (non-accidental) causes in the Sivas province. Daily average concentrations of air pollutants were obtained from two air quality (AQ) monitoring stations, and daily meteorological (air temperature and relative humidity) data were obtained from one meteorological station in Sivas province to determine the effects of air pollution on hospital admissions. It was found to be a significant relationship between air pollution and respiratory hospital admissions in the province. The results of the study showed the relative magnitudes of the risks of cardiovascular diseases and hospital admissions related to air pollutants were as follows: The highest association of each pollutant with cardiovascular diseases was observed for PM₁₀ at lag 4 (ER = 1.74%; 95% CI = 0.95–3.19%), PM_2.5 at lag 2 (ER = 5.12%; 95% CI = 1.39–19.0%), NO₂ at lag 8 (ER = 4.89%; 95% CI = 0.08–288.8%) and SO₂ at lag 5 (ER = 1.21%; 95% CI = 1.10–1.32%). It was seen that short-term exposure to air pollution in Sivas between 2016 and 2019 was positively associated with increasing respiratory hospital admissions. As the first air pollution study to use the generalized linear model (GLM) method in hospital admissions in Sivas, these findings may have implications for local environmental policies and help to combat air pollution.

This study aims to be the first phase of a comprehensive assessment of street canyon air quality in Beşiktaş, Istanbul. The main objective of this initial phase is to develop and systematically evaluate a high-resolution, integrated data infrastructure to support subsequent street-scale air quality modeling (Part B). The study period covers 2016–2023, based on long-term hourly observations. Air quality data (PM₁₀, PM_2.5, O₃, and nitrogen oxides reported separately as NO₂ and total NO_x) were obtained from the Beşiktaş air quality monitoring station. Missing values were filled using seasonal decomposition combined with linear interpolation for NO₂, NO_X, and PM₁₀; for PM_2.5, a multiple linear regression model incorporating PM₁₀, temperature, relative humidity, wind speed, and cyclical time variables was applied. Meteorological variables (temperature, rainfall, relative humidity, wind speed, and direction) were primarily sourced from the Şişli station and gap-filled using linear interpolation. Both datasets were cross-validated for temporal consistency with ground-based observations. Hourly vehicle counts were obtained from the Istanbul Metropolitan Municipality (IMM), while vehicle-type and fuel-category composition were derived from camera-based observations along the corridor. High missing rates in PM_2.5 data and uncertainties regarding vehicle composition are among the main limitations of this study. This study presents a coherent dataset to support street-level air quality modeling and related analyses in Türkiye. Beyond being the first long-term (2016–2023), hourly, street canyon dataset of its kind for Istanbul, this work provides a transferable framework for building model-ready air quality inputs and a practical basis for street-scale dispersion modeling, exposure assessment, and urban air quality management.

Traffic-based air quality proxy models that perform well in compact urban settings have not been evaluated in megacity environments where complex topography, sea–land breeze circulation, and strong meteorological modulation are expected to dominate pollutant transport. We evaluate this transferability for the Bosphorus, an intercontinental corridor between Asia and Europe, using a 17,520-h panel for 2022–2023 that combines hourly traffic counts from six traffic sensors on the 15 Temmuz and Fatih Sultan Mehmet bridges with the closest air quality and meteorological observations. A traffic-proxy baseline (Prophet decomposition with a 9th-degree polynomial for hourly traffic flow) fails uniformly across all bridge × pollutant combinations, yielding negative coefficients of determination. A meteorology-aware machine-learning pipeline (LightGBM, XGBoost, and a bidirectional LSTM-Attention model) trained on traffic, meteorology, temporal indicators, precipitation lags, and short autoregressive terms recovers consistently positive predictive skill for NO₂, NO, NO_x, PM₁₀, PM_2.5, and the derived coarse-particle channel across both bridges. The mean of the twelve per-combination hold-out R² values is ≈0.77 (range 0.58–0.86), yielding a 47–63% reduction in RMSE relative to the baseline; the three model families are practically equivalent in overall performance. SHAP ranks temporal autocorrelation, seasonal contrast, wind direction, and heavy vehicles (rather than total traffic) as the most influential predictors, whereas precipitation-delay characteristics carry diagnostic weight for particle channels. Distribution-free 95% prediction intervals are provided through split-conformal calibration. These results indicate that atmospheric persistence dominates predictive skill, with meteorology and traffic as secondary but systematic modulators.