SUMMERTIME URBAN HEAT ISLAND MITIGATION: PROPOSITIONS BASED ON AN INVESTIGATION OF INTRA-URBAN AIR TEMPERATURE VARIATIONS
ARCHITECTURAL SCIENCE REVIEW, 40(4): 155-164 (in print)
Doctoral Program in Architecture
College of Architecture & Urban Planning
The University of Michigan
2000 Bonisteel Blvd
Ann Arbor, MI 48109-2069, U.S.A.
A normalization technique called “Paired Measurement Program” is utilized to analyze the effects of four land-use/land-cover patterns on intra-urban air temperature variations in Ann Arbor, Michigan, under different atmospheric stability conditions. The aim is to propose broad land-use control strategies for the mitigation of the negative effects of urban heat islands during the summer.
The results show that unstable day-time atmospheric conditions produced the maximum intra-urban air temperature differences, with significant downtown-to-residential as well as residential-to-residential variations. Stable nights lead to significant temperature differences between the downtown location and each of the residential locations, but not between the residential locations. Vegetation shade influenced intra-urban temperature variations at day but not at night. This resulted in an admixture of cool and heat islands at day, but at night a more homogeneous, downtown-centered heat island was the norm. Based on these results, four propositions for the mitigation of the summer heat island effect in mid-latitude cities are made.
UDC: 403(442.2) Keywords: Paired Measurement Program; Urban Heat Island -- Influence of Land-Use, Mitigation.
DTL - Downtown Location.
HDR - High-Density Residential site.
NW2 - Multi-family residential/institutional setting.
LVR - Low-density Vegetated Residential neighborhood.
LOR - Low-density Open-canopy Residential neighborhood.
Inadvertent climate changes induced by urbanization are well documented. Such changes, epitomized by the concept of “urban heat island” (UHI) are usually measured by urban-rural difference method, city traverse method or remote sensing [Ref. 1]. Urban-rural difference method which compares climate data from an urban and a rural weather stations can be further divided into two sub-categories: Time Averaged Method (TAM) where differences in landscape between the two stations are assumed unimportant and Time Rate Change Method (TCM) where climate parameter differences between the stations are related to a measure of urbanization at the urban station [Ref. 2].
Using one or more of these methods, recent research on UHIs has lead to better description of urban climate modifications [Refs. 3, 4 and 5], development of mathematical models of urban climate change [Refs. 6, 7, 8 and 9] and comparison of causes for urban climate modifications based on model simulation [Refs. 10, 11 and 12]. The dominant causes for UHIs identified so far include, heat trapping by urban geometry, alterations to urban thermal properties, changes in vegetation cover and man-made (anthropogenic) heat input [Ref. 13].
This explosion of new knowledge on the theoretical aspects of urban climate change is not well matched by practical applications. In particular, urban designers and planners are yet to utilize the current knowledge base to develop architectural and urban design strategies for the mitigation of the negative effects of UHI [Ref. 14]. This is in part due to weaknesses in current methods. For example, some of the problems associated with remote sensing techniques hinder the detection of air temperature heat island that directly affect human comfort as opposed to surface temperature heat island. These problems include, difficulties in “seeing” the vertical active surfaces, the not so well defined coupling of surface and air temperatures in urban areas and inhomogeniety of urban surfaces leading to a patch work of emissivity and albedo [Refs. 15 and 16]. The problem with urban-rural difference method in general is that it assumes weather over time remains constant. Furthermore, the intra-urban differences are ignored. As Nichol [Ref. 16] points out, it is the intra-urban climatic difference that is of value for urban planners and designers interested in mitigating the negative effects of UHIs. As for the shortcomings with TAM and TCM, they both assume that rural climate is somehow “natural” to the area. In the context of rapid global urbanization, there are very few rural areas remaining with their “natural” climates intact [Ref. 17].
In this light, a “Paired Measurement Program” suggested by Myrup, et al., [Ref. 18] is worth pursuing. Here, microclimate at various urban sites are measured simultaneously along with a control site. The difference between a given site's microclimate and that at the control site is taken to be the “effect” due to urbanization at the said site. Thus the cause for intra-urban variations in climate parameters can be more readily ascertained.
The present study utilizes such a normalization technique to quantify the spatial variability of summer-time air temperature in Ann Arbor, Michigan, a medium-sized mid-latitude city (population 110,000, Latitude 42.3oN). The aim is to identify the effect of two land-use patterns and two land-cover types on intra-urban air temperature differences under different vegetation shade types. Recognizing that intra-urban micro-climate variations are also influenced by macro-scale weather patterns, the study explores the influence of land-use/land-cover under three atmospheric stability types: stable, neutral and unstable.
Two land-use classes (residential and commercial) and two land-cover types (paved and green surfaces) were considered for this study. The permeability of land-surfaces to water, role of industry (including automobiles) in the land-use category and the potential for evapotranspiration were the characteristics of distinction used in the selection of these four variables.
Based on the presence/absence of these variables and the degree of vegetation shade available, it was decided to select five sites in Ann Arbor, with the following land-use/land-cover patterns: |
1. High-density, high-rise, non-residential area with low greenery.
2. Low building density, low-rise, residential area with high greenery.
3. Medium density, mixed residential (some residential, some commercial / institutional) area with a greenery extent between (1) and (2) above.
4 & 5. Areas with similar land-use, building density and greenery, one having a more fully developed vegetation canopy than the other.
A search for locations with these characteristics resulted in the selection of the following sites:
1. Newman's Center, downtown Ann Arbor (DTL): A highly developed commercial area surrounded by buildings as tall as 26 stories accompanied by heavy traffic, high degree of paved areas and little or no greenery.
2. Nystuen Residence (HDR): A high density, medium-rise, older (50 to 75 years old) residential area with low traffic volume and heavy presence of mature vegetation.
3. University of Michigan Family Housing (NW2): An institutional/residential setting with substantial presence of parking lots, as well as significant greenery.
4. Brandle Residence (LVR): A low-density, low-rise, relatively new (less than 20 years old) single-family residential neighborhood with a high degree of vegetation presence that shade a significant portion of the street canyon.
5. Kim Residence (LOR): A low-density, low-rise residential neighborhood, similar in building density and land-use to (4) above, but only a smaller portion of the street canyon shaded by vegetation than at LVR (see Figure 1 for locations).
2.2 Measurement protocol
Air temperature was simultaneously measured at each experimental site and a reference station set up for the study at an open area near the Art & Architecture building of the University of Michigan. In close proximity to the reference station was a full-featured weather station maintained by the College of Architecture and Urban Planning. Data on several climate parameters, including air temperature, humidity, solar radiation, wind speed and global illuminance were available from this source to estimate general climate characteristics during the study period.
The reference station was located near a large open grass field on the northeastern side of town. The surrounding can be classified as low-density institutional land-use, with slightly undulating terrain, significant tree cover and some paved (parking lot) areas. Due to its more or less “rural” like nature, the difference between a location's air temperature and that at the reference station (DT) could be considered to be the “effect” due to urbanization at the said location.
After several initial trials, copper-constantan thermocouple was selected as the measurement probe due to its good performance in the outdoors. The probe was placed in a naturally-ventilated, aluminum foil-covered, insulated wooden box to block direct radiation. Probes used at the reference station and the urban sites were shaded similarly. The shielded probe was sited at an open area approximately 1.5m above ground and was connected to a Campbell Scientific Datalogger (Model 21X). Figure 2 shows a cross section through a typical street canyon where the probe was installed and a detail sketch of the shielding arrangement used.
Static calibration of the thermocouples used an ice bath procedure. Dynamic calibration was performed by running each shielded probe assembly for three days and comparing the temperature readings against the weather station data. The calibration error was ±0.3oC. Both the calibration and site climate data were gathered at 5 second intervals, averaged over a minute period and the minute data were stored in magnetic tape. For the purpose of statistical analysis, the minute data were averaged over an hour.
Data collection spanned a six-week period in June/July, 1995. Average high and low temperatures in June 1995 were 27.8oC and 15.1oC respectively, leading to a monthly average of 21.5oC. This was 2.2oC higher than the 98 year (1897 - 1994) historic average [Ref. 19] for June. July conditions were, average high 29.0oC, average low 17.5oC and average monthly temperature 23.3oC. The monthly average was 1.2oC higher than the historic average. In terms of rainfall, June 1995 was 2.2 cm drier than historical average while July 1995 was 3.6 cm wetter. However, there was little or no rain during the study days, except at HDR where substantial thunder activity was present on two days with heavy rain on one of the days.
Figure 3 shows weather conditions during the study period as measured at the reference station. It can be seen that the early part of the study period (June 8 - 12) was mostly overcast (over 80% cloud cover) and moderately windy (average wind speed = 3.3 m/s). The middle portion of the study period (June 13 - 23) was mostly clear (Average cloud cover = 15%) and somewhat calmer (average wind speed during day = 3.0 m/s). The latter part of the study period (June 24 - July 15) was partly cloudy (average cloud cover = 60%) but calm (average wind speed = 2.8 m/s).
2.3 Atmospheric stability estimation
The magnitude of microclimate modifications in urban areas depend on the atmospheric conditions at the macro-level. Since simultaneous measurements at all five sites were not made, it was necessary to classify the measurement period according to atmospheric conditions so that data from different sites could be compared.
It is said that wind speeds and cloud cover amounts are closely related to the timing and the magnitude of UHIs [Ref. 20]. Among other things, these two parameters also influence atmospheric stability. Fine sunny days with little wind usually lead to unstable atmospheric conditions near ground. Stable atmospheric conditions are generally associated with clear, calm nights while cloudy and windy days tend to produce neutral atmospheric conditions [Ref. 21]. Therefore, data on atmospheric stability near ground can be used to estimate the combined effects of wind speed and cloud cover on the development, timing and the magnitude of UHI.
Although many methods of estimating atmospheric stability near ground exist, the Pasquil-Turner index modified by Karlson [Ref. 22] appears the most relevant for the present study since it utilizes solar radiation and wind speed data only. Solar radiation being heavily influenced by cloud cover, the Modified Pasquil-Turner (MPT) index provides the best estimate of the combined influence of cloud cover and wind on intra-urban microclimate differences.
Karlson’s MPT is given by the following equation:
MPT = Q*/(u)2
where Q* = hourly average net radiation at 1.5 m above ground (Wm-2)
U = hourly average wind speed at 7.4 m above ground (ms-1).
The following MPT values were selected as cut-off points for the three atmospheric stability conditions examined in the present study:
MPT > 30 - Unstable
-10 < MPT < 30 - Neutral / Near neutral
MPT < -10 - Stable.
Data from the College of Architecture and Urban Planning weather station were used to determine the MPT values during the study period.
2.4 Land-cover estimation
Residential and commercial land-use calculations were based upon the 1990 Census Data [Ref. 23]. Streets and paved areas were estimated from digitized aerial photos taken in April 1990 [Ref. 24] using a computer-aided design software (MicroStation). Data from other sources [Refs. 25, 26 & 27] were used to verify the calculated residential land-use fractions. Green area was assumed to be land not covered by residential, commercial/ institutional or streets/paved land-uses. Land-cover/land-use within a 100 m. diameter circle around each of the climate measurement points was counted. Table 1 gives the land-cover estimates for the five sites.
Table 1: Land-use/Land-cover estimates
2.5 Data analysis protocol
During the study period, sun rose between 0557 - 0613 hrs. and set between 2106 - 2116 hrs., local daylight time. In order to completely avoid direct solar radiation, it was decided to use data from 2200 to 0459 hrs. local daylight time, as night-time data leading to a night-time temperature difference (DTN) and 0900 - 1859 hrs. local time, as the day-time data (leading to a day-time temperature difference - DTD). All temperature differences were calculated as site temperature minus reference temperature. Thus, a (-)ve DT indicates the site was cooler than the reference station.
Statistical analysis of the data employed a two-way Analysis of Variance (ANOVA) model of the following form:
Temperature Difference (DT) = constant + Time + Effect + Time*Effect,
where “effect” was the temperature variation caused by one of the following variables: atmospheric stability, vegetation shade or land-use/land-cover type.
The “effect” was considered significant if, the interactions term (Time*Effect) was insignificant and p-value for the “effect” was lower than the statistical level of significance (alpha). It was decided to set a for the study at 0.05%. Acceptance of the significance of “effect” was also contingent upon non-violation of assumption regarding homogeneity of data and normality of residuals. Bartlet's test for homogeneity and Lilliefors probability were the respective tests used for this purpose (see Neter et. al., [Ref. 28] for detailed description of these statistical tests).
Daily maximum temperature during the June study period ranged from 18.9 - 35.0oC. This variation covered the 33rd to 89th percentile of the historic daily June maximum temperature range for Ann Arbor. Daily minimum temperature during the June study period ranged from 7.2 - 20.6oC. This range covered the 34th to 68th percentile of the historic daily June minimum range. The maximum and minimum temperature ranges during the July 1995 study period were 22.8 - 36.7oC (30th - 93rd percentile of historical range) and 8.9 - 23.3oC (12th - 78th percentile) respectively. This shows that the study period covered air temperature ranges most commonly experienced during June and July. Since climatic conditions in June and July are representative of conditions in summer, it can be said that the data collected by the present study well represent summer-time conditions in Ann Arbor.
Figure 4 shows DTD variations during stable atmospheric conditions. The following were the average air temperature differences under stable conditions:
site DTL = +0.72oC
site LOR = -0.20oC
site LVR = -3.43oC
site NW2 = -2.05oC
site HDR = -3.13oC
The downtown site was the warmest (0.92 to 4.15oC warmer than the residential sites), the heavily wooded urban residential site (HDR) and suburban site LVR with fully developed vegetation canopy were the coolest (3.13 and 3.43oC respectively). The low-density residential site (LOR) experienced insignificant DTD (only 0.20oC cooler than the reference site). Among the intra-urban differences, only that between HDR & LVR was insignificant (Fisher’s LSD test p-value = 0.26). Except for site NW2, the hour-to-hour variation in air temperature during day-time was insignificant (p-value 0.08).
At mid-day, the intra urban air temperature differences were as high as 5.5oC. The average maximum day-time heat island under stable atmospheric conditions, was 4.15oC (DTL - LVR).
Statistical tests for day-time temperature differences under stable conditions did not violate assumptions regarding normality of residuals (skewness = 0.03; kurtosis = -0.46; lilliefors probability = 0.253).
Although only three sites experienced unstable conditions on multiple days (sites DTL, LVR & NW2), the patterns were very similar to those produced by stable conditions (see Figure 5). The average air temperature differences during unstable day-time were as follows:
site DTL = +1.47oC
site NW2 = -2.95oC
site LVR = -3.33oC
Here too, DTL was the warmest, maximum day-time UHI was about 4.8oC and hour-to-hour differences were insignificant (p-value = 0.676). But unlike stable conditions, differences among LVR and NW2 under unstable conditions were insignificant (Fisher’s LSD test p-value = 0.25).
Unlike the day time temperature differences, DTN showed a clear downtown-centered heat island (Figure 6).
site DTL = +1.51oC
site LOR = -0.84oC
site LVR = -0.24oC
site NW2 = -0.28oC
site HDR = -0.54oC
All residential sites were cooler than the reference site (from 0.24 to 0.84oC cooler) while the downtown location was up to 1.5 oC warmer. This lead to a maximum night-time air temperature heat island of about 2.4oC during the study period. The intra-urban differences among the four residential sites however, were very small (Fisher’s LSD test yielded p-values between 0.1 - 0.79). Air temperature at site HDR in particular was not significantly different from those at any other residential sites (Fisher’s LSD test p-value = 0.51 - 0.66). The highest night-time intra-urban air temperature difference was observed during early night period (between 2200 -2300 hrs). Unlike day-time, the hour-to-hour variation in air temperature during the night was very significant, particularly at the residential sites (p-value = 0.001). Here too, statistical tests did not violate assumptions regarding normality of residuals (skewness = 0.72; kurtosis = 1.20; lilliefors probability = 0.06).
3.3 Effect of vegetation shade
Site LVR had a more fully developed vegetation canopy than site LOR, but other land-use/land-cover characteristics were similar (see Table 1). A comparison of sites LVR & LOR can therefore reveal the effect of vegetation canopy structure on air temperature. Figs 7 (a) & (b) show such a comparison. During the day, vegetation-covered site LVR was 3.43oC cooler than the reference site while site LOR was only 0.2oC cooler. At night-time however, the trend reversed with LOR being 0.84oC cooler than the reference site and LVR only 0.24oC cooler. Furthermore, LVR rapidly heated up as the night progressed.
3.4 Effect of atmospheric stability
Three sites (DTL, LVR & NW2) experienced stable and unstable day-time conditions on multiple days, among them DTL showed the largest temperature differences (Figure 8). On average, temperature differences at DTL under stable and unstable atmospheric conditions were 0.96 & 1.6oC respectively (significance of difference: p-value = 0.04). At site LVR, the respective averages were -3.43 & -3.04oC (significance of difference: p-value = 0.02); at NW2, -2.66 & -2.95oC (significance of difference: p-value = 0.03). At night, stable and unstable conditions at DTL produced 1.51 and 0.51oC average differences respectively (significance of difference: p-value < 0.001) (see Figure 9).
The variations in air temperature under stable and unstable conditions are also worth noting. At DTL, not only the mean difference was higher than any of the residential sites, but also the standard deviations (0.23 and 0.66oC under stable and unstable conditions respectively). At LVR and NW2 however, variations were much smaller (std. deviations were 0.61 & 0.58 and 0.54 & 0.49oC respectively). Thus, the warmest site (DTL) showed significant differences under stable and unstable conditions, both in the average values and in variations while differences at residential sites were not so clear.
Neutral/near neutral conditions too produced temperature patterns similar to those under stable conditions: downtown was the warmest both at day and night, maximum differences occurred during early night-time (2200 - 2300 hrs), and the hour-to-hour temperature variations during day-time tend to be relatively small. However, lack of sufficient of data preclude definite conclusions being made about neutral atmospheric conditions.
4.0 Urban design implications
The above discussions reveal four underlying themes that have urban design implications: relative temperature effects of vegetation and building density, primary temperature effect of vegetation, temperature variability of downtown areas with atmospheric stability and cool islands vs. heat islands.
Higher vegetation presence was shown to lead to significant cooling, but this need not occur at the expense of higher building densities. For example, site HDR has less than 65% tree cover and 21% paved areas with a residential density of 10.5%. Site LVR on the other hand has much less residential density (4.5%), lower paved areas (12.6%) and higher green cover (over 78%). Yet DTD at HDR and LVR were very similar (-3.13 & -3.43oC respectively under stable conditions). Night-times produced insignificant differences (-0.54 & -0.24oC respectively).
Secondly, the primary effect of vegetation shade is to reduce the day-time temperatures and increase the night-time values. In other words, even as vegetation reduces direct solar penetration during the day, it also inhibits night-time cooling. Thus, higher vegetation cover may not be ideal for urban areas during times of small diurnal variations (e.g. summer-time). Under these conditions, higher presence of vegetation may improve the day time comfort somewhat but this improvement may be more than offset by warm night-time conditions. Furthermore, small diurnal temperature variation during particularly warm spells of summer weather, is known to be harmful to vulnerable sections of the population. This was amply demonstrated by the exceptionally warm weather experienced during the summer of 1995 in the mid-western United States that led to the death of more than 800 people (mostly senior citizens) in the region [Ref. 29]).
Thirdly, it appears that atmospheric stability plays an important role in urban air temperature variations in heavily built-up areas of the city, but not in the less dense sections. Atmospheric stability not only influences the magnitude of temperature differences in the downtown area, but also the hour-to-hour variations. A temperature modulating strategy like planting trees may therefore be more beneficial in the downtown than in the less densely built residential areas.
Finally, urban designers need to pay more attention to the inhomogeniety of air temperature patterns particularly during the day. The general impression created by the concept of UHI -- a downtown centered heat island -- appears to be true only during the night. In contrast, many cool islands exist throughout the city during the day, and their nuclei are related to the vegetation distribution in the locality. Urban design strategies need to take advantage of the positive influence of such cool islands, for example, by manipulating their location with respect to prevailing wind patterns.
The present study explored the effects of vegetation shade and land-use/land-cover type on intra-urban air temperature variations under different atmospheric stability conditions. Unstable day-times show the maximum heat island effect. Under stable conditions, significant temperature differences between the downtown location and each of the residential locations was seen, but not between the residential locations. Under stable conditions however, both downtown-to-residential as well as residential-to-residential differences were significant.
Vegetation shade showed little impact on the intra-urban variations in DTN. However, shaded sites heated up much more slowly than open ones and this lead to cooler day-time temperatures in many parts of the city. Such “cool islands” in residential areas and heat island in the downtown during the day contrast with the night-time situation where a more homogeneous, downtown-centered heat island was found.
It may be worth noting here that regressing land-use/land-cover variables on day- and night-time temperature differences revealed that tree cover had marginally significant correlation to DTN (p-value = 0.061) and to DTD (p-value = 0.11). Such marginal correlation may be due to the coarseness of land cover data used in the present study. Also, the influence of other climate modulating factors, especially the effect of waterbodies, were not considered. Other studies, [Ref. 30] have compared remotely sensed vegetation data with urban/rural temperature differences and showed that vegetation correlates well with temperature differences. If intra-urban temperature differences could be correlated to vegetation density, more robust urban climate change models could be built. This in turn could lead to better prediction of the UHI effects and subsequent development of better UHI mitigation strategies.
A multi-method research approach could help the development of vegetation-based intra-urban temperature variation models. It is envisaged that remote sensing techniques will be used to quantify urban vegetation and paved areas. Such monitoring need not be made on a daily basis and as such, can be carried out on clear days only. Climatic variations in a city can be recorded using a “paired measurement program” such as the one employed in the present study. The combination of these two methods will yield fine resolutioned urban land cover data (as opposed to the rough estimates used here) that could be more satisfactorily linked to intra-urban climate variations.
Acknowledgments The following persons gave permission for setting up temperature probes in their premises: Professors Jong-Jin Kim, Kurt Brandle, John Nystuen and William Stevenson, Pastor, St. Mary’s Church, Ann Arbor. Their help is gratefully acknowledged.
1. GOLDREICH, Y: “Urban Climate Studies in Israel - A Review”, Atmospheric Environment, 29(1995): 467-478.
2. KARL, T.R., H.F. DIAZ & G. KUKLA: “Urbanization: Its Detection & Effect in the United States”, J. Climate, 1(1988):1099-1123.
3. NIEUWOLT, S: “The Urban Microclimate of Singapore”, J. Tropical Geography, 22(1966): 30-37.
4. SANI, SHAM: “Observations on the Effect of a City Form and Functions on Temperature Patterns”, J. Tropical Geography. 36(1973): 60-65.
5. LANDSBERG, H.E: The Urban Climate. New York: Academic Press, 1981.
6. MYRUP, L.O: “A Numerical Model of the Urban Heat Island”, J. Applied Meteorology, 32(1969):896-907.
7. ACKERMAN, T.P: “A Model of the Effects of Aerosols on Urban Climate with Particular Application to the Los Angeles Basin”, J. Atmospheric Sciences, 34(1977): 531-547.
8. CARLSON, T.N. & F.E. BOLAND: “Analysis of Urban-Rural Canopy Using a Surface Heat Flux/Temperature Model”, J. Applied Meteorology, 17(1978): 998-1013.
9. MILLS, G.M: “Simulation of the Energy Budget of an Urban Canyon - Part I: Model Structure & Sensitivity Test”, Atmospheric Environment, B, 27B(1993): 157-170.
10. LEWIS, J.E., & T.N. CARLSON; “Spatial Variations in Regional Surface Energy Exchange Patterns for Montreal, Quebec”, The Canadian Geographer, 33(1989): 194-203.
11. OKE, T.R., G.T. Johnson, D.G. Steyn & I.D. Watson: “Simulation of Surface Urban Heat Islands Under 'Ideal' Conditions at Night. Part 2: Diagnosis of Causation”, Boundary-Layer Meteorology, 56(1991): 339-358.
12. SAILOR, D.J: “Simulated Urban Climate Response to Modifications in Surface Albedo & Vegetative Cover”, J. Applied Meteorology, 34 (1995): 1694-1704.
13. EMMANUEL, R: “A Hypothetical ‘Shadow Umbrella’ for Thermal Comfort Enhancement in the Equatorial Urban Outdoors”, Architectural Science Review, 36(1993): 173-184.
14. DE SCHILLER, S. & J.M. EVANS: “Training Architects and Planners to Design with Urban Microclimates”, Atmospheric Environment, 30(1996): 449-454.
15. ROTH, M., T.R. OKE, & W.J. EMERY: “Satellite-derived Urban Heat Islands from Three Coastal Cities & the Utilization of Such Data in Urban Climatology”, International J. Remote Sensing, 10(1989): 1699-1720.
16. NICHOL, J: “A GIS-Based Approach to Microclimate Monitoring in Singapore's High-Rise Housing Estates”, Photogrammetric Engineering & Remote Sensing, 60(1994): 1225-1232.
17. CHANDLER, T.J: Urban Climatology and its relevance to Urban Design. WMO Technical Note No. 149, Geneva: World Meteorological Organization, 1976.
18. MYRUP, L.O., C.E. MCGINN & R.G. FLOCCHINI: “An Analysis of Microclimatic Variation in a Suburban Environment”, Atmospheric Environment, 27B (1993): 129-156.
19. NCDC SUMMARY OF THE DAY: Hydrodata for Windows, version 3.0, Boulder, CO: Hydrosphere Data Products, 1996.
20. LANDSBERG, H.E: The Urban Climate. New York: Academic Press, 1981.
21. OKE, T.R: Boundary Layer Climates. London: Methuen, 1987.
22 KARLSON, S: “The Applicability of Wind Profile Formulas to an Urban-Rural Interface”, Boundary-Layer Meteorology, 34(1986): 333-355.
23. CENSUS BUREAU: 1990 Census. Washington, D.C.: Bureau of the Census, 1993.
24. PLANNING DEPT: Aerial Photographs - April, 1990. Ann Arbor: Planning Dept., 1990.
25. PLANNING COMMISSION: Comprehensive Housing Affordability Strategy. Ann Arbor: City Council, Planning Dept., Housing Policy Board & Community Development Executive Committee, 1991.
26. PLANNING DEPT: Ann Arbor Downtown Data - Block Data. Ann Arbor, MI.: Planning Department, 1983.
27. PLANNING DEPT: Inventory of Vacant, Industrial & Research Sites, Ann Arbor: Planning Dept., 1984.
28. NETER, J., W. WASSERMAN & M.H. KUTNER: Applied Linear Statistical Models, 3rd ed., Homewood, IL (USA): Irwin, 1990.
29. CHANGNON, S.A., K.E. KUNKEL & B.C. REINKE: “Impacts and Responses to the 1995 Heat Wave: A Call to Action”, Bulletin of American Meteorological Society, 77(1996): 1497-1506.
30. GALLO, K.P., A.L. MCNAB, T.R. KARL, J.F. BROWN, J.J. HOOD & J.D. TARPLEY: “The Use of NOAA AVHRR Data for Assessment of the Urban Heat Island Effect”, J. Applied Meteorology, 32(1993):899-908.
Return to Summary of Main Findings
Go To Research on Urban Climate Modifications
Return to Rohinton Emmanuel's Home Page