SHORT-TERM MATHEMATICAL MODELS FOR AIR POLLUTION

JULIAN DE JESUS BEDOYA VELASQUEZ, Purdue University

Abstract

An extensive air monitoring project collected data including power plant parameters, meteorology, and sulfur dioxide concentrations throughout the State of Indiana. Sigma (theta), the standard deviation of wind direction, was computed by either the sigma meter readings or simply using the instantaneous wind direction readings. The stability class of the atmosphere was determined by using a net radiation scheme in conjunction with the hour of day and wind speed, and in addition by using a scheme based on the values of sigma (theta). These data were used to verify Environmental Protection Agency models PTMTP and RAM. A comparison between model prediction and actual measurements on an hourly, 3-hourly, and daily basis was the first method of verification. The best correlation coefficients correspond to Fairbanks, Edwardsport, and Evansville, in that order. Negative, or very close to zero values were obtained at Indianapolis and Michigan City. The sigma (theta) stability class was more successful with the correlation coefficients than the net radiation method. Models used at Edwardsport, Evansville, and Indianapolis predicted peak levels well, but extremely underpredicted at short distances away from plume centerline. At Fairbanks, models overpredicted maximum peaks and underpredicted low levels. At Michigan City models underpredicted measurements at all levels. Statistical tests at the 90 percentile level rejected the hypothesis of homogeneity of variances and equality of means for all models at all locations, except for RAM urban used in Edwardsport. Prediction of maximum peaks and better prediction of low levels at short distances away from plume centerline were termed the prime variables to be considered in any new short term-model. An improvement on these two variables would bring the models closer to the actual measurements. Two Purdue short-term models use Sigma (theta) to replace the horizontal dispersion coefficient. The first model, PS1, uses vertical dispersion coefficients from Pasquill-Gifford curves, and the McElroy and Pooler curves were used in the second model, PS2. A correction factor was applied to account for the topographic effects of the terrain. A linear relationship between the correction factors for PS1 and roughness length were found, while the corresponding relationship for PS2 factors was found to be exponential. The Purdue short term models showed an improvement over EPA models in the prediction of arithmetic mean, maximum, 90 percentile, and low levels. All F-tests of variances of the models for all locations were very low, ranging from 1.00 to 1.11 indicating that computed and measured values came from like data populations. The F statistical test at the 90 percentile level accepted both PS1 and PS2 at all locations, and the t-test accepted at least one of the models for each location. The Purdue short-term models may be used in replacement of the EPA short-term models if there were a need to account for different topographic effects. The sigma (theta) method was very efficient in determining stability class for both the EPA and Purdue short-term models. The use of sigma (theta) always produce model results closer to the measurements.

Degree

Ph.D.

Subject Area

Civil engineering|Energy

COinS