Planning Examples

Urban Heat Islands

In a study that explores the relationship between land use and the urban heat island effect, Claus Rinner and Mushtaq Hussain provide an excellent example of how urban planners can utilize ANOVA to identify significant differences among groups.

The urban heat island (UHI) effect refers to the tendency of urban areas to have higher surface temperatures in comparison to their rural surroundings. You can imagine that it would be warmer to stand in a parking lot on a hot summer day than it is to stand in a park. The L'HI effect simply expands this observation to the city scale; the surface temperature in an urban environment is higher than in a natural setting.

High temperatures in urban centers have led to serious impacts on human health, including sunburn, heat exhaustion, heat rash, and mortality (Smoyer-Tomic, Kuhn, & Hudson, 2003; Pengelly et al., 2007). Extreme heat waves can be linked to large numbers of deaths in urban environments; the heat wave in Chicago in 1995 killed 500 people over five days, and nearly 70,000 people died of heat-related illnesses in 16 European countries in 2003 (Robine et al., 2008). These past events, coupled with the potential of climate change to increase extreme temperatures, make it important for urban planners to explore the association between different types of urban land uses and surface temperatures.

Table 11.3 The Result ofBonferroni and Tamhane Post-Hoc Test

Multiple Comparisons

Dependent Variable: Intpm

(I) regi on

(f) regio n

Mean Difference (I-J)

Std. Error

Sig.

95 % Confidence Interval

Lower Bound

Upper Bou nd

Bonferroni

Northeast

Midwest

0.82955*

0.24662

0.006

0.1702

1.4889

South

1.08413*

0.22513

0.000

0.4822

1.6861

West

-0.03637

0.24662

1.000

-0.6958

0.6230

Midwest

Northeast

-0.82955*

0.24662

0.006

-1.4889

-0.1702

South

0.25458

0.18836

1.000

-0.2490

0.7582

West

-0.86592*

0.21358

0.000

-1.4370

-0.2949

South

Northeast

-1.08413*

0.22513

0.000

-1.6861

-0.4822

Midwest

-0.25458

0.18836

1.000

-0.7582

0.2490

West

-1.12050*

0.18836

0.000

-1.6241

-0.6169

West

Northeast

0.03637

0.24662

1.000

-0.6230

0.6958

Midwest

0.86592*

0.21358

0.000

0.2949

1.4370

South

1.12050*

0.18836

0.000

0.6169

1.6241

Tamhane

Northeast

Midwest

0.82955*

0.26287

0.021

0.0905

1.5686

South

1.08413*

0.26184

0.001

0.3482

1.8201

West

-0.03637

0.27532

1.000

-0.8035

0.7308

Midwest

Northeast

-0.82955*

0.26287

0.021

-1.5686

-0.0905

South

0.25458

0.16849

0.579

-0.1991

0.7082

West

-0.86592*

0.18877

0.000

-1.3780

-0.3539

South

Northeast

-1.08413*

0.26184

0.001

-1.8201

-0.3482

Midwest

-0.25458

0.16849

0.579

-0.7082

0.1991

West

-1.12050*

0.18733

0.000

-1.6266

-0.6144

West

Northeast

0.03637

0.27532

1.000

-0.7308

0.8035

Midwest

0.86592*

0.18877

0.000

0.3539

1.3780

South

1.12050*

0.18733

0.000

0.6144

1.6266

The Means Plots of Intpm Variable for Four Regions

Figure 11.14 The Means Plots of Intpm Variable for Four Regions

The UHI effect is primarily caused by the built environment in urban areas, where natural areas are replaced with non-permeable and high-temperature surfaces of concrete and asphalt (Maloley, 2009). Factors such as urban form, thermal properties of buildings, and anthropogenic heat sources also affect the magnitude of the L'HI (Taha, Hammer, & Akbari, 2002). Rinner and Hussain hypothesized that the mean temperature for different land uses would be different. To test this, these researchers performed a one-way ANOVA to identify the difference in mean surface temperatures between seven land use types in Toronto, Canada.

The researchers used two main datasets (Figure 11.18). The first was a GIS inventory of the Greater Toronto Area which included land use polygons for seven types of land uses: commercial, government and institutional, open area, parks and recreation, residential, resource and industrial, and water body. The second dataset was a thermal image which measured the surface temperatures with a resolution of 60-m (Figure 11.18).

You can see how an ANOVA test was appropriate; the researchers were working with a continuous dependent variable (surface temperature), and they were interested in the differences between multiple categories of an independent variable (land use type). The dependent variable was normally distributed within land use types, and the variance in the dependent variable among groups was comparable. Therefore a one-way ANOVA was an appropriate test for Rinner and Hussain.

Reading SPSS Data and Drawing a Box Plot

Figure 11.15 Reading SPSS Data and Drawing a Box Plot: R Script and Outputs

ANOVA results showed that the differences between the average temperatures of land use types were statistically significant. The researchers then performed a pairwise post-hoc test to identify' where the significant differences existed, and they elected to use a Bonferonni test because variance between groups was equal. The Bonferroni post-hoc test revealed that only two pairs of land uses did not exhibit statistically significant differences: commercial relative to resource/industrial land uses, and residential land use in relation to government/institutional areas. Average temperatures were significantly higher in commercial and resource/industrial land uses, and lower for parks and recreational land uses as well as water bodies. While these findings were not unexpected, several implications for urban planners emerged.

Building codes could require that new industrial or commercial development include green spaces or reflective roofing. Alternate polices could support tree planting programs for residential front and backyards. These and similar policies

Detecting and Removing Outliers and Drawing a Histogram

Figure 11.16 Detecting and Removing Outliers and Drawing a Histogram: R Script and Outputs

could promote heat mitigation as urban planners attempt to minimize the L'HI effect.

Future research could investigate cities that have implemented policies to reduce the UHI effect, and could use ANOVA to compare average surface temperatures by land use type and by city in order to see if these policies were effective.

ANOVA Test and Post-Hoc Test

Figure 11.17 ANOVA Test and Post-Hoc Test: R Script and Outputs

Land Use Type (top) and Thermal Imaging (bottom) of the City of Toronto

Figure 11.18 Land Use Type (top) and Thermal Imaging (bottom) of the City of Toronto

Source: Rinner & Hussain (2011)

Urban Form and Travel Behavior

In a classic article investigating the relationship between urban form and travel behavior, Susan Handy (1996) provides an example of how planners can use ANOVA to understand the relationship between urban form and travel behavior.

While many studies have attempted to link urban form and travel behavior, Handy hypothesized that high levels of accessibility would be associated with shorter average travel distances, greater variety in destinations, increased trip frequencies, and greater use of non-motorized modes of travel. To test these hypotheses. Handy selected two subregions within the San Francisco Bay Area that contained centers of retail activity: Silicon Valley and Santa Rosa. Within each subregion, she selected one traditional and one modern neighborhood. The traditional neighborhoods are characterized by turn-of-the-century urban form, with rectilinear street patterns, while the modern neighborhoods are characterized by post-World War II urban form, with curvilinear street patterns and numerous cul-de-sacs (Figure 11.19).

These four neighborhoods are the groups that Handy used to compare differences in accessibility and travel behavior. First, she compiled data on the demographics of the neighborhoods, the number and type of commercial establishments, and the type of road network for each neighborhood. Following the physical inventory, she conducted a telephone survey of the 100 residents of each neighborhood that asked residents about their travel choices, destination choices, mode choices, and trip frequencies, giving her continuous dependent variables to compare for her categorical independent variable, the neighborhood in which respondents resided. The data were found to be normally distributed, and variance between the groups was found to be equal. The ANOVA results indicated that there was significantly greater variation between the neighborhoods than within the neighborhoods, for the following multiple survey items: average number of supermarkets within two, five, and ten minutes, minimum time to supermarket, average travel time, percentage of people who walk to usual supermarket, average number of businesses visited last month, and average frequency of trips downtown. Before she could draw conclusions on these results though, she performed a second ANOVA test to ensure that variations in travel behavior were due to the urban form of the different neighborhoods rather than the different demographics in each area.

In the second ANOVA test, she categorized the households in each neighborhood based on the number of adults and children in the house and conducted ANOVA tests to see if the responses on the survey varied according to household type (rather than neighborhood), which they did. However, when she compared F-statistics and associated probabilities, the comparison revealed that for all but two travel behavior variables, neighborhood type had a greater effect on travel than did household type.

The findings of this study revealed not just that accessibility and travel behavior were linked as hypothesized, but different aspects of accessibility affect travel in different ways.

This study shows that ANOVA can be an effective method for investigating the built environment and travel behavior.

Street Networks and Land Use Distributions

Figure 11.19 Street Networks and Land Use Distributions

Source: Handy ( 1996)

Conclusion

This chapter focused on ANOVA with its purpose, history, mechanics, and a practical demonstration to perform ANOVA in SPSS and R. Although this chapter is limited to one type of ANOVA, one-way ANOVA, it gives fundamental understandings of the method and allows the readers to go beyond to other types of ANOVA, such as two-way ANOVA where two categorical independent variables are simultaneously related

Analysis of Variance (AA?01A) 219 to a continuous dependent variable or analysis of covariance (ANCOVA) where the dependent variable is related to one categorical and one continuous independent variable (Rutherford, 2001).

Works Cited

Anderson, D., Sweeney, D., & Williams, T. (2003). Essentials of statistics for business and economics. Mason, OH: Thomson South Western.

Argyrous, G. (2000). Statistics for social and health research. London: Sage Publications.

Bingham, N., & Fry, J. (2010). Regression: Linear models in statistics. London: Springer Undergraduate Mathematics Series.

Cidell, J., & Beata, A. (2009). Spatial variation among green building certification categories: Does place matter? Landscape and Urban Pieinning, 91(3), 142-151.

Devore, J. (2011). Probability and statistics for engineering and the sciences. Boston, MA: Brooks Cole.

Ewing, R., Haliyur, P., & Page, G. W. (1994). Getting around a traditional city, a suburban planned unit development (PUD), and everything in-between. Transportation Research Record, 1466, 53-62.

Fisk, W.J. (2002). How IEQ affects health productivity. Journal-American Society of Heating Refrigerating and Airconditioning Engineer, 44, 56—60.

Greene, J. (2005). Learning to use statistical skills in psychology (3rd ed.). Maidenhead: McGraw-Hill Education.

Handy, S. L. (1996). Understanding the link between urban form and nonwork travel behavior. Journal of Planning Education and Research, /5(3), 183—198.

Hansen, W, G. (1959). How accessibility shapes land use. Journal of the American Planning Institute, 25, 73—76.

Huck, S. (2008). Reading statistics and research. New York, NY: Pearson Publishing.

Maloley, M. (2009). Thermal remote sensing of urban heat island effects: Greater’ Toronto area. In Enhancing resilience to climate change program. Ottawa, ON: Natural Resources Canada.

McGraw Hill Construction. (2010). Green outlook 2011: Green trends driving growth. Retrieved from wavw.ecocosminc.com/img/201 l_McGraw_Hill_Green_Outlook.pdf

Pengelly, L., Campbell, M., Chad, E, Fu, C., Gingrich, S., & Macfarlane, R. (2007). Anatomy of heat waves and mortality in Toronto: Lessons for public health protection. Canadian Journal of Public Health, 98(5), 364-368.

Rinner, C., & Hussain, M. (2011). Toronto’s urban heat island—Exploring the relationship between land use and surface temperature. Remote Sensing, 3(6), 265. http://doi, org/10.3390/rs3061251

Robine, J., Cheung, S. K., Le Roy, S., Oyen, H. V., Griffith, C., Michel, J., & Herrmann, F. R. (2008). Death toll exceeded 70,000 in Europe during the Summer of 2003. Comptes Rendus Biologies, 331(2), 171-178. http://doi.Org/10.1016/j.ci-vi.2007.12.001

Rutherford, A. (2001). Introducing AA0E4 and ANCOVA: A GLM approach. London: Sage Publications.

Smoyer-Tomic, К. E., Kuhn, R., & Hudson, A. (2003). Heat wave hazards: An overview of heat wave impacts in Canada. Natural Hazards, 28(2-3), 463—485.

Sprinthall, R. (2003). Basic statistical analysis. Trenton, NJ: A and В Publishing.

SPSS. (2011). SPSS base 16.0 for Mac user's guide. Chicago, IL: SPSS Inc.

Taha, H., Hammer, H., & Akbari, H. (2002). Meteorological and air quality impacts of increased urban surface albedo and vegetative cover in the greater Toronto area, Canada. Berkeley, CA: Lawrence Berkeley National Laboratory.

Vogt, P. (2005). Dictionary of statistics and methodology: A nontechnical guide for the social sciences (3rd ed.). Thousand Oaks, CA: Sage Publications.

Walford, N. (2011). Practical statistics for geographers and earth scientists. Oxford: Wiley Blackwell.

 
Source
< Prev   CONTENTS   Source   Next >