PART X — ENVIRONMENTAL CONTAMINANTS 



sources distributed uniformly over an 

 area. Current air-quality models ap- 

 ply these formulae to all the sources 

 that contribute to the concentration 

 within the chosen "target area" at the 

 chosen time. They differ in the meth- 

 ods by which they insure that only 

 the essential minimum of computation 

 is carried out. 



Performance of Air 

 Quality Models 



The Gaussian-plume formula has 

 been tested in many field trials in 

 carefully observed weather conditions 

 with controlled sources of a conserved 

 pollutant. Using the standard meth- 

 ods of estimating the diffusion param- 

 eters of <r, and a z , it is found that the 

 formulae yield concentrations that are 

 within a factor of 2 of the observed 

 concentrations in about 75 percent of 

 trials. 



An elaborate short-term air-quality 

 model has been constructed for the 

 state of Connecticut, with the sources 

 of four pollutants specified for areas 

 5,000 feet square and two-hour time 

 resolution for four seasons of the 

 year. There are approximately 5,000 

 separate sources and the program al- 

 lows computation of two-hour aver- 

 age concentrations of each pollutant 

 over any or all of the squares. In 

 specially conducted trials over 25 days 

 with measurements at 30 points, it 

 was found that 45 percent of the com- 

 puted two-hour average concentra- 

 tions of SOl' were within a factor of 

 2 of the measurement at a point with- 

 in the 5,000-foot-square box. Meas- 

 urement of the average concentration 

 over a 5,000-foot square is not prac- 

 ticable, but statistical examination of 

 the space variability of concentration 

 suggested that the output of the model 

 would be within a factor of 2 of the 

 true area average concentration on 70 

 percent of occasions. The correspond- 

 ing figure for a 24-hour average was 

 90 percent. These figures indicate the 



possibilities of the most elaborate of 

 existing short-term air-quality models. 



Existing long-term models also use 

 a Gaussian formula with a vertical 

 diffusion term analogous to i z . The 

 horizontal diffusion term is replaced 

 by statistics of wind speed and direc- 

 tion at each source. There have been 

 no systematic verifications of the per- 

 formance of long-term models applied 

 to multiple sources, but in the original 

 application to a single source about 

 75 percent of the computed seasonal 

 averages were within a factor of 2 of 

 the corresponding observation. 



In the application of both short- 

 and long-term models based on the 

 Gaussian formula, two further elabo- 

 rations are incorporated. The first is 

 an allowance for decay, transforma- 

 tion, or deposition of the pollutant, 

 made by multiplying computed con- 

 centrations by an exponential decay 

 factor, characterized by a "half-life." 

 The second adjustment is for the im- 

 portation of pollutant from the area 

 surrounding the modeled area, for 

 which a detailed source inventory is 

 not available. The long-term models 

 include a uniform "background" 

 term; the short-term models must in- 

 clude a flux of pollutant across the 

 boundary. For example, in the Con- 

 necticut model the New York City- 

 New Jersey source, which at times can 

 dominate the air quality over much of 

 the state, is represented by a uniform 

 line-source about 30 miles long — a 

 submodel which quite accurately sim- 

 ulates observed air quality at the state 

 boundary. 



Deficiencies of Current Models 



Experience with the models shows 

 that, surprisingly, a major source of 

 difficulty is specification of the mean 

 wind field in the short-term models 

 and of wind statistics in the long-term 

 models. Physically, the difficulty 

 arises from the large local variability 

 of measured surface wind, caused by 



small-scale topography and phenom- 

 ena such as sea and lake breezes. 

 Mathematically, the difficulty is to in- 

 sure that mass continuity is observed 

 when adapting three-dimensional phe- 

 nomena to a two-dimensional frame. 

 In operating the Connecticut model, 

 best results were obtained by assum- 

 ing a simple algebraic form for the 

 horizontal streamlines (by inspection 

 of meteorological charts) and modify- 

 ing them to conform to the large-scale 

 (i.e., large compared with the grid 

 size) topography of the state. Specifi- 

 cation of the diffusion terms, particu- 

 larly the horizontal diffusion, was 

 found to be less critical than specifica- 

 tion of the mean wind. 



The decay term has a considerable 

 effect on the output and in the present 

 state of knowledge it can only be 

 specified empirically. For example, in 

 the Connecticut model, it was found 

 that the best fit to observation is ob- 

 tained if a half-life of one to three 

 hours is imposed on emitted sulfur 

 dioxide. The chemistry of sulfur di- 

 oxide in the atmosphere is little un- 

 derstood. There is no theoretical 

 support for the adopted value of the 

 half-life. So far as the source inven- 

 tory is concerned, the indications were 

 that deficiencies were not funda- 

 mental in nature but were due to 

 omissions in compilation and a natural 

 reluctance of those emitting pollutants 

 to disclose the magnitude of their 

 contribution. 



The major theoretical deficiency is 

 the inability of any model based on 

 Gaussian-type formulae to handle the 

 problem of chemically reacting pollu- 

 tants and the production in situ of 

 secondary pollutants — circumstances 

 typical of the production of Los An- 

 geles-type photochemical smog. This 

 type of pollution is not likely to be 

 successfully modeled until concise 

 computational techniques which can 

 handle several simultaneous continu- 

 ity equations have been developed. 

 This, and improved knowledge of the 

 chemistry of urban atmospheres, is the 

 main requisite for further advance. 



336 



