URBAN VEGETATIONAL CHANGE AS AN INDICATOR OF DEMOGRAPHIC TRENDS IN CITIES: THE CASE OF DETROIT

Environment & Planning - B 24(1997): 415-426

Rohinton Emmanuel
College of Architecture & Urban Planning
The University of Michigan

Abstract

Urban vegetation impinges upon the physical environment by positively enhancing the micro-climate, air and water quality. On the other hand, changes to the physical environment as well as the socio-economic conditions of urban dwellers affect urban vegetation health, species abundance and diversity. This two-way vegetation-environment dynamism makes urban vegetation an ideal meter to gauge the environmental health of cities. While studies utilizing such a proposition exist with reference to the physical environment, vegetation's usefulness in monitoring urban social change has received very little attention. This study seeks to find whether changes in urban vegetation can be linked to urban social changes by using Detroit as a case study. Demographic trends in Detroit are analyzed in light of the increasing greenness in the city detected by recent satellite images. Robust relationships between greenness change and demographic factors associated with urban decay (population decline, poverty level increase and vacant unit increase) are found. Models built with a remotely sensed greenness change index and urban decay variables appear to be free of any serious bias. Tests for the validity of such models are also successful. On the basis of this finding, the current study concludes that vegetational trends could be used as indicators of urban socio-economic changes. A vegetation-based urban environmental quality index could therefore be developed to monitor physical and social changes in cities.

Keywords: urban environmental quality, urban decay, remote sensing

Background

Contrary to popular perceptions, there is much vegetation in cities. Morgan (1972) and Moll (1989) estimates that 30 - 36% of surfaces in typical north American cities are vegetation-covered. The leaf and stem area of urban vegetation outnumbers man-made surfaces 4-to-1 (Rowntree, 1984).

This large volume of urban vegetation impinges upon the climatic, air and water quality of urban areas as well as the social and psychological well being of urban dwellers. Some of the recorded climatic effects of urban vegetation include, up to 15oC reduction in surface temperature during hot, summer days in Miami, Fl. (Parker, 1983); up to 25% reduction in building climate-control energy needs in the eastern United States (Heisler, 1986) and 6oC air temperature reductions within 2 mile radii of Mexico City urban parks (Jauregui, 1973). In addition, urban vegetation improves the surface water flow by the quantitative and qualitative regulation of runoff (Rutter, 1972; Shuttleworth, 1989), sequester CO2 and other trace gas pollutants (Nowak, 1994), enhances human thermal comfort (Stark & Miller, 1977; Huang, et al., 1987), adds to the psychological well-being of urban dwellers (Ames, 1980; Ulrich, 1984; Willeke, 1989) and enriches the urban bio-diversity (Hough, 1984).

While urban environment is affected usually positively by vegetation, workers have also recognized that the environment/vegetation dynamics is a two-way cyclical process. The additional stresses associated with urban environment (thermal and air quality stresses) affect the health of urban vegetation. Such environments favor certain non-native, highly stress-resistant plants over native ones. The consensus seems that in the urban areas, environmental effects have a greater influence over vegetation in the short run, while vegetation adapts to and changes the urban environment over a long period of time (Sukopp, 1982; Kowarik, 1990; Stulpnagel, et al., 1990).

This two-way interaction between vegetation and environment enable urban vegetation to be an indicator of the environmental quality of cities. Thomas (1973) reviewed research on using vegetation as an environmental quality monitor. More recently, urban vegetation as a bio-monitor has been employed by Westman & Price, 1988, Nowak & McBride, 1993, Mishra, et al., 1994, Sloof, 1995 (air quality); Gallo, et al., 1993; Suzuki, et al., 1993; Nichol, 1994 (air temperature) and Kawashima, 1994 (surface temperature).

The present study investigates the feasibility of using urban vegetational changes to monitor urban social and demographic changes by taking the case of Detroit as an example. This aspect of urban vegetation use seems missing in current discussions. Although it is taken for granted that urban vegetation is largely human-initiated and therefore influenced by the social and economic circumstances of urban dwellers, the possibility of using vegetational changes as indicators of urban social change remains largely unexplored.

The initial impetus for the current study came from a thematic comparison of two LANDSAT satellite images over Detroit made in autumn of 1975 and 1992. This comparison revealed a greening trend in the heart of the city. It is well known that urban decay has long been the trend in Detroit, resulting in large tracts of vacant land in the central city area (Bruhn & Brahn, 1990). However, the associated greening makes Detroit unique among the world's urban centers for two reasons.

• Detroit seems to have gone through the full cycle of urban demographic change (establishment as a small trade post, relatively slow growth, a period of rapid and chaotic change, decline in population and eventual abandonment). As such, it offers lessons for cities yet to go through the full cycle of transitions;
• Urban decay and vegetation increase in Detroit may be correlated to demographic factors, and if so, may offer lessons for the world's mature urban agglomerates that are also going through similar demographic transitions.

In order to quantify the greenness changes in Detroit, a "Greenness Change Index" (GCI) was developed by comparing greenness amounts from the above-mentioned satellite images. The amount of greenness per pixel was computed on the basis of the spectral reflectances in the near infra-red (NIR) and the visible bands, using a data transformation procedure called "tasseled cap transformation" and the change in greenness was calculated by a process named "change vector analysis". Both these procedures are described below in detail.

It was hypothesized that demographic changes in Detroit would be reflected in its vegetation trends. Major demographic events such as population changes and poverty level status of urban dwellers were expected to have an impact upon quantitative changes in urban vegetation. Specifically, social changes associated with urban decay (population decline, poverty level increase and vacant lot increase) were expected to lead to an increase in urban vegetation. A decrease in vegetation was expected for areas with vibrant urban social fabric.

Tasseled-Cap Transformation

When chlorophyll-containing leaves are exposed to sun, they absorb most of the blue and red light in the visible spectrum (thus appear green to the human eye), but reflect a high amount of near infra-red (NIR) wavelengths. Annual range of reflectance in the visible spectrum is 4.0 - 6.9% for urban vegetation and 8.1 - 9.6% for built-up areas. Reflectance in the NIR wavelengths however, ranges from 12.4 - 34.0% for urban vegetation and only 9.0 - 14.5% for built-up areas (Brest, 1987). Remote sensing platforms comprising of visible and NIR bands utilize this contrast to detect the "greenness" of a surface.

Fig. 1 shows a typical scatter plot of spectral reflections over urban areas in the NIR and the visible (red) bands of the LANDSAT Multispectral Scanner (MSS) platform. The data points appear to fall within a triangular area, elongated along two diagonal axes. Crist & Kauth (1986) named these two axes as "brightness" (indicating variations in soil reflectance) and "greenness" (indicating variations in vegetation cover) components. In the case of urban areas, "brightness" indicates the amount of built-up surfaces. Although LANDSAT Thematic Mapper (TM) platform exhibits a somewhat different data structure, the removal of data from Bands 5 & 7 of the TM platform produces a similar shape (Crist & Kauth, 1986) and make it comparable to the MSS platform.

Rotating the viewing plane of Fig. 1 so as to align the "greenness" axis with the vertical and the "brightness" axis with the horizontal, simplifies the visual interpretation of greenness or the brightness of a data scene. Principal component analysis can determine the two axes themselves and the displacement of a given data point from the "greenness" (or "brightness") axis can thus be estimated. The "greenness" (or "brightness") of a reference green (or built up) surface can be used to determine threshold values above which a given surface could be considered "green" (or built-up). Kauth & Thomas (1976) called this transformation process the "tasseled cap transformation" (TCT) of LANDSAT data for the quantification of greenery. The name comes from the apparent tasseled cap shape of the data structure. The validity of the procedure to quantify greenness was shown by Crist & Cicone (1984) using actual and simulated TM data. Extensive use of TCT has been made in recent years to quantify agricultural as well as urban greenery (see Crist & Kauth, 1986 and Crist et al., 1986 for reviews).

Change Vector Analysis

A comparison of two tasseled cap-transformed data scenes taken over time can reveal the temporal changes in greenness at a given location. Such a comparison, called a "change vector analysis" (CVA) was used by Michalek, et al., (1993) and Luczkovich, et al., (1993) to quantify changes in coral reef formations. CVA computes the magnitude and direction of spectral reflection change per pixel by determining the Euclidean distance between the change space thus: ,

where X1 and X2 are the pixel values at date 1 and date 2 respectively in the ith LANDSAT band (Wagner, et al., 1994).

For the purpose of determining the greenness change between the two images used in the present study, 3 bands for each date (2 visible and one NIR) were used. By color coding each pixel according to the nature (positive or negative) and type (greenness, brightness, etc.) of change, it is possible to create a thematic change vector map - in this case a greenness change vector map of the city of Detroit.

Data Processing

Using the TCT procedure, greenness values of pixels in a satellite data scene taken over Detroit (LANDSAT MSS platform; ID no: 4421-574, scene date: May 10, 1975) and the greenness values of a data scene over essentially the same area but taken in 1992 (LANDSAT TM platform; ID no: 5-020/03070; scene date: May 16, 1992) were calculated. Since the pixel sizes were different for the two platforms, a re-sampling procedure was used to reduce the pixels to 25m x 25m. Aerial photographs of 10m pixel resolution taken over Detroit during the same time period were used for this purpose. Such photographs have been taken at regular intervals to monitor the physical environmental changes in the SEMCOG (South-East Michigan Council of Governments) region for the past 40 years.

A change vector analysis was then performed to determine the type and magnitude of greenness change between 1975 and 1992 in the Detroit area. Pixels with increase in greenness value from 1975 to 1992 were coded green while those with greenness decrease were coded red. No change in greenness between the two images was coded black. Fig. 2 shows portion of the city of Detroit thus color coded. Finally, a greenness change index (GCI) was computed by dividing all the pixels in a given area displaying the color green by the number of red pixels:

GCI = No. of Green Pixels/Total No. of Pixels = No. of Green Pixels (2)
No. of Red Pixels / Total No. of Pixels No. of Red Pixels

GCI values less than 1 indicate a predominance of vegetation loss in a given area; values greater than 1 indicate spatial and temporal increase in greenness.

The social and demographic conditions of the city of Detroit were determined from the 1980 and 1990 decennial census of persons and households (U.S. Bureau of the Census, 1985, 1994). Although the change periods for population and greenness are different (10 and 17 years respectively), the fact that Detroit has experienced continuous population decline since the 1940s and at nearly constant rates (Doxiades, 1966; Emmanuel, 1994) enable one to assume that the comparison of demographic changes between 1980 to 1990 and greenness change from 1975 to 1992 is valid. However, this may under-estimate relationships, if any, between them.

In analyzing the social environmental quality in Detroit, Nystuen, et al., (1995) found that changes in population, level of poverty, presence of minors and vacant housing units were good indicators to urban decay. The current study limited its demographic data collection to changes in population, poverty level and vacant housing units only, as these can be expected to more directly influence changes in urban greenness.

The geographical units of measurement for the dependent variable GCI as well as the independent variables were the 1990 census tracts. Any incompatibility between 1980 and 1990 census tracts were resolved by averaging the variables thus: for tracts that have been broken into n smaller units in 1990 from a larger 1980 tract, the 1990 values were added and the difference between the 1990 amalgamated value and the 1980 value was equally distributed among the n 1990 tracts. For tracts that were joined together in 1990 from two or more 1980 census tracts, the difference between the 1990 value and the 1980 amalgamated value was used. On this basis, it was possible to include 317 of the 321 census tracts in Detroit. Four cases had to be dropped due to lack of data for all three independent variables selected for analysis (changes in population, poverty level and vacant housing units).

Since the purpose of the study was to determine correlation, if any, between demographic variables and greenness change, it was decided to build a model with all the hypothesized variables in the model. In the absence of independent data for validating this model, a prediction model was built using 80% of the selected data-sets (the model-building data set) and the balance cases were used to build a validation model (the validation data set). All models were of the multiple linear regression type. Logarithmic transformation of data was carried out to see if better fit could be achieved.

Parametric statistical procedures were used for model validation. Tests for validating the prediction and the validation models included the following: comparison of two linear regression equations for equal variance, identical regression, equal slope and equal intercept (cf. Neter et al., 1990). On the validation data set, the predicted GCI values using the prediction model were also compared against the observed GCI values. Following tests were carried out to validate the predictive abilities of the model:

• Comparison of Mean Square Prediction Error (MSPE) with the Mean Square Error (MSE) of the prediction model. If the two are not very different, it can be concluded that the model is not seriously biased (Neter, et al., 1990: 466).
• Regressing the predicted GCI on observed GCI, using the validation data set. If a simple linear regression line through the origin and at a slope of 45o can be fitted, more confidence may be placed on the suitability of the prediction model to forecast changes in GCI (i.e. a given increase in the observed GCI results in similar increase in the predicted GCI).
• Comparing the absolute and relative errors against the observed GCIs. For the purpose of these comparisons, absolute and relative errors are thus defined: Absolute Error (AE) = Predicted GCI - Observed GCI (3) Observed GCI Relative Error (RE) = Absolute Error
• Observed GCI These comparisons would reveal systematic prediction errors (under or over-prediction) for a given GCI value and the magnitude of such errors. For the purpose of the current study, the level of statistical significance (a) was 5%.

Results and Discussion

Figures 3(a) - (c) show scatter plots of each of the selected independent variables against the GCI. Since none of the independent variables were highly correlated among themselves (Pearson's correlation < 0.53), a step-wise regression was carried out to find out the best model that could fit some or all of the independent variables to the GCI. Table 1 gives the summary statistics.

Although the relationships are statistically significant (highest p-value = 0.02), the variance in GCI explained by the independent variables is very low (not more 35.1% or R2 = 0.351). Even polynomial or exponential models did not significantly improve the R2 values (R2 increased up to 0.4).

An inspection of Figs. 3(a)-( c) reveals that there could outliers which may be responsible for the low R2 value. A field survey was therefore carried out by the author and others to identify the ground conditions that might have caused these anomalies. Two kinds of conditions could have caused the outliers.

• A drastic decrease in population but not a significant increase in greenness: This was primarily due to the expansion of industrial plants into residential neighborhoods. For example, the expansion of a Chrysler Motor Co., plant caused large exodus of people without compensating increase in greenness (a population decline of 2242 persons [> 70%] and a GCI of less than 0.2 in census tract no. 5176).
• A small increase in greenness in areas with a massive influx of population: In areas that underwent urban renewal, population dramatically increased without significant vegetation change. For example, low-income, multi-family residential development in census tract no. 5208 resulted in a population increase of 1700% over the 1980 value while greenery hardly changed. A housing project for senior citizens in tract no. 5236 also resulted in a similar anomaly.

Furthermore, Fig. 3(a) also indicates that there could be two distinct relationships between population change and GCI: Population decline vs. GCI and population increase vs. GCI. It was therefore decided to split the data set according to the nature of population change (decrease or increase). Data cases with unusually high population changes (> 500%) were also omitted. The resultant data set had 315 cases (loss of 2 data cases) with 237 cases of population decrease and 68 cases of population increase. Population decline cases were then randomly divided into model-building (191 cases) and validation data sets (46 cases). Cases with population increase were separately analyzed.

Comparison of prediction & validation models for population decline cases

Step-wise regression produced the following best fit models from the model-building and validation data sets respectively: (5) (6) All three independent variables are (%) changes in the respective demographic factors over the 1980 values and are log transformed. The GCI is a unit-less greenness fraction also log transformed. The R2 for eqns. (5) and (6) were 0.54 and 0.62 respectively.

Table 2 presents the results of the tests carried out to determine whether the two models were significantly different from each other. It can be seen that the models came from data sets with equal variance (p-value = 0.75) thus validating the random data splitting used. Both the prediction and the validation models had significantly similar regression (p = 0.74) and had statistically identical slopes (p = 0.84) and intercepts (p > 0.5).

These results appear to show that the prediction model may satisfactorily estimate the GCIs in the validation data-set as well. Figures 4(a)-(c) show the results of regressing the observed GCI of the validation data set on predicted GCI using eqn. (5). It can be seen that over- and under-predicted cases are quite evenly matched (27 and 19 respectively). This indicates that there are no systematic prediction errors. The MSPE was 1.036 while the MSE for the prediction model was 1.253. Only four cases had relative errors larger than 1 standard deviation. Thus, the prediction model can be said to be free of any serious bias. Table 3 summarizes the statistical results of the comparison.

Population increase cases

Table 4 shows some of the results of step-wise regression using 68 cases with population increase. Population increase does not appear to have very strong relationship with greenness change. Given the diversity of factors that lead to population increase in sections of Detroit, this was to be expected. Another reason could be the scarcity of data cases (only about 21% of the census tracts in the city exhibited population increase).

Summary

It appears that variables associated with urban decay (population decline, poverty level increase and vacant unit increase) explain a significant portion of variance in the greenness changes in Detroit. The healthy social conditions associated with population increase on the other hand, do not appear to have very strong relationships with greenness change. The prediction and validation models built with population decline data cases were also shown to be statistically similar. However, the model tend to under-predict greenness change at higher GCI values (the slope of predicted vs. observed GCI regression line was less than 1 - see Fig. 4[a]).

All the assumptions regarding normality and homogeneity of residuals were valid. Thus it is appropriate to pool all 237 population decline data cases together and build a final model indicating the relationship between GCI and demographic variables associated with urban decay: (7) This relationship is statistically significant (p = 0.001) and explains over 59% of the variance in GCI. Furthermore, the residuals do not violate assumptions regarding normality or homogeneity.

Implications

The present study indicates that urban vegetational change could be a valid proxy for social changes in cities. Since urban vegetation reflect the physical environmental qualities as well (air, water and climate), it can be used as a holistic environmental indicator. Currently, the lack of such an integrated environmental quality indicator hampers the development of environmental change mitigation policies. Since the magnitude of urban environmental change is usually unknown, it is also difficult to decide on the vigor with which a given mitigation policy option should be applied.

The lack of an integrated environmental change indicator also impedes comparison among urban entities as to the relative status of their social and physical environments. Furthermore, the collection of data required for the establishment of an urban environmental quality index using conventional methods is a costly and time consuming enterprise. An easy-to-develop environmental quality indicator such as urban vegetation change could be an invaluable tool for planning agencies at the national level.

As global urbanization accelerates, monitoring urban environmental changes will become an important concern for both citizens and policy makers. This is especially the case in the rapidly expanding third world cities where the cost of relevant data collection alone is prohibitive. The ready availability of remotely sensed greenness data and transformation procedures such as TCT could simplify the task for these cities as well.

Acknowledgments

Tasseled-cap transformation of LANDSAT MSS and TM data and the process of color coding of pixels from the 1975 and the 1992 satellite images were performed by the Environment Research Institute of Michigan (ERIM). The help of Thomas Wagner of ERIM in gaining access to the transformed images is gratefully acknowledged. The comments and suggestions by an anonymous reviewer was helpful in clarifying the arguments made in the paper.

References

• Ames, R.G., (1980). "The sociology of urban tree planting", J. Arboriculture, 6(5): 120-123.
• Brest, C.L., (1987). "Seasonal albedo of an urban/rural landscape from satellite observations", J. Climate & Applied Meteorology, 26: 1169-1187.
• Bruhn, M.S., & J.F. Brahn, (1990). Survey & recommendations regarding vacant land in the city: Report submitted to the city council of Detroit, Detroit: city planning commission, (unpublished).
• Crist, E.P. & R.C. Cicone, (1984). "Application of the tasseled cap concept to simulated Thematic Mapper data", Photogrammetric Engineering & Remote Sensing, 50(3): 343-352.
• Crist, E.P. & R.J. Kauth, (1986). "The Tasseled Cap de-mystified", Photogrammetric Engineering & Remote Sensing, 52(1): 81 - 86.
• Crist, E.P., R. Laurin & R.C. Cicone, (1986). "Vegetation and soils information contained in transformed thematic mapper data". In, ESA Proceedings of the 1986 International Geo-Science & Remote Sensing (IGARSS '86) Conference, Today's Solutions for Tomorrow's Information Needs, IEEE catalog No. 86CH2268-1, vol 3, pp. 1465-1470, Paris: European Space Agency.
• Doxiades, C., (1966). Emergence & Growth of an Urban Region: The Developing Detroit Urban Area. Detroit: Detroit Edison Co.
• Emmanuel, R., (1994). "A city in transition: urban demographic changes in Detroit & their impact on urban greenness & climate". In, Drake, W.D. & S.L. Arlinghaus, (eds.), Population-Environment Dynamics: Towards Public Policy Strategies, Ann Arbor, MI.: Gryphon Publishing, pp. 77 - 113.
• Gallo, K.P., A.L. McNab, T.R. Karl, J.F. Brown, J.J. Hood & J.D. Tarpley, (1993). "The use of a vegetation index for the assessment of the urban heat island effect", Int'l. J. Remote Sensing, 14(11): 2223-2230.
• Heisler, G.M., (1986). "Effects of individual trees on the solar radiation climate of small buildings", Urban Ecology, 9: 337-359.
• Hough, M., (1984). City Form and Natural Processes, New York: Van Nostrand Reinhold.
• Huang, Y. J., H. Akbari, H. Taha & A.H. Rosenfeld, (1987). "The potential of vegetation in reducing summer cooling loads in residential buildings", J. Climate & Applied Meteorology, 26: 1103- 1116.
• Jauregui, E., (1973). Untersuchungen um Stadtklima von Mexico-Stadt, Doctoral thesis, Rheinischen Friedrich-wilhelms-Universitat. (cited by Oke, T.R., [1989]. op. cit.).
• Kauth, R.J. & G.S. Thomas, (1976). "The Tasseled Cap - a graphic description of the spectral-temporal development of agricultural crops as seen by Landsat". In, Proceedings of a Symposium on Machine Processing of Remotely Sensed Data, Purdue University, West Lafayette, Indiana, 21 June - 1 July, 1976, pp. 4B41-4B51.
• Kawashima, S., (1994). Relation between vegetation, surface temperature & surface composition in the Tokyo region during winter", Remote Sensing of Environment, 50: 52-60.
• Kowarik, I., (1990). "Some responses of flora & vegetation to urbanization in central Europe". In, Sukopp, H., et al., (eds), Urban Ecology: Plants & Plant Communities in Urban Environments, The Hague: SPB Academic Publishing, pp. 45-74.
• Luczkovich, J.J., T.W. Wagner, J.L. Michalek & R.W. Stoffle, (1993). "Discrimination of coral reefs, seagrass meadows & sand bottom types from space: a Dominican Republic case study", Photogrammetric Engineering & Remote Sensing, 59(3): 385-389.
• Michalek, J.L., T.W. Wagner, J.L. Luczkovich & R.W. Stoffle, (1993). "Multispectral change vector analysis for monitoring coastal marine environments", Photogrammetric Engineering & Remote Sensing, 59(3): 381 - 384.
• Mishra, J.K., R. Aarathi & M.D. Joshi, (1994). "Remote sensing quantification & change detection of natural resources over Delhi", Atmospheric Environment, 28(19): 3131-3137.
• Moll, G., (1989). "In search of an ecological landscape". In, Moll, G. & S. Ebenreck, (eds.), Shading our Cities: A Resource Guide for Urban and Community Forests, Washington, D.C.: Island Press, pp. 13-24.
• Morgan, D.L., (1972). Man-Environment Coupling: Physiologic Comfort of Man in an Urban Setting, Ph.D. dissertation, University of California, Davis.
• Neter, J., W. Wasserman & M.H. Kutner, (1990). Applied Linear Statistical Models, 3rd edition, Homewood, IL.: Irwin.
• Nichol, J.E., (1994). "A GIS-based approach to microclimate monitoring in Singapore's high-rise housing estates", Photogrammetric Engineering & Remote Sensing, 60(10): 1225 - 1332.
• Nowak, D.J., (1994). "Atmospheric carbon dioxide reduction by Chicago's urban forest". In, McPherson, E.G., D.J. Nowak & R.A. Rowntree, (eds.), Chicago Urban Forest Climate Project (CUFCP): Final Report. USDA Forest Service Technical Report, NE-186. Upper Darby, PA.: USDA Forest Service, Northeast Experiment Station, pp. 83-94.
• Nowak, D.J. & J.R. McBride, (1993). "Testing microdensitometric ability to determine Monterey Pine urban tree stress', Photogrammetric Engineering & Remote Sensing, 59(1): 89-91.
• Nystuen, J. D., R.W. Marans & R.M. Ryzner, (1995). "Open space & vegetation in a declining city". Paper presented at the 37th Annual Association of Collegiate Schools of Planning (ACSP '95) Conference, October 19-22, Detroit, MI.
• Oke, T.R., (1989). "The micrometeorology of the urban forest", Philosophical Transactions of the Royal Society of London B, 324: 335-350.
• Parker, J., (1983). "Do energy conserving landscape work?" Landscape Architecture, July 1983: 89-90. Rutter, A.J., (1972). Transpiration. London: Oxford University Press.
• Rowntree, R.A., (1984). "Ecology of the urban forest", Urban Ecology, 8:1-11.
• Shuttleworth, W.J., (1989). "Micrometeorology of tropical & temperate forest", Philosophical Transactions of the Royal Society of London, B, 324:299-334.
• Sloof, J.E., (1995). "Lichens as quantitative biomonitors for atmospheric trace element deposition, using transplants", Atmospheric Environment, 29(1): 11-20.
• Stark, T.F. & D.R. Miller, (1977). "Effect of synthetic surfaces & vegetation in urban areas on human energy balance & comfort", Proceedings of a Conference on Metropolitan Physical Environment, Syracuse, NY., 25-29, Aug., 1975. USDA Forest Service Technical Report NE-25, Upper Darby, PA.: Northeastern Forest Experiment Station, pp. 139-151.
• Stulpnagel, A V., (1990). "The importance of vegetation for urban climate", in, Sukopp, H., et al., (eds.), Urban Ecology: Plant & Plant Communities in Urban Environments. The Hague: SPB Academic Publishing Co.
• Sukopp, H., (1982). Nature in Cities. Council of Europe nature & environment series 28. Strasbourg, Germany. Suzuki, K., K. Ohnuma & Y. Yasuda, (1993). "Environmental monitoring by using monthly NVI data & climate data". Paper presented at the 1993 International Geoscience & Remote Sensing Symposium (IGARSS'93) - Better Understanding of Earth Environment, August 18-21, 1993, Tokyo, Japan. Volume III, pp. 1149-1151.
• Thomas, W.A., (1973). Biological Indicators of Environmental Quality. Ann Arbor, MI.: Ann Arbor Science Publishers.
• Ulrich, R.S., (1984). "View through a window may influence recovery from surgery", Science 224: 420-21.
• U.S. Bureau of the Census, (1994). 1990 Census of Population & Housing (CD-ROM version), Washington, D.C.:
• U.S. Bureau of the Census. - - (1985). 1980 Census of Population & Housing, Washington, D.C.: U.S. Bureau of the Census.
• Wagner, T.W., J.E. Colwell & R.A. Shuchman, (1994). Mapping urban metabolism with Landsat data. Ann Arbor, MI.: Environmental Research Institute of Michigan (ERIM).
• Westman, W.E. & C.V. Price, (1988). "Detecting air pollution stress in Southern California using Landsat Thematic Mapper band data", Photogrammetric Engineering & Remote Sensing, 54(9): 1305 - 1311.
• Willeke, D.C., (1989). "The imperative forest". In, Moll, G. & S. Ebenreck, (eds.), Shading our Cities: A Resource Guide for Urban and Community Forests, Washington, D.C.: Island Press, pp. 58-63.

Return to Publications

Return to Resume

Return to Rohinton Emmanuel's Home Page