Investigating the observed sensitivities of air-quality extremes to meteorological drivers via quantile regression
Abstract. Air pollution variability is strongly dependent on meteorology. However, quantifying the impacts of changes in regional climatology on pollution extremes can be difficult due to the many non-linear and competing meteorological influences on the production, transport, and removal of pollutant species. Furthermore, observed pollutant levels at many sites show sensitivities at the extremes that differ from those of the overall mean, indicating relationships that would be poorly characterized by simple linear regressions. To address this challenge, we apply quantile regression to observed daily ozone (O3) and fine particulate matter (PM2.5) levels and reanalysis meteorological fields in the USA over the past decade to specifically identify the meteorological sensitivities of higher pollutant levels. From an initial set of over 1700 possible meteorological indicators (including 28 meteorological variables with 63 different temporal options), we generate reduced sets of O3 and PM2.5 indicators for both summer and winter months, analyzing pollutant sensitivities to each for response quantiles ranging from 2 to 98 %. Primary covariates connected to high-quantile O3 levels include temperature and relative humidity in the summer, while winter O3 levels are most commonly associated with incoming radiation flux. Covariates associated with summer PM2.5 include temperature, wind speed, and tropospheric stability at many locations, while stability, humidity, and planetary boundary layer height are the key covariates most frequently associated with winter PM2.5. We find key differences in covariate sensitivities across regions and quantiles. For example, we find nationally averaged sensitivities of 95th percentile summer O3 to changes in maximum daily temperature of approximately 0.9 ppb °C−1, while the sensitivity of 50th percentile summer O3 (the annual median) is only 0.6 ppb °C−1. This gap points to differing sensitivities within various percentiles of the pollutant distribution, highlighting the need for statistical tools capable of identifying meteorological impacts across the entire response spectrum.