The Air Pollution Division of the State of Pennsylvania has conducted simultaneous measurements of meteorological variables and air quality at Johnstown, Pa. An analysis of the observations for two fall seasons showed that fluctuations in the concentrations can be fairly well explained by the changes in wind speed and fluctuations of vertical air velocities. Wind direction is relatively unimportant, except for the rare east winds, when the air at Johnstown is affected by a major steel plant. This paper presents a simple mathematical model that predicts variations in air pollution from a large number of low-level sources in a narrow valley. This theory and observations are in good agreement. It is likely, therefore, that most of the pollution at Johnstown and similar sites is locally produced. Due to the fact that there usually are no important high-stack sources upwind at Johnstown at present, fumigation is not generally a factor there.