Building and Environment 115 (2017) 269e280

Contents lists available at ScienceDirect

Building and Environment

journal homepage: www.elsevier.com/locate/buildenv

Tree shade coverage optimization in an urban residential environment

Qunshan Zhao a, *, Elizabeth A. Wentz a, Alan T. Murray b
a
Center for Geographical Information Science, School of Geographical Sciences and Urban Planning, Arizona State University, Tempe, AZ 85287-5302, USA
b
Department of Geography, University of California at Santa Barbara, Santa Barbara, CA 93106, USA

a r t i c l e i n f o a b s t r a c t

Article history: Shade provided by trees, shrubs and other vegetation serves as a natural umbrella to mitigate insolation
Received 4 November 2016 absorbed by features of the urban environment, especially building structures. For a desert community,
Received in revised form tree shade is a valuable asset, contributing to energy conservation efforts, improving home values,
16 January 2017
enabling cost savings, and promoting enhanced health and well-being. Therefore, maximizing tree shade
Accepted 31 January 2017
Available online 2 February 2017
coverage is an important component in creating an eco-friendly and sustainable urban environment.
Strategic placement of trees enhances tree shade coverage of buildings. This paper details an optimi-
zation method to simultaneously maximize tree shade coverage on building facades and open structures
Keywords:
Tree shade coverage
and to minimize shade coverage on building rooftops in a 3-dimensional environment. This method
Spatial optimization integrates geographic information systems and spatial optimization approaches for placing trees that
Location modeling provide the greatest potential benefit to a building. A residential area in Tempe, Arizona is utilized to
Solar insolation demonstrate the capabilities of the method. The optimization results show that two trees can provide up
GIS to 22.20 m2 shade coverage at 12:00 across a 54 m2 south-facing facade. This research offers a method to
help homeowners, urban planners, and policy makers to quantitatively evaluate shade coverage from
trees for building structures in a residential environment.
© 2017 Elsevier Ltd. All rights reserved.

1. Introduction the west and southwest of a building for regions in the northern
hemisphere. A simple method to create ample shade involves
The urban heat island (UHI) is the consequence of the thermal planting as many trees as possible on these sides of the building.
properties of the urban fabric that results in higher temperatures in This approach, however, is impractical because of the financial cost
urban areas compared to the surrounding rural areas [1e3]. The of trees as well as water restrictions in many water regulated
UHI exacerbates heat waves during the summer, increases energy communities [19]. Similarly, excessive shading reduces the possi-
consumption, and more importantly, increases the risk of heat- bility of retaining exposed residential rooftops for placing
related morbidity and mortality, especially for the elderly, chil- electricity-generating solar panels [20e23]. So while existing
dren, and disadvantaged groups [4e7]. Well-known UHI mitigation research provides a general guideline on where to locate residential
methods rely on increased vegetation such as shading impervious trees, they fail to consider the position of windows and doors,
surfaces through increased tree coverage, building urban parks residential landscape siting restrictions, and the rooftop solar en-
with lawns and water ponds, and adding green roofs or cool roofs ergy loss from shade coverage [24e28]. The challenge, however, is
on residential and commercial buildings [8e15]. In this research, achieving the maximum benefits of shade at the individual building
we focus on the strategic planning of shade trees in residential structure level with a more quantitative method, something that is
areas, which has been shown to provide significant energy and not fully understood [29,30].
long-term cost savings, to enhance the environmental quality of the The goal of this research is to consider where to optimally and
urban ecosystem, and to promote a range of human health benefits precisely locate shade trees on a residential parcel such that: a) the
[8,16e18]. Intuitively, the benefits of shade are best realized when shading of facade, windows, and doors of home structures is
trees are located on the sunward facing facade of buildings such as maximized and rooftop shade is minimized; b) the shade from
trees to the surrounding structures is considered; and c) spatial
optimization is creatively used to find the best tree locations
* Corresponding author. quantitatively in 3-dimensional (3D) environment. The study is
E-mail addresses: [email protected] (Q. Zhao), [email protected] (E.A. Wentz),
limited to the shade coverage provided by trees and does not
[email protected] (A.T. Murray).

http://dx.doi.org/10.1016/j.buildenv.2017.01.036
0360-1323/© 2017 Elsevier Ltd. All rights reserved.
270 Q. Zhao et al. / Building and Environment 115 (2017) 269e280

consider the dynamics of sensible and latent heat flux that occurs consider the relationship to nearby buildings, additional shade for
through evapotranspiration, diurnal variations in insolation, and windows and doors, and rooftop exposure for solar panel in-
seasonality. While limited in scope, we believe this approach pro- stallations. There are two considerations for nearby buildings and
vides an effective strategy for maximizing the shade of trees on tree placement. Nearby buildings, depending on distance, can
residential structures. We therefore present a 3D spatial optimi- simultaneously provide shade as well as receive shade from target
zation model that identifies optimal tree locations for residential building trees, although little research has examined this dual
structures by integrating geographic information systems (GIS) relationship. Also missing from the literature is tree placement to
with spatial optimization methods to solve this problem as a maximize shade on windows and doors. Windows and doors have
mathematical model. We demonstrate the method on a residential less heat-insulation comparing to facades, so shading the windows
neighborhood in the greater Phoenix metropolitan area of Arizona, by trees or other nearby structures will provide significant energy
where tree shade coverage, water conservation, and solar energy saving to the household comparing to facade [34]. On the other
potential are critical because of the hot and dry conditions. hand, residential building rooftops are the preferred location for
photovoltaic solar panels to generate electricity from direct solar
2. Literature review radiation, shown in multiple geographic locations [35]. Tree canopy
coverage and shade will significantly reduce the photovoltaic effi-
The study described here draws upon literature examining ciency of solar panels [20,21,23].
residential tree shade and spatial optimization in 3D environment.
From the residential tree shade literature, research shows that west 2.2. Spatial optimization in 3D
and east tree shade outside of house open structures provide the
optimal cooling effects and energy reduction on residential homes A challenge in maximizing shade coverage is that the buildings
[25]. The 3D spatial optimization literature guides the research on and trees are 3D objects, where the comparative location of the
how to extend the 2-dimensional (2D) maximum coverage location trees, roof, facade, doors and windows are important components
problem into the 3D space [31]. The following sections elaborate on for insolation remediation. Many real world facility location
these bodies of work. modeling problems have service coverage in the 3D environment
such as camera surveillance or Wi-Fi connection services [31,36,37].
2.1. Impact of tree shade Nevertheless, existing facility location modeling problems are
mostly abstracted and formulated in the 2D environment, such as
Existing research on the impact of tree shade on home struc- the location set covering problem (LSCP) and the maximal covering
tures associates tree shade with energy use savings in a single- location problem (MCLP) [38,39]. To manage the 3D space, these 3D
family house setting. Larger energy savings, up to 54% in some coverage problems were simplified into 2D environment to ease
studies [27], are found with trees located on the west side of a the formulation and solution of the facility location problems [40].
home, followed by trees on the east or southwest [24,25]. These Because of the dimensional simplification, the reliability and ac-
conclusions are similar across different northern hemisphere curacy of optimal facility locations were unavoidably lost projecting
climate zones where both heating and cooling conditions are from a 3D to a 2D environment.
considered. For example, Hwang et al. [26] evaluated the tree shade With the development of 3D computational tools, several at-
effects from a single tree to a single family house during the cooling tempts have been made to appropriately formulate and solve the
and heating season at both northern (Minneapolis and Indian- facility location modeling problems in the 3D environment [31].
apolis) and lower latitude (Charlotte and Orlando) locations. Using Some of this has taken place through a 2.5D surface, such as digital
the distance between the tree and the building through eight car- elevation model (DEM), by using a visibility analysis or viewshed
dinal (E, S, W, N) and inter-cardinal points (NE, SE, SW, NW), they analysis [41]. Goodchild & Lee [42] utilized visibility analysis to
show that trees on the west and east side of the house provided locate the minimum number of viewpoints to observe the entire
more energy conservation than those on the south side during the DEM surface, or to locate a fixed number of viewpoints to maximize
summer followed by the southeast or southwest. the overall visible area on the DEM. This research extended the
The beneficial relationship between tree shade and energy is concept of set-covering problems to the topographic surface, and
well established but there are only general guidelines on tree viewshed analysis was used to derive coverage on the DEM surface
placement strategies and the optimal number of trees. Tree place- rather than the 2D planar surface. However, DEM is not a real 3D
ment strategies emphasize cardinal direction with precision only surface and the coverage derivation by visibility analysis required
specified at the inter-cardinal level [26] and without incorporating extensive computation. These limitations make it difficult to use
the distance from the home structure. This type of information is their method to obtain the optimal coverage in a true 3D envi-
limited when it is infeasible to plant trees in specific cardinal di- ronment. To overcome the computational inefficiency, Kim et al.
rections. Furthermore, the distance trees are planted from the [43] extended Goodchild and Lee's research by only utilizing terrain
house structure, independent of the directionality, can further features (peak, pass and pit) as candidate viewpoints to acquire the
impact the area tree shade on a facade. Similarly, the number of maximal coverage with given number of viewpoints. Their method
planted trees is understudied, with most research focusing on the solved the problems faster and overcame the computational diffi-
impact of a single tree. The starting point for these issues is research culty, but they used the same viewshed method to derive the
such as Simpson & McPherson [24], McPherson et al. [25], Calcer- coverage in 2.5D. Murray et al. [44] found optimal security sensor
ano & Martinelli [28], Huang et al. [32], and Akbari & Taha [33], who placements in a 3D university environment utilizing the MCLP and
examined shading effects on different tree heights, multiple story the backup coverage location problem with visibility analysis. They
buildings, and number of trees. Results are consistent with prior considered the 3D building blocking effects in the coverage deri-
research showing optimal tree placement for energy savings is the vation process, but the coverage was only derived on the ground
east and west side of the buildings. These studies offer a broader surface and did not consider the coverage on campus building fa-
range of design considerations, but they still do not consider the cades. Most recently, Bao et al. [45] applied viewshed analysis to
relationship to neighboring houses, the open features on the derive the watchtower coverage on the DEM, and integrated LSCP
building facade, and a potential for rooftop solar panels. and MCLP solutions to determine the optimal watchtower locations
Design considerations for tree placement additionally need to for forest fire monitoring. To simplify the coverage representation,
 Q. Zhao et al. / Building and Environment 115 (2017) 269e280 271

they used viweshed analysis to derive coverage on the 2D raster evaluated the empty space rather than coverage on 3D objects.
surface. Although their methods integrate different methods to Zhao et al. [47] demonstrated a simple version of shade coverage
improve the efficiency of optimizing watchtower location, the optimization for the single family household in Tempe, AZ. Shade
coverage representation is still limited in the 2D rather than 3D. All coverage was derived on different 3D building structures in the 3D
of the research above demonstrate that visibility analysis or environment. Zhao et al.’s research provides limited details about
viewshed analysis are useful methods to help derive service formulating and solving the facility location modeling problems in
coverage in the 2.5D or 3D environment. However, none of these 3D, requiring more detailed research on service coverage in 3D
existing literature deals with the service coverage on the real 3D objects and decide the best facility locations.
objects. This remains as an obvious research opportunity to extend
this type of research into 3D environment.
To extend existing facility location modeling analysis into 3D, a 3. Methods
range of problems exist such as computational complexity, 3D data
availability, problem size, and model complexity. However, the key 3.1. Study area
question is how to extend the 2D service coverage into the 3D
environment. Besides the visibility analysis, several researchers The study focuses on a parcel with a detached single-family
have attempted to solve the facility location problems with 3D home and the surrounding buildings within a residential neigh-
coverage in the real 3D space [31,37,46]. Lee [31] introduced a 3D borhood in the City of Tempe, Arizona (33.4
N, 111.9
W, Fig. 1).
coverage location model of Wi-Fi access points in an indoor envi- Tempe is a municipality within the greater Phoenix metropolitan
ronment. Euclidean distance in the 3D space was utilized to area in the Sonoran Desert of the U.S. Southwest. The population of
generate the 3D volumetric coverage rather than the 2D circular Tempe in 2010 was more than 160,000 with greater than 40% of the
coverage. The software environment ArcGIS was able to generate residents living in single-family detached dwellings [48]. With
demand nodes and candidate facility sites within the 3D repre- summertime temperatures reaching or exceeding 43
C, heat
sentation, calculate 3D Euclidean distance, and visualize solutions mitigation strategies such as tree shade are essential for reducing
in a 3D environment. Commercial optimization software (CPLEX) heat-related diseases and energy consumption.
successfully solved the problem in seconds with no computational The specific parcel we analyzed is a generic residential parcel in
difficulty. Lee's research provides a successful example to extend a Tempe residential neighborhood where most of the single-family
facility location modeling problems by using 3D volumetric households were built during the 1950s and 1960s. The average
coverage in the 3D GIS environment, however, the 3D volumetric parcel size is 695 m2 and the typical home is single story with an
coverage was all perfectly sphere shape and did not consider the average size of 134 m2. The residential neighborhood has a dense
coverage change by surrounding obstacles. Similar attempts were building arrangement with neighboring structures next to one
made by Amriki & Atrey [37] on bus surveillance system. In their another on the west and east side, except those close to the major
research, they optimized camera locations and orientations in a 3D roads running north-south. This specific neighborhood layout
interior bus space that was simulated by Autodesk 3ds Max. makes it infeasible to plant trees on the west or east side of the
Maximal overall surveillance coverage with a specific number of building to provide shade. Although there are no regulations that
cameras and minimum number of cameras to reach specified specify the type of landscaping, 95% of the parcels in this neigh-
coverage in the bus were presented. They were able to evaluate the borhood contain trees (identified from remotely sensed images),
camera's visible region in 3D while avoiding obstacles, but they which offer some level of shade on the home structures.

Fig. 1. Sample home and parcel in the residential neighborhood of Tempe.

272 Q. Zhao et al. / Building and Environment 115 (2017) 269e280

3.2. Data sources the shadow location, trigonometry principles are used. The formal
trigonometry specifications are shown in equations (1)e(4). In
Two data types are required for the analysis, the specifications of these equations, ðx; y; zÞ represents points from a tree, ε is the solar
the building (e.g., dimensions, location, and facade features) and profile angle, g is the difference between solar azimuth and surface
the specifications of the tree (e.g., tree height, location). The digital azimuth angles, b is the solar altitude angle, and H is the height of
representation of the house structure for the selected parcel in- the roof. All solar angles are calculated based on Duffie and Beck-
volves knowing the building size, shape, roof contour, windows/ man (2013) [51]. Fig. 3 shows that it is possible to mathematically
doors locations, and overall orientation. We used Sketchup and its derive shade coverage associated with 3D object across a range of
Google Map component to create single family houses at the spe- conditions. Fig. 3(a) shows the solar angles and ground shading, the
cific geographical location (Fig. 2). We constructed a 18 m  12 m shading point on the ground is at ðx0; y0; 0Þ. Fig. 3(b) represents the
house, approximately 216 m2 in size, with 4 m height sloped facade shading, the shading point on the facade is ð0; y00 ; z00 Þ.
rooftop, three 2 m  1 m windows, and a 2 m  1.5 m front door on Fig. 3(c) explains the roof shading, the shading point on the rooftop
000 000
the south facade. The house has a multi-faceted roof surface and is is ðx ; y ; HÞ.
positioned with the front of the house facing the south. The area of
z
south roof, south facade and open structures (3 windows and 1 h¼ (1)
door) are 108.5 m2, 45 m2 and 9 m2. The distance between this tanb
structure and nearby buildings is 3 m.
where h is the shadow length on the 2D plane by solar altitude
The digital representation of the tree includes tree size, shape,
angle (b)
and position. The 3D tree plugin in Sketchup was used to create a
theoretical 7 m high, 6 m crown diameter, and 3 m trunk height to z
represent a thornless mature mesquite (Prosopis thornless hybrid x0 ¼ x  ; y0 ¼ y  hsing (2)
tanε
‘AZT™’), a common xeriscape flora found in Tempe residential
neighborhoods. The advantage of our tree model is that we can which calculates the horizontal shadow projection ðx0 ; y0 Þ based on
represent realistic desert trees with low leaf/area index rather than solar profile angle (ε) and azimuth angle (g)
other simple “cylinder-like” or “cone-like” tree models. By using
x0z
this tree model, we can derive a more accurate tree shade on the y00 ¼ y  xtang; z00 ¼ (3)
structure. Although different tree shapes, sizes, species can be x0  x
selected and these parameters would definitely influence the level
which determines the vertical shadow projection, (y00 ; z00 ), over a
of tree shade on different building structures such as rooftops, this
house facade according to horizontal shadow (x0) and azimuth
7 m thornless mesquite is typical of those found in Tempe [49,50]
angle (g)
(see Fig. 2). The challenge, of course, is identifying the best place-
ment of one or more trees to provide shade coverage to this zH
building structure. x000 ¼ x  ; y000 ¼ y  ðx  x000 Þtang (4)
tanε

3.3. Modeling approach which derives the horizontal projection (x000 ; y000 ) on a roof given
building height (H), solar profile angle (ε) and azimuth angle (g).
We utilized GIS and spatial optimization to model the tree shade
coverage optimization problem. GIS tools provide data storage, 3.3.2. Spatial optimization: tree placement
spatial analysis, and 3D topology. Optimization methods are used to Using Church & Murray [52] and the MCLP of Church & ReVelle
abstract the real world situation as a mathematical problem as well [39], we define the following notation:
as solve this problem. This section describes the analytical pro-
cedures we used. i ¼ index of 3D object components;
j ¼ index of potential tree locations;
3.3.1. GIS: surface coverage derivation d ¼ index of extreme heat days;
Spatial topology and trigonometry principles are used to store t ¼ index of extreme heat hours in a day d;
the spatial information and to derive surface coverage. Topological wi ¼ weight of object component i;
data structures in GIS store the location, configuration, and attri- gi ¼ area of object component i;
bute information of 2D and 3D objects. More specifically, to derive p ¼ number of trees to be located;

Fig. 2. 3D building and tree models.

 Q. Zhao et al. / Building and Environment 115 (2017) 269e280 273

Fig. 3. Shade projections from a point (x, y, z) on a 3D object (redraw from Gomez-Munoz et al. (2010) [30]).

fðÞ ¼ shade coverage function relating parameters of an object X XX

using trigonometry equations (1)e(4); Maximize wi Citd (5)
Std ¼ solar angles at time t on day d; i d t
Ni ¼ set of potential tree siting locations that shade object Subject to:
component i;
 
Citd ¼ f gi ; Std ; Xj ; j2Ni ci; t; d (6)
Decision variables are:
X
Xj ¼ p (7)

j
1; if tree located at potential site j
Xj ¼
0; otherwise
Xj ¼ f0; 1g cj (8)
Citd ¼ amount of object component i covered at time t, day d.
This notation allows for the specification of object components,
Citd  0 ci; t; d (9)
such as roof, facade, windows and doors. Accordingly, Citd tracks
shade provided to object component i at time t on day d as a The objective, (5), is to maximize tree shade coverage of
function of surface coverage. Using this notation, the model for 3D different object component i (roof, facade and windows/doors)
object coverage is as follows: during a particular time period and date with a predefined weight
274 Q. Zhao et al. / Building and Environment 115 (2017) 269e280

wi . wi represents the priority of tree shade coverage to different heat hours at given number of summer days. The best tree
building structure i. In general, windows/doors are open structures location can be found by maximizing tree shade coverage (Citd )
and need the most shade to mitigate direct solar radiation in the on building structures with predefined shading weights (wi ).
desert environment, following with building facade. Roof needs 3) To avoid tree crown overlap, the potential tree siting locations
less or no shading because residential roof is always built with (Ni ) within the existing tree crown is eliminated.
heat-insulation materials and is a perfect place to install solar 4) Repeat step 2) and 3) to locate the next tree around the building,
panels to generate electricity from solar energy. Constraints (6) until the potential tree siting location set is empty or locates p
define the amount of coverage that will be provided to object trees.
component i (roof, facade and windows/doors) based upon the tree
locations and solar angles at a specific time period and date. Con-
straints (7) specify the number of trees to be located. Integer re-
3.3.4. Model parameter explanation and simplification
strictions on the siting variables are stipulated in Constraints (8).
Because infinite potential tree locations exist, the simplification
Non-negative restrictions on coverage variables are indicated in
of potential tree siting location set is necessary. Potential tree
Constraints (9).
placement on the residential parcel is summarized based on
landscape design guidelines [53,54]. In the northern hemisphere,
3.3.3. Heuristic landscape design guidelines suggest that trees should be planted
We solve the 3D tree shade optimization problem through a on the south, west, or east of structures. Because of the space
heuristic solution approach for three reasons. First, it is computa- limitation on the west and east side of the house, we limited tree
tionally intensive to calculate shade coverage on different building placement to the south of the building. Further, to avoid unnec-
structures by trigonometry principles we mention above, especially essary tree shade coverage on the rooftops, a minimum distance of
when we have a detailed and complicated 3D tree and building 3 m between the tree and the building is predefined (Fig. 4).
models. Second, there is not an exact method that can be applied to To simplify the solution process, the continuous space was dis-
solve this optimization problem with a nonlinear constraint cretized into 42 potential tree locations as the potential facility
involving a trigonometric function. Third, trees can be located location set (Ni ) (3 m intervals in the east-west direction and 1 m
anywhere in the continuous space resulting in infinite combina- intervals in the south-north direction). Fig. 5 shows half of the
tions of different tree arrangements with multiple trees. Thus, we potential tree locations in the study site. Besides testing the shading
used a greedy-adding algorithm combining with brute-force benefits for the target building, we also derive the shade coverage
(enumeration of all possible candidate sites) method to find a on the two nearby buildings to obtain the shading benefits for the
near-optimal solution for this problem. The detailed steps are: surrounding building structures. We locate two trees (p ¼ 2)
because this is the most common number of trees to be planted in
1) Define the set of potential tree siting locations (Ni ) based on tree the desert city considering the water usage and landscape regula-
height, tree crown diameter, outdoor landscaping codebook, tion, but in general, the spatial optimization method can be used to
and building layouts. locate any number of trees in the 3D environment.
2) Brute-force method is used to locate the first tree by enumer- The weight of object component i (wi ) was defined as 0.7, 0.4
ating all the potential tree locations around the building during and 0.1 for windows/doors, facade and rooftops. Several reasons

Fig. 4. Potential tree placement area in the study site (Plan view).
 Q. Zhao et al. / Building and Environment 115 (2017) 269e280 275

Fig. 5. Potential tree locations in the study site (Plan view).

helped to define these weighting coefficients. The structure com- central parts of the front yard will provide the most shade for the
ponents we assigned as most important for shading were open overall household in regardless of building-tree distance. Fig. 6(b)
structures such as windows and doors. Although these open shows the tree shade coverage surface in the potential tree planting
structures are a small area compared to facades and rooftops, solar area by interpolation in the GIS environment. The results show that
radiation impacts are greatest through windows and doors [55]. facade tree shade area could be reduced to zero if trees are planted
From existing literature, the heat conduction from the sun through near the southern parcel boundary and far away from the buildings.
1 ft2 of facade or roof was only about 2% of the heat that passed This demonstrates that simply following the guidelines, planting
through a window [56]. Thus, when tree shade covers open trees on a specific side of a buildings, could result in little or no
structures, there are greater energy savings. Considering heat shade on the house structure.
conduction and solar radiation, shade coverage on windows/doors The results of the heuristic modeling for the first tree show that
had the highest priority, followed by facade, and rooftops. Further, the best site is at location 4, which is 3 m from the building's south
residential roof was an appropriate location to place solar panels to facade and 9 m from the building west and east facades (Figs. 6(a)
generate solar energy. We penalized the rooftop shading by using a and 7(a)). The accumulated shading time from this single mature
small negative weight. mesquite tree to the central part of the building south facade and
The shade was determined using trigonometry principles open structures on August 15th is up to four hours (Fig. 7(b)). Re-
detailed in section 3.3.1. Sun location and radiation was simulated sults show that the single mature mesquite tree can provide this
in Sketchup. The criteria for measuring shade effects on different shading to the central parts of building facade and open structures
structures of a single-family residence are based on the work of on this day.
Shaviv & Yezioro [57], who proposed the use of a geometrical To locate a second tree with our heuristic method, we first
shading coefficient to express the ratio between shaded and total eliminated the potential facility set based on the first tree location
examined surface areas. We selected the heat period from 9:30 to and landscaping limitation (no tree crown overlap), then
15:30 in a 30 min' interval during four heat days (June 15th, July enumerate all remaining options. To improve the performance of
15th, August 15th, and September 15th) to represent the periods of the heuristic algorithm, we repeated the heuristic algorithm with
greatest insolation [58]. To simplify the optimization criterion, we three different starting conditions (the first tree locates at 3, 4, or 5).
assume the most shading coverage will result in the most cooling The results show that the best near-optimal solution is at location 3
benefits for the building structures in this research. A heuristic and 5 (see Figs. 8 and 9(a)). The accumulated shading time from
approach for solving the optimization model, (5)e(9), was struc- these two trees on August 15th is shown in Fig. 9(b). Two mature
tured based on section 3.3.3. The accumulated weighted shaded mesquite trees can provide up to 6 h shading to the central parts of
area, objective (5), was calculated for each potential location with the building south facade and open structures in this day, and
the given weights (wi ) for windows/doors, facade and rooftops. provide at least one hour shading to the whole building facade. The
top three two-trees siting arrangements are location 3 and 5,
location 4 and 6, and location 2 and 5.
4. Results
A breakdown of component coverage (windows/doors, facade
and rooftops) is summarized in Fig. 10 for the thirteen 30-min time
The optimization results illustrate how tree shade area changes
periods between 9:30 and 15:30 with the average value of June
across different locations and building-tree distances (Fig. 6). From
15th, July 15th, August 15th and September 15th in 2016. Two trees
Fig. 6(a), tree shade coverage significantly decreases when we in-
are located at location 3 and 5, which presents the best near-
crease the distance between the tree and the building, and the
276 Q. Zhao et al. / Building and Environment 115 (2017) 269e280

Fig. 6. Tree shade coverage in one tree scenario.

optimal shading area found by the spatial optimization method in a maximize the building facade and open structure shading, and
two-tree setting. In Fig. 10(a), the tree shade coverage ratio dem- minimize the shade on the building rooftop.
onstrates the shade to different components of the residential
structures. For example, windows/door had more than 35% shade
5. Discussion
coverage ratio during 11:00 to 15:00. This results in a steady solar
radiation deduction provided by these two trees for this single
This study shows that maximizing shade area occurs with trees
family household during the heat hours. The south facade of the
planted 3 m south of central part of the home structure, unlike the
house was covered by shade more than 30% from 10:00 to 15:00.
results from prior studies that measure energy efficiency or con-
Significant blocking effects for solar radiation from tree shade were
sumption from nearby tree shade. The reason behind this differ-
provided. With less solar radiation penetrates the open structures
ence is that the compact urban setting restricts residents to plant
and heats up building facade, the individual household can signif-
trees in the west and east side of the household. When locating
icantly reduce its energy consumption. The south roof coverage
trees in front of the building south facade, the locations of windows
ratio was all less than 7%, which represents a good exposed rooftop
& doors are significant factor to influence the decision making.
for the solar energy potential. From Fig. 10(b), the four-day average
With limited number of trees to be planted for each residential
open structure accumulated shade coverage is 41.96 m2, with the
household, homeowners should focus more on planting shade trees
maximum coverage of 3.82 m2 out of 9 m2 at 12:00. The four-day
in the central area of their south front yard to provide shade (30%e
average accumulated shade coverage is 202.94 m2, with the
35% shade coverage with two trees) for their own open structures
maximum coverage of 18.38 m2 out of 45 m2 at 12:00 as well. The
and facade. Previous research recommend to plant trees at the
four-day average accumulated shade coverage of rooftop is
southwest corner of the building front yard, this research result
80.26 m2, with the maximum coverage of 7.03 m2 out of 108.5 m2 at
shows that it is not always optimal to simply plant trees at the
10:30. By using the spatial optimization method, we successfully
southwest side of the buildings. A quantitative method that
 Q. Zhao et al. / Building and Environment 115 (2017) 269e280 277

Fig. 7. Optimal shading from one tree (August 15th, at location 4).

Accumulated Weighted Shaded Area (Unit: m2)

450
400
350
300
250
200
150
100
50
0
Location Location Location Location Location Location Location Location Location Location Location
4 and 1 4 and 2 4 and 6 4 and 7 5 and 1 5 and 2 5 and 3 5 and 7 3 and 1 3 and 6 3 and 7

Target building Surrounding building

Fig. 8. Shading benefits comparison for two trees (only showing the two tree combination from location 1 to 7).

incorporates neighborhood conditions and building/tree charac- results demonstrate that two desert trees at optimal locations can
teristics is a more reliable way to achieve the maximum shade. still provide a significant amount of shading residential household.
The near-optimal two-tree arrangement (location 3 and 5) To consider the aesthetic design and add the landscaping variety of
provides around 35% shade coverage of open structures and facades residential trees, different tree spacing can be adapted in the resi-
during the peak heat hours at summer months. Since most of the dential yard based on the results in Fig. 8. Besides the tree
previous literature did not recommend to plant trees on the south arrangement at location 3 and 5, location 4 and 6 or location 2 and 5
side of buildings (limited shade coverage during the summer), the also provide significant shade coverage to the residential
278 Q. Zhao et al. / Building and Environment 115 (2017) 269e280

Fig. 9. The best near-optimal shading results from two trees (August 15th, at location 3 and 5).

household. demonstration and implementation of the model using a single-

The breakdown of the building components (windows & doors, family house with its surrounding buildings in Tempe, Arizona.
facade, and rooftops) make it possible to maximize tree shade GIS and spatial optimization techniques were employed to
coverage on the “shade-friendly” building structures such as win- formalize a mathematical model that could be used for identifying
dows & doors. With the weighting coefficients in the optimization optimal placement of the single tree that optimize accumulated
method, the emphasis of tree shade coverage can be easily adjusted shade coverage on building structures. Heuristic was used to solve
depending on different types of building materials and structures. the optimization problem involving trigonometry functions and
On the other hand, rooftop exposure is preserved by minimizing provided the near-optimal solutions of the two trees scenario (real
tree shade coverage on the building rooftops, which is not exam- world scenario) for policy makers and planners. The optimization
ined or achieved in the previous tree shade related research. Be- results show that two trees can provide a maximum of 244.90 m2
sides considering the separated building components, tree shade accumulated shade coverage to a single-story residential house's
coverage to the surrounding buildings are also evaluated. In this south facade and open structures from 9:30 to 15:30 (shade area
particular compact residential neighborhood, tree shade coverage was calculated in every 30 min) on a hot summer day from June to
on the nearby buildings cannot be ignored, especially when September, and the maximum shade coverage is achieved at 12:00
planting trees near the parcel boundaries. However, the results with the shade area of 22.20 m2 in the 54 m2 south facade and open
show little shade coverage on the surrounding buildings. The structures. Optimal tree locations can offer significant energy sav-
reason behind this finding is that shadow length is relatively ings, reduce long-term economic costs and create a healthier living
limited during the greatest insolation hours from 9:30 to 15:30. The environment.
shading benefits to the surrounding buildings need to be further This is the first known attempt to identify the precise location
explored in the future study. and number of trees to maximize tree shade on home structures.
In addition to the specifics of tree shade, this paper also dem- There is, however, more that can be done to extend this work. For
onstrates the way in which a 3D spatial optimization model can example, this study only considers an individual single-family
support the identification of optimal tree locations for providing household and its surrounding buildings. A large residential re-
shade to 3D urban building structures. To implement this model, 3D gion will require automated 3D building extraction and construc-
modeling along with GIS spatial processing techniques are used to tion techniques combining remote sensing and GIS. Also, different
determine the three dimensional geometric properties of struc- tree species, varying growing processes and alternative tree height
tures to be shaded by the optimal location of trees. We provide a and crown size reflect important options for flora. In this research,
 Q. Zhao et al. / Building and Environment 115 (2017) 269e280 279

Average summer month shade coverage ratio (%) at optimal

location
45
40
35
30
25
20
15
10
5
0
09:30 10:00 10:30 11:00 11:30 12:00 12:30 13:00 13:30 14:00 14:30 15:00 15:30

Roof coverage ratio Façade coverage ratio Open structure coverage ratio

(a) Tree shade coverage ratio.

Average summer month shade coverage area (m2) at optimal

location
20
18
16
14
12
10
8
6
4
2
0
09:30 10:00 10:30 11:00 11:30 12:00 12:30 13:00 13:30 14:00 14:30 15:00 15:30

Roof coverage area Façade coverage area Open structure coverage area

(b) Tree shade coverage area.

Fig. 10. Tree shade coverage ratio with the best near-optimal arrangement (location 3 and 5).

we use a 7 m high mature desert tree to represent a common sit- field measurement. The proposed method for carrying out the
uation in the desert setting, however, different tree species will analysis in a 3D environment is an important first step in relating
have different tree height, leaf area index/canopy density, and local level decision making to positive regional and global change.
crown size [59]. Furthermore, all of the tree-related parameters will
change during the tree's growing process [60]. All of these factors Acknowledgements
would influence the final optimization results. Future research can
extend this work to focus on comparing tree shade benefits with This material is based upon work supported by the National
different tree-related parameters at the same tree locations and Science Foundation under Grant SES-0951366, Decision Center for a
arrangements. Desert City II: Urban Climate Adaptation. Any opinions, findings,
and conclusions or recommendations expressed in this material are
6. Conclusions those of the authors and do not necessarily reflect the views of the
National Science Foundation. Qunshan Zhao is partly supported by
Strategic shade provision offers the potential to mitigate the the National Natural Science Foundation of China (No. 51378399
effects of high solar radiation loads on summer days, enabling and 41331175). The authors would like to thank DeltaCodes who
economic, environmental and health related benefits. We build provided an education discount for the Shadow Analysis Sketchup
upon research that links tree coverage with energy savings with plugin and the anonymous reviewers for their valuable comments
higher levels of precision on tree placement. Unlike prior research, and suggestions to improve the manuscript.
we provide specificity beyond the cardinal direction and address
the relationship to nearby structures, shade on windows and doors, References
and retaining the option for rooftop solar panels. Future directions
involve evaluating the microclimate benefits under different tree [1] H.H. Kim, Urban heat island, Int. J. Remote Sens. 13 (1992) 2319e2336, http://
locations and arrangements, such as wind speed/direction and dx.doi.org/10.1080/01431169208904271.
[2] L.M. Gartland, Heat Islands: Understanding and Mitigating Heat in Urban
solar radiation intensity, and quantifying the cooling benefits of Areas, Routledge, 2012.
tree shade through an outdoor urban physical scale model with [3] M. Santamouris, On the energy impact of urban heat island and global
280 Q. Zhao et al. / Building and Environment 115 (2017) 269e280

warming on buildings, Energy Build. 82 (2014) 100e113. in urban planning in hot-arid climatic regions, Landsc. Urban Plan. 94 (2010)
[4] K.L. Bassil, D.C. Cole, Effectiveness of public health interventions in reducing 149e157, http://dx.doi.org/10.1016/j.landurbplan.2009.09.002.
morbidity and mortality during heat episodes: a structured review, Int. J. [31] G. Lee, 3D coverage location modeling of Wi-Fi access point placement in
Environ. Res. Public. Health 7 (2010) 991e1001, http://dx.doi.org/10.3390/ indoor environment, Comput. Environ. Urban Syst. (2015), http://dx.doi.org/
ijerph7030991. 10.1016/j.compenvurbsys.2015.05.002.
[5] P. Bi, S. Williams, M. Loughnan, G. Lloyd, A. Hansen, T. Kjellstrom, K. Dear, [32] Y.J. Huang, H. Akbari, H. Taha, A.H. Rosenfeld, The potential of vegetation in
A. Saniotis, The effects of extreme heat on human mortality and morbidity in reducing summer cooling loads in residential buildings, J. Clim. Appl. Mete-
Australia: implications for public health, Asia. Pac. J. Public Health (2011), orol. 26 (1987) 1103e1116.
http://dx.doi.org/10.1177/1010539510391644, 1010539510391644. [33] H. Akbari, H. Taha, The impact of trees and white surfaces on residential
[6] M.A. McGeehin, M. Mirabelli, The potential impacts of climate variability and heating and cooling energy use in four Canadian cities, Energy 17 (1992)
change on temperature-related morbidity and mortality in the United States, 141e149.
Environ. Health Perspect. 109 (2001) 185e189. [34] H. Safarzadeh, M.N. Bahadori, Passive cooling effects of courtyards, Build.
[7] C.J. Tomlinson, L. Chapman, J.E. Thornes, C.J. Baker, Including the urban heat Environ. 40 (2005) 89e104, http://dx.doi.org/10.1016/j.buildenv.2004.04.014.
island in spatial heat health risk assessment strategies: a case study for Bir- [35] J. Ordo n
~ ez, E. Jadraque, J. Alegre, G. Martínez, Analysis of the photovoltaic
mingham, UK, Int. J. Health Geogr. 10 (2011) 42, http://dx.doi.org/10.1186/ solar energy capacity of residential rooftops in Andalusia (Spain), Renew.
1476-072X-10-42. Sustain. Energy Rev. 14 (2010) 2122e2130, http://dx.doi.org/10.1016/
[8] H. Akbari, M. Pomerantz, H. Taha, Cool surfaces and shade trees to reduce j.rser.2010.01.001.
energy use and improve air quality in urban areas, Sol. Energy 70 (2001) [36] H. Huang, C.-C. Ni, X. Ban, J. Gao, A. Schneider, S. Lin, Connected Wireless
295e310. Camera Network Deployment with Visibility Coverage, 2014, pp. 1204e1212.
[9] C.-R. Chang, M.-H. Li, S.-D. Chang, A preliminary study on the local cool-island [37] K.A. Amriki, P.K. Atrey, Bus surveillance: how many and where cameras
intensity of Taipei city parks, Landsc. Urban Plan. 80 (2007) 386e395, http:// should be placed, Multimed. Tools Appl. 71 (2014) 1051e1085, http://
dx.doi.org/10.1016/j.landurbplan.2006.09.005. dx.doi.org/10.1007/s11042-012-1247-1.
[10] Z.-H. Wang, X. Zhao, J. Yang, J. Song, Cooling and energy saving potentials of [38] C. Toregas, R. Swain, C. ReVelle, L. Bergman, The location of emergency service
shade trees and urban lawns in a desert city, Appl. Energy 161 (2016) facilities, Oper. Res. 19 (1971) 1363e1373.
437e444, http://dx.doi.org/10.1016/j.apenergy.2015.10.047. [39] R. Church, C. ReVelle, The maximal covering location problem, Pap. Reg. Sci. 32
[11] A. Middel, N. Chhetri, R. Quay, Urban forestry and cool roofs: assessment of (1974) 101e118.
heat mitigation strategies in Phoenix residential neighborhoods, Urban For. [40] F. Letourneux, Y. Corre, E. Suteau, Y. Lostanlen, 3D coverage analysis of LTE
Urban Green 14 (2015) 178e186, http://dx.doi.org/10.1016/ urban heterogeneous networks with dense femtocell deployments, EURASIP J.
j.ufug.2014.09.010. Wirel. Commun. Netw. 2012 (2012) 1e14.
[12] J. Gui, P. Phelan, K.E. Kaloush, J.S. Golden, Impact of pavement thermophysical [41] T. Podobnikar, A. Vre cko, Digital elevation model from the best results of
properties on surface temperatures, J. Mater. Civ. Eng. 19 (2007) 683e690, different filtering of a LiDAR point cloud, Trans. GIS 16 (2012) 603e617,
http://dx.doi.org/10.1061/(ASCE)0899-1561(2007)19:8(683). http://dx.doi.org/10.1111/j.1467-9671.2012.01335.x.
[13] J.S. Golden, J. Carlson, K.E. Kaloush, P. Phelan, A comparative study of the [42] M.F. Goodchild, J. Lee, Coverage problems and visibility regions on topo-
thermal and radiative impacts of photovoltaic canopies on pavement surface graphic surfaces, Ann. Oper. Res. (1989) 175e186.
temperatures, Sol. Energy 81 (2007) 872e883, http://dx.doi.org/10.1016/ [43] Y.-H. Kim, S. Rana, S. Wise, Exploring multiple viewshed analysis using terrain
j.solener.2006.11.007. features and optimisation techniques, Comput. Geosci. 30 (2004) 1019e1032,
[14] Q. Zhao, S.W. Myint, E.A. Wentz, C. Fan, Rooftop surface temperature analysis http://dx.doi.org/10.1016/j.cageo.2004.07.008.
in an urban residential environment, Remote Sens. 7 (2015) 12135e12159, [44] A.T. Murray, K. Kim, J.W. Davis, R. Machiraju, R. Parent, Coverage optimization
http://dx.doi.org/10.3390/rs70912135. to support security monitoring, Comput. Environ. Urban Syst. 31 (2007)
[15] Z. Tan, K.K.-L. Lau, E. Ng, Urban tree design approaches for mitigating daytime 133e147, http://dx.doi.org/10.1016/j.compenvurbsys.2006.06.002.
urban heat island effects in a high-density urban environment, Energy Build. [45] S. Bao, N. Xiao, Z. Lai, H. Zhang, C. Kim, Optimizing watchtower locations for
(2015), http://dx.doi.org/10.1016/j.enbuild.2015.06.031. forest fire monitoring using location models, Fire Saf. J. 71 (2015) 100e109,
[16] H. Akbari, Shade trees reduce building energy use and CO2 emissions from http://dx.doi.org/10.1016/j.firesaf.2014.11.016.
power plants, Environ. Pollut. 116 (2002) S119eS126. [46] T.H.D. Dao, Y. Zhou, J.-C. Thill, E. Delmelle, Spatio-temporal location modeling
[17] R. Pandit, D.N. Laband, Energy savings from tree shade, Ecol. Econ. 69 (2010) in a 3D indoor environment: the case of AEDs as emergency medical devices,
1324e1329, http://dx.doi.org/10.1016/j.ecolecon.2010.01.009. Int. J. Geogr. Inf. Sci. 26 (2012) 469e494, http://dx.doi.org/10.1080/
[18] A.V. Parisi, M.G. Kimlin, J.C.F. Wong, M. Wilson, Diffuse component of solar 13658816.2011.597753.
ultraviolet radiation in tree shade, J. Photochem. Photobiol. B 54 (2000) [47] Q. Zhao, E.A. Wentz, A.T. Murray, Shade optimization in a desert environment,
116e120, http://dx.doi.org/10.1016/S1011-1344(00)00003-8. Ext. Abstr. Proc. GISci. 2014 (2014) 118e121. https://www.researchgate.net/
[19] E.A. Wentz, S. Rode, X. Li, E.M. Tellman, B.L. Turner, Impact of Homeowner profile/Qunshan_Zhao/publication/271215424_Shade_Optimization_in_a_
Association (HOA) landscaping guidelines on residential water use, Water Desert_Environment/links/54c2c98c0cf256ed5a8f950c.pdf (Accessed April
Resour. Res. (2016), http://dx.doi.org/10.1002/2015WR018238 n/a-n/a. 22, 2016).
[20] M. Fogl, V. Moudrý, Influence of vegetation canopies on solar potential in [48] US Census Bureau., 2010 Demographic Profile Data, (n.d.). https://factfinder.
urban environments, Appl. Geogr. 66 (2016) 73e80, http://dx.doi.org/ census.gov/faces/tableservices/jsf/pages/productview.xhtml?src¼CF
10.1016/j.apgeog.2015.11.011. (Accessed September 15, 2016).
[21] R. Levinson, H. Akbari, M. Pomerantz, S. Gupta, Solar access of residential [49] Urban tree scientist - central ArizonaePhoenix long-term ecological research,
rooftops in four California cities, Sol. Energy 83 (2009) 2120e2135, http:// Cent. Ariz. Long-Term Ecol. Res. (n.d.). https://sustainability.asu.edu/caplter/
dx.doi.org/10.1016/j.solener.2009.07.016. collaborations/shadetree/ (Accessed June 10, 2016).
[22] Z. Li, Z. Zhang, K. Davey, Estimating geographical pv potential using lidar data [50] Eligible Desert-adaptive Shade Trees, Srpnet. (n.d.). http://www.savewithsrp.
for buildings in downtown san francisco, Trans. GIS 19 (2015) 930e963. com/RD/shadetreetypes.aspx (Accessed June 9, 2016).
[23] T.R. Tooke, N.C. Coops, J.A. Voogt, M.J. Meitner, Tree structure influences on [51] J.A. Duffie, in: John A. Duffie, William A. Beckman (Eds.), Solar Engineering of
rooftop-received solar radiation, Landsc. Urban Plan. 102 (2011) 73e81, Thermal Processes, fourth ed., John Wiley, Hoboken, 2013.
http://dx.doi.org/10.1016/j.landurbplan.2011.03.011. [52] R.L. Church, A.T. Murray, Business Site Selection, Location Analysis, and GIS,
[24] J.R. Simpson, E.G. McPherson, Potential of tree shade for reducing residential Wiley, Hoboken, N.J, 2009.
energy use in California, J. Arboric. 22 (1996) 10e18. [53] City of Tempe, AZ: Zoning and Development Code - Part 4: Development
[25] G. McPherson, J.R. Simpson, P.J. Peper, S.E. Maco, Q. Xiao, E. Mulrean, Desert Standards, 2011. http://www.tempe.gov/home/showdocument?id¼1914
Southwest Community Tree Guide: Benefits, Costs and Strategic Planting, (Accessed June 9, 2016).
2004. http://www.treesearch.fs.fed.us/pubs/47703 (Accessed April 10, 2016). [54] City of Tempe, AZ: Zoning and Development Code - Appendix, 2011. http://
[26] W.H. Hwang, P.E. Wiseman, V.A. Thomas, Tree planting configuration in- www.tempe.gov/home/showdocument?id¼1910 (Accessed June 9, 2016.
fluences shade on residential structures in four US cities, Arboric. Urban For [55] J.A. Wagar, Using vegetation to control sunlight and shade on windows,
41 (2015). http://urbanforestry.frec.vt.edu/documents/articles/Hwang_AUF_ Landsc. J. 3 (1984) 24e35.
2015.pdf (Accessed April 10, 2016). [56] G.M. Heisler, Energy Savings with Trees, 1986. http://www.treesearch.fs.fed.
[27] M. Sawka, A.A. Millward, J. Mckay, M. Sarkovich, Growing summer energy us/pubs/34773 (Accessed December 15, 2016).
conservation through residential tree planting, Landsc. Urban Plan. 113 (2013) [57] E. Shaviv, A. Yezioro, Analyzing mutual shading among buildings, Sol. Energy
1e9, http://dx.doi.org/10.1016/j.landurbplan.2013.01.006. 59 (1997) 83e88.
[28] F. Calcerano, L. Martinelli, Numerical optimisation through dynamic simula- [58] M. Pidwirny, Earth-sun relationships and insolation, Fundam. Phys. Geogr. 2
tion of the position of trees around a stand-alone building to reduce cooling (2006).
energy consumption, Energy Build. 112 (2016) 234e243, http://dx.doi.org/ [59] D. Armson, M.A. Rahman, A.R. Ennos, A comparison of the shading effec-
10.1016/j.enbuild.2015.12.023. tiveness of five different street tree species in Manchester, UK, Arboric. Urban
[29] R. Berry, S.J. Livesley, L. Aye, Tree canopy shade impacts on solar irradiance For 39 (2013) 157e164.
received by building walls and their surface temperature, Build. Environ. 69 [60] M.A. Rahman, D. Armson, A.R. Ennos, A comparison of the growth and cooling
(2013) 91e100, http://dx.doi.org/10.1016/j.buildenv.2013.07.009. effectiveness of five commonly planted urban tree species, Urban Ecosyst. 18
[30] V.M. Gomez-Munoz, M.A. Porta-G ndez, Effect of tree shades
andara, J.L. Ferna (2015) 371e389, http://dx.doi.org/10.1007/s11252-014-0407-7.