INQUIRE 1 2015 16-35 Quantifying the Trade-off Between Landscape Vegetation Height, Surface Temperature and Water Consumption in Single-Family Homes in Tempe, AZ J. Jia, K. Larson and E.A. Wentz School of Geographical Sciences and Urban Planning; Julie Ann Wrigley Global Institute of Sustainability, Arizona State University Introduction Urban sustainability hinges on how resilient populations are to changes in the environment and resource availability. For a city situated in an arid, warm environment, urban sustainability particularly requires the efficient management of water resource as well the mitigation of heat stress (Harlan et al., 2006; Larson et al., 2013). Recent findings suggest that water-intensive landscape features mitigate extreme temperatures associated with the urban heat island effect (UHI). As a result, residential landscapes involve a tradeoff and compromise between water conservation and heat mitigation through urban greening techniques (Larson et al., 2013; Gober et al., 2013). Considering that the Southwestern US is forecast to soon become hotter and drier in the remainder of the 21st century (Karl, Melillo, and Peterson, 2009; Parry et al., 2007) it is imperative that the nuanced implications of vegetation for both urban water-use and heat mitigation are explicitly understood by policy makers and residents if cities are expected to withstand climatic changes (Stabler et al., 2005; Larson et al., 2013). Copyright © 2015 J. Jia, K. Larson and E.A. Wentz : jessica.jia@asu.edu 17 Heterogeneity in the urban environment creates difficulty in assessing which variables impact water consumption and UHI, and, most importantly, to what quantifiable extent. Quantifying the impact that one feature—in this case, vegetation height—has on urban sustainability ideally requires an assessment of all the other factors that play a role in water consumption and heat propagation. Past research has determined that key drivers of urban heating include land cover composition (i.e. materials), land cover spatial form and arrangement, and proximity to nearby heating or cooling factors (Myint et al., 2013; Middel et al., 2014; Stone, 2009). A number of studies have found that urban water consumption depends on social and cultural practices, socioeconomic status, landscape type, urban form and number of household members, among other variables (House-Peters et al. 2011, Wentz et al. 2007, Guhathakurta and Gober 2007; Kontokosta and Jain 2015). While many variables have been linked to water consumption and heat mitigation, several challenges limit findings, include those related to data availability, issues of scale, the impact of socioeconomic and personal choices on consumptive practices, and the methodological difficulties in modeling such complex relationships. The majority of existing research on urban heating and vegetative cooling has been conducted at the neighborhood level or broader scales due to the lack of availability of disaggregated (household-level) water consumption datasets. This study broadens the scope of analysis of vegetative cooling efficiency by introducing vegetation height characteristics to study efficiency at the parcel-level. We ask the principal question—does vegetation of different heights predictably affect water demands and summertime surface temperatures of residential homes? The unique contribution of this study is the incorporation of LiDAR remote sensing imagery that enabled a dataset that included vegetation height (0.600646m/pixel), which was analyzed with MASTER temperature data (approximately 7m/pixel) and household-level water billing data. The null hypothesis is that different vegetation height classes do not affect water consumption and daytime surface temperature. Alternatively, we expect: 1) vegetation greater than 1.5 m in height will reduce daytime 18 surface temperature more than grass coverage, and 2) increased grass coverage will increase household water consumption more than vegetation height. Following an analysis of the potential basic linear relationships between vegetation height and surface temperature and water consumption, this study implies that policy makers should focus on increasing tall tree canopy over other vegetation for cooling surface temperatures. Literature Urban Heating: Previous research on urban heating differs in both the geographic scale of analyses and in the types of heat energy measured. Methodology has ranged from considering in-situ measurements of air temperature (Harlan et al., 2006; Myint et al., 2010), to computer models projecting air temperature (Middel et al., 2014), to surface temperature derived from remotely-sensed thermal imagery (Myint et al., 2013). In some urban heating studies, temperature measurements have been supplanted by energy flux measurements, wherein a net energy release of surface materials comparing the urban to rural heat release differential has been used to quantify the urban heating phenomenon. For example, Stone and Norman (2006) adjusted the thermal emissivity imagery to calculate the “excess flux of thermal energy” within an urbanized area by subtracting the proposed heat of a rural area from that of the urban area. These studies of heat also vary in scale, ranging from the parcel-level of study, to the neighborhood scale, to moving geographic windows of analysis. Some researchers have noted that low-resolution data sources limit the study of UHI drivers, thereby leading to a lack of understanding of land-cover effects (Myint et al., 2013). Across different methodologies, research has found that lowdensity urban development with low amounts of tree canopy cover and high coverage of impervious surface and grass coverage exacerbate urban heating. Local variations in temperature, referred to as microclimates, are 19 caused by an “interactive effect” of vegetation density and non-vegetative surfaces, local climate conditions, water availability, and land cover properties. Both the extent of vegetation cover and vegetation type (i.e. shade tree compared to grass) play an important role (Stabler et al., 2005; Weng et al., 2004). Vegetation coverage and NDVI have a strong relationship with urban temperature, but the percent distribution of vegetation cover alone does not account for changes in maximum air temperatures (Myint et al., 2010). Trees can lower temperature more effectively than grass, and building coverage does not necessarily exacerbate UHI as much as dark impervious surfaces, such as asphalt (Myint et al., 2013). Tree canopy height appears to differentially affect daytime and nighttime cooling. Spronken-Smith and Oke (1998) found that large amounts of tree canopy correlated with relatively cooler daytime temperatures, whereas large amounts of open sky view and dry soils correlated with cooler nighttime temperatures. Similarly, Myint et al. (2013) found that trees and shrubs, as opposed to grass, lower daytime temperatures and increase nighttime temperatures relative to surrounding areas. This leads to the hypothesis that canopy coverage and evaporative cooling lower local daytime temperatures, and open sky view and dry soils lower nighttime temperatures. These studies imply that the moist environment of tree canopy appears to mitigate extreme temperatures. As more information and data collection strategies become available, tree canopy and urban-heating studies can include additional variables to better understand the urban heat mitigation nuances. Water Demand: Research quantifying landscape water use is largely limited because most households do not differentially meter indoor and outdoor water consumption. To address the aggregation of water usage, some researchers have employed “minimum-use methods” as a means of estimation, but these methods tend to overestimate indoor use in the Phoenix metropolitan area (Mayer and DeOreo, 1999). Water demand is a product of a combination of social, behavioral, and mechanical practices. Separating indoor from outdoor water consumption proves difficult, and further separating landscape water demand from psychological practices 20 further complicates understanding how vegetation drives residential water demand. Although past research has cautioned that irrigation practices create a wider disparity in water use variability than homeowner landscape choices (Martin et al., 2004), determining the role of vegetation on water demand remains an area of active research interest. Past studies have shown that a combination of indoor and outdoor variables, such as race and ethnicity, income, household size, household age, lot size, and normalized difference vegetation index (NDVI) determine water consumption (Wentz and Gober, 2007; Turner and Ibes, 2011). Additional vegetation characteristics such evapotranspiration, and characterization of native/non-native plant species have also been shown to affect both water consumption and heat mitigation (Kaplan et al., 2014; Declet-Barreto et al., 2013; Middel et al., 2014). It has been speculated that temperature sensitivity plays a role in landscape water consumption, however, in Phoenix, temperature sensitive water use has been found to be lower than expected, perhaps because consumption is more a function of behavioral irrigation practices that do not respond to variations in weather/climate (Breyer et al., 2012; Balling and Gober, 2007; Martin 2008). Researchers continue to analyze the distinct effect of landscape variables on water consumption, in part to make recommendations for reducing demands. This research aims to explore the possible linear effect of vegetation of different heights on household water consumption in order to further the initiatives in understanding drivers of water demand. Data and Methods The City of Tempe is located near the center of the greater Phoenix metropolitan area. Average temperatures and precipitation (measured 1981-2010) are similar to that of Phoenix, with summertime day temperatures exceeding 100 °F and an average annual precipitation of 9.3 inches (U.S. Department of Commerce, 2010). More than 90% of the parcels analyzed in this study fall within two census block-groups, and all parcels fall within one census tract. Basic social statistics for the census tract encompassing the study area imply that the households studied have a median income near $45,000, are home to slightly more men than women, and are mostly Caucasian (U.S. Census Bureau, 2012). 21 Four basic layers of data were used: Maricopa County Assessor’s Office (MCAO) parcel data (Fig. B.2), Decision Center for a Desert City (DCDC) Quickbird/LiDAR land-cover data (Fig. B.3), City of Tempe household water billing data (Fig. B.4), and Central Arizona-Phoenix Long-Term Ecological Research (CAP-LTER) MASTER overflight temperature data (Fig. B.5). Background information of the datasets is provided in Table B.1. The QuickBird/LiDAR land cover data introduced in this study is the product of previous work funded by DCDC. This land-cover data used QuickBird object-oriented classification to delineate classes of land cover (e.g. buildings, impervious, grass) and LiDAR height data was integrated to differentiate between grass (low height) and three classes of tree height (1.5m-5m; 5m-10m; and >10m) based on natural breaks in the LiDAR data. This data layer was expected to have an accuracy of greater than 85% but may have error skewing land classification of grass and impervious surfaces. The MODIS/ASTER Airborne Simulator (MASTER) is a superspectral sensor developed to support the validation of the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) and Moderate Resolution Imaging Spectroradiometer (MODIS) spaceborne imaging projects. The surface temperature imagery was georeferenced using a 2nd order polynomial with an RMS error of 0.411877 for the daytime data and an RMS error of 0.263945 for the nighttime data. The data layers underwent a series of processing steps in order to obtain numeric attribute values for each parcel. Houses with swimming pools were omitted from the analysis, along with homes exhibiting water consumption that fell outside of the 95th percentile. Water consumption was converted to gallons per day of usage for the month of July 2011, which was chosen to match the temporal scale of the MASTER imagery. Figure 1: The Tempe Study Neighborhood within Metropolitan Phoenix, Arizona Table 1: Variable Attributes Variable Source Water Consumption City of Tempe MASTER Daytime Temperature Night Temperature Parcel Area MASTER Buildings MCAO QuickBird and LiDAR Impervious QuickBird and LiDAR Soil Grass QuickBird and LiDAR QuickBird and LiDAR Trees [1.5m-5m] QuickBird and LiDAR Trees [5m-10m] Trees [>10m] QuickBird and LiDAR QuickBird and LiDAR Spatial Scale Date Type of Information Parcel 7m/pixel 7m/pixel Parcel 0.6 m/pixel 0.6 m/pixel 0.6 m/pixel 0.6 m/pixel 0.6 m/pixel 0.6 m/pixel 0.6 m/pixel July 2011 July 2011 July 2011 2014 March 2008 March 2008 March 2008 March 2008 March 2008 March 2008 March 2008 Gallons per Day Raster Raster Digitized data layer Raster Raster Raster Raster Raster Raster Raster 23 The raw MASTER imagery files provided by CAP-LTER underwent atmospheric correction, conversion to Celsius, and registration prior to analysis. These provided the surface temperature data for the study area, but were subject to slight geographic warping and distortion. This surface temperature data was averaged to obtain a singular mean value for daytime surface temperature and for nighttime surface temperature using the MCAO parcel boundaries to define the geographic extent of each household. Again using the MCAO parcel boundaries as reference, the sum of each land cover class area was taken and appended as an attribute to each parcel. 24 To address our principal question—does vegetation of different heights predictably affect water demands and summertime surface temperatures of residential homes? — we conducted a bivariate analysis and a series of stepwise linear regression models. The bivariate analysis was done using a Pearson’s correlation to examine the significant correlations between independent and dependent variables. Then, two linear regressions were conducted for the dependent variables of Tday and Tnight modeled to the independent variables of land-cover (buildings, impervious surfaces, grass, and trees of 1.5m-5m, 5m-10m, and >10m). Another regression modeled water consumption to the independent variables of land-cover. Although more sophisticated analytical methods could be employed, these methods were chosen in order to represent a preliminary analysis of the variables available in this study. Analysis and Results As detailed in Table C.1, the average parcel is about 680 sq. meters of which the building components occupied about 30% or 210 sq. meters of the area. Approximately 1% or 6.8 sq. meters of each parcel is nonbuilding impervious surface. Approximately 51% of a parcel is overlaid by vegetation, of which grass coverage is the most prevalent coverage, followed by shorter vegetation and taller vegetation, respectively. Grass coverage averages 29% of each parcel, and vegetation measuring 1.5m-5m averages 18% of each parcel. Taller vegetation measuring 5m-10m averages 3% of each parcel, and vegetation measuring greater than 10m averages 0.2% of each parcel. Excepting the low percentage of impervious material coverage, the design of the study parcels is consistent with medium to low density developments in Tempe, Arizona. 25 Table 2: Descriptive Statistics (n=196) Water Consumption (GPD) 55 1162 412 Standard Deviation 260 Daytime Temperature (°C) 43 56 51 2.5 Nighttime Temperature (°C) 19 24 22 0.75 Parcel Area 320 1100 680 120 Buildings (m2) 53 430 210 53 Impervious (m ) 0 110 6.8 17 Soil (m2) 0 390 120 99 Grass (m ) 0 620 200 120 Trees [1.5-5m] (m2) 0 590 120 112 2 0 230 23 44 2 0 140 1.8 14 Variable Minimum 2 2 Trees [5-10m] (m ) Trees [>10m] (m ) Maximum Mean The bivariate correlations shown in Table 2 suggest that larger parcels tend to have larger building area, more grass, and more vegetation of 1.5m-5m height. Parcels with more vegetation of 1.5m-5m height and of 5m-10m height tend to have less soil and less grass, and correlate to lower daytime surface temperature. These parcels with taller vegetation also correlate with higher water consumption and higher nighttime surface temperature. Parcels with higher soil coverage correlate to higher daytime temperature and lower water consumption. Buildings and grass correlate to decreased nighttime surface temperature but show insignificant correlation with daytime surface temperature. Impervious surfaces and vegetation taller than 10m do not show correlation with temperature, perhaps because tall trees cover such a small percentage of parcel area. Table 3 provides the linear regression models incorporating the three most significant land cover predictors of parcel average daytime surface temperature (Tday). Vegetation of 1.5m-5m height significantly decreased Tday around 0.008 °C (0.0144 °F) for every additional 1 square meter of coverage. Taller vegetation is expected to have a stronger impact on daytime temperature; among the parcels, an additional 1 sq. meter of vegetation of 5m-10m of height decreases Tday about 0.01 °C (0.0180 °F). Soil slightly increased daytime temperatures, about 0.004 °C (0.0072 °F) increase for every additional 1 square meter of soil. The total regression model incorporating vegetation of 1.5m-5m, vegetation of 5m-10m 26 and soil coverage represented about 20% of the variation in Tday among the parcels studied. Table 3: Correlation Coefficients, key variables Table 4: Land-Cover Determinants of Average Daytime Surface Temperature (n=196) Model (Adjusted R2) 1 (0.143) Variables (Constant) Trees [1.5m-5m]* 2 (0.187) 3 (0.204) (Constant) t Unstandardized Coefficients B Standard error 217.679 51.922 0.239 -5.796 -0.008 0.001 Standardized Beta* -0.384 217.530 52.120 0.240 Trees [1.5m-5m]* -5.373 -0.008 0.001 -0.351 Trees [5m-10m]* -3.383 -0.012 0.004 -0.221 (Constant) 132.294 51.415 0.389 Trees [1.5m-5m]* -4.036 -0.006 0.002 -0.285 Trees [5m-10m]* -2.814 -0.010 0.004 -0.187 0.166 0.004 0.002 0.166 Soil* * p < 0.05. Independent variables that were not significant: Building, Impervious, Grass, and Trees [>10m]. 27 Table 5 provides the stepwise linear regression models incorporating the three most significant predictors of parcel average nighttime surface temperature (Tnight). The addition of 1 sq. meter of building decreased Tnight by about 0.005 °C (0.0090 °F), and an increase of 1 square meter of grass decreased nighttime surface temperature by about 0.004 °C (0.0072 °F). An increase of 1 square meter of tree canopy of 5m10m height increased nighttime surface temperature by about 0.002 °C (0.0036 °F). The model incorporating buildings, grass, and trees 5m-10m height accounted for approximately 30% of the variation in nighttime temperature. Table 5: Land-Cover Determinants of Average Nighttime Surface Temperature (n=196) Model (Adjusted R2) 1 (0.113) 2 (0.253) 3 (0.290) Variables t Unstandardized Coefficients B Standard error 23.010 0.206 Standardized Beta* (Constant) 111.487 Buildings* -5.083 -0.005 0.001 (Constant) 108.503 23.670 0.218 Buildings* -6.459 -0.006 0.001 -0.405 Grass* -6.101 -0.002 0.000 -0.383 (Constant) 104.916 23.442 0.223 Buildings* -6.077 -0.005 0.001 -0.376 Grass* -5.343 -0.002 0.000 -0.336 3.325 0.004 0.001 0.207 Trees [5m- -0.343 10m] * *p < 0.05. Independent variables that were not significant: Impervious, Soil, Trees [1.5m5m] and Trees [>10m] Results comparing the dependent variable of water consumption to the independent variables of land-cover (i.e., building, impervious, soil, grass, trees [1.5-5m], trees [5-10m], trees [>10m]) were not reported here due to finding a lack of significant relationship with water consumption. 28 Discussion Determinants of Urban Heating: Vegetation cover of taller heights reduced daytime surface temperature more effectively than lower vegetation cover (Table C.3). Specifically, a 1 sq. meter increase in vegetation of 5m-10m of height decreased daytime temperature by about 0.01 °C, whereas vegetation 1.5m-5m of height decreased daytime surface temperatures by about 0.008 °C. This implies that taller vegetation, such as tree canopy, provides a better cooling environment than shorter vegetation, likely because taller vegetation is associated with a larger overall extent of canopy coverage in a parcel. Tall vegetation likely represents the crowns of existing tree canopy rather than discrete trees of tall height. In other words, the classification method does not distinguish between individual trees or shrubs, and as a result, houses with vegetation of 5m-10m height in this study are more likely to also have vegetation of 1.5m-10m height (Table C.2) because tree crowns are surrounded by branches of lower height. The aggregate heat mitigation from total tree canopy coverage, regardless of height, may be responsible for the magnitude of the relationship between vegetation of 5m-10m height and lower daytime temperature (Table C.3). As a whole, these results were consistent with Stone and Norman (2006) who determined that parcel surface temperature decreased with size of tree canopy coverage per parcel, and with Middel et al. (2015) who suggested that the relationship between percent canopy cover and air temperature reduction is linear at the neighborhood level. Our findings specifically suggest that taller tree canopy provides for better shading and advective cooling than shorter vegetative cover, but this could be due to the relationship between tall canopy and overall canopy coverage. Vegetation height differentially affects nighttime surface temperatures. Grass reduced nighttime temperatures, and tall tree canopy of 5m-10m height increased nighttime surface temperatures. This reversal of the effects of tree canopy on temperature for daytime versus nighttime may be because uncovered surfaces cool more rapidly than surfaces under tree canopy, since canopy coverage prevents heat release of below-canopy 29 surfaces (Myint et al., 2013; Spronken-Smith and Oke, 1998). These results align with past findings, namely that landscape designs which exhibit lower daytime surface temperatures often exhibit higher nighttime surface temperatures (Nichol, 2005). Results from our study support past studies by implying that vegetation of 5m-10m in height serves as a microclimate buffer between rapid heating and cooling in the residential environment, cooling daytime temperatures and conserving nighttime heat. Impervious surfaces and building coverage did not correlate as strongly as we expected with average parcel surface temperatures. We hypothesize that variable surface reflectance properties of rooftops (such as rooftop angle and albedo) contribute to the lack of correlation between building coverage and daytime surface temperature. Increased building area did linearly predict lower nighttime surface temperatures (Table 5), and this is speculated to be a result of indoor temperature from air conditioning reducing rooftop temperatures faster than the surrounding landscape surfaces (Myint et al., 2013). Given the findings of a strong relationship between impervious surfaces and temperature in previous studies (Myint et al., 2013; Stone and Norman, 2006) a positive correlation is expected for impervious surfaces and daytime surface temperature. We hypothesize that our study did not correlate impervious coverage with surface temperature because driveways in this dataset are misclassified as soil instead of impervious surfaces – thereby underrepresenting impervious coverage. In the descriptive statistics we can see that impervious surface area is well below what should be expected for at least a singular parcel driveway. These descriptive statistics report less than 1% of parcel area to be impervious, yet driveway surface is estimated to comprise an average of 18% of singlefamily residential lots (Stone, 2004). This hypothesis of impervious classified as soils is supported by visual inspection of the land cover imagery (Figure 3), and supported by the significant positive correlation found between soil and daytime surface temperature (Table 4). The large disparity between our land cover imagery and known impervious surface area warrants a closer analysis of the limitations of the land classification methods used in this study, and should be further analyzed in future studies. Determinants of Water Consumption: The linear regression that evaluated land-cover determinants of water consumption was inconclusive and omitted from this study. The lack of significant relationships between land-cover variables and water consumption is likely due to the absence of social and outdoor determinants of water consumption in our analysis. Past studies have shown that a combination of social and outdoor variables, such as race, income, household size, household age, lot size, and NDVI determine water consumption (Wentz and Gober, 2007; Turner and Ibes, 2011). Additionally, this study area experienced a high standard deviation for household water usage – about 63% the value of the average household water usage (Table 2). Our findings support the conclusion that the relationship between water input and cooling efficiency is non-linear due to the feedback effects among water demand, irrigated vegetation, and the urban heat island effect, along with the lack of explicit datasets (Guhatharkurta and Gober, 2010; Middel et al., 2012; Middel et al., 2011). Future studies on land-cover determinants of water consumption are warranted to examine these relationships more closely. Limitations and Future Research: Our study was limited by the scope of analysis and the type and quality of the data available. We recommend that future studies address variables such as spatial form and arrangement, albedo, NDVI, evapotranspiration, air-temperature, among other variables, which have been shown to affect both water consumption and heat mitigation (Kaplan et al., 2014; Turner and Ibes, 2011; DecletBarreto et al., 2013; Middel et al., 2014). Future temperature studies could also use different methods to determine temperature for each parcel, such as air temperature measurements. We also recommend that future investigation of vegetation height water demand analyze strictly outdoor water consumption data. Indoor water-use variables (number of showers, laundry, additional indoor water use variables) and consumer demographics (number of occupants in household, income, additional social variables) associated with household water consumption obscure the predictive influence of vegetation height classes on water consumption. 30 Our land-cover analyses were largely limited by observable landcover classification errors. In addition to the under-classification of impervious surfaces and over-classification of soil, we also observed the tendency to misclassify vegetation coverage. Some tree canopy is misclassified as soil or as grass, and non-irrigated vegetation (weeds) is commonly classified as grass. The inclusion of weeds in grass coverage in particular could explain the lack of significant relationship between grass coverage and water consumption. Turf landscaping should strongly influence water consumption (Wentz and Gober, 2007), but did not exhibit any regressive relationship in this study likely because grass coverage included more than irrigated turf. This would be probable for the vegetation dataset because the data was gathered in March, meaning that spring growth of land vegetation would be high for grass-like weeds. We recommend that future studies be conducted with more robust accuracy in land cover classification. Conclusions More research is needed to better understand the urban sustainability trade-offs of residential landscape designs. By analyzing more specific contributions of land cover characteristics in the form of distinct vegetation height, we have determined that vegetation height differentially affect surface temperature at the parcel-level. Vegetation of 5m-10m height correlated to mitigation of extreme temperatures – lowering daytime surface temperatures and raising nighttime surface temperatures. Vegetation of 1.5m-5m height lowered daytime surface temperatures to a lesser magnitude than vegetation of taller height. Results imply that planners and landscape designers should consider strategically arranging buildings and vegetation to maximize shading and cooling benefit. Specifically, we recommend that planners place emphasis on cultivating tree canopy taller than 5m of height in order to mitigate extreme urban temperatures. Policy makers and planners can evaluate the amount of tall tree canopy in single-family residential landscapes and modify plans to consider vegetation height and its contribution to surface cooling of residential properties. 31 These recommendations for policy support of taller tree height elicit a word of caution when considering the wider urban sustainability initiative, especially in an environment where water is scarce. The implications of vegetation height on water consumption remain unclear. Given the tradeoff between water conservation and heat mitigation, residents and municipalities seeking to reduce water consumption will need to incorporate a wider set of variables than vegetation in order to predictably reduce demand. Future research should, ideally, consider additional vegetation attributes (e.g., native vs. non-native trees) in order to find a landscaping solution that is both water efficient and heat mitigating. The context of land cover conditions (e.g. the type and placement of vegetation cover relative to the built structures, the angle of the sun, and the surrounding environment) might be more important than the universal endorsement of one particular land-cover (or vegetation height height) over another. Land-use planners and policy makers should consider the effect of vegetation in relation to factors such as shade angle, density of development, and NDVI properties of vegetation. These context-specific recommendations necessitate the continued study of potential benefits of policy recommendations of landscaping types. While policy initiatives aimed at increasing urban sustainability must address the pros and cons that accompany applying rudimentary regulations on a wide diversity of landscaping choices and consumption practices, research initiatives must address the advantages and disadvantages of oversimplifying or complicating analysis of the urban landscape water-temperature relationship. Quantifying key drivers to heat propagation and water consumption necessitates the admission that many limitations still exist. Addressing the these limitations is necessary, particularly for contributing to urban sustainability studies by incrementally exploring the potential interrelationships that may directly exist between vegetation characteristics, urban heating, and water consumption. 32 Acknowledgements Special thanks to Philip Tarrant, Qunshan Zhao, Chao Fan, and the City of Tempe Public Works Department. References Breyer, B., Chang, H., and Parandvash G.H. (2012) Land-use, temperature, and single-family residential water use patterns in Portland, Oregon and Phoenix, Arizona. Applied Geography 35 (1-2): 142-151. Chang, H.J., Parandvash, G.H., and Shandas V. (2010) Spatial Variations of Single-Family Residential Water Consumption in Portland, Oregon. Urban Geography 31(7): 953-972. Chow, W.T.L., and Brazel, A.J. (2012) Assessing xeriscaping as a sustainable heat island mitigation approach for a desert city. Building and Environment 47:170-181. Declet-Barreto, J., Brazel, A.J., Martin, C.A., Chow, W.T.L, and Harlan, S.L. (2013) Creating the park cool island in an inner-city neighborhood: heat mitigation strategy for Phoenix, AZ. Urban Ecosystems 16(3): 617635. Gober, P., Brazel, A.J., Quay, R., Myint, S., Grossman-Clarke, S., Miller, A., and Rossi, S. (2010) Using watered landscapes to manipulate urban heat island effects: How much water will it take to cool Phoenix? Journal of the American Planning Association 76(1): 109-121. Gober, P., Larson, K.L., Quay, R., Polsky, C., Chang, H., and Shandas, V. (2013) Why land planners and water managers don’t talk to one another and why they should! Society & Natural Resources 26(3): 356364. Guhathakurta, S., and Gober. P. (2007) The impact of the Phoenix urban heat island on residential water use. Journal of the American Planning Association 73(3): 317-329. Guhathakurta, S., and Gober, P. (2010) Residential land use, the urban heat island, and water use in phoenix: A path analysis. Journal of Planning Education and Research, 30(1): 40-51. Harlan, S. L., Brazel, A. J., Prashad, L., Stefanov, W. L., and Larsen, L. (2006) Neighborhood microclimates and vulnerability to heat stress. Social Science & Medicine, 63(11): 2847-2863. 33 Harlan, S.L., Yabiku, S., Larsen, L., and Brazel, A.J. (2009) Household water consumption in an arid city: Affluence, affordance, and attitudes. Society and Natural Resources 22:691-709. House-Peters, Lily A., and Chang H. (2011) Urban water demand modeling: Review of concepts, methods, and organizing principles. Water Resources Research 47(5). Kaplan, S., Myint, S.W., Fan, C., and Brazel, A.J. (2014) Quantifying outdoor water consumption of urban land Use/Land cover: Sensitivity to drought. Environmental Management, 53(4): 855-864. Karl, T. R., Melillo, J.M., and Peterson, T.C. (2009) Global climate change impacts in the United States. New York, NY: Cambridge University Press. Kontokosta, C.E., and Jain, R.K., (2015) Modeling the determinants of large-scale building water use: implications for data-driven urban sistanability policy. Sustainable Cities and Society 18: 44-55. Larsen, L., and Harlan, S. L. (2006) Desert dreamscapes: Residential landscape preference and behavior. Landscape and Urban Planning, 78(1–2): 85-100. Larson, K., Polsky, C., Gober, P., Chang, H., and Shandas, V. (2013) Vulnerability of water systems to the effects of climate change and urbanization: A comparison of phoenix, arizona and portland, oregon (USA). Environmental Management, 52(1): 179-195. Martin, C.A. (2008). Landscape sustainability in a Sonoran Desert City. Cities and the Environment. 1(2): Article 5, 16 Mayer, P.W., & DeOreo,W. B. (1999). Residential end uses of water. Denver, Colorado: AWWA Research Foundation. Middel, A., Brazel, A.J., Gober, P., Myint S.W., Chang, H., and Duh, J.D. (2011) Land cover, climate, and the summer surface energy balance in Phoenix, AZ, and Portland, OR. International Journal of Climatology, 32(13): 2020-2032. Middel, A., Brazel, A.J., Kaplan, S., and Myint, S.W. (2012) Daytime cooling efficiency and diurnal energy balance in Phoenix, Arizona, USA. Climate Research 54(1): 21-34. Middel, A., Häb, K., Brazel, A. J., Martin, C. A., and Guhathakurta, S. (2014) Impact of urban form and design on mid-afternoon 34 microclimate in phoenix local climate zones. Landscape and Urban Planning, 122(0):16-28. Middel, A., Nalini, C., and Quay, R. (2015) Urban forestry and cool roofs: Assessment of heat mitigation strategies in Phoenix residential neighborhoods. Urban Forestry & Urban Greening, 14(1): 178-186. Myint, S.W., Brazel, A., Okin, G., and Buyantuyev, A. (2010) Combined effects of impervious surface and vegetation cover on air temperature variations in a rapidly expanding desert city. GIScience & Remote Sensing, 47(3): 301-320. Myint, S.W., Wentz, E., Brazel, A., and Quattrochi, D. (2013) The impact of distinct anthropogenic and vegetation features on urban warming. Landscape Ecology, 28(5): 959-978. Nichol, J. (2005). Remote sensing of urban heat islands by day and night. Photogrammetric Engineering & Remote Sensing, 71(5): 613-621. Parry, M., Canziani, O., Palutikof, J., van der Linden, P., and Hanson, C. (2007) Climate change 2007: Impacts, adaptation and vulnerability. Contribution of working group II to the fourth assessment report of the intergovernmental panel on climate change. Cambridge, UK: Cambridge University Press. Spronken-Smith, R. A., & Oke, T. R. (1998). The thermal regime of urban parks in two cities with different summer climates. International journal of remote sensing, 19(11): 2085-2104. Stabler, L. B., Martin, C. A., and Brazel, A. J. (2005) Microclimates in a desert city were related to land use and vegetation index. Urban Forestry & Urban Greening, 3(3–4): 137-147. Stone Jr, B. (2004) Paving over paradise: How land use regulations promote residential imperviousness. Landscape and Urban Planning, 69(1): 101-113. Stone Jr, B. and Norman J. M. (2006) Land use planning and surface heat island formation: A parcel-based radiation flux approach. Atmospheric Environment 40: 3561–3573. Turner, V.K., and Ibes, D.C. 2011. The impact of homeowners associations on residential water demand management in Phoenix, Arizona. Urban Geography 32(8):1167-1188. U.S. Census Bureau (2012) 2008-2012 American Community Survey 5-Year Estimates. Retrieved from http://ftp2.census.gov/ 35 U.S. Department of Commerce (2010) Heat in the Southwest. Retrieved from http://www.wrh.noaa.gov/psr/general/safety/heat/ Weng, Q., Lu, D., & Schubring, J. (2004). Estimation of land surface temperature–vegetation abundance relationship for urban heat island studies. Remote sensing of Environment, 89(4): 467-483. Wentz, E.A., and Gober, P., 2007, Determinants of small-area water consumption for the city of Phoenix, Arizona. Water Resources Management 21:1849-1863. Yuan, F., and Marvin E. B. (2007) Comparison of impervious surface area and normalized difference vegetation index as indicators of surface urban heat island effects in Landsat imagery. Remote Sensing of Environment 106 (3): 375-386.