Impact of shade on outdoor thermal comfort—a seasonal field study in Tempe, Arizona
Shade plays an important role in designing pedestrian-friendly outdoor spaces in hot desert cities. This study investigates the impact of photovoltaic canopy shade and tree shade on thermal comfort through meteorological observations and field surveys at a pedestrian mall on Arizona State University’s Tempe campus. During the course of 1 year, on selected clear calm days representative of each season, we conducted hourly meteorological transects from 7:00 a.m. to 6:00 p.m. and surveyed 1284 people about their thermal perception, comfort, and preferences. Shade lowered thermal sensation votes by approximately 1 point on a semantic differential 9-point scale, increasing thermal comfort in all seasons except winter. Shade type (tree or solar canopy) did not significantly impact perceived comfort, suggesting that artificial and natural shades are equally efficient in hot dry climates. Globe temperature explained 51 % of the variance in thermal sensation votes and was the only statistically significant meteorological predictor. Important non-meteorological factors included adaptation, thermal comfort vote, thermal preference, gender, season, and time of day. A regression of subjective thermal sensation on physiological equivalent temperature yielded a neutral temperature of 28.6 °C. The acceptable comfort range was 19.1 °C–38.1 °C with a preferred temperature of 20.8 °C. Respondents exposed to above neutral temperature felt more comfortable if they had been in air-conditioning 5 min prior to the survey, indicating a lagged response to outdoor conditions. Our study highlights the importance of active solar access management in hot urban areas to reduce thermal stress.
Outdoor thermal comfort is a complex function of atmospheric conditions and physical, physiological, psychological, and behavioral factors. These conditions and factors induce a subjective integrated response, thermal sensation, which has been the focus of many human biometeorology studies (Chen and Ng 2012; Johansson et al. 2014). Previous research has concentrated on identifying the factors that determine thermal comfort and breaking down their relative importance for thermal sensation using mixed methods that combine subjective and objective thermal assessments (e.g., Spagnolo and de Dear 2003; Eliasson et al. 2007; Kántor et al. 2012; Yin et al. 2012; Krüger et al. 2013; Pearlmutter et al. 2014). While indoor thermal comfort studies are usually conducted in climate-controlled conditions and can draw on several existing guidelines and standards (Johansson et al. 2014), the assessment of outdoor thermal comfort in cities is more challenging, as thermal conditions are less stable. Urban areas are heterogeneous and encompass various urban forms (type, density, and arrangement of buildings), surface materials, and landscapes, creating local scale and microscale climates that vary widely across space and time (Erell et al. 2012; Stewart and Oke 2012; Middel et al. 2014). Several studies have investigated thermal comfort in the context of urban form, focusing on street canyons or sky view factor (Johansson and Emmanuel 2006; Ali-Toudert and Mayer 2007; Pearlmutter et al. 2007; Mayer et al. 2008; Lin et al. 2010; Holst and Mayer 2011; Lee et al. 2014). Although the relationship between thermal comfort and the built environment tends to be strong, environmental factors, including meteorological conditions, generally only account for half of the variance in thermal sensation (Nikolopoulou et al. 2001; Nikolopoulou and Steemers 2003). The other 50 % can be attributed to a dynamic human parameter, which is composed of personal characteristics, i.e., age and gender; physiological factors such as weight and fitness level; psychological factors that include past experience, expectations, adaptation, thermal history, perceived control, and esthetic appreciation; and behavioral aspects such as clothing insulation, metabolic rate, time of exposure, and choice of location (e.g., Nikolopoulou and Lykoudis 2006; Vanos et al. 2010; Chen and Ng 2012; Klemm et al. 2015). All of these factors must be addressed in order to fully understand the integrated subjective thermal sensation response.
This study aims to quantify the impact of shade on subjective thermal sensation in a hot desert city—Tempe, Arizona—using subjective and objective comfort measures to address the environmental and non-environmental factors that impact thermal comfort. The importance of shade for reducing thermal stress in hot climates has already been emphasized by several authors (Johansson and Emmanuel 2006, Lin et al. 2010; Vanos et al. 2016). Our study objective is threefold: (1) examine the impact of shade on thermal comfort, perception, and perceived temperature; (2) investigate the relationship between atmospheric conditions and subjective thermal sensation; and (3) identify the most important drivers of outdoor thermal comfort in hot dry environments.
To quantify the thermal benefits of shade and investigate the relationship between perceived comfort and meteorological conditions outdoors, we conducted an objective and subjective assessment of thermal conditions through seasonal on-site meteorological observations and concurrent field surveys in Tempe, Arizona. Our assessment included sun-exposed locations as well as artificially shaded and tree shaded sites. We performed t tests to compare seasonal subjective thermal sensation in shaded and non-shaded locations and analyzed people’s air temperature estimates. Through regression analysis, we determined the physical drivers of thermal comfort. In a subsequent factorial ANCOVA, we examined how subjective thermal sensation varies by non-climatic factors after controlling for meteorological conditions. We then calculated physiological equivalent temperature (PET) from field observations and survey responses to determine neutral temperature, acceptable comfort range, and preferred temperature. Finally, we investigated the impact of air-conditioning on subjective thermal stress during pre-monsoon summer.
Downtown Tempe is home to the main campus of Arizona State University (ASU), a public university spread across four campuses in the Phoenix metropolitan area. ASU’s Tempe campus is about 2.6 km2, consists of broad pedestrian malls, and can be classified as open midrise LCZ. The Memorial Union, located in the heart of the Tempe campus, serves as community center for the ASU population and is a place of social interaction and gathering spot. The Memorial Union building offers student support amenities, restaurants, and services more than 14,000 people every day during the semester. The north and west exits lead to an expansive paved pedestrian mall, a walk-only zone from 8:00 to 16:00 h that used to have little vegetation, few mature trees, and little shade. In 2013, ASU partnered with a local utility provider and a solar energy company to cover the mall with three solar canopy structures to transform the open space. The installation was completed in May 2014, utilizing 1380 photovoltaic solar panels to cover 3330 m2 of land. The canopy structure now produces 397 kW DC and shades most of the pedestrian mall in front of the Memorial Union, including an outdoor dining area and a stage for outdoor events.
Experimental design and meteorological measurements
Sensor specifications and measurement height for stationary and handheld observations
LASCAR Electronics EL-USB-2+ (shielded)
−35° to +80 °C
± 0.3 °C
0 to 100 % RH
± 2.0 % RH
−10° to +55 °C
± 0.5 °C
0 to 100 % RH
± 3.0 % RH
−10° to +55 °C
± 1.4 °C
± 0.7 °C
0.6 to 60.0 ms−1
Larger of 3 % of reading, least significant digit or 20 ft/min
−40° to 510 °C
± 2.0 °C
Matrix Mk 1-G Pyranometer
Solar radiation (incoming and outgoing shortwave)
0.35 to 1.15 μm
± 5 %
Field survey design
Concurrent with the seasonal meteorological measurements in June, November, January, and April, we conducted questionnaire surveys under and near the photovoltaic canopies between 7:00 and 18:00 h. The surveys were designed to be transversal, i.e., each respondent only participated once, and could be completed in 3–5 min. Although the surveys were administered randomly, the set-up was quasi-experimental, because the respondents were mainly ASU students, staff, and faculty. The questionnaire covered personal characteristics, psychological and environmental factors, contextual information, and self-reported thermal perception, affective evaluation of comfort, and preference. First, respondents were asked to disclose their health-related mood on a 5-point scale: very bad (0), bad (1), fair (2), good (3), or very good (4). To assess the level of physical and cultural thermal adaptation, we collected information on the time of residency in Arizona. Adaptation was coded into 4 climate familiarity categories: just moved here (not familiar), have experienced a summer in the desert before (somewhat familiar), have lived here for 5 years (familiar), have lived here for >5 years or moved here from another hot dry environment (very familiar). Subjects indicated the reason for being at the Memorial Union (passing by, attending a class, meeting someone, lunch/resting) as a measure of perceived control. To survey thermal perception, we collected subjective thermal sensation votes (TSV) on a semantic differential 9-point scale, which is particularly suitable for extreme environments: very cold (−4), cold (−3), cool (−2), slightly cool (−1), neutral (0), slightly warm (+1), warm (+2), hot (+3), and very hot (+4). Perceived comfort was evaluated on a 4-point scale from comfortable (0) to very uncomfortable (3). Subjects rated their thermal preference on a 7-point scale, ranging from much cooler (−3) to neither warmer nor cooler (0) to much warmer (+3). All subjective judgment scales we employed comply with ISO 10551 (1995). The last part of the survey requested personal characteristics (gender and age group), clothing information, and details about the respondents’ activity level and location 5 and 30 min prior to the survey (short-term and long-term thermal history). Respondents were also asked to estimate the current air temperature in the sun and in the shade. Finally, subjects noted their sun exposure (full sun, shaded by the solar canopy, or shaded by a tree) and the time of survey completion so that the responses could be linked to meteorological observations.
with globe emissivity ε = 0.95, globe diameter D = 0.0254 m, and the globe’s mean convection coefficient 1.1∙108Va0.6[ms-1] (Thorsson et al. 2007). Each survey response was linked to observed meteorological conditions and Tmrt either in full sun, under the solar structure, or in tree shade based on location, time, and date of the response. The self-reported short-term and long-term thermal history was recoded into binary variables indicating if the subject was exposed to air-conditioning (AC) 5 and 30 min before taking the survey. We converted clothing responses to clothing insulation units (clo) according to ISO 9920 (2007) and calculated the metabolic rates in Wm−2 (ISO 8996 2004) based on reported activities. In order to compare subjective thermal sensations to actual measured thermal conditions, we chose PET as biometeorological index (Mayer and Höppe 1987). PET has been widely used in outdoor conditions and allows us to compare our results to other thermal comfort studies (Johansson and Emmanuel 2006; Lin 2009; Hwang et al. 2011; Chen and Ng 2012; Kántor et al. 2012; Makaremi et al. 2012). We calculated individual PET values for each subject from meteorological observations, Tmrt, clothing level, metabolic rate, and personal information using the MEMI model (Höppe 1999) implemented in Rayman (Matzarakis et al. 2007, 2010).
Figure 2 in the supplemental materials illustrates daily minimum, maximum, and mean air temperature and average relative humidity recorded by the stationary reference sensors at the Memorial Union between June 1, 2014 and May 31, 2015. The recorded sun-exposed shielded maximum air temperature was up to 2 °C higher than maximum air temperature in the shade. This relationship is reversed at night, with warmer minimum air temperature under the solar canopy and under trees, indicating a slight heat retention (up to 1 °C). The weather conditions during the selected field work days were clear and calm. Wind speed was low, averaging 0.6 ms−1 in the summer, 0.3 ms−1 in the fall, 1.1 ms−1 in the winter, and 0.5 ms−1 in the spring (Table 2, supplemental materials). On field work days in June 2014, air temperature reached 43.0 °C and globe temperature peaked at 51.7 °C in the sun, while relative humidity (water vapor pressure) was as low as 11.0 % (7.9 hPa). Weather conditions on November 7, 2014 (fall), January 22, 2015 (winter), and April 2, 2015 (spring) were milder, with maximum air temperature of 30.8, 19.3, and 30.6 °C; maximum globe temperature of 44.7, 32.8, and 43.3 °C; and an average daytime relative humidity (water vapor pressure) of 20.0, 15.3, and 16 % (8.5, 7.0, and 7.5 hPa).
Personal characteristics of survey participants
Summer (N = 306)
Fall (N = 364)
Winter (N = 338)
Spring (N = 276)
Contextual and personal factors covered in the survey; frequency distribution of survey responses for nominal and ordinal variables, mean responses for interval variables
Summer (N = 306)
Fall (N = 364)
Winter (N = 338)
Spring (N = 276)
Class at Memorial Union
Thermal perception (thermal sensation vote)
Neither warmer nor Cooler
Shaded (solar structure)
Short-term thermal history
No AC (5 min ago)
AC (5 min ago)
Long-term thermal history
No AC (30 min ago)
AC (30 min ago)
Metabolic rate (5 min ago)
Metabolic rate (30 min ago)
Air temperature estimate
Impact of shade on thermal comfort
(a) Independent samples t test for thermal sensation votes of artificially and naturally shaded respondents. (b) Independent samples t test for thermal sensation votes of shaded and sun-exposed respondents (**p < .001). We assume that the semantic differential 9-point TSV scale has interval properties, meaning that distances between points on the scale are equal. A non-parametric independent samples Mann-Whitney U test confirmed the t test results
Drivers of thermal comfort
Previous outdoor thermal comfort literature has shown that thermal comfort is influenced by physical, psychological, physiological, and behavioral factors (Chen and Ng 2012). We investigated which factors are the most significant drivers of subjective thermal sensation, using meteorological observations and survey responses as independent variables. First, we used multiple regression analysis to identify the meteorological drivers for variations in TSV. Independent variables to explain TSV included observed air temperature, water vapor pressure (derived from relative humidity and air temperature), surface temperature, incoming and outgoing shortwave radiation, WBGT, and globe temperature. The linear combination of meteorological variables was significantly related to TSV, F(7,1271) = 218.64, p < 0.0001, with R2 = 0.55. Globe temperature was the only significant predictor of TSV in the regression of observed meteorological variables, emphasizing the importance of the radiative environment for outdoor thermal comfort in hot and dry climates (Table 3, supplemental materials). In a separate regression between globe temperature and TSV, globe temperature explained 51 % of the variance in subjective thermal sensation, F(1,1277) = 1353.62, p < 0.0001 (Fig. 3, supplemental materials). These results are in agreement with previous studies that found a better correlation of thermal sensation with globe temperature than air temperature in Europe (Nikolopoulou and Lykoudis 2006) and a stronger effect of MRT on thermal comfort than air temperature in Malaysia (Makaremi et al. 2012). MRT, which can be derived from globe temperature or measured with three-dimensional short- and long-wave radiation sensors, has been identified as the most important variable for outdoor thermal comfort by various authors (e.g., Ali-Toudert and Mayer 2007; Mayer et al. 2008; Lee et al. 2013; Lee et al. 2014). Our results strongly suggest that, in hot and dry climates, solar access is more important for thermal comfort than humidity, as humidity levels are usually low (except during monsoon season). Therefore, measures such as WBGT and Heat Index (HI) are less suitable to predict thermal discomfort and heat stress in regions with low humidity levels, such as Arizona.
Results for a factorial ANCOVA with TSV as dependent variable and globe temperature as covariate; factors are categorical survey variables and additional interaction terms that were significant in separate ANCOVAs
Sum of squares
Time of day
Shaded or sun-exposed
AC or no AC (5 min ago)
AC or no AC (30 min ago)
Metabolic rate (5 min ago)
Metabolic rate (30 min ago)
Season * globe temperature
Thermal comfort * globe temperature
Time of day * globe temperature
Thermal preference * globe temperature
Adaptation * shaded or sun-exposed
Time of day * shaded or sun-exposed
Thermal comfort * shaded or sun-exposed
Season * shaded or sun-exposed
Neutral temperature, acceptable comfort range, preferred temperature
To determine the year-round acceptable outdoor thermal comfort range for respondents at the Memorial Union, we used a direct assessment of comfort through the thermal comfort vote. Similar to calculating neutral temperature, thermal comfort votes were averaged for each PET bin and plotted. Using a second-order polynomial curve fit, the curve segment that corresponds to a mean thermal comfort vote of <0.5 represents acceptable thermal comfort conditions (Fig. 4b). Survey results yielded a thermal acceptable range between 19.1 and 38.1 °C. In other climates, upper boundaries of acceptable outdoor conditions were found to be lower, e.g., 21.3–28.5 °C in Taiwan (Lin 2009) and 19–26 °C in Tel Aviv (Cohen et al. 2013).
Impact of air-conditioning on thermal stress in the summer
Discussion and conclusions
Linking field survey responses to meteorological observations, we examined the seasonal impact of shade on outdoor thermal comfort, compared subjective and objective comfort measures, and investigated how various environmental and non-environmental factors impact subjective thermal sensation. We used a Kestrel 4400 Heat Stress Meter to obtain Tmrt from Tg. The Kestrel has a 25.4-mm black powder coated copper globe and therefore overestimates Tmrt, especially when exposed to the sun, because it absorbs too much short wave radiation (e.g., Kántor and Unger 2011). Due to its small size, convective heat loss increases with higher wind speed, but the response time is significantly reduced compared to standard black globe thermometers (D = 150 mm). As wind speed was low when we conducted our field work (0.3 to 1.1 ms−1), convective heat loss was minimal. Our case study design limits the validity of our results to calm, clear conditions. The majority of days in Tempe exhibit these conditions, with 80–90 % possible sunshine throughout the year (Table 1, supplemental materials). Furthermore, our results are biased towards a healthy undergraduate student body and should not be generalized to more vulnerable populations, such as the elderly and children.
A regression of binned PET values and mean thermal sensation votes showed that respondents felt neither warm nor cold at 28.6 °C. This neutral temperature was found to be lower in humid and temperate climates. Our analysis of subjective comfort yielded a year-round acceptable outdoor thermal comfort range of 19.1–38.1 °C. Interestingly, the upper boundary of this range corresponds to the “triple digits” Fahrenheit air temperature threshold (38.1 °C = 100.6 °F), which is commonly used by the media and the general public in Arizona to denote the beginning and end of the heat season. While people seem to feel comfortable outdoors in a wide range of conditions, the temperature they prefer is 20.8 °C, as determined by probit analysis. This temperature is representative of air-conditioned environments, indicating that respondents are conditioned to indoor environments, because they are exposed to AC most of the day during the summer. Although short-term exposure to AC was not a significant non-meteorological factor in a year-round thermal comfort analysis (as opposed to adaptation level, gender, thermal comfort vote, thermal preference, season, and time of day), it significantly reduced TSV in the summer when conditions are warmer than neutral temperature. Exposure to AC prior to being outdoors lowered TSV by about half a point on the semantic differential 9-point scale, pointing to a lagged response to heat exposure. These results contribute to the discussion of increased thermal stress for vulnerable populations with no access to AC.
In a seasonal analysis, shade increased thermal comfort significantly in the spring, summer, and fall. Shade reduced TSV by 1 point on the semantic differential 9-point scale, improving subjective thermal sensation from hot to warm in the summer and from slightly warm to neutral in the transitional seasons. A multiple regression of TSV on the physical drivers of thermal comfort further emphasized the importance of solar access for thermal sensation. Globe temperature, the integrative measure of air and radiant temperature, was the only statistically significant meteorological predictor of TSV, explaining 51 % of the variation. These findings confirm results from previous studies showing that air temperature alone is not a comprehensive indicator of thermal comfort or stress, because it does not accurately represent the significant variation of thermal conditions in urban environments (e.g., Ali-Toudert and Mayer 2007; Mayer et al. 2008; Lee et al. 2014). Complex shading patterns from buildings and trees modify solar access at the pedestrian level. Therefore, perceived thermal conditions can vary several degrees in the shade and sun, as is evident from the survey respondents’ perceived air temperature estimates. While respondents in direct sun consistently overestimated air temperature, people in the shade underestimated it. Our results show that globe temperature, representative of the radiative environment and solar access, is a prime determinant of thermal comfort and stress in hot dry climates, outperforming indices such as WBGT or the heat index.
Thermal sensation responses did not significantly vary by shade type, suggesting that artificial and natural shade are equally efficient in mitigating heat stress in hot dry climates. Survey results reveal that the human body cannot resolve meteorological differences between shade types when humidity levels are low. This major finding opens up new avenues for active shade management strategies in hot dry climates to mitigate heat stress on citizens. Exposure to extreme heat in desert cities is a hazard of particular concern due to health risks, and it is expected to further increase in the future with projected rapid urbanization and more intense, more frequent, and longer lasting heat waves. Mitigating outdoor thermal stress through photovoltaic canopy shade is especially valuable in dry regions, because photovoltaic structures do not require irrigation and offer the co-benefit of electricity production with high solar potential. Our study did not take into account the esthetics of natural shade, which were found to be significant in recent studies (Klemm et al. 2015); we also did not consider other benefits of trees, such as storm water retention or wildlife habitat. In this context, artificial shade structures cannot replace natural shading, especially in urban green spaces and recreational areas. However, photovoltaic canopies offer a viable shade alternative in desert urban spaces where tree mortality is high or other tree benefits are considered secondary, such as parking lots, bus stops, and pedestrian malls, to create high quality public realm through climate sensitive design.
This research was supported by ASU Lightworks, PowerParasol®, the Julie Ann Wrigley Global Institute of Sustainability, and the International Graduate School IRTG 2057 (DFG, German National Science Foundation). Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the sponsoring organizations. We extend a special thanks to Robert Boscamp and Kevin White for their assistance and for providing us with the opportunity to conduct this research project. We would also like to thank Courtney Russell and Benjamin Mackowski for downloading meteorological data from stationary sensors that provided background weather conditions for this study.
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.