840 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 53

Effects of Urbanization on the Temperature Inversion Breakup in a Mountain

Valley with Implications for Air Quality

ANGELA M. RENDON, JUAN F. SALAZAR, AND CARLOS A. PALACIO

El Grupo de Ingeniera y Gesti

on Ambiental, Escuela Ambiental, Facultad de Ingeniera, Universidad
de Antioquia, Medelln, Colombia

VOLKMAR WIRTH AND BJORN BROTZ

Institute for Atmospheric Physics, Johannes Gutenberg-University Mainz, Mainz, Germany

(Manuscript received 14 May 2013, in final form 2 January 2014)

ABSTRACT

Many cities located in valleys with limited ventilation experience serious air pollution problems. The
ventilation of an urban valley can be limited not only by orographic barriers, but also by urban heat island
induced circulations and/or the capping effect of temperature inversions. Furthermore, land-use/-cover
changes caused by urbanization alter the dynamics of temperature inversions and urban heat islands, thereby
affecting air quality in an urban valley. By means of idealized numerical simulations, it is shown that in
a mountain valley subject to temperature inversions urbanization can have an important influence on air
quality through effects on the inversion breakup. Depending on the urban area fraction in the simulations, the
breakup time changes, the cross-valley wind system can evolve from a confined to an open system during the
daytime, the slope winds can be reversed by the interplay between the urban heat island and the temperature
inversion, and the breakup pattern can migrate from one dominated by the growth of the convective boundary
layer to one also involving the removal of mass from the valley floor by the upslope winds. The analysis
suggests that the influence of urbanization on the air quality of an urban valley may lead to contrasting and
possibly counterintuitive effects when considering temperature inversions. More urban land does not nec-
essarily imply worse air quality, even when considering that the amount of pollutants emitted grows with
increased urbanization.

1. Introduction and the atmosphere (Arnfield 2003; Grimmond 2007;

Kanda 2007). Therefore, land-use/-cover changes in-
Many cities located in valleys with limited ventilation
duced by urbanization modify the dynamics of local
experience serious air pollution problems of concern for
winds (e.g., Lee and Kim 2008), the urban heat island
public health (Edgerton et al. 1999; Romero et al. 1999;
(UHI) (e.g., Chen et al. 2006), and the local temperature
Panday and Prinn 2009). The transport of pollutants out
inversions, with impacts on air quality and implications
of an urban valley atmosphere can be limited not only
for urban planning (Romero et al. 1999; Janhall et al.
by orographic barriers, but also by urban heat island
2006; Rizwan et al. 2008; Allen et al. 2011).
induced circulations (Haeger-Eugensson and Holmer
Although the term urbanization may refer to different
1999; Savijarvi and Liya 2001) and/or the presence of
processes of land-use/-cover change, here we use it only
temperature inversions near the surface (Janhall et al.
to refer to an expansion of a homogeneous urban area
2006; Yao and Zhong 2009). On a local scale, the char-
(further details are given below). Therefore, as a sim-
acteristics of the urban land have a large influence on the
plification, we are not considering processes such as
heat and momentum transfer between the land surface
population densification or any spatial heterogeneity
within the urban area.
The temperature inversion is a common phenomenon
Corresponding author address: Angela M. Rend

on, GIGA, Escuela
Ambiental, Facultad de Ingenier
a, Universidad de Antioquia
in the lower atmosphere of urban areas and mountain
(UdeA), Calle 70, No. 52-21, Of. 20-414, Medell
n, Colombia. valleys (Oke 1995; Whiteman 2000). Although the temper-
E-mail: angela.rendon@udea.edu.co ature inversion is sometimes defined in terms of absolute

DOI: 10.1175/JAMC-D-13-0165.1

2014 American Meteorological Society

APRIL 2014
N ET AL.
RENDO 841

atmospheric temperature (e.g., Jacobson 2005, p. 54), traffic are exacerbated by temperature inversions and
here we consider that an inversion occurs when the po- strongly controlled by the complex topography of the
tential temperature increases with height above the terrain (Wallace et al. 2010). The numerical simulations
surface (i.e., when the potential temperature gradient by Colette et al. (2003) show that the lifetime of the
is positive; Whiteman 1982). The atmospheric layer inversion layer is different in valleys with different
within which the potential temperature gradient is pos- widths and depths, and that the topographic shading
itive is termed an inversion layer. can delay the BTI. According to the simulations by
The process of destruction of the inversion layer is Anquetin et al. (1998), the BTI also varies with season,
defined as the breakup of the temperature inversion because in summer the more effective solar radiation
(BTI). Over flat, homogeneous terrain, inversions are speeds up the BTI. This seasonal variation may alterna-
destroyed predominantly by the upward growth of a tively be interpreted, in a more general way, as a conse-
convective boundary layer (CBL) from the ground. In quence of increasing the heating of the surface (as it is in
a valley, however, inversions are destroyed not only by summer), which could be a result of land-use/-cover
a CBL that develops over valley surfaces, but also as a changes (such as those caused by urbanization). This is
result of the removal of air from the base of the inversion consistent with the simulations of Bader and McKee
by the upslope flows that develop over heated sidewalls, (1985), which show that decreasing the surface heating
thus inducing subsidence warming over the valley center rate in a valley delays the BTI, although the development
(Whiteman 1982; Zoumakis and Efstathiou 2006). The of the structure of the boundary layer proceeds similarly.
interplay between the growth of a CBL and the removal This latter study also shows that as the valley floor widens,
of air from the base of the inversion leads to three dif- the slope effects become increasingly less important until
ferent patterns of BTI in mountain valleys (Whiteman the valley atmosphere behaves similarly to boundary
and McKee 1982; Whiteman 1982). These BTI patterns layers over flat terrain.
will be described and considered in section 4. Given the dependence of the BTI on the topography
The dynamics of the temperature inversion (forma- of the terrain, the present study is limited in that we did
tion and breakup) is mainly driven by the heating of the not investigate different topographical settings. An
surface, which is largely influenced by urbanization. analysis of this will be given in a forthcoming paper.
These dynamics can exert strong effects on the air However, studying the influence of urbanization on the
quality of urban areas, especially when they are located temperature inversion dynamics through its effects on
in complex terrain such as mountain valleys. A number the surface heating is critical for understanding the in-
of studies confirm the association between the occur- terplay between land-cover change, local meteorology,
rence and persistence of temperature inversions and the and air quality. Cities obviously produce air pollution,
presence of high concentrations of air pollutants over but what has been less well studied is the impact of the
urban areas. For instance, one of the worst national air boundary layer dynamics produced by the surface en-
pollution episodes in the United States took place in ergy balance on this pollution (Masson 2006). In this
the city of Logan in the Cache Valley of Utah and was regard, here we show that in an urban valley subject to
largely caused by the trapping of pollutants owing to a temperature inversions, land-use/-cover changes result-
strong inversion (Malek et al. 2006). Also, heavy air pol- ing from urbanization, which include changes in the size
lution events experienced in Santiago, Chile (Rutllant and of the UHI, have an important influence on air quality
Garreaud 1995), and the Lanzhou urban valley, China through effects on the BTI. To show this, we carried
(Chu et al. 2008), have been associated with low-level through and analyzed idealized simulations with an
temperature inversions. In several Nordic urban sites, EulerianLagrangian numerical model (EULAG) to
winter temperature inversions are a major cause for ex- study the BTI in a mountain valley with varying frac-
ceeding air-quality legislation thresholds for traffic- tions of urban and rural land cover, as well as the
related pollutants. In such situations the BTI exerts transport of a passive tracer released from the urban
a strong influence on air quality through the effects on the surface. Although idealized, our model experiment is
rates of vertical mixing and the consequent distribution of loosely based in a reference urban valley, the Aburr
a
urban aerosols (Janhall et al. 2006; Olofson et al. 2009). valley in the Colombian Andes, where urbanization has
Previous studies have shown that the dynamics and been progressing rapidly and the urban area (Medell
n
effects of the temperature inversion vary depending on and its neighboring cities) has expanded beyond the
several factors, especially those associated with the to- valley floor toward the sidewalls. Our analyses examine
pography of the terrain and the surface energy balance. various effects resulting from the urban land expansion.
For instance, in the city of Hamilton, Ontario, Canada, The remainder of the paper is organized as follows:
the air quality impacts of emissions from industry and the model and the experiment are described in section 2,
842 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 53

results are presented in section 3 and discussed in section valley every 20% between 0% and 100%, and then we
4, and the conclusions are drawn in section 5. compared the results between the six cases. In all cases,
the urban area is homogeneous and grows symmetrically
from the center of the valley toward the sloping side-
2. Model description and experiment design
walls. It is important to bear in mind that a 40% or
a. Model description greater urban area fraction implies that urbanization
spreads to the valley sidewalls (e.g., Fig. 2 in section 3,
The numerical model employed in this study has been
where the thick outline of the orography marks the ur-
thoroughly documented in the literature, mainly in
ban area). All simulations started with a condition of
studies by P. K. Smolarkiewicz and others (Prusa et al.
temperature inversion and were performed during the
2008 and references therein). EULAG is a nonhydrostatic
daytime between 0600 local standard time (LST) and
model for simulating fluid flows across a wide range of
1800 LST.
scales and geophysical flows, whose name refers to its
The urban and rural areas are distinguished only by
capability of solving the underlying anelastic equations
differences in the specified heat flux. However, we tested
within either an Eulerian (flux form) or a Lagrangian
the sensitivity of our results to the inclusion of roughness
(advective form) framework. Different variants of
effects under the assumption that the urban areas are
EULAG have been applied to investigate orographic
rougher than the rural ones, and found no significant
flows (Welch et al. 2001; Piotrowski et al. 2010), flows
differences. In these test runs the drag coefficients were
around urban buildings (Smolarkiewicz et al. 2007),
0.01 and 0.001 for the urban and rural land, respectively.
turbulence (Smolarkiewicz and Prusa 2004), cloud dy-
We assume that the sensible heat flux from urban areas
namics (Andrejczuk et al. 2004), and global atmospheric
is higher than that from rural ones, which agrees with the
circulation (Smolarkiewicz et al. 2001), among other
idea that the urban energy balance is typically charac-
topics generally related to the fluid dynamics of com-
terized by a high Bowen ratio owing to the thermal
pressible or incompressible fluids. EULAG has been
properties and impermeability of urban surfaces like
found to simulate the evolution of a daytime valley wind
concrete or asphalt (Offerle et al. 2006; Nadeau et al.
system, with results qualitatively similar to those of nine
2009). For instance, Offerle et al. (2006) reported values
other mesoscale models [see the intercomparison study
of the Bowen ratio for an urban area ranging from ;0.5
by Schmidli et al. (2011)].
to ;4.5 depending on the fraction of the vegetated
EULAG admits several optional formulations of
surface. The idealized diurnal cycle of sensible heat flux
the equations of motion. Here, we are concerned with
was defined through the function
valley-scale thermal flows, so we use EULAG as a non-
hydrostatic model that solves the anelastic equations of
 
1 2p
motion in terrain-following coordinates, under the Qh (x, t) 5 Q r (x) 1 2 cos t , (1)
2 T
Ogura and Phillips (1962) approximation of the atmo-
sphere. The model employs an explicit predictive equation
where Qh(x, t) is the spacetime varying flux of sensible
for turbulent kinetic energy (TKE hereinafter) and uses
heat (J m22 s21); Qr(x) is a maximum reference value
a local Deardorff-type TKE closure scheme derived
depending on whether the surface at x is urban [Qr(x) 5
for large eddy simulations for which the length scale
300 J m22 s21] or rural [Qr(x) 5 90 J m22 s21]; t is the
is based on the grid spacing (Deardorff 1980). Surface
time since the initial time (0600 LST), ranging from 0 to
drag is accounted for through the usual bulk formula
12 h; and T is the total time of the simulation (T 5 12 h).
with a drag coefficient (CD 5 0.01). Furthermore, we
Note that the sensible heat flux peaks at noon.
modify the EULAG code to introduce a time- and space-
There is no dependence of Qh(x, t) on y (the along-
varying forcing that represents the diurnal cycle of sensible
valley direction), because of the symmetry and orien-
heat flux over the mountain valley (see section 2b).
tation of the valley (topography is described below in
section 2c). We assume that the valley is oriented east
b. Experiment design
west and located at 08 latitude, so that the sun follows
The model experiment performed in this study is a path along the valley, and there is no topographic
based on the general idea that the comparison of paired shading. The Coriolis effect is neglected because of the
simulations with different representations of land cover spatial scale of our study and the location of the valley.
reveals the environmental effects of land-cover change The sensible heat function [Eq. (1)] prescribes a sym-
(Bonan 2008). We performed simulations with different metric sensible heat flux throughout the simulations.
land-cover distributions involving two classesurban This is in good agreement with previous simulations
and rural landvarying the urban area fraction in the of thermally driven valley winds by Serafin and Zardi
APRIL 2014
N ET AL.
RENDO 843

(2010a), where the symmetry of the surface forcing is may be raised as well. Neither chemical reactions nor
substantially maintained during the simulation time (see the radiative effects of pollutants in the atmosphere are
their Fig. 1), although they did not prescribe sensible considered, and we do not include moisture in our sim-
heat but simulated it through a bulk relationship. ulations at this stage, because our main concern is re-
Equation (1) involves the assumption that the shape of lated to thermally driven winds on a dry atmosphere,
the diurnal cycle of sensible heat is similar to that of net and to the BTI, which is mainly driven by heating pro-
solar radiation (see, e.g., Grimmond and Oke 2002; cesses (Whiteman and McKee 1982). This approach is
Matzinger et al. 2003). Values produced by Eq. (1) are similar to that of Anquetin et al. (1998), Colette et al.
on the same order of magnitude as those reported in the (2003), and Serafin and Zardi (2010a). We will leave
literature for different urban and rural areas (e.g., consideration of moist processes and the intraurban
Wesson et al. 2001; Grimmond and Oke 2002; Jung et al. spatial heterogeneity to future work.
2011) and can be considered as being representative of
our reference urban valley (the Aburr
a valley). c. Model setup
To obtain insights into the air quality impacts, we
1) TOPOGRAPHY AND SIMULATION DOMAIN
simulate the transport of a passive tracer released from
the urban area, by assuming that this area acts as a uni- The model experiments are carried out in a three-
form and continuous source, that is, there is no spatial or dimensional domain of 25 600 m in the x direction (across
temporal heterogeneity of emissions. Urban areas can the valley) and 6400 m in the y direction (along the valley),
be considered one of the major sources of air pollution extending from the surface to a height of 6400 m (z di-
as a result of multiple factors such as the emission of rection). Figure 2, described below in section 3, clarifies
gases by the industrial (Rosenzweig et al. 2010) or the this geometry by showing the x and z directions in the
transport sectors (Colvile et al. 2001). Hence, when the cross section of the valley. The extension in the y di-
percentage of urban land cover increases, the amount of rection (normal to the plane shown in Fig. 2) allows us to
pollutants released to the atmosphere is also expected to capture the dynamics of large convective eddies inside
increase. the valley, as well as to reduce the influence of the
In summary, we assume that urbanization inherently boundary effects in the mouths of the valley on the rest
leads to an areal expansion of both the UHI and the of the domain. The depth of the domain (z) permits us
emission area, which is consistent with the idea that, for to represent the evolution of vertical motions and
instance, as the urban land expands so does the number thermal plumes over the valley. The dimensions of the
of cars in the city. Intraurban variations of pollution modeling domain are similar to that of Anquetin et al.
(Wilson et al. 2005) and surface energy fluxes (Offerle (1998), Colette et al. (2003), and Serafin and Zardi
et al. 2006) are still very complex issues that are strongly (2010b), and they are representative of our reference
dependent on the particularities of each city. Therefore, valley.
we regard our assumption of intraurban homogeneity as Following Rampanelli et al. (2004), the idealized to-
the simplest possibility, although other assumptions pography is represented by the function

8
> 1, jx 2 x0 j . sx 1 y x
>
>
>
<1
 
pjx 2 x0 j 2 y x
z(x) 5 h 1 2 cos , y x , jx 2 x0 j , sx 1 y x , (2)
>
> 2 sx
>
>
:
0, jx 2 x0 j , yx

where z(x) is the height of the terrain above the valley boundaries far from the top of the sidewalls. The
floor (m), h is the maximum depth of the valley (h 5 average steepness of the sidewalls is 12.5%, and
1000 m), x0 is the x coordinate of the center of the valley the transitions between them and the plateaus and
(x0 5 12 800 m), sx is the width of the sloping sidewalls the valley floor are smooth in order to prevent abrupt
(sx 5 8000 m), and yx is the half-width of the valley floor changes in wind direction. The vertical grid is set
(y x 5 2000 m). Equation (2) represents a symmetrical with terrain-following coordinates stretched with a
cross section, constant along the y direction. In ad- minimum grid size of 25 m at the surface and a maxi-
dition, two 2800-m-length plateaus are included in mum grid size of 121 m at the top of the simulation
the modeling domain in order to have the lateral domain.
844 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 53

TABLE 1. Summary of the model setup. only minor quantitative changes with respect to a single
cross section in the center of the valley. The plots do not
Parameter Value
show the entire simulation domain but only a part of it
Simulation domain (nx, ny, nz) 512, 128, 128
around the valley, which is the selected region of interest.
Horizontal grid resolution Dx, Dy (m) 50
Vertical grid resolution Dz (m) 25121 We first consider the BTI as described by consecutive
Time step Dt (s) 2 vertical profiles of u in the center of the valley (section
Initial simulation time (LST) 0600 3a), then we consider several features of the cross-valley
Final simulation time (LST) 1800 wind system and its relationship with BTI (section 3b),
Time of simulation (h) 12
and finally we give insights into the impacts of the cir-
Initial BruntV ala frequency (s21)
ais 0.019
culation on the transport of pollutants released from the
urban surface (section 3c).
2) INITIAL AND BOUNDARY CONDITIONS
a. Temperature inversion breakup
The simulations are initialized with a temperature
inversion profile that is constant over the whole valley The time variation of vertical u profiles in the center of
with a BruntVaisala frequency of N2 5 0.019 s21, corre- the valley for different percentages of urban land cover
sponding to a vertical potential temperature gradient of is depicted in Fig. 1, where the inversion layer is evident.
11 K km21 in the lowest 800 m above the valley floor, This layer is destroyed faster (i.e., the BTI is acceler-
which represents a strong temperature inversion. Above ated) as the urban area fraction increases, which is in
the top of the inversion we assume neutral stratification agreement with the simulations by Bader and McKee
(u/z 5 0 for z . 800 m). We assume that the atmo- (1985), Anquetin et al. (1998), and Colette et al. (2003),
sphere is at rest at the beginning of the simulations. showing that decreasing the surface heating rate in a
These initial conditions are similar to those used in valley delays the BTI. Note that the inversion layer is
previous studies of the temperature inversion in ideal- finally destroyed at a time when the height of the CBL
ized mountain valleys by Anquetin et al. (1998) and grows to the height of the top of the inversion layer, thus
Colette et al. (2003). producing a constant potential temperature atmosphere
Since we are concerned with the local valley winds, we in the valley (see, e.g., Fig. 2 of Whiteman 1982).
neglect the influence of synoptic winds, under the as- Interestingly, there are only small differences be-
sumption of weak synoptic conditions. To represent this, tween the sequences of u profiles corresponding to the
all the lateral boundary conditions are defined as peri- highest percentages (80% and 100%) of urban land
odic. The bottom and the top of the domain are treated cover (Figs. 1e,f), thus suggesting that there is an upper
as impermeable boundaries. A Rayleigh damping layer percentage of urban land above which the BTI is not
of 2000 m is included below the top of the domain, to further accelerated by augmenting the urban area frac-
avoid unrealistic reflection of upward-propagating in- tion. We attribute this to the location of the ruralurban
ternal gravity waves aloft (Warner 2011). The model fringe relative to the altitude of the top of the inversion
setup is summarized in Table 1. layer: for the highest urban area fractions the former is
above the latter, so in both cases all of the land below the
top of the inversion layer is urban and, therefore, the
3. Temperature inversion breakup and
heating of the surface is the same. Furthermore, any
wind fields in the valley
additional heat from the urban land above the inversion
In this section we present the simulations of the BTI top is distributed over a very large volume and, there-
process in the valley (see section 2 for details) under fore, it is hardly noticeable. We interpret this as a satu-
different urban and rural land distributions: recall that ration effect in our model setup, which occurs when the
the urban area fraction in the valley varies from 0% to heat released from the surface is distributed either in
100% every 20%. Our results show the time variation of the atmospheric column above the inversion layer or in
the potential temperature u profile in the center of the the entire atmospheric column after the BTI. This effect
valley, the horizontal u and vertical w wind fields, and cannot be generalized to all situations because it may
the resolved TKE in the valley cross section, during the change if, for instance, the conditions above the inver-
daytime from 0600 to 1800 LST. Our analyses are fo- sion were not neutral.
cused on the dynamics of the atmosphere in the valley For the lowest percentages of urban land (#20%;
cross section (the x direction), so all of the results are Figs. 1a,b), although the depth of the inversion layer
averaged along the valley (the y direction). We numer- decreases throughout the daytime, a temperature in-
ically verify that this averaging procedure does not af- version remains until the end of the simulations. This
fect the qualitative behavior of our results; there are is in agreement with observations of the temperature
APRIL 2014
N ET AL.
RENDO 845

FIG. 1. Vertical u profiles taken every hour from 0600 to 1800 LST in the center of the valley. Each panel corresponds to a different urban
area fraction (ranging from 0% to 100%) and shows the evolution in time from the initial inversion profile on the left to the final profile on
the right. The arrow indicates the profile corresponding to 1200 LST, and the horizontal dotted line indicates the altitude of the top of the
valley.

inversion in mountain valleys subject to weak thermal The rural valley produces relatively weak upslope winds
forcing, where a temperature inversion can persist prevailing during the daytime with a maximum speed
throughout the day and can even remain in the valley for around 0.5 m s21. These upslope flows are compensated
several days (Savov et al. 2002). by horizontal winds below the top of the inversion layer,
thus forming a closed circulation within the CBL. The
b. Wind fields
persistence of upslope winds during the daytime is the
The BTI is intrinsically related to the dynamics of typical condition in mountain valleys (Whiteman 2000,
thermal cross-valley winds. The components of the p. 187). These results are an example of a confined cir-
mean wind field in the valley cross section for different culation system within a valley (e.g., Li et al. 2009), which
urban area fractions are depicted every 3 h in Fig. 2 can be caused by a temperature inversion (Yao and Zhong
(vertical wind component w) and Fig. 3 (horizontal wind 2009). Closed circulations within a CBL have also been
component u). These wind fields are characterized by observed in regions with steep terrain (Reuten et al. 2005).
various features that are described below. The atmosphere in the urban valleys is subject to
In all simulations (rural and urban valleys), the flow stronger vertical mixing than that of the rural valley, as
field is initially symmetric owing to the symmetry of the a result of the higher heating of the urban surface. This
topography and of the initial conditions. Symmetry is produces differences in the slope wind patterns. In the
substantially maintained on the slopes throughout the 40% urbanized valley, downslope winds appear close to
daytime but not above the valley floor because of the the ruralurban fringe, as distinct from the prevailing
development of randomly spaced thermals. The develop- upslope winds in the rural and more urbanized valleys
ment of these thermals is in good agreement with the (Figs. 2di and 3di). This difference can be explained
simulations of thermally driven winds in a wide valley by by effects related to both the urban heat island and the
Serafin and Zardi (2010a). temperature inversion. The urban heat island in the core
Although the initial atmospheric conditions and the of the 40% urbanized valley, resulting from the differ-
topography are the same in all valleys, the slope winds ential heating of the surface (urban or rural), creates
evolve differently depending on the urban area fraction. buoyant thermals that induce air convergence from the
846 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 53

FIG. 2. Consecutive longitudinal mean fields of the cross-valley w in the (left) rural, (center) 40% urbanized, and (right) 80% urbanized
valleys. Isentropes are shown every 1 K. The thick (thin) outline of the orography marks the urban (rural) area.

surrounding rural areas. It is well known that the dif- urban area is close to the mountaintop, while for 20% or
ferential heating of the surface promotes the formation 40% of urban land this boundary is close to or within the
of rising thermals (Jacobson 2005, p. 22). Furthermore, valley floor. When the sidewalls are highly urbanized,
thermal convection in the core of the valley, along with the heating of the surface enhances thermal convection
the presence of the inversion layer that restricts the and upslope flows, in turn strengthening the transport
vertical free convection, causes downslope compensating of air from the bottom of the valley to the atmosphere
winds below the inversion layer and close to the rural above the inversion layer. As a consequence, the com-
urban fringe. The heat islands effect on creating down- pensating flow responsible for the downslope winds in
slope winds toward the city near the surface was also found the 20% and 40% urbanized valleys does not appear in
in simulations of local winds in a valley city by Savijarvi the 80% urbanized valley.
and Liya (2001). As will be shown in section 4, downslope Upslope winds detach from the ridgetops and create
winds also appear in the 20% urbanized valley. thermal plumes in the rural and urban valleys. The in-
In contrast to the downslope winds found in the 20% tensity of the plumes is nearly the same at 0900 LST in
and 40% urbanized valleys, upslope winds prevail all cases (Figs. 2ac), but evident differences appear at
throughout the daytime in the 80% urbanized valley. noon: while the rural and the 40% urbanized valleys
We attribute this difference to effects related to the produce upward motions in the mountaintops with
extension of urban land along the sidewalls of the valley. maximum speeds of around 0.3 m s21 (Figs. 2d,e), in the
Note that for 80% of urban land the outer edge of the more urbanized valley thermal plumes reach maximum
APRIL 2014
N ET AL.
RENDO 847

FIG. 3. As in Fig. 2, but for u.

updraft speeds of about 1.0 m s21 (Fig. 2f). The differ- of about 2.0 m s21, in a layer approximately 700 m above
ence between the more urbanized valley and the other the mountaintop level (i.e., z ;1700 m). These winds
two valleys suggests that the extra heat produced from advect warm air and thermal energy from the plumes
the extended urban land is resulting in more energetic toward the valley center, in turn producing air sub-
plumes. On the other hand, the similarity between the sidence. This compensating flow is similar to that found
rural and the 40% urbanized valleys can be related to by Serafin and Zardi (2010a), and it is in agreement with
the downslope winds appearing in the second case. The the fact that the ascending motions related to upslope
higher heating of the surface in the 40% urbanized flows and thermal plumes must be compensated by de-
valley, in comparison to the rural one, does not result in scending flows elsewhere, because of mass continuity.
more energetic thermal plumes because the extra energy However, in contrast to that study, we also found that air
is expended in the circulation involving the downslope subsidence does not reach the valley core since the de-
winds, as a result of the interplay between the urban heat scending winds deviate toward the mountaintops as
island and the temperature inversion. Moreover, such a consequence of the inversion layer, which restricts the
extra energy results in more and larger plumes within downward motions. Hence, we found horizontal winds
the CBL of the 40% urbanized valley than in the rural diverging from the valley center just above the inversion
one. layer, so that a circulation cell is evident above such
Thermal plumes detaching from the mountaintops are a layer (Figs. 3df). The formation of this circulation cell
linked to compensating horizontal winds blowing to- results from the effect of the inversion layer in decou-
ward the center of the valley with maximum intensities pling the wind field above and below it, as the temperature
848 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 53

FIG. 4. As in Fig. 2, but for resolved TKE.

inversion restricts both the growth of the CBL and and maximum vertical wind speeds of around 1 m s21,
subsidence from the atmosphere aloft. A similar decou- while the temperature inversion persists. In the 80%
pling effect was observed over the urban area of Milan, urbanized valley (Fig. 2, right), thermals reach maxi-
Italy (see Fig. 2 of Argentini et al. 1999). mum vertical wind speeds around 1 m s21 and a vertical
Figure 2 also shows the development of a CBL at the extent of 700 m before the BTI (at 1200 LST), and the
floor of all of the valleys, exhibiting a structure of con- development of the CBL at the valley floor continually
vective circulations composed by updrafts (thermals) elevates the bottom of the inversion layer until the
and downdrafts. This structure agrees well with the well- breakup process is completed (at 1500 LST).
known structure of the CBL arising when turbulence The mean field of the resolved TKE in the cross sec-
generated by buoyancy due to upward heat flux from the tion of the valleys is depicted in Fig. 4. This figure con-
surface dominates relative to turbulence generated by firms that the flow field in the valley floor is dominated
mean shear (Schmidt and Schumann 1989). Since the by thermal turbulence driving the growth of the CBL,
CBL is capped by the inversion layer, its growth is di- while in the slopes it is dominated by the slope flow. This
rectly linked to the BTI. In the rural valley (Fig. 2, left) is in agreement with simulations showing that turbu-
thermals never exceed a vertical extent of 500 m, while lence is better developed above a valley floor than over
exhibiting maximum vertical wind speeds not greater the slopes (Serafin and Zardi 2010a), and also confirms
than 0.4 m s21. In contrast, thermals in the 40% urban- that turbulence is enhanced over the urban areas. The
ized valley (Fig. 2, center) reach a vertical extent be- TKE above the floor of the rural valley peaks at around
tween 700 m (at 1200 LST) and 1500 m (at 1500 LST), 0.3 m2 s22, while in the urban valleys such peak value is
APRIL 2014
N ET AL.
RENDO 849

more than tripled (;1.0 m2 s22). The weak turbulence in surface flux is given by 17 kg21 h21 km22. At each time
the rural valley can be explained by considering the in- step, a corresponding amount of tracer is added to the
terplay of two factors: the low heating of the rural sur- lowest grid box. The resulting mixing ratio x is then
face that diminishes the available energy to promote advected by the flow according to Dx/Dt 5 0, where D/Dt
buoyancy and thermal convection, and the persistent denotes the material derivative. The chosen surface flux
flow divergence at the slopes, which continually removes is based on measurements of annual carbon emissions
the TKE produced on the surface. from an urban area of about 360 km2 in Colombias
Turbulence is more efficiently generated by buoyancy Aburr
a river valley (Toro et al. 2001). Note that the
above the urban areas, so it is better developed over the amount of pollutant released to the atmosphere grows
urban valleys that we studied than over the rural one. linearly both with time and with the size of the urban
This is in good agreement with the well-known fact that area. Note also that the rural valley is not included in
cities increase the production of thermal turbulence Fig. 5 because the only source of pollution considered is
compared to their surrounding rural areas, as a result of the urban area.
the increase in the surface heat flux and changes in sta- Figure 5 confirms that the temperature inversion
bility accompanying the urban heat island (Oke 1995; restricts the transport of pollutants out of the valley
Baklanov 2002). Better development of turbulence can atmosphere, as previously reported in a number of ex-
enhance the BTI because of the effect of the CBL on the perimental and modeling studies (e.g., Anquetin et al.
bottom of the inversion layer. 1999; Savij arvi and Liya 2001; Berge et al. 2002; Reuten
A comparison between Figs. 4h and 4i reveals an ef- et al. 2005). While a temperature inversion persists, the
fect of temperature inversion on turbulence within the mixing ratio of pollutants in the valley atmosphere
valley atmosphere. Turbulence remains substantially reaches higher values than those after the BTI. For in-
confined within the core of the less urbanized valley, stance, in the 80% urbanized valley, the mixing ratio
while in the more urbanized example the TKE gener- reaches higher values during the morning (Figs. 5b,d)
ated at the surface is continually removed by free con- than those in the afternoon when the inversion layer has
vection after the BTI. already been destroyed (Fig. 5f), and therefore the
Figures 24 also show that the urban area fraction pollutant is distributed over a larger volume than that of
influences whether the diurnal valley circulation is a the stable core (below the inversion top). This is anal-
confined or an open system, that is, whether the ogous to the saturation effect that was discussed in
circulation in the valley atmosphere is decoupled (con- section 3a when considering that the heat released from
fined system) or coupled (open system) to that in the the surface tends to be distributed in the entire atmo-
overlying atmosphere. For instance, Fig. 3 shows that spheric column after the BTI. Note that in the 40% ur-
for low extents of urban land (Fig. 3, center), the valley banized valley a temperature inversion persists throughout
circulation is confined to the valley during almost the the daytime so the pollutant remains substantially trapped
entire daytime, while for high extents of urban land within the valley atmosphere (Fig. 5, left). This is consis-
(Fig. 3, right) there is a transition from a confined to an tent with the aforementioned difference between a con-
open system during the afternoon (Figs. 3fl). Decou- fined and an open circulation in the valley, depending on
pling is caused by the effect of the inversion layer in the presence of the inversion layer (section 3b).
restricting vertical motions (see, e.g., Kalthoff et al. The maximum mixing ratios are found in the slopes
1998; Zhong et al. 2004). The transition from a confined of the 80% urbanized valley during the morning, before
to an open system in a valley atmosphere is also ex- the BTI (Fig. 5b). The occurrence of such extreme
pected to be dependent on some topographic features values over the slopes can be related to three factors.
(Li et al. 2009); however, as mentioned in the in- First, as opposed to the less urbanized valley (Fig. 5, left)
troduction, in this study we did not consider different in the more urbanized one (Fig. 5, right), the sidewalls
topographical settings. are mostly covered by urban land that releases pollut-
ants to the atmosphere. Second, upslope winds advect
c. Transport of pollutants
polluted air from the valley floor to the sidewalls,
To consider the air quality implications, we simulate and the presence of the inversion layer restricting the
the transport of a passive tracer continuously emitted transport of mass out of the valley produces an accu-
from the urban area, under the influence of the BTI mulation of this polluted air over the slopes. Third, the
(further details are given in section 2b). Figure 5 shows atmospheric column below the inversion layer is shorter
the mixing ratio of the tracer for different urban area along the sidewalls than on the valley floor, so over the
fractions (40% and 80%). We assumed that the tracer is slopes the pollutant is mixed with relatively less air.
emitted only from the urban area, and in these areas the Numerical simulations by Lehner and Gohm (2010; see
850 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 53

FIG. 5. Passive tracer mixing ratio (mg kg21) in the (left) 40% urbanized and (right) 80% urbanized valleys (see text for
details). Isentropes are shown every 1 K. The thick (thin) outline of the orography marks the urban (rural) area.

their Fig. 3) show a similar process of accumulation of (darker areas in Fig. 5c). This may be an example of a
polluted air over the slope of a valley as a consequence smog trap (i.e., a situation in which air pollutants tend
of the upslope flow interacting with an inversion layer. to accumulate), which is likely to occur in a valley city as
Finally, a comparison between Figs. 5c and 5d shows a result of the interplay between the slope flow, a tem-
an effect of the slope wind patterns on the transport of perature inversion, and an urban heat island, according
the passive tracer. In the 40% urbanized valley a large to Savij
arvi and Liya (2001). However, here it is worth
amount of the pollutant is trapped near the ruralurban noting that the same mechanisms causing the smog trap
fringe (Fig. 5c), in the region where the downslope winds do not appear in our simulations when the sidewalls are
associated with a closed circulation below the inversion highly urbanized. In the 80% urbanized valley (Fig. 5d),
layer appear (recall Fig. 3e). This closed circulation can most of the tracer is removed from the valley floor and
trap air pollutants within a relatively small volume distributed along the slopes by the prevailing upslope
APRIL 2014
N ET AL.
RENDO 851

FIG. 6. Heights of the bottom (dotted line) and top (solid line) of the inversion layer in the center of the valley for different urban area
fractions (ranging from 0% to 100%).

winds (recall Fig. 3f), even before the complete de- the BTI patterns proposed by Whiteman (1982) and
struction of the inversion layer. Whiteman and McKee (1982). They distinguish three
different patterns depending on the relative magnitude
of two processes whose interplay causes the breakup: (i)
4. Discussion
the rise of the bottom of the inversion layer resulting
To better understand the likely effects of urbanization from the growth of the CBL and (ii) the descent of the
on the BTI in a valley and their implications on air top of the inversion layer caused by the removal of air
quality, in this section we conceptually integrate the from the bottom of the valley by the upslope winds, and
results presented above by considering changes in the BTI the air subsidence from the upper atmosphere. In pat-
pattern as a result of changes in the urban area fraction tern 1, the BTI is dominated by the rise of the bottom of
(section 4a), and we discuss the possible air quality impli- the inversion layer, while in pattern 2 the BTI is mainly
cations of such alterations (section 4b). caused by the descent of the top of the inversion layer. In
Before continuing, it is important to note two things. pattern 3, the BTI is caused by the combined effect of
First, our simulations do not include different topog- both the ascent of the bottom of the inversion layer and
raphical settings, although the influence of the topog- the descent of its top.
raphy of the terrain on the BTI has been well recognized We find that the BTI pattern is sensitive to the per-
(see references in the introduction section). Second, our centage of urban land. Figure 6 shows the temporal
model runs use one uniform value of roughness length evolution of the heights of the bottom and top of the
for the entire model domain. Obviously, in a more re- inversion layer obtained from the potential temperature
alistic model setup one would expect the roughness profiles in the center of the valley showed in Fig. 1. In all
length to increase as one passes from the rural to the cases the growth of the CBL continuously reduces the
urban part of the domain. Correspondingly, we did some depth of the inversion layer, but it is not always com-
additional model runs where we accounted for increased pletely destroyed. For low percentages of urban land
surface friction over the urban area (no figure shown). It (#40%; Figs. 6ac) the height of the top of the inversion
turned out that the sensitivity of our results (such as layer does not substantially change, and the reduction of
those shown in Figs. 2 and 3) to the treatment of the drag the inversion layer is dominated by the growth of the
coefficient is very small. This is consistent with a recent CBL, so this can be described as pattern 1. In contrast,
scaling analysis by J. Schmidli and R. Rotunno (2013, for high percentages of urban land ($60%; Figs. 6df),
personal communication) indicating that surface friction the breakup follows pattern 3, meaning that the inver-
plays a secondary role in thermally driven valley wind sion layer is reduced not only by the rise of its bottom,
systems. but also by the descent of its top.
BTI pattern 2 was not found in any of the studied
a. Breakup patterns
cases; that is, the descent of the top of the inversion layer
The effects of urbanization on the BTI in a valley may does not dominate the BTI in any case. This may be
be summarized and understood within the framework of explained by two factors. First, the presence of the urban
852 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 53

FIG. 7. Idealized sketch of the evolution in time of the BTI process and the mean cross-valley circulation depicting the inversion layer (dashed
lines), the main wind flows (black arrows), and the resolved TKE field (shading), for the (left) rural, (center) 40% urbanized, and (right) 80%
urbanized valleys. Isentropes (solid lines) are shown every 1 K. The thick (thin) outline of the orography marks the urban (rural) area.

area in all valleys but the rural one increases the heating of urban land, based on the results presented in section 3
of the valley floor, thus enhancing the growth of a CBL, and on our understanding of the BTI process. This
which is a process that does not substantially occur in scheme includes a sketch of the inversion layer (dashed
pattern 2. Second, the occurrence of pattern 2 implies lines), the isentropes (solid lines), the main wind flows
not only that the development of a CBL over the valley (black arrows), and the resolved TKE (shading). Figures
floor is weak, but also that the upslope winds are strong 2 and 3 do not always show the updraft region over
enough to substantially remove air from the base of the the mountaintops as sketched in Fig. 7, simply because
inversion. In our simulations, the upslope winds are not sometimes the associated thermals lie outside the plot-
that strong in the rural valley due to the low heating of ted domain. Similarly, the convergence zone above the
the surface, and the development of the CBL is not that mountaintops is not completely shown in Fig. 3 but is
weak in the urban valleys because of the high heating of present in our results and sketched in Fig. 7. It is worth
the urban area. noting that the dashed lines bounding the inversion
The differences between the breakup pattern for low layer have been sketched so as not to exactly match the
and high percentages of urban land can be explained by isentropes, but only with the purpose of improving the
changes in the dynamics of the wind fields associated visualization.
with the urban area fraction. To clarify and discuss In the rural valley the height of the top of the inversion
such differences, we introduce Fig. 7, which shows a sche- layer remains approximately constant at ;800 m (above
matic representation of the BTI for different percentages the valley floor), while that of the bottom rises from
APRIL 2014
N ET AL.
RENDO 853

FIG. 8. Consecutive longitudinal mean fields of the (left) cross-valley w and (center) u wind speed components, and (right) the passive
tracer mixing ratio (mg kg21) in the 20% urbanized valley. Isentropes are shown every 1 K. The thick (thin) outline of the orography marks
the urban (rural) area.

;200 m (at 0900 LST) to ;450 m (at 1800 LST), as a previously. In both valleys, the interplay between the
result of the buoyant production of turbulence at the urban heat island and the presence of the inversion layer
surface and the consequent development of a CBL. produces downslope winds, thus restricting the removal
Pattern 1 implies that the removal of air from the valley of air from the valley floor. Recall from section 3 that the
floor, as well as the subsidence above the inversion layer, circulation associated with these downslope winds leads
are weak. The airmass transport out of the rural valley to an accumulation of air pollutants below the inversion
atmosphere is weak because the low heating of the layer (smog trap). Moreover, weak upslope flows along
surface produces weak upslope flows, and the persistent the upper half of the slopes (Figs. 2e,h and 3e,h; Figs. 8a,
inversion layer restricts vertical motions. The effect of b,d,e) are linked to weak air subsidence from above the
the inversion layer in decoupling the circulation above inversion layer, since the ascending motions related to
and below it is evident. The lines (isentropes) limiting the former are compensated by the descending motions
the inversion layer in the cross section of the valley are related to the latter. Similarly to the case in the rural
approximately horizontal, as proposed in the conceptual valley, in the low urbanized valleys a temperature in-
model of Whiteman (1982) and Whiteman and McKee version remains throughout the daytime. The presence
(1982). At the end of the simulation, an inversion layer is of a shallow inversion layer between 1500 and 1800 LST
still present and calm winds prevail. in the 20% and 40% urbanized valleys is evidenced in
In low urbanized valleys (20% and 40% percentages Figs. 5e,g and 8f,i, where the tracer remains capped by
of urban land) the BTI also follows pattern 1, but the such a layer.
mechanisms driving this pattern differ from those in the In contrast to the rural valley, in low urbanized valleys
rural valley. Figure 8 summarizes the results obtained turbulence is more efficiently produced over the urban
for the 20% urban area fraction, which were not shown areas, and the associated growth of the CBL substantially
854 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 53

reduces the depth of the inversion layer. The surface the more urban land, the worse the air quality, even
heterogeneity in land cover (urban versus rural land) when considering that the amount of pollutants emitted
may lead to nonuniform elevation of the bottom of the grows with the size of the urban area. This is because of
inversion layer caused by the CBL, so that the isentropes the strong effects that a temperature inversion can exert
that limit the inversion layer may differ from horizontal on the air quality of an urban valley, particularly through
ones, as suggested in Fig. 7h. More significant variations effects on the exchange of air between the atmosphere
of these isentropes from horizontal ones may arise when above and below the inversion layer. At this point it is
considering a southnorth-oriented valley where the important to remember that we have defined urbaniza-
symmetry of the surface heating is lost, as a result of tion as the expansion of a homogeneous urban area with
effects such as those of topographic shading. We are respect to both heating and emissions. Therefore, the
currently working on including topographic shading in situation described here could become even more com-
our analyses and plan to report on it in a subsequent plex if the intraurban variations of emissions were con-
study. sidered, but we did not address this issue in the present
In highly urbanized valleys ($60% percentages of study.
urban land) the BTI follows pattern 3, which implies that A possibly counterintuitive effect of urbanization on
both the growth of the CBL and the removal of air from the air quality of a valley city is apparent when com-
the valley floor exert significant effects on the breakup. paring both columns in Fig. 5. Although peak mixing
Consequently, the top of the inversion layer descends ratios occur for the higher percentages of urban land
while the bottom ascends. As a result of the high urban during the temperature inversion (Fig. 5d), higher urban
area fraction, most of the valley surface, including area fractions do not necessarily imply higher mixing
a significant part of the sidewalls, is highly heated, thus ratios during the whole day (cf. Figs. 5g and 5h), al-
giving rise to an efficient buoyant production of turbu- though the expansion of the urban area implies an in-
lence on the valley floor, and inducing stronger upslope crease in the emission of pollutants. This can be explained
winds as compared to those in low urbanized valleys. by the effects of the temperature inversion on the venti-
Hence, a CBL develops above the valley surface, while lation of the valley, which in turn is related to the urban
air is continually removed from the valley floor by the area fraction. Under the assumption of homogeneity of
upslope winds, in turn enhancing air subsidence from the urban area, a temperature inversion may persist
the upper atmosphere. The combined action of these during the whole day in a less urbanized valley, so that
processes leads to the complete destruction of the in- air pollutants may remain trapped for longer periods of
version layer between 1200 and 1500 LST. After that, time than in a more urbanized valley, owing to the better
most of the pollutant is transported out of the valley ventilation of the latter after the destruction of the in-
atmosphere and may be diluted at larger scale. Further version layer. It would be interesting to examine this
implications on air quality are considered in the fol- idea with empirical evidence from historical data of air
lowing section. quality covering different percentages of urban land in
Our results regarding the dependence of the BTI a mountain valley. We could not find such data, so we
pattern on the percentage of urban land are in reason- leave this as an open question.
able agreement with the study by Whiteman et al. (2004) Taha et al. (2010) performed a model study in three
of valleys exhibiting different BTI patterns as a conse- U.S. cities and found that UHI mitigation measures as-
quence of differences in factors affecting the sensible sociated with lowering surface temperatures can lead to
heat flux from the surface. These authors compared the reductions in air pollution. Our findings are in striking
BTI pattern in two valleys, one with moist conditions contrast to this study, since, in our simulations, a weaker
and the other with dry conditions, and they found that urban heat island associated with lower percentages of
the breakup in the former follows pattern 2 while in the urban land or lower heating of the surface does not
latter it follows pattern 3. They attributed this difference necessarily result in lower air pollution throughout the
in the BTI pattern to the lower or higher heating of the daytime. In the 40% urbanized valley we found an urban
surface in the moist or dry valley, respectively. heat island effect concurrent with a temperature in-
version and low air quality, while the absence of a tem-
b. Air quality implications
perature inversion after the BTI in the 80% urbanized
The influence of urbanization on the air quality of an valley is concurrent with better air quality (recall Fig. 5).
urban valley may exhibit contrasting effects when con- This is due to the effects of the temperature inversion in
sidering a temperature inversion. This influence does restricting the valley ventilation and the consequent
not appear to be as straightforward as it seems, meaning trapping of air pollutants, which are not considered in the
that keeping all properties of the urban land constant, simulations by Taha et al. (2010). The authors acknowledge
APRIL 2014
N ET AL.
RENDO 855

that the extent to which urban areas can effectively We want to stress that in this paper we are not sug-
improve local air quality through UHI mitigation de- gesting that urbanization is good for air quality, since
pends on numerous factors, among which one could land-use/-cover change (e.g., urbanization) has poten-
consider a temperature inversion. Furthermore, we study tially detrimental, or at least uncertain, environmental
the transport of a passive tracer, without taking into ac- effects at multiple spatial and temporal scales (Foley
count any chemical reaction or the possible effects of et al. 2005). Our concern is to investigate the local effects
pollution on radiation (see e.g., Fan and Sailor 2005). of urbanization at the valley scale and, particularly,
However, we still consider that our findings with respect those related to the temperature inversion and its in-
to the interplay between the transport of pollutants re- fluence on the transport of pollutants out of an urban
leased to the atmosphere from a valley city, and the valley. Further research on this topic is needed to pro-
presence and eventual destruction of an inversion layer vide better information for decision-making processes
are plausible, and that they are in reasonable agreement related to urban planning, which is of particular impor-
with previous studies about the impacts of the temper- tance in several medium-sized and megacities in Latin
ature inversion on air quality over urban and/or complex America (Romero et al. 1999), many of which are located
terrain, as discussed above. in the mountain terrain of the Andes (e.g., Medell
n in the
In sections 3 and 4a we described a circulation below Aburr
a river valley in Colombia).
the inversion layer that can lead to a smog trap situation
associated with the appearance of downslope winds. We
5. Summary and conclusions
showed that such a downslope flow does not occur in our
simulations when the sidewalls are highly urbanized, We have shown through idealized simulations that in
that is, for high percentages of urban land ($60%). an urban valley subject to temperature inversions, land-
However, the interplay between the upslope flows, the use/-cover changes resulting from urbanization, which
temperature inversion, and the emissions from urban include changes in the size of the UHI, can have an
areas over highly urbanized valleys may lead to an ac- important influence on air quality through effects on the
cumulation of air pollutants over the sidewalls and be- BTI. Despite the completely idealized framework
low the inversion layer, which might be considered as adopted, the present study provides evidence about
another type of smog trap. While a temperature inver- a series of physical processes that are likely to occur in
sion remains, prevailing upslope flows tend to accumu- nature and agrees well with previous studies. In sum-
late air pollutants over the slopes. This is why peak mixing mary, the main effects of urbanization on the BTI and
ratios are found over the slopes in the highly urbanized the accompanying air quality implications are related to
valleys and before the complete destruction of the in- (i) changes in the breakup time, (ii) alterations of the
version layer (recall Fig. 5). This may be relevant to thermal cross-valley winds and turbulence dynamics,
understanding the features of the spatial variability of (iii) modifications of the breakup pattern, and (iv) im-
air quality over highly urbanized valleys along the di- pacts on the ventilation of the urban valley.
urnal cycle. The percentage of urban land affects the breakup
An important aspect of our model experiment is that time, that is, the time that it takes for the inversion layer
as the urban land expands beyond the valley floor, the to be completely destroyed. We find that the larger the
sloping sidewalls are increasingly heated. Heating the urban area fraction, the faster the BTI. However, there
sidewalls will generally enhance the BTI and the venti- are no noticeable differences between the breakup times
lation of the valley through strengthening of the upslope for high-percentage urban land cases ($80% in our
winds. A comparison between the valleys with 40% and simulations), thus suggesting that there is an upper limit
20% urban area fractions (left panels of Fig. 5 and above which the BTI is not further accelerated by in-
right panels of Fig. 8) shows that the air quality is worse creasing the urban area fraction. This seems to be at-
in the former valley throughout the daytime. This re- tributable to the location of the ruralurban fringe
sult, along with our previous observation that higher relative to the altitude of the top of the inversion layer,
urban area fractions do not imply worse air quality since it determines whether or not the atmosphere be-
during the whole day (recall the comparison between low the inversion layer is all urban and hence subject to
the 40% and 80% urbanized valleys, based on Fig. 5), a similar heating. This conclusion is only valid under
suggest that the extent to which urbanization adversely conditions where the vertical transport of heat above the
affects air quality depends strongly on the possible ex- inversion is substantially unrestricted.
pansion of the urban area beyond the valley floor toward Several changes in the thermal cross-valley wind sys-
the sidewalls. We think that this problem deserves fur- tem, associated with changes in the dynamics of the
ther research. temperature inversion, can occur depending on the
856 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 53

percentage of urban land. We found that (i) changes in As a final remark, it is stressed that we are not sug-
the urban area fraction may result in a confined or open gesting that increasing the urban area fraction of a
diurnal cross-valley circulation, that is, cross-valley cir- mountain valley would lead to better air quality but,
culation cells decoupled (confined system) or coupled rather, are pointing out that the urbanization processes
(open system) to the overlying atmosphere, due to of a mountain valley subject to temperature inversions
a capping effect of the inversion layer; (ii) lower per- may lead to nonevident, and possibly counterintuitive,
centages of urban land (#40% in our simulations) result impacts on air quality at the valley scale. Better under-
in reversed cross-valley circulations producing down- standing of these topics is of particular importance for
slope winds, as a consequence of the interplay between urban planning in growing cities located in mountain
the UHI in the core of the valley and the presence of the terrains, such as several Latin American cities in the
inversion layer restricting vertical motions; (iii) upslope Andes. We acknowledge that urbanization has poten-
winds and thermal convective plumes rising off the top tially detrimental, or at least uncertain, environmental
of the valley sidewalls may be intensified as the urban area effects on multiple spatial and temporal scales, and that
fraction grows, thus strengthening the removal of air from further research into these topics is needed to support
below the inversion layer; and (iv) turbulence is more ef- decision-making processes related to urban planning.
ficiently produced by buoyancy over the urban areas and,
therefore, the expansion of urban land enlarges the area Acknowledgments. We thank S. Serafin for insightful
above which the CBL is substantially developed. comments during the early development of this study.
The BTI pattern in an urban valley is sensitive to the We also thank Editor Joseph Charney and the anony-
percentage of urban land. Under null (rural valley) or mous reviewers for their constructive comments. The
low percentages (#40% in our simulations) of urban computations for this study were performed on the su-
land, the reduction of the inversion layer is dominated percomputer Blizzard at the German Climate Com-
by the rise of its bottom owing to the growth of the CBL puting Center (Deutsches Klimarechenzentrum). AMR
(pattern 1), while when the percentage of urban land is thanks the German Academic Exchange Service, DAAD,
high ($60% in our simulations), the inversion layer is for supporting her as a visiting doctoral student in
reduced by both the rise of its bottom and the descent of Germany through Short-Term Scholarship A/10/77699,
its top (pattern 3). Such descent is caused by the removal and the Colombian Administrative Department of Sci-
of air from the bottom of the valley by the upslope ence, Technology and Innovation, COLCIENCIAS, for
winds, as well as the air subsidence from the upper at- funding her doctoral studies through the research project
mosphere. This process of the descent of the top of the on de procesos f
sicos en la baja atm

Caracterizaci
osfera
inversion layer does not dominate the BTI (pattern 2) in del Valle de Aburr
a (C

odigo 1115-48925515, Contrato
any of the studied cases. 458-09) and grant Francisco Jos
e de Caldas.
The influence of urbanization on the air quality of an
urban valley may lead to contrasting and possibly
counterintuitive effects when considering a temperature REFERENCES
inversion. It does not seem to be necessarily true that, Allen, L., F. Lindberg, and C. Grimmond, 2011: Global to city scale
keeping all properties of the urban land constant (as- urban anthropogenic heat flux: Model and variability. Int.
suming that emissions expand uniformly with urban land J. Climatol., 31, 19902005, doi:10.1002/joc.2210.
cover), more urban land implies worse air quality, even Andrejczuk, M., W. Grabowski, S. Malinowski, and P. Smolarkiewicz,
2004: Numerical simulation of cloud-clear air interfacial mixing.
when considering that the amount of pollutants emitted
J. Atmos. Sci., 61, 17261739, doi:10.1175/1520-0469(2004)061,1726:
grows with the size of the urban area. The mixing ratio of NSOCAI.2.0.CO;2.
pollutants in the valley atmosphere is sensitive to the Anquetin, S., C. Guilbaud, and J. Chollet, 1998: The formation and
percentage of urban land, not only because of changes in destruction of inversion layers within a deep valley. J. Appl.
the amount of pollutant emitted, but also because of the Meteor., 37, 15471560, doi:10.1175/1520-0450(1998)037,1547:
effects that urbanization can exert on the dynamics of TFADOI.2.0.CO;2.
, , and , 1999: Thermal valley inversion impact on
the temperature inversion, and the consequent effects the dispersion of a passive pollutant in a complex mountain-
on the exchange of air between the atmosphere above ous area. Atmos. Environ., 33, 39533959, doi:10.1016/
and below the inversion layer. In the simulations pre- S1352-2310(99)00137-5.
sented in this study, peak mixing ratios occur for the Argentini, S., G. Mastrantonio, and F. Lena, 1999: Case studies of
higher percentages of urban land when the temperature the wintertime convective boundary-layer structure in the
urban area of Milan, Italy. Bound.-Layer Meteor., 93, 253267,
inversion is present, but higher percentages of urban doi:10.1023/A:1002057414039.
land do not imply higher mixing ratios throughout the Arnfield, A., 2003: Two decades of urban climate research: A re-
daytime and everywhere in the valley. view of turbulence, exchanges of energy and water, and the
APRIL 2014
N ET AL.
RENDO 857

urban heat island. Int. J. Climatol., 23, 126, doi:10.1002/ Jung, M., and Coauthors, 2011: Global patterns of land-atmosphere
joc.859. fluxes of carbon dioxide, latent heat, and sensible heat derived
Bader, D. C., and T. B. McKee, 1985: Effects of shear, stability and from eddy covariance, satellite, and meteorological ob-
valley characteristics on the destruction of temperature in- servations. J. Geophys. Res., 116, G00J07, doi:10.1029/
versions. J. Climate Appl. Meteor., 24, 822832, doi:10.1175/ 2010JG001566.
1520-0450(1985)024,0822:EOSSAV.2.0.CO;2. Kalthoff, N., H. Binder, M. Kossmann, R. V ogtlin, U. Corsmeier,
Baklanov, A., 2002: The mixing height in urban areasA review. F. Fiedler, and H. Schlager, 1998: Temporal evolution and spatial
Proc. Workshop on Mixing Height and Inversions in Urban variation of the boundary layer over complex terrain. Atmos.
Areas, Toulouse, France, European Cooperation in Science Environ., 32, 11791194, doi:10.1016/S1352-2310(97)00193-3.
and Technology Action 715, 928. Kanda, M., 2007: Progress in urban meteorology: A review.
Berge, E., A. Karppinen, J. Kukkonen, M. Kltzow, and L. Slrdal, J. Meteor. Soc. Japan, 85B, 363383, doi:10.2151/jmsj.85B.363.
2002: Simulations of wintertime inversions in northern Euro- Lee, S.-H., and H.-D. Kim, 2008: Effects of regional warming due
pean cities by use of NWP-models. Proc. Workshop on Mixing to urbanization on daytime local circulations in a complex
Height and Inversions in Urban Areas, Toulouse, France, basin of the Daegu metropolitan area, Korea. J. Appl. Meteor.
European Cooperation in Science and Technology Action Climatol., 47, 14271441, doi:10.1175/JAMC1504.1.
715, 7588. Lehner, M., and A. Gohm, 2010: Idealised simulations of daytime
Bonan, G., 2008: Forest and climate change: Forcings, feedbacks, pollution transport in a steep valley and its sensitivity to
and the climate benefits of forests. Science, 320, 14441449, thermal stratification and surface albedo. Bound.-Layer
doi:10.1126/science.1155121. Meteor., 134, 327351, doi:10.1007/s10546-009-9442-y.
Chen, X.-L., H.-M. Zhao, P.-X. Li, and Z.-Y. Yin, 2006: Remote Li, Y., R. Smith, and V. Grubisic, 2009: Using surface pressure
sensing image-based analysis of the relationship between ur- variations to categorize diurnal valley circulations: Experi-
ban heat island and land use/cover changes. Remote Sens. ments in Owens Valley. Mon. Wea. Rev., 137, 17531769,
Environ., 104, 133146, doi:10.1016/j.rse.2005.11.016. doi:10.1175/2008MWR2495.1.
Chu, P. C., Y. Chen, S. Lu, Z. Li, and Y. Lu, 2008: Particulate air Malek, E., T. Davis, R. S. Martin, and P. J. Silva, 2006: Meteorological
pollution in Lanzhou China. Environ. Int., 34 (5), 698713, and environmental aspects of one of the worst national air pollu-
doi:10.1016/j.envint.2007.12.013. tion episodes (January, 2004) in Logan, Cache Valley, Utah, USA.
Colette, A., F. Chow, and R. Street, 2003: A numerical study of Atmos. Res., 79, 108122, doi:10.1016/j.atmosres.2005.05.003.
inversion-layer breakup and the effects of topographic shad- Masson, V., 2006: Urban surface modeling and the meso-scale
ing in idealized valleys. J. Appl. Meteor., 42, 12551272, impact of cities. Theor. Appl. Climatol., 84, 3545, doi:10.1007/
doi:10.1175/1520-0450(2003)042,1255:ANSOIB.2.0.CO;2. s00704-005-0142-3.
Colvile, R., E. Hutchinson, J. Mindell, and R. Warren, 2001: The Matzinger, N., M. Andretta, E. Gorsel, R. Vogt, A. Ohmura, and
transport sector as a source of air pollution. Atmos. Environ., M. Rotach, 2003: Surface radiation budget in an alpine valley.
35, 15371565, doi:10.1016/S1352-2310(00)00551-3. Quart. J. Roy. Meteor. Soc., 129, 877895, doi:10.1256/qj.02.44.
Deardorff, J. W., 1980: Stratocumulus-capped mixed layers derived Nadeau, D., and Coauthors, 2009: Estimation of urban sensible
from a three-dimensional model. Bound.-Layer Meteor., 18, heat flux using a dense wireless network of observations. En-
495527, doi:10.1007/BF00119502. viron. Fluid Mech., 9, 635653, doi:10.1007/s10652-009-9150-7.
Edgerton, S. A., and Coauthors, 1999: Particulate air pollu- Offerle, B., C. Grimmond, K. Fortuniak, and W. Pawlak, 2006:
tion in Mexico City: A collaborative research project. Intraurban differences of surface energy fluxes in a central Eu-
J. Air Waste Manage. Assoc., 49, 12211229, doi:10.1080/ ropean city. J. Appl. Meteor. Climatol., 45, 125136, doi:10.1175/
10473289.1999.10463915. JAM2319.1.
Fan, H., and D. Sailor, 2005: Modeling the impacts of anthropo- Ogura, Y., and N. Phillips, 1962: Scale analysis of deep and shallow
genic heating on the urban climate of Philadelphia: A com- convection in the atmosphere. J. Atmos. Sci., 19, 173179,
parison of implementations in two PBL schemes. Atmos. doi:10.1175/1520-0469(1962)019,0173:SAODAS.2.0.CO;2.
Environ., 39, 7384, doi:10.1016/j.atmosenv.2004.09.031. Oke, T., 1995: The heat island of the urban boundary layer:
Foley, J., and Coauthors, 2005: Global consequences of land use. Characteristics, causes and effects. Wind Climate in Cities,
Science, 309, 570, doi:10.1126/science.1111772. J. E. Cermak et al., Eds., Kluwer Academic, 81107.
Grimmond, C., and T. Oke, 2002: Turbulent heat fluxes in urban Olofson, K., P. Andersson, M. Hallquist, E. Ljungstr om, L. Tang,
areas: Observations and a local-scale urban meteorological D. Chen, and J. Pettersson, 2009: Urban aerosol evolu-
parameterization scheme (LUMPS). J. Appl. Meteor., 41, 792 tion and particle formation during wintertime tempera-
810, doi:10.1175/1520-0450(2002)041,0792:THFIUA.2.0.CO;2. ture inversions. Atmos. Environ., 43, 340346, doi:10.1016/
Grimmond, S., 2007: Urbanization and global environmental j.atmosenv.2008.09.080.
change: Local effects of urban warming. Geogr. J., 173, 8388, Panday, A. K., and R. G. Prinn, 2009: Diurnal cycle of air pollution
doi:10.1111/j.1475-4959.2007.232_3.x. in the Kathmandu Valley, Nepal: Observations. J. Geophys.
Haeger-Eugensson, M., and B. Holmer, 1999: Advection caused by the Res., 114, doi:10.1029/2008JD009777.
urban heat island circulation as a regulating factor on the noc- Piotrowski, Z. P., M. J. Kurowski, B. Rosa, and M. Z. Ziemianski,
turnal urban heat island. Int. J. Climatol., 19, 975988, doi:10.1002/ 2010: EULAG model for multiscale flowsTowards the
(SICI)1097-0088(199907)19:9,975::AID-JOC399.3.0.CO;2-J. petascale generation of mesoscale numerical weather predic-
Jacobson, M., 2005: Fundamentals of Atmospheric Modeling. tion. Parallel Processing and Applied Mathematics, Lecture
2nd ed. Cambridge University Press, 813 pp. Notes in Computer Sciences, R. Wyrzykowski et al., Eds.,
Janhall, S., K. Olofson, P. Andersson, J. Pettersson, and M. Hallquist, Springer, 380387.
2006: Evolution of the urban aerosol during winter temperature Prusa, J., P. Smolarkiewicz, and A. Wyszogrodzki, 2008: EULAG,
inversion episodes. Atmos. Environ., 40, 53555366, doi:10.1016/ a computational model for multiscale flows. Comput. Fluids,
j.atmosenv.2006.04.051. 37, 11931207, doi:10.1016/j.compfluid.2007.12.001.
858 JOURNAL OF APPLIED METEOROLOGY AND CLIMATOLOGY VOLUME 53

Rampanelli, G., D. Zardi, and R. Rotunno, 2004: Mechanisms of Taha, H., S. Chang, and H. Akbari, 2010: Meteorological and air
up-valley winds. J. Atmos. Sci., 61, 30973111, doi:10.1175/ quality impacts of heat island mitigation measures in three US
JAS-3354.1. cities. Lawrence Berkeley National Laboratory Rep. LBL-
Reuten, C., D. Steyn, K. Strawbridge, and P. Bovis, 2005: Obser- 44222, 51 pp.
vations of the relation between upslope flows and the con- Toro, M. V., J. J. Ram
rez, R. A. Quiceno, and C. A. Zuluaga, 2001:
vective boundary layer in steep terrain. Bound.-Layer Meteor., C
alculo de la emisi
on vehicular de contaminantes atmos-
116, 3761, doi:10.1007/s10546-004-7299-7. f
ericos en la ciudad de Medell
n mediante factores de emisi
on
Rizwan, A., L. Dennis, and C. Liu, 2008: A review on the generation, Corinair. Rev. Acodal, 191, 4249.
determination and mitigation of urban heat island. J. Environ. Wallace, J., D. Corr, and P. Kanaroglou, 2010: Topographic and
Sci. (China), 20, 120128, doi:10.1016/S1001-0742(08)60019-4. spatial impacts of temperature inversions on air quality using
Romero, H., M. Ihl, A. Rivera, P. Zalazar, and P. Azocar, 1999: mobile air pollution surveys. Sci. Total Environ., 408, 5086
Rapid urban growth, land-use changes and air pollution in 5096, doi:10.1016/j.scitotenv.2010.06.020.
Santiago, Chile. Atmos. Environ., 33, 40394047, doi:10.1016/ Warner, T., 2011: Numerical Weather and Climate Prediction.
S1352-2310(99)00145-4. Cambridge University Press, 526 pp.
Rosenzweig, C., W. Solecki, S. Hammer, and S. Mehrotra, 2010: Welch, T., P. Smolarkiewicz, R. Rotunno, and B. Boville, 2001:
Cities lead the way in climate-change action. Nature, 467, 909 The large-scale effects of flow over periodic mesoscale
911, doi:10.1038/467909a. topography. J. Atmos. Sci., 58, 14771492, doi:10.1175/
Rutllant, J., and R. Garreaud, 1995: Meteorological air pollution 1520-0469(2001)058,1477:TLSEOF.2.0.CO;2.
potential for Santiago, Chile: Towards an objective episode Wesson, G., K. Gabriel, and L. Chun-Ta, 2001: Sensible heat flux
forecasting. Environ. Monit. Assess., 34, 223244, doi:10.1007/ estimation by flux variance and half-order time derivative
BF00554796. methods. Water Resour. Res., 37, 23332343, doi:10.1029/
Savij
arvi, H., and J. Liya, 2001: Local winds in a valley city. Bound.- 2001WR900021.
Layer Meteor., 100, 301319, doi:10.1023/A:1019215031007. Whiteman, C., 1982: Breakup of temperature inversions in deep
Savov, P., T. Skakalova, I. Kolev, and F. Ludwig, 2002: Lidar in- mountain valleys: Part I. Observations. J. Appl. Meteor., 21, 270
vestigation of the temporal and spatial distribution of atmo- 289, doi:10.1175/1520-0450(1982)021,0270:BOTIID.2.0.CO;2.
spheric aerosols in mountain valleys. J. Appl. Meteor., 41, 528541, , 2000: Mountain Meteorology: Fundamentals and Applica-
doi:10.1175/1520-0450(2002)041,0528:LIOTTA.2.0.CO;2. tions. Oxford University Press, 355 pp.
Schmidli, J., and Coauthors, 2011: Intercomparison of mesoscale , and T. McKee, 1982: Breakup of temperature inversions in
model simulations of the daytime valley wind system. Mon. deep mountain valleys: Part II. Thermodynamic model. J. Appl.
Wea. Rev., 139, 13891409, doi:10.1175/2010MWR3523.1. Meteor., 21, 290302, doi:10.1175/1520-0450(1982)021,0290:
Schmidt, H., and U. Schumann, 1989: Coherent structure of the con- BOTIID.2.0.CO;2.
vective boundary layer derived from large-eddy simulations. , B. Pospichal, S. Eisenbach, P. Weihs, C. Clements, R. Steinacker,
J. Fluid Mech., 200, 511562, doi:10.1017/S0022112089000753. E. Mursch-Radlgruber, and M. Dorninger, 2004: Inversion
Serafin, S., and D. Zardi, 2010a: Daytime heat transfer processes breakup in small Rocky Mountain and Alpine basins. J. Appl.
related to slope flows and turbulent convection in an idealized Meteor., 43, 10691082, doi:10.1175/1520-0450(2004)043,1069:
mountain valley. J. Atmos. Sci., 67, 37393756, doi:10.1175/ IBISRM.2.0.CO;2.
2010JAS3428.1. Wilson, J. G., S. Kingham, J. Pearce, and A. P. Sturman, 2005: A
, and , 2010b: Structure of the atmospheric boundary review of intraurban variations in particulate air pollution:
layer in the vicinity of a developing upslope flow system: A Implications for epidemiological research. Atmos. Environ.,
numerical model study. J. Atmos. Sci., 67, 11711185, 39, 64446462, doi:10.1016/j.atmosenv.2005.07.030.
doi:10.1175/2009JAS3231.1. Yao, W., and S. Zhong, 2009: Nocturnal temperature inversions in
Smolarkiewicz, P. K., and J. M. Prusa, 2004: Forward-in-time dif- a small, enclosed basin and their relationship to ambient at-
ferencing for fluids: Simulation of geophysical turbulence. mospheric conditions. Meteor. Atmos. Phys., 103, 195210,
Turbulent Flow Computation, D. Drikakas and B. J. Geurts, doi:10.1007/s00703-008-0341-4.
Eds., Springer, 279312. Zhong, S., C. Whiteman, and X. Bian, 2004: Diurnal evolution of
, L. Margolin, and A. Wyszogrodzki, 2001: A class of non- three-dimensional wind and temperature structure in Cal-
hydrostatic global models. J. Atmos. Sci., 58, 349364, ifornias Central Valley. J. Appl. Meteor., 43, 16791699,
doi:10.1175/1520-0469(2001)058,0349:ACONGM.2.0.CO;2. doi:10.1175/JAM2154.1.
, R. Sharman, J. Weil, S. Perry, D. Heist, and G. Bowker, 2007: Zoumakis, N., and G. Efstathiou, 2006: Parameterization of in-
Building resolving large-eddy simulations and comparison version breakup in idealized valleys. Part II: Thermodynamic
with wind tunnel experiments. J. Comput. Phys., 227, 633653, model. J. Appl. Meteor. Climatol., 45, 609623, doi:10.1175/
doi:10.1016/j.jcp.2007.08.005. JAM2354.1.