Effect of Urbanization on the Thermal Structure in the Atmosphere

by R. VISKANTA, R. 0. JOHNSON, and R. W. BERGSTROM, JR., School of Mechanical Engineering, Purdue University, West Lafay- ette, Ind. 47907. The senior author is professor of mechanical engineering; Johnson is a graduate assistant, now with the Division of Engineering Design., Tennessee Valley Authority, Knoxville, Tenn. 37902; and Bergstrom is a postdoctoral fellow, now a research associate at NASA Ames Research Center, Moffett Field, Calif. 94035.

ABSTRACT.-An unsteady two-dimensional transport model was used to study the short-term effects of urbanization and air pollution on the thermal structure in the urban atmosphere. A number of simulations for summer conditions representing the city of St. Louis were performed. The diurnal variation of the surface temperature and thermal structure are presented and the influences of various parameters are discussed.

IT IS ESSENTIAL in air-pollution sources and radiatively participating forecasting to understand how urban- pollutants have not been sufficiently ization and the urban area modify the studied, and their quantitative influences atmospheric environment not only in the are not completely understood. immediate vicinity of the city, but also Observational programs and mathe- for a considerable distance downwind. matical modeling are needed to gain Modification of the environment takes understanding of the urban environ- on many forms and includes alteration ment. Unfortunately, the very nature of the wind, temperature, and water- of the urban environment necessitates vapor proflles as well as the surface extensive measurements over large dis- temperature and the injection of gaseous tances and long periods of time, which and particulate pollutants into the are not only difficult but also costly. atmosphere. Therefore mathematical models can be During the past few years, many employed to advantage to help fill the serious atempts have been made to ob- observational gap by numerically simu- serve and explain the microclimatic ef- lating the transport processes in the fects over and around urban areas. The atmosphere. To the extent that the best-documented and least-questioned niathematical model simulates the real climatic effect is the urban influence on atmosphere, it can then become a valu- temperature in the atmosphere. The able tool for use in micrometeorological urban heat-island phenomena is clearly weather prediction, forcasting of pol- a result of the modification of surface lution episodes, urban planning, inter- and atmospheric parameters by urbani- pretation of field data, and identification zation, which in turn leads to an altered of pollutants by means of remote sens- energy balance. The possible causes of ing methods. the heat island are well recognized In addition to the above, numerical (Peterson 1969), but the individual ef- simulations can also be used as a guide fects such as physical and radiative for observational programs such as the property differences between urban and Regional Air Pollution Study (RAPS) rural areas, flow changes caused by the sponsored by the U. S. Environmental roughness elements, man-made heat Protection Agency (EPA) for the St. Louis Metropolitan area. The main ad- (natural) atmosphere where the mete- vantage of a numerical simulation lies orological variables are considered to be in its ability to predict what will happen time-independent ; (2) the polluted at- for any given changes in urban para- mosphere (the planetary boundary layer, meters, boundary, or initial conditions. PBL), where the meteorological vari- In this paper we will describe the ables such as the horizontal, vertical, short-term effects of urbanization on the and lateral wind velocities, temperature, thermal structure in the atmosphere of water vapor, and pollutant concentra- an urban area, using an unsteady two- tions are functions of time, height, and dimensional transport model. The em- distance along the urban area; (3) the phasis is on the potential effects of soil layer, where the temperature is as- urbanization and air pollution on the sumed to be a function of deptli and thermal structure in the urban planetary time only (the soil and interfacial radia- boundary layer. As a specific example, tion lsroperties are allowed to vary in results of numerical simulations of the the horizontal direction) ; and (4) the city of St. Louis for summer conditions lithosphere, where the temperature is are discussed. Results of similar ex- assumed to be constant during the simu- periments have been reported (Pnndolfo lation period. The atmosphere is as- et n2. 1971, Atzvuter 1972, Bornstein sumed to be cloud-free, and no variation 1972, Wagner and Yu 1972, and Atzoater of topogralshy is accounted for along the 1974). urban area. In the polluted urban planetary bound- NUMERICAL MODEL ary layer the transport of momentum, Physical Model and Assumptions energy, and species takes place by verti- In the unsteady two-dimensional cal and horizontal advection as well as transport model (fig. I), the earth- by vertical and horizontal turbulent dif- atmosphere system is assumed to be fusion. In addition, radiative energy is composed of four layers : (1) the free transported in the solar (short-wave)

Figure I.-Schematic representation of the urban environment and of influences analyzed by the urban boundary layer model. u, v, 8, p, C,, C,, C, SPECIFIED

FREE I ATMOSPHERE

ATMOSPHERE I I PLANETARY and thermal (long-wave) portions of where t, x, and z are the time, hori- the spectrum. The interaction of both zontal, and vertical coordinates ; u and w natural atmospheric constituents and are the horizontal and vertical velocity gaseous as well as particulate pollutants components ; +i (x,z,t) is the dependent with solar and thermal radiation is ac- variable "conserved ;" Si (x,z,t) is the ap- counted for. The planetary boundary propriate source term; and KXi and K,i layer and soil layer are coupled by are suitable turbulent diffusivities. The energy and species balances at the at- independent variables and their corres- mosphere-soil interface. The horizontal ponding sources are listed in table 1. A variation of the urban parameters such general discussion of the conservation as man-made heat and pollutant sources, equations was given by Plate (1971), surface solar albedo (reflectance) and and Johnson (1 975) presented a detailed thermal emittance, surface roughness, derivation of the equations used in the thermal diffusivity and conductivity of model. the soil, and Halstead's soil moisture The boundary conditions prescribed at parameter are arbitrarily prescribed the edge of the outer flow (top of the functions of position along the urban PBL) and at the soil surface are the area. It has been observed (Stern et nl. following : at the edge of the outer flow, 1972) that the man-made heat and pol- +i is specified ; at the interface, u = v = lutant sources, for example, vary during w = 0 ; and the surface temperature (at the diurnal cycle. However, more real- any x) is predicted from an energy istic modeling of these sources during balance the diurnal cycle must await more com- plete observational data.

Model Equations The numerical model is based on the conservation equations of mass, momentum, energy, and species. The equations used to describe the planetary boundary layer can be written in a general form as , - PI a+i a+i a+i a In this equation, the first two terms -+u-f W-=-- K,i- account for absorption of solar and at ax az ax 2) ( thermal radiation, the third term repre- sents thermal emission, the fourth and fifth terms account for sensible and latent heat transfer by molecular and [I] turbulent diffusion, the sixth term rep-

Table I .-Dependent variables and source terms

1 u horizontal velocity f (v-v,) geostrophic deviation 2 v lateral velocity f (ug-U) geostrophic deviation 3 fj potential temperature aF/az, q radiative and man-made heat sources 4 C, water-vapor concentration C,water vapor emission 5 Ci concentration of pollutant aerosol CI pollution emission 6 C2 concentration of gaseous pollutant cz pollution emission resents heat conduction into the ground, ployed for the prediction of the dif- and the final term is the man-made sur- fusivities in the transition layer under face heat flux. The water vapor con- stable conditions. centration at the surface is prescribed Stable conditions were assumed to by Halstead's moisture parameter (Pnn- exist when the average Richardson num- dolfo et (11. 1971) by the expression ber in the lowest 25 m of the atmosphere was greater than zero. Where this con- (x,O) =M(x) Cw,sat[T(x,O) c, I + dition was reached, Pandolfo's model [I-M (x) IC, (ZIP) [31 was used only near the surface, while where M is the moisture parameter, the polynomial was employed in the Cw,s,t is the water-vapor concentration transition layer. Otherwise, Pandolfo's at saturated conditions, and z, is the model was used throughout the entire first grid point above the surface. For PBL. If the Richardson number ex- a prescribed pollutant flux at the sur- ceeded the critical value, it was reset to face, m,,, a species balance at the inter- this value so that unreasonable diffusiv- face, yields the condition ity values would not be predicted.

Radiative Transfer Model The radiative transfer model used has been discussed in detail elsewhere At the upwind boundary, the meteoro- (Bergstrom and Visknntn 1973b), so logical variables are predicted from the only a summary of it is included here. unsteady one-dimensional model of Berg- The urban atmosphere is considered to strom and Viskanta (1973a). Initially, be cloudless, plane-parallel, and consist- +i is specified everywhere and is as- ing of two layers: (1) the free atmos- sumed to be independent of the hori- phere, and (2) the urban PBL where zontal coordinate x. the pollutants are concentrated. The earth's surface was considered to emit Turbulent Diffusivities and reflect radiation as prescribed func- Specification of turbulent diffusivities tions of wavelength. Since the atmos- for an urban atmosphere in connection pheric gases and particles absorb, emit, with numerical modeling of the PBL is and scatter radiation, the radiative a very difficult task and has been dis- transfer between the free atmosphere cussed in a recent review (Oke 1973n). and the PBL is coupled. However, no The semi-empirical equations developed consideration is given to individual point by Pandolfo et al. (1971 ) were initially sources of pollutants. Since multidimen- employed. The decay of turbulence in sional radiative transfer is complex, it the upper part of the PBL was pre- is assumed that the transport of radia- scribed by following Blackadar's (1962) tion can be approximated by a quasi- formulation. However, in the hours be- two-dimensional field based on the fore sunrise, when the atmosphere be- vertical temperature, water-vapor, and came quite stable, unrealistically deep pollutant distributions at several pre- surface inversions resulted and the determined horizontal positions. The Richardson numbers were found to ex- radiative fluxes in the atmosphere are ceed the critical value. then evaluated at a few prescribed In these situations the diffusivities horizontal locations and linear or non- predicted by Pandolfo's eddy diffusivity- linear interpolation is then used to de- Richardson number correlations were termine the radiative fluxes between not applicable. Therefore, the cubic these locations. polynomial developed by O'Brien (1970) The radiative fluxes and flux diver- and used by Bornstein (1972) was em- gences are evaluated by dividing the entire electromagnetic spectrum into employed in the horizontal direction and solar (0.3 < A < 4 pm) and thermal in the soil layer. (4 < A < 100 pm) portions. The com- The number of grid points and their putational details can be found in Berg- spacing in the vertical and horizontal strom and Viskanta (1973h). Total directions can be varied. The results re- emissivity data for water vapor and car- ported have been obtained by using 22 bon dioxide were used, and scattering nodes in the vertical direction and 17 in was neglected in predicting radiative the horizontal direction. The first verti- transfer in the thermal part of the spec- cal grid point was located at 5 m from trum. It was assumed that the influence the surface and the horizontal grid spac- of gaseous pollutants could be confined ing was 1.5 km. The program required to the 8 to 12 pm spectral region due to practically the entire high-speed memory the relative opacity of the H,O and CO, capacity (138,000/150,000 bytes octal) bands. Ethylene or sulfur dioxide were of the National Center for Atmospheric considered to be representative pollu- Research CDC-7600 digital computer. tants. The spectral absorption and scat- The computational time was approxi- tering characteristics of the aerosol in mately 8 minutes for a 24-hour simula- a polluted atmosphere were taken from tion period. the model developed by Bergstrom (1 972). RESULTS AND DISCUSSION Numerical Method of Solution Numerical Experiments The alternating - direction - implicit The numerical model has been tested, (ADI) method (Ronche 1972) was em- and a number of numerical experiments ployed to solve the transport equations have been performed, using the city of [I]. The selection of suitable grid spac- St. Louis as an example. Because of ing, appropriate finite-diff erence ap- the length and scope of the paper, it is proximations for the spatial derivatives possible to include only some selected and the algorithm itself, and tests for results for the temperature in the at- convergence are discussed in detail by mosphere. The experiments were de- Johnson (1 975). To improve the resolu- signed to simulate the thermal structure tion near the surface, a logarithmic- and pollutant dispersion in the urban uniform grid spacing was chosen in the atmosphere. The urban area was mod- vertical direction. The logarithmic spac- eled by varying the appropriate surface ing extended to about 1 km from the parameters between rural and urban earth surface, and from there to the top values in the horizontal direction. The of the PBL (- 2 km) the spacing was value of these interface parameters are uniform. A uniform spacing was also given in table 2. A horizontal distribu-

Table 2.-Numerical values of interface parameters for simulations (Johnson, 1975) Upwind Urban Parameter rural Center Solar albedo 0.18 0.12 Thermal emittance .90 .95 Soil thermal conductivity (W/mK) .1 .5 Soil thermal diffusivity (107 m2/s) 1 2.5 Surface roughness (m) .2 1.0 Moisture availability parameter .1 .05 Lower soil boundary temperature (K) 295.5 295.5 Urban heat source (W/m2) 2 20 Aerosol pollutant source (rg/m2~) 0 2.5 or 5 Pollutant gas source (ug/m2s) 0 2.5 or 5 tion was established by selecting the those given by Lettau and Davidson for values of parameters at the rural and the O'Neill Great Plains Turbulence urban center locations and, for lack of Study. The pollutant-concentration pro- any better data or information, a Gaus- files were initialized to a constant back- sian distribution curve was then fitted ground value of 50 pg/m3. between the urban and rural locations by specifying the standard deviation of Components of the Energy a suitably chosen mean value. The data Budget at the Surface given in table 2 are representative values The surface temperature can be con- over the city. sidered as the forcing function of the The simulations were started at 1200 model. Thus it is important to under- solar time and continued for a 24-hour stand the variation of the energy budget period. The initial temperature profiles components along the urban area. These used were taken from Lettau and David- are presented in figures 2 and 3 at 2400 son (1957) for 24 August 1954 and were of the first day and 1200 of the second imposed over the entire area (no x- day. Inspection of figure 2 shows that variation). The initial horizontal and the emitted (q,.) and the absorbed ther- lateral velocity fields were either identi- mal (q,,) fluxes are the dominant com- cal to or decreased by a factor of 2 of ponents. The turbulent (q,), the latent

Figure 2.-Variation of the surface balance flux components (qt-turbulent, q,-latent, q,-emitted, q,s-absorbed solar, qat-absorbed thermal, q,-ground conduction, Q-urban heat source) at the surface at 2400 of the first day; Gaussian distribution, radiatively nonparticipating, u,=6 m/s and v,=4 m/s.

qe 400 7 Figure 3.-Variation of surface energy balance flux corn onents a+ 1200 of the second day: Gaussian distribution, raJatively nonparticipating, u,=6 rn/s and v,=4 rnJs. r

(q,), the ground conduction (q,), and surface temperature drops sharply in the man-made urban heat (Q) fluxes are the late afternoon, reaches a minimum significantly smaller. However, at noon just before sunrise (between 0500 and (fig. 3) the absorbed solar flux (q,,,) is 0600), and rises sharply in the morning. the largest term in the energy balance During the first few hours of the simula- and the man-made heat source (Q) is tion, an initial transient is noted in the the smallest. At the urban center (x = surface temperature. This is due to the 10.5 km) the soil heat conduction term fact that the assumed initial velocity, (q,) amounts to only about 10 percent temperature, and water-vapor concen- of the absorbed solar flux. The turbulent tration profiles were far away from the (q,) and the latent (q,) fluxes are quite quasi-steady solution induced by the sensitive to surface gradients and show diurnal cycle, and it took about 2 to 3 significant variation along the urban hours for the system to adjust. The sur- area. face temperatures at the downwind residential location (node 12) are slightly Surface Temperature higher than at the upwind residential The diurnal variation of the surface (node 4) locations. This is attributed to temperatures at four positions along the the heating of the air as it flows over urban area are shown in figure 4. The the warm city. The maximum surface temperature difference (about 4OC) between the surface temperatures are con- tween the urban center and upwind siderably smaller for higher wind speeds rural locations occurs just before sun- (Johnson 1.975). rise. On the other hand, for a simulation A comparison of the surface tempera- with higher wind speeds (u, = 12 m/s tures for the Gaussian and rectangular and v, = 8 m/s) , the maximum difference distributions of urban man-made heat occurs in the evening at 1900 and re- sources is illustrated in figure 6. The mains almost constant throughout the results show that the difference between night (Johnson 1975). the two results is only about 1°C, and The effect of radiatively participating the maximum occurs early in the morn- pollutants as measured by a local sur- ing (0500) when the man-made urban face-temperature difference (tempera- heat source is a significant component ture with participating pollutants/temp- of the energy budget. At noon (1200) erature without) is illustrated in figure the surface temperatures in the city 5. The temperature differences result differ by only about O.l°C, indicating from an interaction between many com- the predominance of the radiant en- plex phenomena, and therefore individ- ergy transfer in the interfacial energy ual influences are not easily identified. balance. The surface temperature difference is largest during the night and reaches a Temperature Structure maximum before sunrise 0500 and is The isopleths of the two-dimensional smallest at noon. The differences be- potential temperature fields at 6-hour

Figure 5.-Local surface temperature difference (simulation with radiatively participating ollutants minus simulation with radiatively nonparticipating poP lutants) along the city; Gaussian distribution, ethylene (C2H4)-pollutant gas, u,=6 m/s, v,=4 m/s, m, =m2=2.5 pg/m2s. Figure 6.-Comparison of surface temperatures for Gaussian and rectangular distributions of urban heat and pollutant sources; radiatively participating, ethylene (CzH4)-pollutant gas, up6m/s, v,=4 m/s, ml=mz=2.5 pg/mZs.

---- RECTANGULAR GAUSSIAN

intervals for a simulation with radia- son 1969, Oke 1973b). Just after sun- tively nonparticipating pollutants are rise (0600) the breakup of the stable shown in figure 7. At 1800 the atmos- layer was noted. The surface inversion phere is nearly adiabatic, especially in breaks up rapidly after sunrise due to the upwind rural area, with a thermal heating of the surface by absorption of plume having a temperature of about solar radiation, and by 0900 all traces 305 K forming at a height of about 100 of the inversion have disappeared. Simi- n~ downwind of the city center (node 8, lar types of diurnal variation of thermal x = 10.5 km). Note that the last digit structure have been found for other denoting the temperature of the iso- simulations under different conditions therm at 1800, 2400, and 0600 hours has (Johnson 1975). been truncated. However, the plume is The maximum temperature difference not felt downwind, and the upwind and between simulations with radiatively downwind surface temperatures are participating and nonparticipatng pollu- nearly identical. A surface temperature tants occurs at the urban center late at inversion develops at night and is seen night (fig. 8). The temperature differ- to be deeper over the rural area than ences are seen to be largest near the over the city. This is indicative of the surface and are confined to the lowest nocturnal heat island, which decreases 600 meters of the PBL. When the pol- the stability of the atmosphere. The lutant gas is assumed to have the radia- magnitude of this heat island is larger tive properties of sulfuy dioxide (SO,), at night than during the day. This tjrpe the temperature structure predicted is of behavior is well documented (Petrr- practically the same as that for the Figure 7.-lsopleths of potential temperature (in K); Gaussian distribution, radiatively nonparticipating, up6 m/s, v,=4 m/s.

0 6 12 18 0 6 12 18 24 x (km) (a) t = 18:OO (b) t = 24:OO Figure 8.-Comparison of thermodynamic temperatures at the urban center; Gaussian distribution, ethylene (C2H4)-pollutant gas, up6 m/s, v,=4 m/s, ml=m2=2.5 pg/mzs.

NONPARTICIPATING ------PARTICIPATING

-E '400 -

200 - t=18:00

radiatively nonparticipating pollutant whereas for the rectangular distribution gas. This is due to the fact that SO1 is during the night it occurred downwind a weakly radiating gas. The results are of the center. The results presented in consistent with those obtained with the the figure show that there is a double one-dimensional model (Bergstrom and peak in AT,,-,.. The first peak is noted at Viskr~nta1973n). The net effect of the about 2000. It arises due to the more city and the human activity on the rapid cooling at the upwind rural area temperature distribution is illustrated than in the city. The maximum heat- in figure 9. The results show that the island intensity occurs just before sun- presence of radatively participating pol- rise and has a magnitude of about 4OC. lutants dampens the amplitude of the For the population of the urban area diurnal temperature variations and that and wind speeds of the simulation this for a relatively short (24-hour) simula- value is in good agreement with the tion the effects are confined primarily observations and empirical correlation below 600 m. of Oke (1973b). For a simulation with lower wind speeds (u, -- 2.4 m/s and Urban Heat Island v, -- 1.6 m/s), which are not reported The urban heat island is a well-known here, the maximum heat-island intensity and accepted fact (Peterson 1969, Oke reached about 10°C. This is also in good 1973b). The urban heat-island itensity agreement with the nighttime and day- (maximum difference between upwind time observed temperature excesses be- rural and highest urban surface temp- tween the urban and rural locations erature, at,-,) predicted is shown in (Oke 1973b, DeMarrais 1975). figure 10. For the assumed Gaussian distribution of man-made heat sources, CONCLUSIONS the maximum surface temperature dif- As a result of a limited number of ference occurred at the urban center, numerical simulations that have been Figure 9.-Comparison of perturbation potential temperatures (temperature in the cit minus the temperature at the upwind rural location) at the ur r,an center for radiatively nonparticipating (a) and radiatively participating (b) simulations; Gaussian distribution, ethylene (C2H,)-pollutant gas, up6 m/s, v,=4 m/s, ml=m2=2.5 pg/m2s.

@' (to 63' (to

Figure 10.-Comparison of maximum urban upwind rural surface temperature differences; ethylene (C,H,)-pollutant gas, ug=6 m/s, vg=4 m/s, ml=m2=2.5 rg/m2s.

NONPARTICIPATING ------PARTICIPATING A

" 12:OO 16:OO 20:OO 24:OO 04:OO 08:OO 12:OO TIME perforn~ed for selected meteorological LITERATURE CITED conditions during the summer, the fol- Atwater, Marshall A. 1972. THERMALCHANGES INDUCED BY URBAN- lowing generalizations can be made : IZATION. Am. Meteorol. Soc. Conf. on Urban Environ. and 2d Conf. on Biometeorol.: 1. The unsteady two-dimensional trans- 153-158. Boston. port model predicts an urban heat- At;$a7~rk~~~~~1d",.ANGESINDUCED BY POLLu- island intensity that is sensitive to TANTS FOR DIFFERENT CLIMATIC REGIONS. Am. wind speed and latent energy trans- Meteorol. Soc. Symp. on Atmos. Diffusion and Air Pollut. : 147-150. Boston. port through Halstead's moisture Rergstrom, Robert W., Jr. ~h~ intensity of the ur- 1972. PREDICTIONSOF THE SPECTRAL ABSORP- TION AND EXTINCTION COEFFICIENTS OF AN ban heat island reached a magnitude URBAN AIR POLLUTION AEROSOL MODEL. Atmos. of about 4OC near sunrise and about be^^^^^^; ~o~:~~2&8;Jr., and 1.4OC at noon. Raymond Viskanta. 2. F~~the conditions the radia- 1973a. MODELINGTHE EFFECTS OF GASEOUS AND PARTICULATE POLLUTANTS IN THE URBAN tively participating pollutants are ATMOSPHERE. PART I. THERMAL STRUCTURE. J. Appl. Meteorol. 12 : 901-913. in Rergstrom, R. W., Jr., and K. Viskanta. the urban heat island compared to 1973b. PREDICTIONOF THE SOLAR RADIANT other factors. During the night, air FLUX AND RATES IN A ATMOSPHERE. Tellus 25 : 486-498. pollution does increase the surface Blackadar, Alfred K. temperatureby about 1.250~at the 1962. THE VERTICAL DISTRIBUTION OF WIND AND TURBULENT EXCHANGE IN A NEUTRAL urban center. ATMOSPHERE. J. Geophys. Res. 67: 3095-3102. 3. The role of radiatively participating BO~&$~$$~~~&SIONAL,NON-STEADY NUM- pollutants in the formation of the ERICAL SIMULATIONS OF NIGHTTIME FLOWS OF thermal is relatively small A STABLE PLANETARY BOUNDARY LAYER OVER A structure ROUGH WARM CITY. Am. Meteorol. Soc. Conf. and it is greatest late at night. The on Urban Environ. and 2d Conf. on Bio- meteorol. : 89-94. Boston. largest effect of pollutants was to de- DeMarrais, Gerard A. crease the stability at night and hence 1975. NOCTURNALHEAT ISLAND INTENSITIES increase turbulent transport near the ~~&~~~$C~e~~he~~~S~~4:0135ytl.NG surface particularly at heights below Johnson, Robert 0. 500 m. 1975. THE DEVELOPMENT OF AN UNSTEADY TWO-DIMENSIONAL TRANSPORT MODEL IN A 4. The feedback mechanism between POLLUTED URBAN ATMOSPHERE. M.S. thesis, Purdue University. 209 p. ~~~~~~~~~9 structure, Lettau, Heinz H., and Ben Davidson, editors. and dispersion has the potential of 1957. EXPLORINGTHE ATMOSPHERE'S FIRST MILE. VOL. I AND 11. Pergamon Press, New being significant in modifying pol- York. 578 p. lutant concentrations under more O'Brien, James J. stable conditions and higher pollutant 1970. ON THE I'ERTICAL STRUCTURE OF THE EDDY EXCHANGE COEFFICIENT. J. Atmos. Sci. loadings. However, the magnitudes 27 : 1213-1215. of the changes are dependent on the Ok~i~~A URBAN coupling between the radiative prop- 1968-1972. Univ. Brit. Columbia Dep. Geogr., erties of air pollution and the buoy- o~~~""~ ancy-enhanced turbulence. Due to 197313. CITY SIZE AND URBAN HEAT ISLAND. Atmos. Environ. 7 : 769-779. the uncertainties in predict- Pandolfo, Joseph P., Marshall A. Atwater, ing either of these two quantities, the and Gerald E. Anderson. magnitude of these changes is un- 1971. PREDICTIONBY NUMERICAL MODELS OF TRANSPORT AND DIFFUSION IN AN URBAN certain. BOUNDARY LAYER. Cent. for Environ. and Man Final Rep. Hartford, Conn. 139 p. Peterson, James T. 1969. THE CLIMATE OF CITIES: A SURi7EY OF RECENT LITERATURE. Air. Pollut. Control Adm., Publ. AP-59. 48 p. Washington, D. C. Plate, Erich J. 1971. AERODYNAMICCHARACTERISTICS OF AT- MOSPHERIC BOUNDARY LAYERS. U.S. Atomic Energy Comm. Div. Tech. Inf. Oak Ridge, Tenn. 190 p.

75 Roache, Patrick J. Acknowledgments. This work was sup- 1972. COMPUTATIONALFLUID DYNAMICS. Her- ported by the Division of Meteorology, mosa Publishers, Albuquerque, N. M. 434 p. U.S. Environmental Protection Agency, Stern, Arthur C., Henry C. Wholers, Richard W. Boubel, and William P. Lowry. under Grant No. R801102 and R803514. 1972. FUNDAMENTALSOF AIR POLLUTION. Computer facilities were made available by Academic Press, New York. 492 p. Purdue University and the National Center Wagner, Norman K., and Tsann-wang Yu. for Atmospheric Research. The interest and 1972. HEATISLAND FORMATION: A NUMERICAL support of J. T. Peterson, grant project EXPERIMENT. Am. Meteorol. Soc. Conf. on officer, is acknowledged with sincere Urban Environ. and 2d Conf. on Biome- thanks. The authors also thank A. Venka- teorol. : 83-88. Boston. tram for his contributions to this effort.