Abstract
Daily precipitation amounts from a 20-station network within 50 km of a major steelworks at Dunaujváros have been analysed in two ways for a decade preceding construction of the works in the late 1950's and a decade immediately following. The “classical” method is by examination of isohyets of precipitation totals. The second method employs statistical testing of stratified inter-decadal differences at each station using Student's t test. Stratifications are by season (June and December) and by synoptic weather types [five precipitation-producing types chosen from a catalog of 13 published by Péczely (1957) in Hungary].
Post-industrial maps showing isohyets of mean monthly totals for June and for December both exhibit relative maxima to the northeast and the southeast of the steelworks at distances between 40 and 55 km. Aggregation of interdecadal differences across synoptic types and seasons exhibits greater numbers of “large” values of t, both positive and negative, than would be expected by chance. These two results by themselves suggest that an effect on precipitation due to the steelworks has been detected.
The pre-industrial isohyetal map of monthly totals for June exhibits a distinctive maximum 50 km to the northeast of the steelworks, and post-industrial isohyetal maps for individual synoptic types show no spatially coherent maxima or minima “downwind” of the steelworks. Within synoptic types, inter-decadal differences considered statistically “large” are neither spatially coherent in general nor of consistent sign between seasons in the single type having several large differences in both seasons.
The authors conclude, on balance, that no effect due to the steelworks has been detected, and then examine the implications of that conclusion. It is also argued that this comparison of methods of analysis points up the need for both synoptic stratification and the statistical testing of temporal differences in attempts to detect anthropogenic effects on precipitation amount using standard climatological data.
