Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

Quarterly Journal of the Royal Meteorological Society Q. J. R. Meteorol. Soc.

139: 16431657, July 2013 B

A coupled energy transport and hydrological model for urban


canopies evaluated using a wireless sensor network
Zhi-Hua Wang,a,b * Elie Bou-Zeidb and James A. Smithb
a
School of Sustainable Engineering and the Built Environment, Arizona State University, Tempe, AZ, USA
b
Department of Civil and Environmental Engineered, Princeton University, NJ, USA
*Correspondence to: Z. H. Wang, School of Sustainable Engineering and the Built Environment, Arizona State
University, Tempe, AZ 85287, USA. E-mail: zhwang@asu.edu

We propose a new surface exchange scheme coupling the transport of energy and
water in urban canopies. The new model resolves the subfacet heterogeneity of urban
surfaces, which is particularly useful for capturing surface exchange processes from
vegetated urban surfaces, such as lawns or green roofs. We develop detailed urban
hydrological models for surfaces consisting of either natural (soil and vegetation) or
engineered materials with water-holding capacity. The coupling of energy and water
transport enables us to parametrize surface evaporation from different urban facets
including soils, vegetation and water-holding engineered surfaces. The new coupled
model is evaluated using field measurement data obtained through a wireless sensor
network deployed over the Princeton University campus. Comparison of model
prediction and measured results shows that the proposed surface exchange scheme
is able to predict widely varying surface temperatures for each subfacet with good
accuracy. Different weather conditions and seasonal variability are found to have
insignificant effect on the model performance. The new model is also able to capture
the subsurface hydrological processes with reasonable accuracy, particularly for
urban lawns. The proposed model is then applied to assess different mitigation
strategies of the urban heat island effect.
Key Words: urban canopy model; urban hydrology; urban heat island; land-atmosphere interaction;
water-holding pavements

Received 13 July 2011; Revised 8 August 2012; Accepted 9 August 2012; Published online in Wiley Online Library
30 October 2012
Citation: Wang Z-H, Bou-Zeid E, Smith JA. 2013. A coupled energy transport and hydrological model
for urban canopies evaluated using a wireless sensor network. Q. J. R. Meteorol. Soc. 139: 16431657.
DOI:10.1002/qj.2032

1. Introduction the performance of mesoscale numerical models (Cuenca


et al., 1996; Masson, 2006; Champollion et al., 2009; Hamdi
Urban areas are home to more than half the worlds popu- et al., 2010; Flagg and Taylor, 2011).
lation today. Being humanitys engine of creativity, wealth In the past decade, numerous urban canopy models
production and economic growth, rapid urbanization has (UCM) focusing on the surface energy transport or on
also emerged as the source of many adverse environmental atmospheric flow dynamics in the lower urban atmosphere
impacts due to anthropogenic stressors (Fernando, 2010) have been developed. Unlike the slab model with a flat
such as the heat island effect, emissions of greenhouse gases, surface representation of urban areas (e.g. Grimmond and
production of pollutants, etc. Understanding of urban land Oke, 2002), these physically based models explicitly capture
surface processes, such as transport of energy, water, trace most of the complex energy transport inside urban canopies
gases and pollutants, is key to solving many urban environ- by representing the urban area as big canyons (Johnson
mental problems. In addition, accurate parametrizations of et al., 1991; Oke et al., 1991), while parametrizing other
urban surface exchange schemes can substantially improve processes such as turbulent exchanges over the building


c 2012 Royal Meteorological Society
1644 Z. H. Wang et al.

facets. Existing UCMs can be categorized broadly into two difficulty also lies in the coupling of a hydrological model
types: the single-layer (e.g. Masson, 2000; Kusaka et al., 2001; with an energy transport model that captures all the essential
Lee and Park, 2008; Oleson et al., 2008) and the multilayer physics of urban canopies.
schemes (e.g. Martilli et al., 2002; Dupont et al., 2004; Kondo In this article we present a new surface exchange model
et al., 2005). The former group of UCM usually focuses on that couples the transport of energy and water inside urban
parametrization of surface energy budgets and has been canopies. The model resolves the surface heterogeneity
adopted in the Weather Research and Forecasting (WRF) of each urban facet (subfacet heterogeneity). A spatially
model (Chen et al., 2011), whereas the latter group attempts analytical scheme for computing urban surface temperatures
to also better represent the flow dynamics induced in urban and soil heat fluxes is adopted in this model. We also
canopies, although the flow is still parametrized rather than implement detailed urban hydrological models for both
resolved as for example in direct (Coceal et al., 2006) or natural and engineered materials. Predictions by the model
large-eddy (Bou-Zeid et al., 2009) turbulence simulations. are evaluated through comparison with field measurements
The performance of current UCM schemes is generally using a wireless sensor network deployed on the Princeton
good in reproducing surface temperatures and sensible University campus. Results from the comparison highlight
heat budgets. However, they are inevitably inadequate in the importance of coupling the hydrological model for
capturing the dynamics of the urban water budget and the capturing the subsurface hydrological processes and for
latent heat due to the lack of realistic urban hydrological more accurate prediction of evaporation, particularly from
models in the UCM (Grimmond et al., 2010, 2011) and vegetated surfaces. The new model is also applied to a case
the oversimplification of the representation of hydrological study investigating the effect of green roofs on the mitigation
processes. While the importance of linking the transport of urban heat island (UHI) effect.
of water and energy is increasingly recognized for urban
studies (Bedient and Huber, 1992; Mitchell et al., 2008), the 2. Urban energy balance model
modelling practice is still lagging behind, partially due to
the complexity of the physics of water transport, particularly Our new urban canopy model adopts the most common
for unsaturated soils. Lee and Park (2008) and Lee (2011) single-layer big canyon representation for urban areas,
proposed a model, in which vegetation (tall trees and grass) similar to those developed by Masson (2000), Kusaka
was incorporated into the UCM, but the hydrological model et al. (2001) and Chen et al. (2011), as shown in Figure 1.
was relatively crude and their study lacked evaluation of the The alternative representation as regular building arrays
water transport scheme in soil. is also straightforward to implement and would mainly
An additional dimension of complexity in urban involve modifying the radiative component of the model
hydrological models is the characterization of the hydraulic (Kanda et al., 2005). Note, however, that in the new model,
properties of soils and vegetation in urban areas. Field and each urban facet is subdivided into different material types
laboratory measurements of urban soil properties, such as to incorporate the surface heterogeneity, as illustrated
those of lawns and parks, are rarely found in the literature. in Figure 1. For example, roofs can be a mixture of
On the other hand, all hydrological models rely heavily on conventionally covered roofs, or green roofs with planted
accurate characterization of soil properties such as water vegetation; ground surfaces can consist of different fractions
diffusivity and hydraulic conductivity, particularly for the of asphalt, concrete pavements, and lawns; wall materials
unsaturated soils (Hanks, 1992; Chen and Dudhia, 2001; can be a combination of brick and glass. Also note here
Loosvelt et al., 2011). For example, evaporation from bare that this is the first single-layer UCM that explicitly resolves
soil surfaces is more strongly influenced by the soil-water subfacet heterogeneity, thus allowing, for example, direct
content and subsoil physics than by atmospheric demand investigation of the effect of green roofs on the UHI intensity.
when the soils are unsaturated, which is most commonly In addition, the number of subfacets for each urban facet is
found in natural environment (De Bruin and Holtslag 1982; not fixed but rather flexible to include as many variations as
Schlunder, 1988; Chen et al., 1996; Shokri et al., 2008). possible for representing real urban surface configurations.
Parametrization of urban vegetation further complicates The energy balance equation solved by the UCM for the
numerical models, by introducing heterogeneous quantities whole atmospheric layer occupied by buildings is given by:
such as roughness lengths and stomatal resistance (Niyogi
and Raman, 1997). Sensitivity studies showed that accurate Rn + HF = H + LE + G, (1)
description of the urban vegetation fraction is of great
importance for model performance (Loridan et al., 2010; where Rn = S +L S L is the net radiation with S and
Wang et al., 2011b). L denoting shortwave and longwave radiative components
Unlike their natural counterparts (bare soils or vegeta- respectively; downward and upward arrows denote the
tion), hydrological models for urban pavements or impervi- downwelling and upwelling components respectively; HF
ous roofs should be capable of resolving the water-holding is the anthropogenic heat flux (generation or release of
capacity of impervious media, which regulates the evapo- heat by anthropogenic activities per unit area); and H, LE
ration for the major fraction of urban surfaces (Yamanaka and G are the sensible, latent and conductive heat fluxes
et al., 1997; Andersen et al., 1999; Hagishima et al., 2005; respectively. We ignore the advection component since the
Borselli and Torri, 2010; Nakayama and Fujita, 2010; Abahri canyon is assumed infinite (very large). In addition, its
et al., 2011). Therefore, urban hydrological models, in addi- effect in a realistic finite canyon is implicitly included by
tion, need to incorporate the unique feature of evaporation direct measurements of the air temperature above the urban
arising from surfaces with engineered materials. Despite canopy as input to the UCM. Small advective effects may
the existence of numerous hydrological models for natural exist inside the street canyon due to small-scale subfacet
surfaces, it remains a challenge to develop an adequate but heterogeneity present in our model, but the canyon surface
relatively simple hydrological model for built surfaces. The can realistically be assumed to be statistically homogeneous


c 2012 Royal Meteorological Society Q. J. R. Meteorol. Soc. 139: 16431657 (2013)
Urban Canopy Energy Transport and Hydrology Modelling 1645

z angle sun and canyon orientation can , with

cos sun = (cos z sin sin )cscz sec ; (3)


Ta Ta Ta
za
and z is the solar zenith angle (Stull, 1998)
RESR HR Hcan
cos z = sin sin cos cos cos t . (4)

Here is latitude (positive north), the solar declination


TR2 TR1 and t the solar hour angle (24 h period), and
zR
TW1  
tUTC 2 (dn dr )
Tcan GR t = ; = r cos , (5)
zT h
12 dy
GW

TW2 TB where is the longitude (positive west); iUTC is the


Coordinated Universal Time in hours; r is the latitude
soil heat storage TG1 TG2 TG3 of the Tropic of Cancer (23.5 = 0.410 rad); dn is the day
w r number of the year (since 01 January) in the Northern
Hemisphere; dr is the day of the summer solstice (173) in
Figure 1. Schematic of the resistance network of energy transport in the the Northern Hemisphere; and dy is the average number of
single-layer urban canopy model developed and used in this study (Wang
et al., 2011a). Dimensionless w, r and h denote the normalized canyon
days per year (365.25).
width, roof width and building height respectively; za is the reference Assuming all surfaces are Lambertian (with isotropic
height in the atmospheric layer; zR is the height of the building; zT is scattering and reflection), for each urban facet, be it roof,
the representative height in the street canyon; subscripts a, R, W, G and wall or ground, the net shortwave radiation at each subfacet
can denote the atmosphere, roof, wall, ground and canyon, respectively;
numeric subscripts denote numbering of heterogeneous subfacets. This
is computed as (based on the formulae for homogeneous
figure is available in colour online at wileyonlinelibrary.com/journal/qj urban facets in Kusaka et al., 2001):

SR,k = (1 aR,k )(SD + SQ ), (6)


lshadow
over relatively small scales ( 100 m), such that these effects SD 2h + SQ FWS (1 aW,k )
are minor. The UCM computes surface temperatures and +SD wlshadow aG FWG
turbulent fluxes averaged over the sunlit and the shaded SW,k = (1 aW,k ) +S F + S lshadow a F , (7)
w
Q WG D 2h W,k WW
fractions. Heat conduction through building enclosures
+SQ FWS FWW aW,k
(roofs and walls) is determined using an imposed interior

building temperature. An adiabatic boundary is assumed for w lshadow


SG,k = (1 aG,k ) SD + SQ FGS +
heat conduction through ground surfaces at a sufficiently w
large depth. The UCM is driven by atmospheric forcing
lshadow
including air temperature, pressure, humidity, wind speed SD aW FGW +SQ FWS aW FGW , (8)
and downwelling shortwave/longwave radiation. In the 2h
following sections, we present the detailed parametrization where S and S are the direct and the diffuse solar radiation
D Q
for heat budgets in Eq. (1). received by a horizontal surface respectively; a is the surface
albedo; subscripts S, R, W and G denote sky, roof, wall
2.1. Radiation budgets and ground facet respectively; numeric subscript k = 1, 2,
3,. . . denotes the index of each subsurface of a facet; and Fij
The presence of buildings has a significant effect on the are the sky view factors computed as (Harman et al., 2004):
distribution and redistribution of radiation inside the urban
canyon, due to shading and radiative trapping (reflection).
2
h h
Although an infinite number of radiative reflections can be FSG = FGS = 1 + , (9)
resolved analytically in a street canyon with homogeneous w w
 w 2 w
facets (Harman et al., 2004), extension of the algorithm to
an urban canyon with subfacet heterogeneity (see Figure 1) F WW = 1 + , (10)
h h
is not trivial and involves expensive matrix operations with
FGW = 0.5(1 FGS ), (11)
large dimensions. It has been verified that a two-reflection
model is sufficient for typical urban surface temperatures FWG = FWS = 0.5(1 FWW ). (12)
and thermal properties (Wang, 2010) and is adopted in this
article. Note that in Eqs (7) and (8), the quantities aG and aW are the
Inside the urban canyon we define the normalized shadow equivalent albedos of the ground and wall facets respectively.
length lshadow as (Kusaka et al., 2001): Derivation of the equivalent albedos is as follows. Consider
diffuse solar radiation, ij , originating from facet j and
 absorbed by facet i (indices i, j can represent wall, ground,
h tan z sin n lshadow < w
lshadow = , (2) roof and sky). If facet j is heterogeneous and consists of N
w lshadow w subsurfaces, it can be written as
 N 
where h and w are the normalized building height and 
canyon width respectively, r is the normalized roof width, ij = (1 ai ) fk aj,k Fji SQ = (1 ai )aj Fji SQ . (13)
with r +w = 1; n is the difference between the solar azimuth k=1


c 2012 Royal Meteorological Society Q. J. R. Meteorol. Soc. 139: 16431657 (2013)
1646 Z. H. Wang et al.


N 
N
practical, particularly when UCMs are employed to study
Therefore aG = fk aG,k and aW = fk aW,k where fk is
k=1 k=1 a relatively long period or a mix of canyons with random
the area fraction of the kth subsurface in a given ground or orientations.
wall facet. Hereafter, an equivalent quantity with an overbar
is defined in the same manner as aG , unless otherwise 2.2.1. Fluxes inside the canyon
specified. It is also noteworthy that in Eq. (13), all the
N subsurfaces are assumed to be uniformly distributed It can be seen from Figure 1 that there are two resistance
in space, such that the view factor Fji is the same for networks for turbulent fluxes in urban canopies: one
each subfacet of facet j, relaxing this assumption would inside the canyon where canyon facets (walls and ground)
significantly complicate the problem. exchange energy with the canyon air and the other from
Similarly, using two reflections between urban facets, the the roof/canyon-top to the atmospheric layer. Inside the
net longwave radiation at each subfacet can be calculated as: canyon, sensible heat for each subfacet is parametrized as

LR = R (L TR4 ), (14) cp a (TG,k Tcan )



HG,k = , (19)
LW,k = W,k (FWS L + G FWG TG4 + 4
W,k FWW TW,k RESG
cp a (TW,k Tcan )
TW,k
4
) HW,k = , (20)

RESW
+ W,k (1 G )FGS FWG L + 2(1 G )W,k FGW
4
FWG TW,k where RES is the aerodynamic resistance; cp is the specific

heat of air; a is the density of air; and subscripts a and
+ W,k (1 W,k )FWS FWW L + (1 W,k ) can denote air and canyon, respectively.
G FWG FWW TG4 (15) Parametrization of latent heat flux inside the urban
canyon is more complex. First of all, we assume that the
+ W,k W,k (1 W,k )FWW FWW TW,k
4
contribution to latent heat from the wall is negligible since
LG,k = G,k (FGS L + 2W FGW TW
4
TG,k
4
) (16) the water-holding capacity of the wall is insignificant, that is

+ 2G,k (1 W )FWS FGW L + (1 W )
LEW,k = 0, (21)
4
G,k FGW FWG TG,k ,
+ 2G,k W (1 W )FWW FGW TW
4 Next we consider the ground facet consisting of
engineered pavements with water-holding capacity, bare soil
where is the emissivity, the StefanBoltzmann constant and vegetation. For engineered (concrete, brick or asphalt)
and T the surface temperature. surfaces,

2.2. Turbulent heat fluxes 0 for w = 0
LEG,eng = a Lv (qG,eng qcan ) , (22)
RESG for w > 0
Turbulent transport of energy depends largely on the wind
profile in urban canopies. Above the roof, the mean wind
profile follows the common log-law but varies exponentially where Lv is the latent heat of vaporization, q is the specific
within the canyon (Masson, 2000). The horizontal (along humidity, superscript star denotes the saturated specific
canyon) and vertical wind speed, integrated over 2 to humidity at a given surface temperature and w is the actual
average over all possible wind-canyon relative orientations, depth of retained water film/layer on the engineered surfaces
are determined by for which a prognostic equation, presented later, is solved.
Note that Masson (2000) developed specific evaporation
  schemes for roofs and ground surfaces partially covered

h/3
2 1 h ln z0,town with snow, similar to the formulation for water-holding
Ucan = exp   |Ua |, (17) pavements in Eq. (22).
4 w ln
z+h/3
z0,town For natural surfaces, namely bare soils or vegetation, the

Wcan = Cd |Ua |, (18) parametrization scheme for latent heat follows Brutsaert
(2005):
where Ua is the wind velocity, usually measured at the first
atmospheric model level, z0,town = 1.2 m is the roughness LEnat = Lv e Ep,nat , (23)
length of the urban area,
z = za zR is the height of
the first atmospheric model level above the roof (or where e is a reduction factor as a function of soil-water
measurement height for off-line implementations), with content. For vegetation, e = 1 and the dependence of the
za the height of the atmospheric model level and zR the evapotranspiration on soil-water availability is incorporated
height of the roof. The drag coefficient, Cd , is computed using the stomatal resistance in Eqs (25) and (26). On the
from the temperature and humidity in and above the other hand, for bare soil,
canyon (to account for stability following the formulation
of Mascart et al. (1995)), and from z0,town . Note that in r
e = , (24)
Eq. (17) integration over 360 is performed to eliminate the s r
dependence of Ucan on the direction of Ua , assuming that the
probability of occurrence of all wind directions with respect with the volumetric soil-water content, s the soil-water
to canyon orientation is the same. The angular average is content at saturation and r the reference soil-water content


c 2012 Royal Meteorological Society Q. J. R. Meteorol. Soc. 139: 16431657 (2013)
Urban Canopy Energy Transport and Hydrology Modelling 1647

at which evaporation is suppressed. And the potential turbulent heat fluxes from roofs and canyons,
evaporation rate is again given by the resistance method, as

NR
a (qG,nat qcan ) Hu = r fR,k HR,k + wHcan , (32)
Ep,nat = for bare soil
RESG k=1
a (qG,nat qcan ) 
NR
Ep,nat = for vegetation, (25) LEu = r fR,k LER,k + wLEcan , (33)
RESG + RS
k=1
where Rs is the stomatal resistance of vegetation. Here we
use the parametrization scheme based on a meteorological where NR is the number of subsurfaces on the roof facet.
approach (cf. physiological approach), by relating the The canyon temperature and humidity, Tcan and qcan are
stomatal resistance to meteorological variables including computed by the energy and humidity balance inside the
solar radiation, soil-water availability, vapour pressure canyon, by requiring that
deficit and air temperature (Noilhan and Planton, 1989;
2h  
W N G N
Niyogi and Raman, 1997):
Hcan = fW,k HW,k + fG,k HG,k , (34)
w
Rs = Rs,min FSR Fq Fe FT /LAI, (26) k=1 k=1


NG
where Rs,min is the minimum stomatal resistance depending LEcan = fG,k LEG,k , (35)
on the vegetation type; LAI is the leaf area index; and k=1
FSR , Fq , Fe and FT are the adjusting factors for the solar
radiation, soil-water content, vapour pressure deficit and where NW and NG are the number of subsurfaces of wall and
temperature, respectively (detailed in Appendix A). ground, respectively. Combining with Eqs (28), (29) and the
The aerodynamic resistance between canyon facets and fluxes inside the urban canyon in section 2.2.1, Eqs (34) and
air was formulated by Rowley et al. (1930) and Rowley and (35) can be solved as
Eckley (1932), based on in situ measurements (Masson,
2000): Ta /REScan + (2h/w)(TW /RESW ) + TG /RESG
Tcan = ,
 1/REScan + (2h/w)(1/RESW ) + 1/RESG
RESG = RESW = (11.8 + 4.2 Ucan 2 + W 2 )1 . (27) (36)
can
qa /REScan + qG /RESG
2.2.2. Fluxes from canyon and rooftop to the atmosphere qcan = . (37)
1/REScan + 1/RESG
From the canyon air to the atmospheric layer, turbulent 2.3. Surface temperatures and conductive heat fluxes
fluxes are given by
Consider a finite one-dimensional domain z [0, d], where
cp a (Tcan Ta )
Hcan = , (28) for roof or wall, d is the thickness and d for ground.
REScan Surface temperatures and conductive heat flux of urban
Lv a (qcan qa ) facets are computed by solving the one-dimensional heat
LEcan = , (29) conduction equation using Greens function approach
REScan
(Wang et al., 2011a; Wang and Bou-Zeid, 2012). The
and from the roof top to the atmospheric layer the turbulent solutions are spatially analytical in terms of convolution
heat exchange is given by: integrals, which can be evaluated numerically though
temporal discretization, and are capable of reproducing
cp a (TR Ta ) smooth spatial variation of temperatures and heat fluxes
HR = , (30) inside solid media:
RESR
Lv a (qR qa )  t
LER = e , (31)
RESR T(z, t) = T 0 + f1 (t )dg(z, )
0
 t
The reduction factor in Eq. (31) follows from Eq. (22)
f2 (t )dg(d z, ), (38)
for conventional roofs (paved with engineered materials) 0
and Eq. (24) for green roofs. Aerodynamic resistances  t
T
RESR and REScan are computed using MoninObukhov G(z, t) = k = k f1 (t )dg (z, )
similarity theory. To reduce the computation cost needed z 0
 t 
for numerical iterations, here we use a set of closed-form
+ f2 (t )dg (d z, ) , (39)
fluxprofile relations derived by Mascart et al. (1995), as 0
shown in Appendix B.
where f1 and f2 are the heat flux input at the exterior (exposed
2.2.3. Total turbulent heat flux and related canyon to sun) and interior boundaries respectively; T0 is the initial
temperature and humidity temperature profile inside the solid (assumed uniform); k is
the thermal conductivity; g(x, t) is the fundamental (Greens
The total turbulent heat fluxes from the urban area to function) solution of one-dimensional heat diffusion with
the atmospheric layer are then given by the summation of homogeneous boundary conditions; and g = g/x is the


c 2012 Royal Meteorological Society Q. J. R. Meteorol. Soc. 139: 16431657 (2013)
1648 Z. H. Wang et al.

spatial derivative of g. For walls and roofs with finite


thickness d, Greens function solution is given by:

  
z 2

t d 2d  1
g(z, t) = + 3 1 1 2 (40)
kd 6k d k n=1 n2
 n z    n 2 
cos exp t ,
d d

where = k/c is the thermal diffusivity and n is a


summation index. For a semi-infinite soil domain, such
as in the case of the urban ground facet, Greens function
solution is given by



2 (t/ ) z2 z z
g(z, t) = exp erfc .
k 4t k 2 t
(41)

where erfc() is the complementary error function. The


exterior boundary condition at z = 0, for all urban facets
with solar exposure is
Figure 2. Schematic hydrological models for (a) natural surfaces and

T  (b) engineered pavements in urban areas. The soil in (a) is vertically
f1 = k = Rn H LE. (42) discretized to capture the variation of hydraulic properties of unsaturated
z z=0 soils. In (b), dw is the maximum water-holding depth of porous
engineered materials. This figure is available in colour online at
wileyonlinelibrary.com/journal/qj
The interior boundary condition for the wall and the roof
is
3.1. Prognostic equation for volumetric water content
T(d, t > 0) = TB , (43)
In the hydrological model of natural surfaces, the prognostic
where TB is the building interior temperature and the heat equation for the volumetric soil-moisture content is
flux at the interior surface can be solved using Eq. (38) by given by the one-dimensional vertical diffusive form of
setting z = d. For ground facet, the interior boundary at the Richards equation, derived from Darcys law (Hanks,
z is adiabatic, that is 1992):


T  nat nat
f2 = k =0 for ground. (44) = D + K + F , (45)
z z t z z

The integrals in Eqs (38) and (39) are then discretized in where D is the soil-water diffusivity, K is the hydraulic
time and the two equations can be solved numerically conductivity and F represents source and sink terms; that
at each time step. Details of numerical algorithm and is, F = P +QF R E, with precipitation P, evaporation
implementation of the spatially-analytical scheme of surface E for bare soil (or evapotranspiration ET for vegetation),
temperatures and heat fluxes were discussed by Wang at el. surface runoff R and anthropogenic water QF . Soil moisture
(2011a). is calculated by vertically discretizing the soil layer, see
Figure 2(a), and rewriting Eq. (45) in the form:
3. Urban hydrological model
1 1
= (P + QF E R Qw,12 ) for layer 1, (46)
Evaporation and transport of water/moisture are inade- t d1
quately resolved in existing urban canopy models (Grim- k 1
mond et al., 2010). This can be attributed to the lack of = (Qw,k1k Qw,kk+1 ), for layer k > 1,
t dk
simple but adequate urban hydrological models. In particu- (47)
lar, evaporation from engineered pavements, such as asphalt
and concrete, and its cooling effect have long been ignored where d , k = 1, 2, 3, . . . , is the thickness of the kth sublayer.
k
in most urban models (Nakayama and Fujita, 2010). How- Q
w,kk+1 is the water infiltration through the interface
ever, some models account for it in alternative ways, for between the kth and (k + 1)th soil layers:
example Masson (2000) included snow cover on impervi-
ous pavements. In this section, we detail the integration of
k k+1
hydrological models for both natural (bare soil and vegeta- Qw,kk+1 = Dk,k+1 + Kk,k+1 , 1 k < N,
tion) and engineered (concrete, asphalt, gravel, etc.) surfaces (dk + dk+1 )/2
in the UCM. (48)


c 2012 Royal Meteorological Society Q. J. R. Meteorol. Soc. 139: 16431657 (2013)
Urban Canopy Energy Transport and Hydrology Modelling 1649

with averaged hydraulic properties defined as where Ks is the saturation hydraulic conductivity, the
soil tension function, s the saturation soil suction, and b a
dk +dk+1 dk +dk+1 fitting parameter dependent on soil type.
Dk,k+1 = , Kk,k+1 = .
dk /Dk +dk+1 /Dk+1 dk /Kk +dk+1 /Kk+1
(49) 3.3. Anthropogenic heat and water
The lower boundary condition is assumed to be a zero-flux Urban models involve unique sources of heat and moisture
surface due to human activities. The so called anthropogenic
Qw,NN+1 0, (50) stressors (Fernando, 2010) result from sources of emission
including human metabolism, vehicles, heating, ventilation
where N is the total number of discrete sublayers. and air conditioning (HVAC) in buildings, industry and
The surface evaporation term in Eq. (46) can be evaluated power plants. Anthropogenic heat and moisture inputs
from the latent heat defined in section 2.2 for each urban vary significantly both in space and time, of which
subfacet consisting of bare soil or vegetation. Note that in measurements are not readily available in most cities.
section 2.2, the evaporation (latent heat) is expressed as a Grimmond et al. (2010), based on the results of an
function of soil moisture content, thus these two quantities international urban energy balance models comparison
are strongly coupled and the solutions of the two equations project, concluded that the omission of anthropogenic water
require numerical iteration. is a main contributor to the inadequacy of current urban
For engineered pavements, it is assumed that there exists a canopy models in predicting evaporation and latent heat
water-holding layer above the impervious datum, as shown budget. Masson (2000) pointed out that anthropogenic heat
in Figure 2(b). The prognostic equation for the actual depth release does not directly modify the surface energy balance,
of water retention w can be written as but rather affects the atmospheric budgets.
 A review of methods for estimating anthropogenic heat
w P Ep , if w < dw and moisture in the urban environment indicated that a
eng = , (51)
t 0, if w dw combination of the inventory method and the simplified
building energy method is needed at the current stage of
where eng is the surface porosity of engineered materials; practice (Sailor, 2011). Due to limited measurement data
Ep is the potential evaporation rate, which can be computed available, we do not explicitly incorporate the anthropogenic
from Eq. (22); and dw is the maximum water-holding depth components HF and QF in the energy balance and
of a engineered subfacet, which is set as 50 mm, 5 mm and hydrological equations, respectively. More importantly, for
5 mm for the gravel roof, the concrete pavement and the the offline UCM driven by field measurements, the inclusion
asphalt pavement, respectively. of anthropogenic sources is similar to an increase in the local
atmospheric forcing, hence its effect is implicitly lumped
3.2. Hydraulic properties for unsaturated soils into the measurement of the atmospheric temperature and
specific humidity by eddy covariance stations.
Determination of water infiltration within the soil layer, as
shown in Eq. (48), is dictated by the soil-water diffusivity 4. Model evaluation
D and the hydraulic conductivity K, which are highly
nonlinearly related to soil moisture. Both D and K can vary In this section, we evaluate the new UCM using field
by several orders of magnitude even for very small changes measurements by a mixed network of wireless and
in soil moisture, particularly when the unsaturated soil is conventional sensor nodes deployed over the campus of
relatively dry (van Genuchten, 1980). The high nonlinearity Princeton University under the Sensor Network Over
of unsaturated hydraulic properties, together with the large Princeton (SNOP) project. The sensing instruments of
variation of soil moisture from the soil surface to the flux- relevance in this study include: (i) a wireless network
free boundary, precludes analytically tractable solutions to consisting of 12 Sensorscope stations (nodes) (see
the prognostic equation in (45), compared with the solutions http://SNOP.princeton.edu for the details); and (ii) a
in section 2.3 for temperatures and heat fluxes. conventional eddy-covariance (EC) station deployed on
It was shown by Cuenca at al. (1996) that for bare soils, the the roof of a building. Each Sensorscope station includes a
parametrization of soil water has a significant effect on the TNX infrared thermometer (ZyTemp) for measuring surface
partitioning of surface energy into sensible and latent heat. temperatures, an EC-TM probe (Decagon Devices, Inc)
Although there are numerous existing and ongoing research for measuring volumetric water contents and temperatures
efforts focusing on the determination of the functional of soil, and a solar radiation sensor (Davis). The EC
relation between the soil-water content and the unsaturated station contains a Campbell Scientific three-dimensional
soil hydraulic properties, in this study, we adopt the widely sonic anemometer (CSAT3, measuring at 20 Hz frequency),
used empirical model developed by Clapp and Hornberger an open-path infrared gas analyser (LI-7500 from Licor
(1978) and Cosby et al. (1984) Biogeosciences), a temperature and relative humidity probe

2b+3 (HMP45C from Vaisala), an infrared surface temperature

K( ) = Ks , (52) recorder (IRR-P from Apogee Instrument), a wind monitor
s (05103 R.M. Young from Campbell), and a four-component
radiometer (NR01 from Hukseflux). Data are collected
D( ) = K( ) , (53) with a Campbell Scientific CR3000 data logger. The


b instrumentation and the map of deployment of the roof
s
( ) = s , (54) EC station and the Sensorscope stations are shown in
Figure 3. The spatial scale in this study ranges from the


c 2012 Royal Meteorological Society Q. J. R. Meteorol. Soc. 139: 16431657 (2013)
1650 Z. H. Wang et al.

Figure 3. The Sensor Network Over Princeton instrumentation: (a) the roof top eddy covariance (EC) station, (b) a typical Sensorscope wireless station
and (c) map of deployment of the EC station and the wireless stations (0112) in the study area (stations 0105 and 12 on roof level, stations 0611 on
ground level). The study area ( 400 m 400 m) of the UCM is circumscribed within the yellow dotted lines. This figure is available in colour online at
wileyonlinelibrary.com/journal/qj

building-resolving scale ( 10 m, i.e. footprint of individual Table 1. Input canyon dimension and surface parameters.
Sensorscope station) to the small neighbourhood scale
( 500 m, i.e. the entire study area). Variable Symbol Value
Lists of input parameters for canyon dimension and
material properties are shown in Tables 1 and 2, respectively. Roof level (building height) (m) zR 18.9
Reference height of atmospheric measurements (m) za 23.23
In Table 1, the location ( and ) and dimensional canyon
Normalized building height (-) h 0.3
parameters (zR , za , dR and dW are directly measured from Normalized roof width (-) r 0.4
the typical street and building dimensions in our study Normalized road width (-) w 0.6
area. In addition, normalized canyon dimensions (h, r Thickness of roof (m) dR 0.5
and w), fractions of subfacet surfaces (fasp , fcon and fveg ), Thickness of wall (m) dW 0.3
street canyon orientation (can ) and roughness lengths for Fraction of asphalt pavement on ground (-) fasp 0.5
momentum transfer (zm,R and zm,can ), are also estimated for Fraction of concrete pavement on ground (-) fcon 0.2
Fraction of vegetation on ground (-) fveg 0.3
the study area from direct inspection and aerial photographs
Roughness length for momentum for roof (m) zm,R 0.01
of Google Earth . Roughness lengths for heat transfer is Roughness length for heat for roof (m) zh,R 0.001
empirically related to the ones for momentum transfer Roughness length of momentum for canyon (m) zm,can 0.05
by zh = zm /10 for both roof and canyon. Values of all Roughness length of heat for canyon (m) zh,can 0.005
parameters in Table 1 have been tested in Wang et al. (2011a) Street canyon orientation (rad) can /4
for the same study area and yielded surface temperature Latitude (positive north) (rad) 0.7043
predictions with reasonable accuracy (general discrepancy Longitude (positive west) (rad) 1.3029
< 2.5 C as compared with measurements).


c 2012 Royal Meteorological Society Q. J. R. Meteorol. Soc. 139: 16431657 (2013)
Urban Canopy Energy Transport and Hydrology Modelling 1651

Table 2. Input material thermal/hydraulic properties. subfacets. The UCM predicts surface temperatures for
four subfacets with significantly different material thermal
Parameter Symbol Value properties, namely gravel roof, asphalt pavement, concrete
Roof surface albedo aR 0.30 pavement and lawn, with good accuracy as compared
Wall surface albedo aW 0.25 with the field measurements. Figure 5 shows that it also
Albedo of asphalt, concrete, aG 0.15, 0.40, 0.15 reproduces the net radiation over the roof well.
vegetated (A, C, V) ground Note that in Figure 4(b), we have measurements of
Roof surface emissivity R 0.95 the grass temperature from four Sensorscope stations
Wall surface emissivity W 0.95 deployed at different locations, while there are only two
Emissivity of (A, C, V) ground G 0.95, 0.98, 0.93
stations measuring the surface temperature of concrete and
Thermal conductivity of roof kR 1.00
(W K1 m1 ) asphalt pavement surfaces (one per surface type). As the
Thermal conductivity of wall kW 1.30 UCM computes surface temperatures averaged over the
(W K 1 m1 ) sunlit and the shaded fractions in the canyon, there is
Thermal conductivity of (A, C, kG 1.20, 1.80, 2.00 a higher discrepancy between the UCM prediction and
V) ground (W K1 m1 ) measurements of concrete and asphalt temperatures, in
Heat capacity of roof (MJ CR 2.00 both the magnitude and phase of variation, compared with
K1 m3 ) grass. This discrepancy, as verified in Wang et al. (2011a),
Heat capacity of wall (MJ K1 CW 1.20
m3 )
cannot be eliminated by calibration of the thermal properties
Heat capacity of (A, C, V) CG 1.00, 2.40, 1.30 for the asphalt or concrete pavements, but is expected
ground (MJ K1 m3 ) to decrease if we have more measurement points for the
Volumetric water content at s 0.47 same type of surfaces that cover both sunlit and shaded
saturation (m3 m3 ) fractions in the canyon. On the other hand, it is clear
Saturation soil suction (m) s 0.355 that by averaging the experimental data over four different
Fitting exponent for unsatu- b 5.33 locations, the difference between sunlit and shaded areas is
rated soil (-)
smoothed out in the measured grass temperature (Tveg
Hydraulic conductivity at sat- Ks 3.38E-06
uration (m s1 )
in Figure 4(b)), in a manner similar to the averaged
Reference soil-water content r 0.15 UCM parametrization; and the agreement between model
(m3 m3 ) predictions and measurements is better.
Porosity of roof gravel (-) eng 0.3 The statistics including the root-mean-square error
(RMSE), the mean bias error (MBE) and the coefficient
of determination (R2 ) (see Grimmond et al. (2010) for
The soil hydraulic properties in Table 2 are adopted definition) of the comparisons between model predictions
from Chen and Dudhia (2001) for sandy loams, which and field measurements, for both test periods, are
are typical for soils used in urban lawns (Craul, 1985). summarized in Table 3. In the test period from 4 to 9
Values of thermal properties in Table 2 are first selected May 2010, predictions of surface temperatures are of good
from practical ranges of corresponding solids/soils reported accuracy in general, with RMSE around 2 C, MBE around
in the literature (e.g. Brutsaert, 2005; ASHRAE, 2009), 1 C and R2 around 0.97 (except for concrete pavement with
and then calibrated to give optimal output for the 24 h R2 of 0.90). The RMSE values for Rn is 16 W m2 , which
period on 5 May 2010. Thermal parameters that have is statistically better than the mean RMSE value of 30 W
been tuned during the calibration process are shown in m2 of a group of 32 urban canopy schemes evaluated by
bold in Table 2. Once determined, values of these thermal Grimmond et al. (2011) at the final run (with most detailed
parameters (shown in Table 2) are fixed for the rest of the inputs).
test period to assess the robustness of the model and the The soil volumetric water content (VWC) was measured
calibration. Also note that values of emissivity are selected as a lumped value along the first 5 cm of the soil using
from University of CaliforniaSanta Barbaras emissivity EC-TM probes. The measured soil moisture is sensitive to
library (http://www.icess.ucsb.edu/modis/EMIS/html/em. variation of soil temperature. Thus VWC measurements
html) and not subject to calibration. In addition we note that are corrected by multiple regression analysis with diurnal
UCM predictions are insensitive to emissivity (Wang et al., measurement of 3 days (with each day from May, July and
2011b). The initial distributions of temperatures/moistures October of 2010 respectively to cover seasonal variability)
of the soil and solid media are assumed to be uniform across (Cobos and Campbell, 2007). The corrected VWC is given
the entire depth, with the values initialized as the measured by:
surface temperatures/moistures by the Sensorscope stations
for each subfacet respectively. corrected (in%) = 0.943measured 8.502 102 Tsoil +3.936.
The first evaluation test consists of a period of 6 days: (55)
49 May 2010, which cover different weather conditions,
including clear (5 and 7 May), cloudy (4 and 6 May), The R2 value for the linear regression is 0.95, indicating
overcast (9 May) and lightly rainy (8 May) days. Results of a good quality of the statistical analysis. In Figure 6 the
comparison between UCM predictions and measurements model prediction of the volumetric soil-water content is
of surface temperatures, net radiation, and soil-water compared with the measurements by the wireless stations.
content are shown in Figures 46. It is clear from Figure 4 Note that there was a precipitation event on 3 May 2010,
that the new UCM, with subfacet resolution and an that is on the day immediately before our study period.
advanced scheme for computation surface temperature The soil moisture was hence replenished prior to the test
(Wang et al., 2011a) is able to capture the differences in period. For the first 3 days, 46 May 2010, when the soil
surface temperature evolution for heterogeneous urban saturation was relatively high (> 60%), the model predicts


c 2012 Royal Meteorological Society Q. J. R. Meteorol. Soc. 139: 16431657 (2013)
1652 Z. H. Wang et al.

(a) 80
TR measured
TR UCM
60

Temperature (C)
40

20

0
0 24 48 72 96 120 144
UTC time (h)

(b) 80
Tasp measured
Tcon measured
Tveg measured
60 Tasp UCM
Tcon UCM
Temperature (C)

Tveg UCM

40

20

0
0 24 48 72 96 120 144
UTC time (h)

Figure 4. Comparison of urban canopy model predictions and measurements of surface temperatures for (a) gravel roof and (b) heterogeneous ground
with asphalt and concrete pavements and vegetation, during 49 May 2010. This figure is available in colour online at wileyonlinelibrary.com/journal/qj

800
Rn measured
600 Rn UCM
Net radiation (W m2)

400

200

200
0 24 48 72 96 120 144
UTC time (h)

Figure 5. Comparison of urban canopy model predictions and measurements of net radiation over roofs, on 49 May 2010. This figure is available in
colour online at wileyonlinelibrary.com/journal/qj

the soil moisture with good accuracy. The overall statistics The evaluation above illustrates the good performance
in Table 3 shows that model is capable of predicting the of the new UCM for a period of 6 days under varying
near surface soil-water content with good accuracy, with weather conditions, including clear, cloudy, overcast and
RMSE and MBE values < 0.5% m3 m3 . Note that in the light rainy days. To investigate the applicability of the model
current setting, we have only one lumped measurement to different seasons, a second test period was selected with
of the soil moisture of urban lawns through the first few three relatively hotter and drier days, namely 2729 July
centimetres of the soil surface layer. To further evaluate 2010. Input parameters identical to those used in the first
the urban hydrology model proposed in this article, it is test period in May were applied to this case study. Results of
desirable to have a vertical measurement of the soil moisture comparison between UCM predictions and measurements
profile from the surface through an adequate depth of the of surface temperatures, net radiation over roof and soil-
soil, preferably down to a depth of 2 m. water content are shown in Figure 7. It can be seen that the


c 2012 Royal Meteorological Society Q. J. R. Meteorol. Soc. 139: 16431657 (2013)
Urban Canopy Energy Transport and Hydrology Modelling 1653

Table 3. Summary of the statistics of comparisons between model predictions and field measurements for both test periods.

Variable 049 May 2729 July


RMSE MBE R2 RMSE MBE R2
TR ( C) 2.5 1.7 0.97 3.1 2.6 0.97
Tasp ( C) 2.1 1.0 0.98 2.4 0.6 0.98
Tcon ( C) 2.5 0.5 0.90 2.9 1.0 0.83
Tveg ( C) 1.3 0.4 0.97 2.7 1.9 0.93
Rn (W m2 ) 16.0 4.8 0.99 20.1 9.0 0.99
soil (m3 m3 ) 0.005 0.002 0.99 0.003 0.001 0.99

0.50 effectively reduce surface temperature through evaporative


cooling. Note that the results in Figure 8 were obtained
Soil water content (m3 m3)

measurement
UCM by taking the atmospheric forcing and surface parameters
0.40 for 49 May 2010, the same as in our first case study. An
interesting observation is that during the first 3 days of
the simulation, soil moisture retained by roof vegetation
0.30 was relatively high, thus sufficient evaporation occurred and
kept the surface temperature of the green roof lower than
that of a roof with aR = 0.5. During the last 3 days, when soil
0.20 moisture decreased rapidly, the effectiveness of a green roof
24 0 24 48 72 96 120 144
UTC time (h) in reducing roof temperature was significantly decreased.
Particularly in the last 48 h, the surface temperature of the
Figure 6. Comparison of urban canopy model predictions and measure- green roof is comparable to a roof with aR = 0.1, indicating
ments of volumetric soil-water content, on 49 May 2010. This figure is
available in colour online at wileyonlinelibrary.com/journal/qj that the evaporative cooling effect of a green roof is nearly
suppressed for dry soil.
An alternative way to assess the UHI mitigation by
accuracy of UCM predictions, when applied to a different evaporative cooling is through the partitioning of latent
season and meteorological conditions, are maintained for and sensible fluxes, for example by using the Bowen ratio
all the quantities compared with field measurements. The H/LE for the entire urban area (not limited to the roof
statistics of this test period are similar to those of the first fraction). In Figure 8(b), a time series of the Bowen ratio
period, as shown in Table 3. was plotted by comparing a green roof with a conventional
roof with aR = 0.3. Again it is clear that for the first 4 days of
5. Case study for urban heat island mitigation simulation, the green roof is effective in reducing the Bowen
ratio, by redistributing sensible heat to latent heat through
A useful feature of the UCM, coupled with an urban evaporative cooling effect. As the soil-moisture content
hydrological model, lies in its ability to assess different approaches the residual value, evaporation through a green
strategies for mitigation of the well-known UHI effect in roof is greatly reduced through an increase in stomatal
cities, such as green (i.e. vegetated) roofs or white (highly resistance, and the lower albedo of the green roof in this
reflective) roofs. As our new UCM is able to capture comparison results in higher sensible heat fluxes towards
subfacet heterogeneity of building roofs, together with the the end of our study period.
hydrological model for urban soils, it can be readily extended The simple case study presented here highlighted a couple
to study the effect of increasing surface albedo (white roof) of interesting features of the new UCM developed in
versus evaporative cooling (green roof) in reducing roof this article. First, by coupling a relatively detailed urban
surface temperatures as well as in redistributed partitioning hydrological model to the energy balance model, the new
of energy fluxes for UHI mitigation. Here we present a case UCM is capable of physically simulating urban surface
study comparing such an option for our first evaluation exchange processes, in particular, those associated with
period. vegetated surfaces in urban areas. Therefore, it is applicable
In Figure 8(a), roof temperatures are evaluated using to assess different primary UHI mitigation strategies, such
various surface albedos, that is aR = 0.1, 0.5 and as green versus white roofs. Second, vegetated surfaces, not
0.9 respectively. An increase of surface albedo induced limited to green roofs, are in general very effective in reducing
significant reduction in roof surface temperatures, under urban surface temperature and redistributing turbulent heat
all weather conditions. Temperature reduction as large as transport through evaporative cooling, provided they are
30 C around noon can be accomplished by increasing aR well irrigated. Otherwise, the relatively low surface albedo
from 0.1 to 0.9. This implies that white roofs, with a high of green grass and other short vegetation (0.150.25,
surface albedo, are effective in reducing UHI by reducing according to Brutsaert, 2005) increases heat absorption
sensible heat transfer from the roof to the atmosphere and and has a potential exacerbating effect on the UHI. This
reduced heat storage inside the urban canopies. This also also is partially illustrated in our measurements where
has a positive impact on the cooling loads of the building towards the end of the first test period, the differences
since a significant fraction of the incoming solar radiation is in peak temperatures between the vegetated and concrete
simply reflected back to the atmosphere in this case. On the ground surfaces are significantly reduced as the soil moisture
other hand, a green roof, with a relatively low surface albedo and evaporation from the green surfaces are reduced,
aR = 0.15 (cf. aR 0.20.4 for a conventional roof), can although ground vegetated surfaces maintain more available


c 2012 Royal Meteorological Society Q. J. R. Meteorol. Soc. 139: 16431657 (2013)
1654 Z. H. Wang et al.

(a) 80 (b) 80
Tasp measured
TR measured
Tcon measured
TR UCM
60 Tveg measured
60
Temperature (C)

Temperature (C)
Tasp UCM
Tcon UCM
40 Tveg UCM
40

20
20

0
0 24 48 72 0 24 48 72
UTC time (h) UTC time (h)

(c) 800 (d) 0.40


Rn measured measurement

Soil water content (m3 m3)


600 Rn UCM UCM
Net radiation (W m2)

0.35

400
0.30
200

0.25
0

200 0.20
0 24 48 72 0 24 48 72
UTC time (h) UTC time (h)

Figure 7. Comparison of urban canopy model predictions and measurements of (a) roof temperature, (b) ground surface temperatures, (c) net radiation
over the roof and (d) volumetric soil-water content, on 2729 July 2010. This figure is available in colour online at wileyonlinelibrary.com/journal/qj

moisture compared with the usually thin green roof energy transport and hydrological model. It includes
covers. general subfacet heterogeneity such that complex urban
surface configurations can be resolved realistically, the
6. Concluding remarks capability of explicit inclusion of green roofs being an
important application. Although subfacet heterogeneity can
Physically based surface exchange schemes for mass be treated more generally using three-dimensional urban
and energy transport in urban canopies are com- energy balance models (Krayenhoff and Voogt, 2007),
plex due to the wide range of processes controlling the proposed UCM retains the numerical simplicity by
urban landsurfaceatmosphere interactions. Accurate using the one-dimensional big canyon representation
parametrization of the latent heat exchanges is particu- of urban areas and is more appropriate for coupling
larly difficult due to the lack of adequate resolution of urban to large-scale atmospheric models such as WRF. The
hydrological processes. As urban field measurements and new UCM also improves upon current urban models
modelling studies increasingly confirm the importance of through the inclusion of simple but sufficient hydrological
the inclusion of vegetation cover in urban canopy schemes processes, not only for the urban vegetation cover, but
for improving model performance (Grimmond et al., 2011), also for bare soils and porous engineered surfaces. These
the coupling of water/moisture transport schemes from hydrological schemes were developed and fully coupled to
urban surfaces to the energy transport schemes remains energy transport schemes. The skill of the new scheme
a significant modelling challenge. The widely used urban was tested against experimental data collected through
canopy schemes, for example those adopted in the WRF a wireless sensor network deployed over the campus of
model (including both single-layer and multilayer models), Princeton University. Surface temperatures as well as net
are still not very realistic in terms of their physical rep- radiation measurements over roofs were captured very well
resentation of these coupled transport processes, despite because they largely depend on surface parameters that
the continuous improvement in model parametrizations we calibrated for. Testing the model for a period with
and the increasing model complexity in the past decade. very different meteorological conditions also yielded good
The only exception is the vegetated UCM developed by matching of these parameters, underlining the robustness
Lee and Park (2008) and Lee (2011), including a very of the calibrated model. The model performance was
brief description of the hydrological modelling for tall found to be insensitive to changes in weather conditions
trees and a grass surface inside urban canyon while leav- or seasonal variability. Predictions of the volumetric soil-
ing the subsurface transport of water unrepresented in the water content were also in good agreement with field
model. measurements. It highlighted the models capability of
In this article we address these challenges by developing capturing subsurface water transport for urban lawns,
a new urban surface exchange scheme with a coupled although some divergence towards the end of the modelling


c 2012 Royal Meteorological Society Q. J. R. Meteorol. Soc. 139: 16431657 (2013)
Urban Canopy Energy Transport and Hydrology Modelling 1655

(a) 60
aR = 0.1 dark
aR = 0.5 grey
aR = 0.9 white

Roof surface temperature (C)


aR = 0.15 green
40

20

0
0 24 48 72 96 120 144
UTC time (h)

(b) 8
7 Bo, conventional roof
Bo, green roof
6
5
Bowen ratios

4
3
2
1
0
1
2
0 24 48 72 96 120 144
UTC time

Figure 8. Comparison of (a) roof surface temperatures with various surface albedos and a green roof (aR = 0.15), and (b) Bowen ratio with conventional
(aR = 0.3) and green roof; both cases are driven by the atmospheric forcing on 49 May 2010 and with the same canyon properties as used in our
evaluations test cases. This figure is available in colour online at wileyonlinelibrary.com/journal/qj

period suggests that periodic reinitialization of the soil With the improved schemes implemented in this article it
moisture, either through measurements or automatically is also possible to apply the new UCM to many applications
after a storm, might be beneficial. Also note that in this in urban environmental studies with different problem
article we mainly focused on short urban vegetation, i.e. settings. One interesting example, as we illustrated through
lawns and green roofs. In the future development of this the case study, is the assessment of potential mitigation
model, the effect of tall trees on energy balance (shading in strategies of UHI using green or white roofs. Our case
urban canyon) as well as on hydrological processes will be study illustrates that although white roofs are consistently
incorporated. beneficial, green roofs need to be watered regularly to keep
While Grimmond et al. (2010, 2011), by evaluating the their cooling effect active.
performance of a group of 32 urban energy balance models,
found that there was no general improvement of the Acknowledgements
latent heat prediction from knowing more details about
This work is supported by the US National Science
urban vegetation and the simple models even consistently
foundation (NSF) under Grant No. CBET-1058027, by the
performed better, the statistical results in this study show that
Greater Philadelphia Innovation Cluster (GPIC) for Energy-
the proposed scheme improves the model performance in
Efficient Buildings, by the Andlinger Center for Energy
terms of capturing the surface and subsurface characteristics and the Environment at Princeton University, and by the
of urban canopies, in particular, the soil-water content. This Mid-Infrared Technology for Health and the Environment
suggests that the new physically based UCM, by including (MIRTHE) NSF center. We are also grateful to Sensorscope
separate subfacets for urban vegetation and coupling the (http://www.sensorscope.ch/), the manufacturers of the
urban energy balance model with hydrological models, has Sensor Network over Princeton stations, for their assistance,
the potential to better parametrize the evaporation process especially in data management. The authors would also like
and latent heat flux with better capturing of surface and to thank the editor, Professor Mark Baldwin, the associate
subsurface physics, and therefore represents a promising editor handling the manuscript and two anonymous
direction for UCM development. reviewers for their constructive comments.

c 2012 Royal Meteorological Society Q. J. R. Meteorol. Soc. 139: 16431657 (2013)
1656 Z. H. Wang et al.

Appendix A References
Abahri K, Belarbi R, Trabelsi A. 2011. Contribution to analytical and
Adjusting factors for the stomatal resistance numerical study of combined heat and moisture transfer in porous
building materials. Build. Environ. 46: 13541360.
Andersen CT, Foster IDL, Pratt CJ. 1999. The role of urban surfaces
In this article, the parametrization of adjusting factors for (permeable pavements) in regulating drainage and evaporation:
the stomatal resistance in Eq. (26) was originally proposed development of a laboratory simulation experiment. Hydrol. Process.
by Noilhan and Planton (1989) and later extended by Niyogi 13: 597609.
and Raman (1997) for use in atmospheric models. It consists ASHRAE. 2009. 2009 ASHRAE Handbook: Fundamentals, Chapter 26.
American Society of Heating, Refrigerating and Air-Conditioning
of the following equations: Engineers Inc; 22 pp.
Bedient PB, Huber WC. 1992. Hydrology and Floodplain Analysis, 2nd
1+f SD 2 edn. Addison-Wesley; 692 pp.
FSR = withf = 0.55 , (A1)
f + Rs,min /Rs,max SD,lim LAI Bou-Zeid E, Overney J, Rogers BD, Parlange MB. 2009. The effects of
building representation and clustering in large eddy simulations of
1, 2 > 0.75s flows in urban canopies. Boundary-Layer Meteorol. 132(3): 415436.
2 r
F = 0.75s r , r < 2 < 0.75s , (A2) Borselli L, Torri D. 2010. Soil roughness, slope and surface storage
relationship for impervious areas. J. Hydrol. 393: 389400.
0, 2 < r Brutsaert W. 2005. HydrologyAn introduction. Cambridge University
Fe = 1 0.025[e (Ts ) ea ], (A3) Press; 605 pp.
Champollion C, Drobinski P, Haeffelin M, Bock O, Tarniewicz J,
FT = 1 0.0016[25 Ta ] . 2
(A4) Bouin MN, Vautard R. 2009. Water vapour variability induced by
urban/rural surface heterogeneities during convective conditions. Q.
J. R. Meteorol. Soc. 135: 12661276.
In the above equations, Rs,max is the maximum stomatal Chen F, Dudhia J. 2001. Coupling an advanced land surface-hydrology
resistance, and is set as 5000 s m1 following Jacquemin model with the Penn State-NCAR MM5 modeling system. Part
and Noilhan (1990); SD,lim is the radiation limit at which I: Model implementation and sensitivity. Mon. Weather Rev. 129:
photosynthesis is assumed to start (e.g. 30 W m2 for a forest 569585.
Chen F, Mitchell K, Schaake J, Xue YK, Pan HL, Koren V, Duan QY, Ek M,
and 100 W m2 for a crop); 2 is the deep soil-water content Betts A. 1996. Modeling of land surface evaporation by four schemes
at 1 m below the surface; e (Ts ) is the saturated vapour and comparison with FIFE observations. J. Geophys. Res.-Atmos. 101:
pressure at the surface temperature Ts ; ea is the atmospheric 72517268.
vapour pressure; and Ta is the atmospheric temperature Chen F, Kusaka H, Bornstein R, Ching J, Grimmond CSB, Grossman-
Clarke S, Loridan T, Manning KW, Martilli A, Miao S, Sailor D,
in C. Salamanca F, Taha H, Tewari M, Wang XM, Wyszogrodzki A, Zhang C.
2011. The integrated WRF/Urban modeling system: development,
Appendix B evaluation, and applications to urban environmental problems. Int. J.
Climatol. 31: 273288.
Clapp RB, Hornberger GM. 1978. Empirical equations for some soil
Closed-form formulae for aerodynamic resistance hydraulic properties. Water Resour. Res. 14: 601604.
Cobos D, Campbell C, 2007. Correcting Temperature Sensitivity of ECH 2 O
The aerodynamic resistance is computed using Soil Moisture Sensors. Application Note #800-755-2751, Decagon
Devices, Inc.: Pullman; 7 pp.
MoninObukhov similarity theory, where the stability coef- Coceal O, Thomas TG, Castro IP, Belcher SE. 2006. Mean flow and
ficients of Mascart et al. (1995) are adopted: turbulence statistics over groups of urban-like cubical obstacles.
Boundary-Layer Meteorol. 121: 491519.
C Cosby BJ, Hornberger GM, Clapp RB, Ginn TR. 1984. A statistical
RESR = , (B1) exploration of the relationships of soil moisture characteristic to the
Ua a2 (
Z)Fh (
Z
physical properties of soils. Water Resour. Res. 20: 682690.
C Craul PJ. 1985. A description of urban soils and their desired
REScan = , (B2)
(Ua Ucan )a (za zT )Fh (za zT )
2 characteristics. J. Arboriculture 11: 330339.
Cuenca RH, Ek M, Mahrt L. 1996. Impact of soil water property
2 parameterization on atmospheric boundary layer simulation. J.
a2 (z) = , (B3) Geophys. Res. Atmos. 101: 72697277.
[ln(z/z0m )]2

De Bruin HAR, Holtslag AAM. 1982. A simple parameterization
ln(z/z0m ) 1
bRib of the surface fluxes of sensible and latent heat during daytime
if Rib 0 compared with the PenmanMonteith Concept. J. Appl. Meteorol. 21:
Fh (z) = 1+Ch Rib ,
ln(z/z0h ) 1
if Ri > 0
16101621.

1+b Rib b Dupont S, Otte TL, Ching JKS. 2004. Simulation of meteorological fields
within and above urban and rural canopies with a mesoscale model
(B4) (MM5). Boundary-Layer Meteorol. 113: 111158.
Fernando HJS. 2010. Fluid dynamics of urban atmospheres in complex
terrain. Annu. Rev. Fluid Mech. 42: 365389.
= ln(z0m /z0h ), (B5) Flagg DD, Taylor PA. 2011. Sensitivity of mesoscale model urban
boundary layer meteorology to the scale of urban representation.
Ch = 3.2165 + 4.3431 + 0.536 0.0781 ,
2 3
(B6) Atmos. Chem. Phys. 11: 29512972.
= ln(z0m /z0h ), (B7) Grimmond CSB, Oke TR. 2002. Turbulent heat fluxes in urban areas:
Observations and a local-scale urban meteorological parameterization
ph = 0.5802 0.1571 + 0.0327 0.0026 ,
2 3
(B8) scheme (LUMPS). J. Appl. Meteorol. 41: 792810.

Ph Grimmond CSB, Blackett M, Best MJ, Barlow J, Baik JJ, Belcher SE,
Bohnenstengel SI, Calmet I, Chen F, Dandou A, Fortuniak K,
2 ln(z/z0m ) z
Ch = a b Ch , (B9) Gouvea ML, Hamdi R, Hendry M, Kawai T, Kawamoto Y, Kondo H,
ln(z/z0h ) z0h Krayenhoff ES, Lee SH, Loridan T, Martilli A, Masson V, Miao S,
Oleson K, Pigeon G, Porson A, Ryu YH, Salamanca F, Shashua-Bar L,
where is the von Karman constant, C = 0.74, b = 9.4, b Steeneveld GJ, Tombrou M, Voogt J, Young D, Zhang N. 2010. The
= 9.4 and Rib is the bulk Richardson number defined by international urban energy balance models comparison project: first
results from phase 1. J. Appl. Meteorol. Climatol. 49: 12681292.
gz
T Grimmond CSB, Blackett M, Best MJ, Baik JJ, Belcher SE, Beringer J,
Rib (z) = . (B10) Bohnenstengel SI, Calmet I, Chen F, Coutts A, Dandou A, Fortuniak K,
TU 2 (z) Gouvea ML, Hamdi R, Hendry M, Kanda M, Kawai T, Kawamoto Y,


c 2012 Royal Meteorological Society Q. J. R. Meteorol. Soc. 139: 16431657 (2013)
Urban Canopy Energy Transport and Hydrology Modelling 1657

Kondo H, Krayenhoff ES, Lee SH, Loridan T, Martilli A, Masson V, Masson V. 2000. A physically-based scheme for the urban energy budget
Miao S, Oleson K, Ooka R, Pigeon G, Porson A, Ryu YH, Salamanca F, in atmospheric models. Boundary-Layer Meteorol. 94: 357397.
Steeneveld GJ, Tombrou M, Voogt JA, Young DT, Zhang N. 2011. Masson V. 2006. Urban surface modeling and the meso-scale impact of
Initial results from phase 2 of the international urban energy balance cities. Theor. Appl. Climatol. 84: 3545.
model comparison. Int. J. Climatol. 31: 244272. Mitchell VG, Cleugh HA, Grimmond CSB, Xu J. 2008. Linking urban
Hagishima A, Tanimoto J, Narita K. 2005. Intercomparisons of water balance and energy balance models to analyse urban design
experimental convective heat transfer coefficients and mass transfer options. Hydrol. Process. 22: 28912900.
coefficients of urban surfaces. Boundary-Layer Meteorol. 117: 551576. Nakayama T, Fujita T. 2010. Cooling effect of water-holding pavements
Hamdi R, Termonia P, Baguis P. 2010. Effects of urbanization and made of new materials on water and heat budgets in urban areas.
climate change on surface runoff of the Brussels Capital Region: a case Landscape Urban Plan. 96: 5767.
study using an urban soilvegetationatmosphere-transfer model. Niyogi DS, Raman S. 1997. Comparison of four different stomatal
Int. J. Climatol. DOI: 10.1002/joc.2207. resistance schemes using FIFE observations. J. Appl. Meteorol. 36:
Hanks RJ. 1992. Applied soil physics: Soil Water and Temperature 903917.
Applications, 2nd edn. Springer-Verlag; 176 pp. Noilhan J, Planton S. 1989. A simple parameterization of land
Harman IN, Best MJ, Belcher SE. 2004. Radiative exchange in an urban surface processes for meteorological models. Mon. Weather Rev. 117:
street canyon. Boundary-Layer Meteorol. 110: 301316. 536549.
Jacquemin B, Noilhan J. 1990. Sensitivity study and validation of a Oke TR, Johnson GT, Steyn DG, Watson ID. 1991. Simulation of surface
land surface parametrization using the HAPEX-MOBILIHY data set. urban heat islands under ideal conditions at night, Part 2: diagnosis
Boundary-Layer Meteorol. 52: 93134. of causation. Boundary-Layer Meteorol. 56: 339358.
Johnson GT, Oke TR, Lyons TJ, Steyn DG, Watson ID, Voogt JA. 1991. Oleson KW, Bonan GB, Feddema J, Vertenstein M, Grimmond CSB.
Simulation of surface urban heat islands under ideal conditions 2008. An urban parameterization for a global climate model. Part I:
at night, Part I: theory and tests against field data. Boundary-Layer Formulation and evaluation for two cities. J. Appl. Meteorol. Climatol.
Meteorol. 56: 275294. 47: 10381060.
Kanda M, Kawai T, Kanega M, Moriwaki R, Narita K, Hagishima A. Rowley FB, Eckley WA. 1932. Surface coefficients as affected by wind
2005. A simple energy balance model for regular building arrays. direction. ASHRAE Trans. 39: 3346.
Boundary-Layer Meteorol. 116: 423443. Rowley FB, Algren AB, Blackshaw JL. 1930. Surface conductances as
Kondo H, Genchi Y, Kikegawa Y, Ohashi Y, Yoshikado H, Komiyama H. affected by air velocity, temperature and character of surface. ASHRAE
2005. Development of a multi-layer urban canopy model for the Trans. 36: 429446.
analysis of energy consumption in a big city: Structure of the urban Sailor DJ. 2011. A review of methods for estimating anthropogenic heat
canopy model and its basic performance. Boundary-Layer Meteorol. and moisture emissions in the urban environment. Int. J. Climatol.
116: 395421. 31: 189199.
Krayenhoff ES, Voogt JA. 2007. A microscale three-dimensional urban Schlunder EU. 1988. On the Mechanism of the constant drying rate
energy balance model for studying surface temperatures. Boundary- period and its relevance to diffusion controlled catalytic gas phase
Layer Meteorol. 123: 433461. reactions. Chem. Eng. Sci. 43: 26852688.
Kusaka H, Kondo H, Kikegawa Y, Kimura F. 2001. A simple single- Shokri N, Lehmann P, Vontobel P, Or D. 2008. Drying front and
layer urban canopy model for atmospheric models: Comparison with water content dynamics during evaporation from sand delineated by
multi-layer and slab models. Boundary-Layer Meteorol. 101: 329358. neutron radiography. Water Resour. Res. 44: W06418.
Lee SH. 2011. Further development of the vegetated urban canopy Stull RB. 1988. An Introduction to Boundary Layer Meteorology. Kluwer
model including a grass-covered surface parameterization and Academic Publishers; 670 pp.
photosynthesis effects. Boundary-Layer Meteorol. 140: 315342. Van Genuchten MT. 1980. A closed-form equation for predicting the
Lee SH, Park SU. 2008. A vegetated urban canopy model hydraulic conductivity of unsaturated soils. Soil Sci. Soc. Am. J. 44:
for meteorological and environmental modelling. Boundary-Layer 892898.
Meteorol. 126: 73102. Wang ZH. 2010. Geometric effect of radiative heat exchange in concave
Loosvelt L, Pauwels VRN, Cornelis WM, De Lannoy GJM, Verhoest NEC. structure with application to heating of steel I-sections in fire. Int. J.
2011. Impact of soil hydraulic parameter uncertainty on soil moisture Heat Mass Transfer 53: 9971003.
modeling. Water Resour. Res. 47: W03505. Wang ZH, Bou-Zeid E. 2012. A novel approach for the estimation of soil
Loridan T, Grimmond CSB, Grossman-Clarke S, Chen F, Tewari M, ground heat flux. Agric. Forest Meteorol. 154155: 214221.
Manning K, Martilli A, Kusaka H, Best M. 2010. Trade-offs and Wang ZH, Bou-Zeid E, Smith JA. 2011a. A spatially-analytical scheme
responsiveness of the single-layer urban canopy parametrization for surface temperatures and conductive heat fluxes in urban canopy
in WRF: An offline evaluation using the MOSCEM optimization models. Boundary-Layer Meteorol. 138: 171193.
algorithm and field observations. Q. J. R. Meteorol. Soc. 136: 9971019. Wang ZH, Bou-Zeid E, Au SK, Smith JA. 2011b. Analyzing the sensitivity
Martilli A, Clappier A, Rotach MW. 2002. An urban surface exchange of WRFs single-layer urban canopy model to parameter uncertainty
parameterisation for mesoscale models. Boundary-Layer Meteorol. using advanced Monte Carlo simulation. J. Appl. Meteorol. Climatol.
104: 261304. 50: 17951814.
Mascart P, Noilhan J, Giordani H. 1995. A modified parameterization Yamanaka T, Takeda A, Sugita F. 1997. A modified surface-resistance
of fluxprofile relationship in the surface layer using different approach for representing bare-soil evaporation: Wind tunnel
roughness values for heat and momentum. Boundary-Layer Meteorol. experiments under various atmospheric conditions. Water Resour.
72: 331344. Res. 33: 21172128.


c 2012 Royal Meteorological Society Q. J. R. Meteorol. Soc. 139: 16431657 (2013)

You might also like