Sarajevo is among Europe’s most polluted capitals, with winter particulate matter often exceeding health-based guidelines. We analyzed hourly PM 10 and PM 2.5 and meteorology from two stations in Sarajevo representing a central urban basin location and a suburban lowland/airport area over over 2021-2023 to quantify how meteorological drivers shape both levels and high-risk events across heating and off-heating seasons. Methods combined controlled Granger-screened ARX for PM levels and autologistic-ARX for exceedance hours, and month–hour summaries of daily maximum |ΔPM| and of daily exceedance peaks. Level models explained approximately 85–90% of hour-to-hour variability, while exceedance models showed near-perfect ranking performance (validation AUC = 0.984–0.994). In winter, the strongest retained drivers were wind speed at immediate lags and temperature at short horizons with negative level scale effects, while relative humidity and wind direction were weaker and less stable, and off-season retention was sparse with smaller effects. Winter extremes occurred in the evening at Bjelave and near midnight at Butmir, and winter magnitudes were two to four times larger than summer, consistent with heating-season emissions combined with stable nocturnal boundary layers in a valley setting. Given the observational design and lack of source tracers, these patterns are interpreted as suggestive rather than causal attribution. The evidence supports clean-heat transitions, building-insulation upgrades, and inversion-episode management as priority measures, while future work should add source tracers and household energy data to better constrain source contributions. • Heating season accounts for 73–96% of PM exceedance days • •ARX models explain about 85–90% of hourly PM variability • •Exceedance prediction AUC reaches 0.984 to 0.994 • •Wind speed immediate and temperature short lags drive PM • •Evening and midnight peaks dominate winter exceedance timing
Aganović et al. (Sun,) studied this question.