The atmospheric mixing state and emission rates play decisive roles in public exposure to urban air pollution. This study utilizes atmospheric radon measurements taken with the SM200 “stability monitor,” which reflect changes in the atmospheric mixing state, to evaluate and forecast air quality. Using six months (March–August 2016) of atmospheric radon measurements in Jinhua, China, we classify the nocturnal atmospheric stability conditions into four distinct categories, “well-mixed”, “weakly stable”, “moderately stable”, and “most stable”, by applying a modified radon-based stability technique. We calculate the atmospheric self-cleaning ability index (ASI) and evaluate it with the four-category stability scheme, and the results confirm that the atmospheric radon measurements reliably represent the atmospheric mixing state. Analyzing PM2.5, PM10, SO2, NO2, CO, and O3 measurements from three nearby stations during the campaign, we find that the pollutant concentrations and air quality index (AQI) values assigned using the aforementioned stability scheme are consistent with the defined atmospheric mixing states. We subsequently demonstrate that the modified radon-based stability method is suitable for targeting the most unfavorable air quality conditions and determining where the emissions originated. Finally, we propose a simple ASI-based model for predicting regional severe air pollution.