ACPAtmospheric Chemistry and PhysicsACPAtmos. Chem. Phys.1680-7324Copernicus PublicationsGöttingen, Germany10.5194/acp-17-7261-2017Sensitivity of local air quality to the interplay between small- and large-scale circulations: a large-eddy simulation studyWolf-GrosseTobiastobias.wolf@nersc.nohttps://orcid.org/0000-0001-6004-2374EsauIgorhttps://orcid.org/0000-0003-4122-6340ReuderJoachimhttps://orcid.org/0000-0002-0802-4838Nansen Environmental and Remote Sensing Center, 5006 Bergen, NorwayGeophysical Institute, University of Bergen, 5007 Bergen, NorwayTobias Wolf-Grosse (tobias.wolf@nersc.no)16June20171711726172765October201624October201611May201715May2017This work is licensed under the Creative Commons Attribution 3.0 Unported License. To view a copy of this licence, visit https://creativecommons.org/licenses/by/3.0/This article is available from https://acp.copernicus.org/articles/17/7261/2017/acp-17-7261-2017.htmlThe full text article is available as a PDF file from https://acp.copernicus.org/articles/17/7261/2017/acp-17-7261-2017.pdf
Street-level urban air pollution is a challenging concern for
modern urban societies. Pollution dispersion models assume that the
concentrations decrease monotonically with raising wind speed. This
convenient assumption breaks down when applied to flows with local recirculations such as
those found in topographically complex coastal areas. This study looks at a
practically important and sufficiently common case of air pollution in a
coastal valley city. Here, the observed concentrations are determined by the
interaction between large-scale topographically forced and local-scale
breeze-like recirculations.
Analysis of a long observational dataset in Bergen, Norway, revealed that
the most extreme cases of recurring wintertime air pollution episodes were
accompanied by increased large-scale wind speeds above the valley. Contrary
to the theoretical assumption and intuitive expectations, the maximum
NO2 concentrations were not found for the lowest 10 m ERA-Interim wind
speeds but in situations with wind speeds of 3 m s-1. To explain
this phenomenon, we investigated empirical relationships between the
large-scale forcing and the local wind and air quality parameters. We conducted
16 large-eddy simulation (LES) experiments with the Parallelised Large-Eddy Simulation Model (PALM) for atmospheric and oceanic flows. The LES
accounted for the realistic relief and coastal configuration as well as for
the large-scale forcing and local surface condition heterogeneity in Bergen.
They revealed that emerging local breeze-like circulations strongly enhance
the urban ventilation and dispersion of the air pollutants in situations
with weak large-scale winds. Slightly stronger large-scale winds, however, can
counteract these local recirculations, leading to enhanced surface air stagnation.
Furthermore, this study looks at the concrete impact of the relative
configuration of warmer water bodies in the city and the major transport
corridor. We found that a relatively small local water body acted as a
barrier for the horizontal transport of air pollutants from the largest
street in the valley and along the valley bottom, transporting them
vertically instead and hence diluting them. We found that the stable
stratification accumulates the street-level pollution from the transport
corridor in shallow air pockets near the surface. The polluted air pockets
are transported by the local recirculations to other less polluted areas
with only slow dilution. This combination of relatively long distance and
complex transport paths together with weak dispersion is not sufficiently
resolved in classical air pollution models. The findings have important
implications for the air quality predictions over urban areas. Any
prediction not resolving these, or similar local dynamic features, might not
be able to correctly simulate the dispersion of pollutants in cities.
Introduction
Urban air pollution is a major concern for urban dwellers and city managers
(Baklanov et al., 2007). Typically, urban air
pollution is thought of in connection with large industrial areas and
megacities with high local emissions (e.g. Zhang et al., 2012). Air pollution
episodes can, however, also affect smaller cities
(e.g. Junk et al., 2003; Schicker and Seibert, 2009; Grange et al., 2013). In areas with comparatively low emissions, slow
removal and long accumulation of locally emitted pollutants are usually
responsible for the occurrence of air pollution episodes. For cities in
mountainous areas, this can be caused either by a local circulation trapping
the pollutants or by local stagnation within the surrounding mountains
(Rotach et al., 2004; Steyn et al., 2013). The
most prominent examples of such local stagnation are typically observed in
stably stratified cold air pools trapped in valleys and other orographic
depressions (e.g. Reeves and Stensrud, 2009; Hoch et al., 2011; Sheridan et al., 2014; Hughes et al., 2015).
The local trapping and stagnation conditions are notoriously difficult to
study, as their interaction with the controlling large-scale circulation in
the free atmosphere can be complex and nonlinear (e.g. Whiteman and Doran,
1993; Zängl, 2003). Interactions within the cold air pools can be further complicated by other
local dynamic effects, such as breeze-like circulations (e.g. Lareau et al., 2013) or up- and down-slope winds
(Fernando, 2010). This complexity challenges numerical weather
prediction (NWP) models, which often cannot properly resolve and hence
predict air pollution episodes, in particular those developing during the
periods with strong stable stratification in the lower atmosphere and
mountainous terrain (e.g. Berge et al., 2002; Fay and Neunhäuserer, 2006;
Baklanov et al., 2011). Thus, the problem could be partially related to an
insufficient spatial model resolution with respect to the relevant
topographic features. But partially, it should be also related to the
breakdown of inadequate turbulent mixing schemes for the atmospheric
stably stratified boundary layer (Atlaskin
and Vihma, 2012; Fernando and Weil, 2010; Mahrt, 2014; Zilitinkevich et al.,
2015). These and other inaccuracies in the air quality assessment and
modelling motivated the Pan-Eurasian Experiment (PEEX) community to include
air pollution in the stably stratified boundary layers in the list of
research priorities addressing the environmental and socio-economic
challenges of the northern high latitudes (Lappalainen et al., 2016).
Large-eddy simulations (LESs) can resolve much of the relevant turbulence
dynamics. The computational costs of high-resolution LESs over larger urban
areas are, however, too high to be used operationally. While this might
change in the near future (Schalkwijk et al., 2015), LES
models are today mostly used for the simulation of specific scenarios or for
process studies in order to gain a deeper understanding of the flow features
at hand (Bergot et al., 2015; Esau, 2007, 2012; Patnaik and Boris, 2010). Increased local
knowledge and general understanding of the relevant processes
(e.g. Glazunov et al., 2016) can then help
the forecasters to improve air quality predictions based on NWP models, as
suggested by Steyn et al. (2013).
LES models are more and more used for the study of the air flow in model
domains with realistic urban topography. One could name a few relevant urban
studies. For example, Letzel et al. (2008) and Nakayama et
al. (2012) simulated the flow on the neighbourhood-scale in Tokyo, Japan.
Esau (2012) studied the dispersion of a passive tracer
over central Paris, France. Keck et al. (2014) used an LES model for
an assessment of the effect of a new densely built-up artificial island in
Macau, China. Park et al. (2015a, b) studied the street-level
ventilation in Seoul, South Korea.
So far however, most of the published studies were conducted under neutral
or slightly convective conditions. While the use of LES for the simulation
of the stably stratified conditions has become more common in the recent
years, only few studies run LESs of the urban stably stratified atmospheric boundary layer (ABL). Cheng
et al. (2010) and Tomas et al. (2016) studied a
stratified flow over arrays of surface mounted boxes.
Xie et al. (2013) studied the effect of
stable stratification on the dispersion conditions over a kilometre-scale
domain in London. Chen et al. (2011) developed a
coupled modelling system, which includes the mesoscale Weather Research and Forecasting (WRF) and LES models, and
Wyszogrodzki et al. (2012) applied
this modelling system to study convective and weakly stably stratified flows
over a kilometre-scale domain in Oklahoma City. A review of urban ABL
modelling in a number of different settings was published by
Barlow (2014). Nevertheless, to our knowledge there are no
high-resolution studies devoted to the simulation of the stably stratified
urban ABL in a domain relevant to urban management and decision-making.
This study is intended to reduce the above-mentioned knowledge gap. It is based on
a new set of fine-resolution LESs of the topography- and breeze-induced
circulation for the realistic topography of a larger part of Bergen, Norway.
Strongly stable stratification of the lowermost atmospheric layers has been
recognized as a precursor for the accumulation of air pollutants and
subsequent air quality hazards in Bergen. The city centre is embedded in a
coastal valley ending at a large sea inlet called Byfjorden. Periods of high
air pollution are observed during persistent temperature inversions,
indicating the stably stratified conditions inside the valley
(Wolf et al., 2014; Wolf and Esau, 2014).
Observations of the typical circulation and temperature conditions above and
within the valley are used to drive the LES runs. We would like to focus on
the interaction between the large-scale circulation above the valley,
modified by the local topography, and a local land breeze between the cold
land and the warm fjord under inversion conditions. This interaction is
leading to the strongest stagnation and hence accumulation of pollutants
inside the valley. The simulations should further serve as a proof of
concept for the added value of high-resolution LESs for the analysis of urban
air pollution dispersion in complex mountain terrain.
This study is structured as follows. In Sect. 2 we will present an
observational dataset that motivated this work. In Sect. 3 we will
outline the applied methodology, and the results will be presented in Sect. 4 and summarized and discussed in the last section.
NO2 concentrations at the two air quality reference stations
in Bergen against the 10 m wind speeds (ws) over the Bergen Valley from EraI
and from local measurements inside the valley between 2003 and 2013. EraI
data are available every 3 h. NO2 concentrations and the local
measured winds represent hourly means. The NO2 concentrations are the
3-hourly maxima and the local wind data are the 3-hourly means around the
3 h time steps. Only wintertime data (November–February) are included. The
edges of the boxes are the 25th and 75th percentiles. Lines inside the boxes
show the median. The maximum whisker length is 1.5 times the distance
between the 25th and 75th percentiles or the maxima and minima of the data.
Outliers with higher or lower concentration values than that are shown as
circles. The lower panels show the number of simultaneously valid wind and
pollution measurements within each wind-speed bin.
Observational perspective on air pollution in coastal cities
For coastal cities, land–sea breeze circulations can have a strong impact on
the local circulation. This can lead to a transport of pollutants from high
emission areas onwards to inhabited areas (Gariazzo et al., 2007) or a closed
recirculation and hence accumulation of pollutants (Crosman and Horel, 2016;
Lo et al., 2006; Rimetz-Planchon et al., 2008). Most of these works have in
common that the background circulation reduces the relevance of the local
effects (Crosman and Horel, 2010). For coastal mountain
cities there is, however, also the possibility of an interaction between
local circulations such as local breezes and cold air pools
(Holmer et al., 1999; Lareau et al., 2013).
These local circulations can be either dominating the local circulation
forced by the large-scale flow or interacting with it. Both can lead to a
deviation from the usual situation of a weaker background flow leading to
higher local accumulation of air pollutants.
Figure 1 illustrates this based on the distribution of the NO2
concentrations against wind speeds above and inside the Bergen Valley. Wind
speeds inside the valley are measured on a mast on top of the Geophysical
Institute (GFI; see Sect. 3.1) at 12 m above the roof
(approximately 50 m height). The air flow at this height is unobstructed
by other buildings. The measurements should therefore be representative of
the air flow in the middle of the valley. Quality-controlled hourly data for
the entire period between 2003 and 2013 were readily available online
(Norwegian Meteorological Institute, 2016). The expected
logarithmic decrease of the NO2 concentrations with increasing wind
speeds inside the valley is visible for both an urban background (UB) and
high-pollution (HP) reference station (compare to Sect. 3.1;
data downloaded from Norwegian Institute for Air Research, 2016). For
the surface (10 m) wind speeds from the ERA-Interim (EraI) reanalysis
product, however, such a clear decrease does not exist. Due to the low
resolution of EraI, the 10 m winds are not influenced by small-scale
topographic features such as the Bergen Valley. They should therefore
represent the large-scale flow that is only modified by the Norwegian
topographic features on scales larger than a few tens of kilometres. The
maximum NO2 air pollution is found for wind speeds around 3 m s-1.
This maximum in the concentrations at some non-zero background wind
speed suggests an interaction of the background flow with a local forcing
under the cold air pool conditions. Some combination of all three
circulation features then leads to a maximum stagnation over the valley bottom.
Thus, we observe that any air quality prediction based on meteorological
fields from models that are not resolving this local forcing could fail to
resolve the highest local air pollution concentrations. We assume that the
most relevant local forcing in the Bergen Valley is the breeze circulation
caused by the temperature difference through the land–sea interface. The
relevance of the interplay between the larger-scale circulation, the local
topography, and the local forcing is therefore assessed in this study.
Model experimentsGeographical description of the simulation domain
Bergen, Norway, is located at the Norwegian west coast (60.4∘ N,
5.3∘ E). The central part of the city is located in a curved valley
with a minimum width at the valley base of approximately 1 km (Fig. 2).
The surrounding mountain tops are between 344 and 642 m high. Inside
the valley there are a number of water bodies aside the fjord. Only the
largest ones close to the city centre are explicitly treated in this study.
Since 2003 two measurement stations have been monitoring the street-level
(2 m above surface) air quality in the Bergen Valley: a high-pollution
reference station that is located nearby one of the busiest crossroads
in the city and an urban background reference station. The valley
favours frequent wintertime ground-based temperature inversions
(Wolf et al., 2014), leading to the exceedances of air quality
thresholds for NO2 and PM2.5, especially in areas with high
traffic. In contrast to some cold air pools in large valleys
(e.g. Zhong et al., 2001), the cold air pools
connected to air quality hazards in the Bergen Valley are caused by
ground-based temperature inversions. They cannot exist without persistent
long-wave (LW) radiative cooling such as during fair-weather events in
wintertime with little to no solar insolation and clouds. Temperature
inversions in the Bergen Valley usually appear in connection with a specific
circulation pattern both inside and outside the valley. Wolf et
al. (2014) analysed the 10 min wind measurements from an automatic
meteorological station (AMS) on top of GFI together with measurements of the
vertical temperature profile above GFI. The study identified almost
exclusively down-valley winds during measurements with temperature
inversions, while during measurements without temperature inversions both
up-valley and down-valley channelled flows existed. This, together with the
previously mentioned south-easterly background flow above the Bergen Valley
during high air pollution events, gave reason to assume that the preferred
wind direction inside the valley simply follows the large-scale wind
direction above. In this case the mountain to the south-east of the Bergen Valley would shelter the valley. This is, however, too simplistic a picture
since the sheltering mountains would represent a backward-facing step for
the approaching winds, which can lead to very complicated flow regimes
(Mohamad and Viskanta, 1995).
Topographic map of the Bergen area represented in the model
simulations. Blue colour indicates water. The white contour line marks the
areas 1 and 2 selected for averaging of the wind and passive tracer
concentration. The red contour line marks the area of passive tracer
emissions in most of the experiments. Both lines are overlapping along the
coast. The yellow line represents the main street along the valley,
eventually disappearing into a tunnel after having crossed the valley. White
arrows indicate the positions of the wind measurements (AMS), and the urban background (UB) and
high pollution (HP) reference stations for NO2 air pollution. The area for averaging of
the 2 m temperature over the fjord in Table 1 is marked as area 3; averaging
of the 2 m temperature over the city centre is done separately over areas 1
and 2. The black arrow in the upper right corner indicates the 110∘ geostrophic wind direction used for the simulations.
During persistent wintertime temperature inversions in the Bergen Valley,
the local lakes often freeze over. The fjord, however, remains mostly ice-free, causing large temperature contrasts between the relatively warm fjord
surface and the cold land surface, possibly leading to the local breeze
circulation inside the valley. The large water body roughly in the middle of
the domain is connected to the fjord only though a narrow channel. It
therefore has a lower salinity than the fjord and tends to freeze over with
a thin layer of ice during persistent temperature inversion episodes. It is
consequently treated as a lake later on.
The model
In the experiments for this study we used the Parallelized Large-Eddy
Simulation Model (PALM) for atmospheric and oceanic flows, (Maronga et al., 2015;
Raasch and Schröter, 2001). PALM solves the finite-difference numerical
realization of the non-hydrostatic, filtered, incompressible Navier–Stokes
equations in their Boussinesq approximated form. The model employs a 1.5-order closure using a subgrid-scale TKE balance equation
(Deardorff, 1980; Moeng and Wyngaard, 1988;
Saiki et al., 2000). Advection terms are computed using the fifth-order scheme
after Wicker and Skamarock (2002).
The model time step is adjusted dynamically. The incompressibility condition
is satisfied with a predictor–corrector method, using the Temperton fast
Fourier transformation to solve the Poisson pressure equation. Spatial
discretization is done on an Arakawa type C grid. Topographic features of
the urban area are simplified as ground-mounted boxes. Topographic input
needs to be specified as a separate input file that fits the horizontal
model grid. Vertical discretization of the surface height onto the model
grid is done automatically by the model.
Modifications to the PALM code
PALM runs either with Neumann surface temperature boundary condition (BC),
where the kinematic surface heat flux (Hs) is prescribed, or with
Dirichlet BC, where the surface temperature (Ts) is fixed. In the case
of the Dirichlet BC, heat fluxes on horizontal surfaces are calculated by
assuming a Prandtl layer. In order to be able to simulate the potential
breeze effect in the valley, we added the possibility to use mixed Dirichlet
and Neumann BCs in the model. Consistent with the large heat capacity of
water, we kept Ts constant over the water-covered parts of the
computational domain. This allows for the development of the typical pattern
of organized turbulence over the water surface, as it has been repeatedly
found for breeze circulations, e.g. induced by arctic leads
(Esau, 2007; Lüpkes et al., 2008). Over the
land surface, we kept Hs constant in order to represent the effect of LW
radiative cooling. While specifying negative Hs can be problematic for
LES studies (Basu et al., 2008), it was necessary here to
incorporate the inhomogeneities in surface temperature due to the large
differences in surface elevation and the land–sea interface. A description
of our modifications to PALM for the usage of heterogeneous surface boundary
conditions is given in the Supplement. In order to be able to
study the dispersion of pollutants from different emission sources, we used
the same approach for reading heterogeneous fields for the surface flux of a
passive tracer Fs. To avoid unphysical recycling of the passive
tracer due to periodic boundary conditions, we set the passive tracer to 0
at the lateral boundaries of the computational domain.
Domain
The simulations are done for the realistic topography of the city of Bergen,
Norway. For this, we received laser scanning data from the municipality of
Bergen for a 5 km square around Bergen city hall. In the choice of the
final domain size, we tried to include the mountains directly surrounding
the inner part of the city, while attempting to keep the computational
domain as small as possible. At the lateral boundaries, we used periodic BC.
The Bergen Valley is open towards the north-west and south-west (not visible
in Fig. 2 due to the small domain used for the simulations). In order to
allow for a more realistic free flow along the valley axis, we created an
artificial channel at the northern end of the domain. While necessary in
order to avoid an unnatural stagnation in the southern part of the valley,
this channel might also alter the circulation over the fjord at the northern
boundary of the computational domain. For future studies this should be
avoided by using a larger north–south extent of the domain. At the lateral
boundaries we used a 1000 m wide buffer zone, both in the x and
y directions. In this buffer zone, surface elevations are linearly interpolated
in order to guarantee a smooth transition through the periodic boundaries.
Polygons of all water bodies in Bergen were provided by Bergen municipality.
From this, we produced the input files for the areas with constant Hs
over land and constant Ts for grid boxes that were covered by more than
50 % of water. As a simplification we ignored most freshwater lakes.
The final domain consists of 1024 × 1024 grid nodes in the horizontal
x and y directions including the buffer zones, and 128 levels in the
vertical z direction. The model resolution is 10 m for each coordinate axis
in the lower 750 m of the computational domain. Above 750 m the grid
is vertically stretched by 1 % for each additional grid level. The
total domain size is Lx=Ly= 10 240 m and Lz= 1451 m – well
above the highest mountaintop at 650 m. We smoothed the topography
with a running mean filter over three grid cells in both lateral directions.
The extent of the fjord in this setup is roughly 4 km in the E–W
direction and 3 km in the N–S direction. The western boundary of the
fjord is representative for the location of a large island closing the
fjord, except for a 700 m wide passage. To the north, the fjord extends
in reality much further than in our setup. The artificially set northern
boundary therefore reduces the fjord's extent. A comparison to
Esau (2007) shows, however, that 2–4 km is the size where the
drag of air from warm arctic leads stagnates at its maximum, meaning that
the convergence in the N–S direction should be sufficient to cause a
realistic drag of air in the valley.
It should be noted that a 10 m resolution is clearly too coarse
to resolve the circulation in street canyons (Letzel et
al., 2008, 2012). It is also too coarse for the simulation of the stably
stratified ABL over flat and homogeneous surfaces
(Beare et al., 2006; Mason and
Derbyshire, 1990). The focus of this study, however, is on the effect of the
larger topographic features (valley width above 1 km) and the convection
over the fjord on the circulation within the valley. Both forcings should be
sufficiently reproduced with the chosen resolution. An analysis of the
kinematic heat fluxes of sensitivity test simulations with 5 and 20 m
resolutions in addition to the 10 m resolution chosen here confirmed this.
The smoothing to a 10 m resolution from the laser-scanning-point clouds
implicitly includes a surface displacement in the case of strongly variable
topography. No further displacement height is included in the simulations
as the non-resolved roughness elements are small compared to the resolved
mountain topography. The unresolved roughness length in LESs of urban areas
and strong topography is an interesting question itself but goes beyond the
scope of the work presented here. We have therefore chosen a constant
roughness length across the entire domain of 0.5 m. A test with roughness
length of 2 m showed qualitatively similar results with some quantitative
differences. We therefore assume the roughness length to be a relevant
parameter that should be studied further but assume the overall conclusions
of this publication not to be dependent on it.
Numerical experiments
The numerical experiments are listed in Table 1. All experiments share a
common basic setup: the latitude was 60.38∘ N, corresponding to
Bergen. At initialization, the surface potential temperature was 273.15 K
with constant potential temperature up to 650 m, and a constant
potential temperature gradient of 5.5 × 10-3 K m-1 above that level.
This is the mean potential temperature gradient above Bergen during high-pollution cases derived from the EraI data set. Hs over land was chosen
as -0.025 K m s-1, corresponding to approximately 25 W m-2.
This is in agreement with heat fluxes found from observational
studies (Brümmer and Schultze, 2015; Nordbo et al.,
2012). In addition to the cases with mixed BC, we also simulated a test case
with a constant Hs of 0 K m s-1 over the entire domain,
representing a situation with neutral stratification.
All simulations were run for 12 h in order to reach a quasi-equilibrium
state. During this time we kept the surface potential temperature over most
water bodies at 273.15 K. The surface potential temperature of the
elongated lake in the north-east of the Bergen Valley was 273.89 K.
This is the potential temperature corresponding to a temperature of
0 ∘C at the height of the lake surface (approximately 75 m). The
temperature chosen over the fjord is given in Table 1 for each simulation.
For cases 1 through 12 we included emission of a passive tracer over the
entire urban area into the model simulations. As emission rate we chose an
arbitrary value of 1 (kg m-2 s-1). In order to mimic the
actual extent of the build-up city area, we only allowed for emissions over
land-covered grid cells with surface elevation below 70 m in the
original input file. For the last three cases we used the same emission
strength, but only for the area covered by the largest street in the valley
(see Fig. 1). The exact geographic location of this street was provided to
us by Bergen municipality.
PALM simulations are usually driven with an imposed geostrophic wind
profile. For the geostrophic winds in our model experiments we used three
different scenarios, illustrated in Fig. 3. The profile with the lowest wind
speeds follows the mean of the wind-speed profiles from EraI above Bergen
during days with high-NO2 air pollution (at least one hourly mean
measurement exceeding 200 µg m-3 at the high-traffic reference
station). Because of the varying height of the EraI model levels, we
linearly interpolated the wind profiles between the nearest model levels
to 410, 450, 600, 800, 1000, 1200, and 2000 m height before
averaging. Since EraI has a rather low resolution, the lowest grid layer
over Bergen is at approximately 410 m, depending on surface pressure.
In 4 out of the total 45 high-pollution days in the measurement record,
the lowest model layer was centred above 410 m. For these cases, no
wind speed was available at the 410 m height. As the lowest point in the
PALM domain is the sea surface, it was, however, necessary to specify a wind
speed below 410 m height. We chose to use the wind speed from 410 m in
EraI at 100 m height in our PALM simulations and 0 m s-1 wind
speed at sea level. The mean wind direction profile during all high-pollution cases changed from 100∘ at the lowest levels to
120∘ higher up. For simplicity, we kept the wind direction constant
at 110∘. The two higher wind-speed scenarios follow vertical wind-speed gradients of 1.5 and 2 times the mean gradient for all height levels
above 100 m.
a Simulation with Neumann BC and Hs= 0 K m s-1
over entire domain. b Calculated from a linear extrapolation of the
potential temperature gradient between the two lowest grid points above topography.
Absolute temperature calculated with reference pressure 1000 hPa. The mean
temperatures over land contain areas up to 70 m surface elevation. In inversion
conditions this results in a higher mean temperature than only for the lowest
areas. c Emissions only from the largest street in the Bergen Valley.
Geostrophic wind profile scenarios used for the model simulations.
The black line shows the mean EraI wind profile over the Bergen Valley for
all high-pollution cases (they start only at 410 m due to the
low horizontal resolution of EraI). Above 450 m the red and
black lines are overlapping.
ERA-Interim data
EraI data are available from the ECMWF archive (Dee et
al., 2011; ECMWF, 2016). The resolution of EraI (T255) is too coarse to
resolve any of the relevant features of the Bergen Valley. We therefore used
the EraI wind speed and direction for the specification of the background
winds in PALM. We downloaded data at a horizontally interpolated grid of
0.25∘ resolution and used the mean over the two grid boxes centred
at 5.5∘ E, 60.25∘ N and 5.5∘ E,
60.5∘ N that represented best the location of Bergen. For
calculating the daily mean fields, we used EraI data at a 3-hourly
resolution from a combination of the analysis steps at 06:00, 12:00, 18:00, and
00:00 UTC plus the 3 and 9 h lead time forecasts of the 00:00 and 12:00 UTC analyses.
ResultsMain features of the Bergen Valley circulation
The setup used here was chosen to study the potential effect of the
breeze-induced circulation on the dispersion of pollutants in the Bergen Valley under the conditions of typical wintertime temperature inversions.
Here we will highlight the most relevant features and briefly compare them
to the results of the observational study conducted by Wolf
et al. (2014) in order to better understand the circulation in the valley. By
this we will also investigate the potential and limitations of the chosen
setup for the proposed flow interactions. We use case 3 with the mean wind-speed profile during all high-air-pollution cases as a baseline. The fjord
temperature of 2.5 ∘C should be realistic for typical persistent
wintertime temperature inversions.
The 55 m wind field from case 3 together
with Hs (K m s-1). The figure shows the mean fields over
the last 15 min of the 12 h simulation time. Wind
vectors point into the flow direction. The wind vector in the upper right
corner indicates the scale of the wind vector length. It shows a wind speed
of 3 m s-1. The 55 m topographic line is indicated by the thick black dashed line. The water–land
interface is shown by the thick solid line.
After 12 h all simulations are in quasi-steady-state conditions. The
resulting 2 m temperatures over the fjord and area 1 and 2 of the city
are given in Table 1. The 2 m temperature is calculated as a linear
extrapolation of the temperature curve between the first two vertical
grid levels above the topography at 5 and 15 m. While the absolute
temperatures over the different areas are irrelevant for this study, their
differences are an indicator for the breeze circulation. By design, the
difference between the air temperatures over the fjord and the temperatures
over the city are increasing with increasing fjord surface temperatures,
consequently applying the forcing for a breeze circulation. In addition, the
air temperature over the interior part of the valley is, except for case 12,
lower than the air temperature over the outer part of the valley. This could
exert another breeze-like forcing between these two parts of the valley. The
dependence of the breeze effect on the prescribed fjord surface temperature
decreases over time. This allows simulations based on the same wind profile
to converge. The reason is that, over land, the heat flux is fixed instead of the
absolute temperature. The land surface temperature therefore slowly adapts
to the fjord surface temperature due to the advection of relatively warmer
air from the recirculation within the valley and the periodic BC. The latter
is caused by an imbalance in the Hs across the entire domain through
different fjord surface temperatures. Neglecting the constant temperature
over the smaller water bodies, after very long simulation times in principle
the atmosphere should reach an equilibrium circulation that is equal for all
different fjord surface temperature scenarios. How fast this equilibrium is
established depends on the total volume of the computational domain and the
land and sea fractions. No convergence was visible for any of the wind-speed
scenarios simulated here.
Vertical profiles for temperature and wind speed and wind
direction (lower panels) at three locations in the Bergen Valley (marked in
the upper panels). The figures show the 15 min mean data
corresponding to case 3 presented in Fig. 4. All profiles are horizontal
averages over the areas indicated in the top panels. The blue areas indicate
water surface. The brown shading gives the surface elevation in metres. The AMS
is located in the centre of area 2. The geostrophic wind profile for this
scenario is indicated as a black line in the lower middle panel.
Figure 4 shows the wind field from case 3 on the vertical level centred at
55 m together with Hs. This is the first vertical level above the
AMS used in Wolf et al. (2014; marked in Fig. 2). The
south-easterly, down-valley mean flow above GFI is reproduced in our
simulation. The mean Hs over all water bodies is 143 W m-2.
Maximum values of up to 1000 W m-2 for a few grid nodes are located in
direct proximity to the coast in areas with the strongest seaward flow, as a
result of the temperature contrast between the land and the fjord surface.
Over the fjord, Hs is not simply decreasing towards the middle of the
fjord, but reaches its minimum in the areas of flow convergence. These are,
for case 3, organized in the form of two convergence lines.
Figure 5 shows local profiles of temperature, wind speed, and wind direction
over and around the large water body in the middle of the valley. Area 2 in
the plot is centred over GFI. The south-easterly flow is visible up to a
height of 95 m. For areas 1 and 2 there is a gradual eastwards
rotation of the wind direction, likely caused by a combination of the
proximity to the warm sea inlet and the local topography. Between
95 and 105 m height, the wind direction jumps from easterly to
north-westerly before rotating back to mostly easterly wind between 300 and
400 m. Over area 3, the wind direction remains almost constant
for the lowest parts of the ABL before gradually turning to the same
north-westerly wind around 300 m.
Case 0 (Fig. 6) with missing surface temperature and heat flux
heterogeneity, showed a northerly up-valley flow, both above the fjord and
inside the valley. The down-valley flow at the valley bottom was absent. The
up-valley flow seems to be a persistent feature of the Bergen Valley
topography under the given geostrophic wind profile, to some degree
balancing the south-easterly down-valley flow that is caused by the
convergence over the fjord. The flow in the valley is therefore not simply
following the upper-air wind distribution, but is clearly locally forced.
Same as Fig. 4 but for case 0 with Neumann BC and neutral ABL.
The north-westerly flow higher up in the valley ABL is only visible for the
simulations with 270.65, 273.15, and 275.65 K fjord surface
temperature, and for case 11 with 278.15 K fjord surface temperature and
wind speed scenario 3. For the simulations with higher fjord surface
temperatures it is not detectable and might be masked by the south-easterly
flow from the convergence over the fjord. A reverse flow above a breeze
circulation is usually associated with the returning branch of the breeze
circulation. An increase in the strength of the land-breeze should therefore
also result in an increase of the return branch. Since this is not the case for our simulations here with increasing fjord surface temperatures indicates
that the north-westerly flow higher up in the valley ABL is not the return
branch of the land-breeze above the valley bottom.
The modelled temperature profiles in Fig. 5 show inversions up to 135 m
height split into two separate inversions with two closely adjacent maxima
for areas 1 and 2. The top of the lowest inversion was at 75 m and the
highest at 135 m. For area 3, there is only one inversion ending in
between the two maxima from area 1 and 2. The results of the measurements
with a passive microwave temperature profiler (MTP-5HE from Attex) on the
rooftop platform of GFI, presented in Wolf et al. (2014),
provide the possibility for a comparison with observed inversion heights in
the Bergen Valley. This instrument does not measure the vertical profiles
directly above the instrument, but averages over variable horizontal
distances based on the scanning angle. The measured temperature profiles are
therefore rather comparable to a combination of the profiles along the
horizontal measurement path, with the lower height levels including a larger
horizontal distance along the measurement direction than the higher height
levels. Areas 1–3 are roughly placed along the measurement path of the
microwave radiometer, and should therefore be appropriate locations for a
comparison. Due to the limited vertical resolution, the angular scanning
microwave radiometer smoothes out fine structures, e.g. two closely
adjacent maxima. For the other simulated cases, the inversion heights were
rather similar, varying between 85 and 145 m. The observed inversion
depths in the Bergen Valley are typically ranging between 70 and 270 m,
with the majority of observations between 70 and 220 m.
Inversion episodes lasting longer than 2 h were on average most
frequently 170 m deep. This indicates that our LES simulations somewhat
underestimate the inversion depth. The modelled maximum inversion strength
in the order of 1–3 K is in accordance to the observations
(Wolf et al., 2014). Since the simulations presented here are
rather idealized, it is likely that relevant physical processes for a more
realistic representation of the inversions, such as large-scale subsidence
or LW radiation divergence in the atmosphere, are not fully considered.
Hoch et al. (2011) suggested a cooling of the inversion
top together with a simultaneous heating of the air layers above and below.
This could be a possible mechanism for further inversion growth. In
addition, the temperature profile above the inversion is usually weakly
stable, whereas our simulations show a well mixed profile almost to the top
of the computational domain. This is caused by the application of the
periodic boundary conditions in the model simulations. A nudging of the mean
potential temperature gradient above the valley could reduce this problem,
but is not available in the model setup we used. Improving the
representation of both processes in the model is expected to result in a
growth of the inversion depth. Furthermore, the flow along the valley bottom
through the southern domain boundary might be overestimated in our
simulations. Even though this flow feature might to some extent also exist
in reality, due to the large water bodies further south in the Bergen Valley
causing convergence and hence draining of air out of the city centre, this
could reduce the potential for cold air pooling. Finally, the relatively low
spatial resolution for the simulation of the stably stratified ABL might
also negatively impact the representation of the temperature inversions in
the Bergen Valley.
However, our simulations produced ground-based inversions higher than the
55 m model level. This means that they should be able to give us
relevant information on the mean flow around the height of the AMS and
below, the levels most relevant for the dispersion of locally emitted air pollutants.
Wind fields at 55 m height and passive tracer
concentration 2 m above the surface. The top, middle, and bottom
panels show the results for wind speed scenario 1, 2, and 3, respectively.
All data are means over the last four output steps of the 12 h
simulation. Each output step is an average of 15 min. Colour
and wind-speed scales are the same in all panels. Darker colour means higher
tracer concentrations. The domain is cut off at the right boundary since the
topography here was mostly above 55 m. The wind vector in the
upper right corner indicates a wind speed of 5.2 m s-1. The
water–land interface is shown by the black solid line.
The interplay between the local and the larger-scale conditions
The range of selected fjord surface temperatures and geostrophic wind speeds
allows investigation of the interplay between the south-easterly down-valley
flow, triggered by the convection over the fjord, and the north-westerly
up-valley flow, forced by the flow above the valley, as well as of the effect of this interplay on the
dispersion of pollutants inside the valley. Figure 7 shows the wind field at
55 m height and the terrain following concentration of the surface-emitted passive tracer at 2 m above the ground, calculated the same way
as the 2 m temperature before. The runs with fjord temperatures of
2.5 ∘C or more show clear signs of flow convergence in the wind field
over the fjord and a distinct signature of prevailing down-valley flow. For
the leftmost panels with 0 ∘C fjord surface temperature, this
convergence line is pushed all the way towards the coastline, and the flow in
the exterior part of the Bergen Valley is even reversed towards an up-valley
flow. For case 9, the convergence line is masked by this up-valley flow at
the 55 m height level and is therefore only visible for the lower-level
wind fields. The results show a gradual movement of the convergence line
with decreasing fjord surface temperatures and increasing geostrophic wind
speeds inwards towards the city centre. This is caused by a weakening of the
convergence over the colder fjord. The overall flow pattern becomes more and
more dominated by the up-valley recirculation, especially for the scenarios
with higher wind speeds. It should, however, be mentioned that the outflow
out of the artificially generated channel at the northern border of the
computational domain seems to interact with the up-valley flow, enhancing it
and hence pushing the convergence line towards the coast. Up-valley flows
are rarely observed during temperature inversions in the valley, but
they are visible in our simulations. While the coldest fjord surface
temperatures considered here are also rarely observed, possibly explaining this
lack of observations of the up-valley flow, it could also be artificially
enhanced in our simulations by this domain boundary effect. This is,
however, a persistent feature of all simulations. It might therefore shift
the balance between the up-valley circulation, forced by the flow above the
valley, and the breeze circulation towards lower geostrophic wind speeds and
higher fjord temperatures, respectively, but is not expected to change the
conclusions on the existence of the balance itself.
The effect of this on the circulation inside the valley is summarized in
Fig. 8 in terms of the horizontal mean of the passive tracer concentration
2 m above the ground and the wind speed and direction at the 55 m
height level and 10 m above the ground. For each of the given wind-speed
scenarios, there is a combination of geostrophic winds and local breeze
circulations, that leads to a maximum in the stagnation in the exterior part
of the valley (area 1, Fig. 2). For the winds, this is visible as a plateau
for the wind speed in scenarios 2 and 3, and a turning of the wind direction
both at the 55 m height level and 10 m above the ground. Based on
this, it can be assumed that there would be a minimum in the local wind
speeds at intermediate fjord temperatures between 0 and
2.5 ∘C For scenario 1, with the lowest considered wind speeds,
this balancing is not yet reached for the 10 m wind, while the wind at
the 55 m height level is already rotated towards an up-valley flow. This
indicates, however, that the rotation of the winds at 10 m above the
surface would occur at fjord surface temperatures below 0 ∘C To
investigate this further we conducted a test simulation with a fjord surface
temperature of -2.5 ∘C. This scenario, however, led to an
unrealistic maximum in convergence over all other water bodies, except for
the fjord. While the absolute temperatures of the fjord and land surface are
not relevant, the relative temperature differences are. During the winter
the freshwater bodies are usually colder than the fjord. The constant
temperature of 0 ∘C over the freshwater bodies in our simulations
therefore causes an unrealistic circulation in the valley at fjord surface
temperatures below 0 ∘C.
Mean passive tracer concentration (blue) and wind speed at
55 m (red, dashed) and 10 m above the surface
(red, solid) for area 1 (a1) and 2 (a2), as indicated in Fig. 1 for the
three wind-speed scenarios as indicated in Table 1. The wind vectors indicate the mean
wind direction in the area at the 55 m height level (top panels) and
10 m above the ground (bottom panels). Vectors point into the
direction of the flow. All data are averaged over the last four output steps of
the 12 h simulations, corresponding to Fig. 7.
For the interior part of the valley (area 2, Fig. 2) a balance, like for the
exterior part of the valley, is not visible. The set of simulations with the
lowest fjord surface temperatures of 0 ∘C show both the lowest
wind speeds and maximum pollutant concentrations, and no reversal of the wind
direction is seen for any of the simulations at the height of the GFI AMS.
One reason for this distinctly different behaviour of the interior part of
the valley is the colder temperature there. The temperature difference
between the 2 m surface air temperatures over area 1 and 2, at least for
the lower fjord surface temperatures, is similar to the temperature
difference between the surface air temperatures over area 3 and 1 (see Table 1
for comparison). The different land surface temperatures therefore exert
an additional forcing on the air in the interior part of the valley.
Furthermore, as the large water body in the middle of the valley is warmer
than the surrounding land, it causes another centre of flow convergence
similar to the fjord. It enhances an up-valley flow over area 1, while it
weakens it for area 2. This effect is increasing with decreasing fjord
surface temperature due to the changes in land surface temperature between
the different model simulations. Especially for the cases with 0 ∘C
fjord surface temperature, however, the convergence over the water
body in the middle of the valley is most likely overestimated, since the
brackish water lake usually cools off much faster than the fjord, resulting
in a smaller temperature difference between the lakes and the surrounding
land. A thin layer of ice, as it is typical on this water body, utterly
reduces this effect.
The impact of the wind circulation on the dispersion of the passive tracer
is two-fold. Locally, more stagnant conditions in area 1 lead to an
increased accumulation of pollutants in this region. The concentration over
area 2 is in general higher than over area 1, except for case 7. This part
of the valley is more protected from the geostrophic winds than area 1. In
addition, the wind direction plays a more important role for area 1. An
up-valley flow automatically leads to a lower pollutant concentration there,
since air from the fjord, with considerably lower tracer concentrations,
will be transported inland.
Analysis of single road contribution
For many valleys the main pollutant emissions over land come from single transit
roads. The same is the case for the NO2 emissions in the Bergen Valley. The main emission source for traffic-emitted pollutants is the
transit road marked in Fig. 1. Emissions from a reduced area give insight
into the efficiency of the horizontal dispersion for the most relevant areas
that was obscured by the areal emissions assessed before. Therefore, we
repeated cases 9 to 11 with tracer emissions only from this road. The wind
fields, together with the passive tracer concentrations 2 m above the
surface, are shown in Fig. 9. We chose to use sc 3 for the wind speeds in
order to see the full range of conditions, from a down-valley flow to
maximum stagnation.
The up-valley flow in case 13 transports the passive tracer away from the
city centre. As the emissions from the street here are only scaling with the
area covered by the street, ignoring traffic density and pattern for the
emission factors, the large interchange road is causing the highest density
of emissions. However, the high tracer concentrations are not transported to
other places in the valley. They are rather caught by the convection from
the lake in the middle of the valley, which also serves as an effective barrier
for the tracer transport from this street. This is, to a lesser
degree, also visible for the other two cases. Emissions from area 2
contribute less to the pollution directly north of the lake. This indicates
that even relatively smaller interior water bodies could improve the urban
ventilation through driving convective and local-scale circulations.
For the other two cases, the mean down-valley flow transports the tracer
along the valley axis. While for case 14 there is still an accumulation
visible because of a relative stagnation, the down-valley flow is
sufficiently strong and keeps the passive tracer concentrations at a low
level over large parts of the city centre for case 15.
Same as Fig. 7 but for the cases with emissions only from the main
street inside the Bergen Valley. The colour coding for the concentration is
the same in all three panels, but different from Fig. 7. Shading is chosen to show the horizontal distribution of the pollutants and not the maximum
values over the street.
Summary and discussion
In this study, we run a set of large-eddy simulations with PALM to
evaluate the role of local circulations and their sensitivity to the
interplay between external large-scale forcing and local forcing due to
heterogeneities in the surface conditions. Specifically, we addressed the
sea–land temperature difference, the large-scale wind speed and the imposed
static stability of the lower atmosphere for typical conditions leading to
high air pollution in Bergen, Norway.
Urban settlements in mountainous terrain at high latitudes are especially
prone to adverse effects of temperature inversions on air quality. A
lack of solar heating in winter and topographic constraints on the low-level
atmospheric circulation can lead to an accumulation of air pollutants near
the surface. At the same time, large sea–land surface temperature
differences create local and meso-scale circulations in coastal cities,
which can partially compensate or even overwhelm the low-level circulations
forced by the large-scale atmospheric flow. Will
the circulations due to the surface heterogeneity impact the urban
ventilation and hence the air quality, and to what degree? The outcome depends not only on the
case-specific geographical features of the terrain and the specific
emissions in the city, but also on more universal physical mechanisms and
scalings. Such dependencies have already been extensively studied for flat
surfaces and more regular heterogeneities, e.g. those related to sea-ice
fractures or leads in the Arctic Ocean (Esau, 2007) or to
idealized surface plot patterns (van Heerwaarden et al., 2014).
Published studies on urbanized mountain valleys
often did not have sufficiently fine resolution to reproduce interactions
between the flow above and the local flow features inside the valley.
Moreover, they have not addressed these interactions in the stably
stratified boundary layers, when air pollution may be particularly high.
The conducted joint analysis of the NO2 concentration and
meteorological parameters for the Bergen Valley revealed an unexpected
build-up of air pollution under synoptic situations with significant
non-zero large-scale winds. There was a distinct maximum in the distribution
of observed NO2 concentrations against the ERA-Interim surface wind
speeds at around 3 m s-1. This behaviour is inconsistent with the
usually assumed monotonic concentration–wind-speed dependencies and the
faster depletion of valley cold air pools with increasing wind speeds
(e.g. Zängl, 2003; Lareau and Horel, 2015). The behaviour
of monotonically decreasing concentrations against wind speeds is, however,
recovered for the actually measured surface winds. This indicates that some
sufficiently strong local circulations emerge near the surface that are able
to counteract the large-scale winds, but are not resolved in ERA-Interim.
We therefore studied the physical mechanisms, dynamics, and sensitivities to
the surface features for those circulations with a set of 16 PALM scenarios
(Table 1) for the realistic terrain surrounding the city.
The simulations showed that, either the local circulation, forced by the
large-scale flow, or the locally forced breeze circulation could dominate
the dispersion of air pollution in the lower valley atmosphere. The maximum
pollutant trapping under prevailing inversion conditions was dependent on
the exact interplay of the three circulation features, i.e. large-scale
flow, topographic steering, and breeze circulation. The simulations with the
lowest fjord surface temperatures showed a mean up-valley flow dominated by
the topographically steered recirculation of the large-scale flow. The
pollution emitted from urban activities was represented by a passive tracer
emitted from the land surface over the urban area. The up-valley winds
therefore caused the advection of tracer-free air from over the fjord into
the city centre. The simulations with the higher fjord surface temperatures
showed a mean down-valley flow dominated by the breeze circulation. For the
highest fjord surface temperatures this lead to an efficient depletion of
the tracer emitted over the urban area. For the simulations with
intermediate temperatures, however, both circulation features balanced each
other, leading to local stagnation and an accumulation of the tracer.
Perhaps one of the most interesting implications of this study is the
possibility to analyse pollution scenarios for a specific area induced by
concrete emission sources. Such inverse diffusion problems are frequently
solved through a Green function method for regular domains
(Lin and Hildemann, 1996), but for irregular
domains, the direct simulations remain more computationally efficient. The
simulations demonstrated that the strongly localized concentrations are
rather sensitive to small-scale convective sources such as interior lakes.
These effectively work as barriers for the dispersion of pollutants near the
ground. An approach to predicting local air quality without resolving such
local features will not be able to simulate the pollutant dispersion pattern correctly.
References for all data sets used in this publication are
given in the text. All data sets are available on request from the stated sources.
The ERA-Interim and topographic data sets have limitations for availability to
the public due to restrictions on commercial use. Relevant contact persons are
noted in the acknowledgements.
The Supplement related to this article is available online at https://doi.org/10.5194/acp-17-7261-2017-supplement.
The authors declare that they have no conflict of interest.
Acknowledgements
This study was funded by the GC Rieber foundation. We thank Bergen
Municipality (Endre Leivestad) for provision of the topographic input data, the land–sea mask,
and the position of roads in the Bergen Valley. We further thank the
Norwegian Meteorological Institute and the Geophysical Institute at the
University of Bergen (Anak Bhandari) for providing wind measurement data, the European Centre
for Medium-Range Weather Forecasts for the ERA-Interim data, and the Norwegian
Institute for Air Research (Rita Våler) for the air quality measurement data. Sigfried Raasch
and Björn Maronga from the Institute of Meteorology and
Climatology at the University of Hanover were helpful in adapting and
applying the PALM code. CPU time was provided through the Norwegian Supercomputing
Project (NOTUR II grant number nn2993k).
Edited by: V.-M. Kerminen
Reviewed by: two anonymous referees
ReferencesAtlaskin, E. and Vihma, T.: Evaluation of NWP results for wintertime
nocturnal boundary-layer temperatures over Europe and Finland, Q. J. Roy.
Meteorol. Soc., 138, 1440–1451, 10.1002/qj.1885, 2012.Baklanov, A., Hänninen, O., Slørdal, L. H., Kukkonen, J., Bjergene, N.,
Fay, B., Finardi, S., Hoe, S. C., Jantunen, M., Karppinen, A., Rasmussen, A.,
Skouloudis, A., Sokhi, R. S., Sørensen, J. H., and Ødegaard, V.: Integrated
systems for forecasting urban meteorology, air pollution and population exposure,
Atmos. Chem. Phys., 7, 855–874, 10.5194/acp-7-855-2007, 2007.Baklanov, A. A., Grisogono, B., Bornstein, R., Mahrt, L., Zilitinkevich, S.
S., Taylor, P., Larsen, S. E., Rotach, M. W., and Fernando, H. J. S.: The
Nature, Theory, and Modeling of Atmospheric Planetary Boundary Layers, B.
Am. Meteorol. Soc., 92, 123–128, 10.1175/2010BAMS2797.1, 2011.Barlow, J. F.: Progress in observing and modelling the urban boundary layer,
Urban Clim., 10, 216–240, 10.1016/j.uclim.2014.03.011, 2014.Basu, S., Holtslag, A., Wiel, B., Moene, A., and Steeneveld, G.-J.: An
inconvenient “truth” about using sensible heat flux as a surface boundary
condition in models under stably stratified regimes, Acta Geophys., 56,
88–99, 10.2478/s11600-007-0038-y, 2008.Beare, R. J., Macvean, M. K., Holtslag, A. A. M., Cuxart, J., Esau, I.,
Golaz, J.-C., Jimenez, M. A., Khairoutdinov, M., Kosovic, B., Lewellen, D.,
Lund, T. S., Lundquist, J. K., Mccabe, A., Moene, A. F., Noh, Y., Raasch, S.,
and Sullivan, P.: An Intercomparison of Large-Eddy Simulations of the Stable
Boundary Layer, Bound.-Lay. Meteorol., 118, 247–272, 10.1007/s10546-004-2820-6, 2006.Berge, E., Walker, S.-E., Sorteberg, A., Lenkopane, M., Eastwood, S.,
Jablonska, H. I., and Ødegaard Køltzow, M.: A Real-Time operational
Forecast Model for Meteorology and Air Quality during Peak Air Pollution
Episodes in Oslo, Norway, Water Air Soil Pollut. Focus, 2, 745–757,
10.1023/A:1021378005618, 2002.Bergot, T., Escobar, J., and Masson, V.: Effect of small-scale surface
heterogeneities and buildings on radiation fog: Large-eddy simulation study
at Paris-Charles de Gaulle airport, Q. J. Roy. Meteorol. Soc., 141,
285–298, 10.1002/qj.2358, 2015.Brümmer, B. and Schultze, M.: Analysis of a 7-year low-level temperature
inversion data set measured at the 280 m high Hamburg weather mast,
Meteorol. Z., 24, 481–494, 10.1127/metz/2015/0669, 2015.Chen, F., Kusaka, H., Bornstein, R., Ching, J., Grimmond, C. S. B.,
Grossman-Clarke, S., Loridan, T., Manning, K. W., Martilli, A., Miao, S.,
Sailor, D., Salamanca, F. P., Taha, H., Tewari, M., Wang, X., Wyszogrodzki,
A. A., and Zhang, C.: The integrated WRF/urban modelling system: Development,
evaluation, and applications to urban environmental problems, Int. J.
Climatol., 31, 273–288, 10.1002/joc.2158, 2011.Cheng, W. C., Liu, C.-H., and Leung, D. Y. C.: Large-eddy simulation of
turbulent transport in urban street canyons in different thermal
stabilities, in: The Fifth International Symposium on Computational Wind
Engineering (CWE2010), Chapel Hill, North Carolina, USA, available at:
ftp://ftp.atdd.noaa.gov/pub/cwe2010/Files/Papers/064_Liu.pdf
(last access: 12 June 2017), 2010.Crosman, E. T. and Horel, J. D.: Sea and Lake Breezes: A Review of Numerical
Studies, Bound.-Lay. Meteorol., 137, 1–29, 10.1007/s10546-010-9517-9, 2010.Crosman, E. T. and Horel, J. D.: Winter Lake Breezes near the Great Salt
Lake, Bound.-Lay. Meteorol., 159, 439–464, 10.1007/s10546-015-0117-6, 2016.Deardorff, J. W.: Stratocumulus-capped mixed layers derived from a
three-dimensional model, Bound.-Lay. Meteorol., 18, 495–527, 10.1007/BF00119502, 1980.Dee, D. P., Uppala, S. M., Simmons, A. J., Berrisford, P., Poli, P.,
Kobayashi, S., Andrae, U., Balmaseda, M. A., Balsamo, G., Bauer, P.,
Bechtold, P., Beljaars, A. C. M., van de Berg, L., Bidlot, J., Bormann, N.,
Delsol, C., Dragani, R., Fuentes, M., Geer, A. J., Haimberger, L., Healy, S.
B., Hersbach, H., Hólm, E. V., Isaksen, L., Kållberg, P.,
Köhler, M., Matricardi, M., McNally, A. P., Monge-Sanz, B. M.,
Morcrette, J.-J., Park, B.-K., Peubey, C., de Rosnay, P., Tavolato, C.,
Thépaut, J.-N., and Vitart, F.: The ERA-Interim reanalysis: configuration
and performance of the data assimilation system, Q. J. Roy. Meteorol. Soc.,
137, 553–597, 10.1002/qj.828, 2011.ECMWF: MARS, the ECMWF archive, available at:
https://software.ecmwf.int/wiki/display/UDOC/MARS+user+documentation,
last access: 30 September 2016.
Esau, I.: A Health Damage Pattern Due to Street-Level Pollution in the
Central Paris Area Estimated With a Turbulence-Resolving Model, in:
Sustainable Environmental Design in Architecture, vol. 56, edited by:
Rassia, S. T. and Pardalos, P. M., Springer, New York, NY, 307–324, 2012.Esau, I. N.: Amplification of turbulent exchange over wide Arctic leads:
Large-eddy simulation study, J. Geophys. Res., 112, D08109, 10.1029/2006JD007225, 2007.Fay, B. and Neunhäuserer, L.: Evaluation of high-resolution forecasts with
the non-hydrostaticnumerical weather prediction model Lokalmodell for urban air
pollutionepisodes in Helsinki, Oslo and Valencia, Atmos. Chem. Phys., 6,
2107–2128, 10.5194/acp-6-2107-2006, 2006.Fernando, H. J. S.: Fluid Dynamics of Urban Atmospheres in Complex Terrain,
Annu. Rev. Fluid Mech., 42, 365–389, 10.1146/annurev-fluid-121108-145459, 2010.Fernando, H. J. S. and Weil, J. C.: Whither the stable boundary layer?,
B. Am. Meteorol. Soc., 91, 1475–1484, 10.1175/2010BAMS2770.1, 2010.Gariazzo, C., Papaleo, V., Pelliccioni, A., Calori, G., Radice, P., and
Tinarelli, G.: Application of a Lagrangian particle model to assess the
impact of harbour, industrial and urban activities on air quality in the
Taranto area, Italy, Atmos. Environ., 41, 6432–6444, 10.1016/j.atmosenv.2007.06.005, 2007.Glazunov, A., Rannik, Ü., Stepanenko, V., Lykosov, V., Auvinen, M., Vesala,
T., and Mammarella, I.: Large-eddy simulation and stochastic modeling of
Lagrangian particles for footprint determination in the stable boundary layer,
Geosci. Model Dev., 9, 2925–2949, 10.5194/gmd-9-2925-2016, 2016.Grange, S. K., Salmond, J. A., Trompetter, W. J., Davy, P. K., and Ancelet, T.:
Effect of atmospheric stability on the impact of domestic wood combustion to
air quality of a small urban township in winter, Atmos. Environ., 70, 28–38,
10.1016/j.atmosenv.2012.12.047, 2013.Hoch, S. W., Whiteman, C. D., and Mayer, B.: A Systematic Study of Longwave
Radiative Heating and Cooling within Valleys and Basins Using a Three-Dimensional
Radiative Transfer Model, J. Appl. Meteorol. Clim., 50, 2473–2489,
10.1175/JAMC-D-11-083.1, 2011.Holmer, B., Haeger-eugensson, M., Holmer, B., and Haeger-eugensson, M.:
Winter land breeze in a high latitude complex coastal area, Phys. Geogr.,
20, 152–172, 10.1080/02723646.1999.10642674, 1999.Hughes, J. K., Ross, A. N., Vosper, S. B., Lock, A. P., and Jemmett-Smith, B.
C.: Assessment of valley cold pools and clouds in a very high-resolution
numerical weather prediction model, Geosci. Model Dev., 8, 3105–3117,
10.5194/gmd-8-3105-2015, 2015.Junk, J., Helbig, A., and Lüers, J.: Urban climate and air quality in
Trier Germany, Int. J. Biometeorol., 47, 230–238, 10.1007/s00484-003-0162-6, 2003.
Keck, M., Raasch, S., Letzel, M. O., and Ng, E.: First Results of High
Resolution Large-Eddy Simulations of the Atmospheric Boundary Layer, J. Heat
Isl. Inst. Int., 9, 39–43, 2014.Lappalainen, H. K., Kerminen, V.-M., Petäjä, T., Kurten, T., Baklanov,
A., Shvidenko, A., Bäck, J., Vihma, T., Alekseychik, P., Andreae, M. O.,
Arnold, S. R., Arshinov, M., Asmi, E., Belan, B., Bobylev, L., Chalov, S.,
Cheng, Y., Chubarova, N., de Leeuw, G., Ding, A., Dobrolyubov, S., Dubtsov, S.,
Dyukarev, E., Elansky, N., Eleftheriadis, K., Esau, I., Filatov, N., Flint, M.,
Fu, C., Glezer, O., Gliko, A., Heimann, M., Holtslag, A. A. M., Hõrrak, U.,
Janhunen, J., Juhola, S., Järvi, L., Järvinen, H., Kanukhina, A.,
Konstantinov, P., Kotlyakov, V., Kieloaho, A.-J., Komarov, A. S., Kujansuu, J.,
Kukkonen, I., Duplissy, E.-M., Laaksonen, A., Laurila, T., Lihavainen, H.,
Lisitzin, A., Mahura, A., Makshtas, A., Mareev, E., Mazon, S., Matishov, D.,
Melnikov, V., Mikhailov, E., Moisseev, D., Nigmatulin, R., Noe, S. M., Ojala,
A., Pihlatie, M., Popovicheva, O., Pumpanen, J., Regerand, T., Repina, I.,
Shcherbinin, A., Shevchenko, V., Sipilä, M., Skorokhod, A., Spracklen,
D. V., Su, H., Subetto, D. A., Sun, J., Terzhevik, A. Y., Timofeyev, Y.,
Troitskaya, Y., Tynkkynen, V.-P., Kharuk, V. I., Zaytseva, N., Zhang, J.,
Viisanen, Y., Vesala, T., Hari, P., Hansson, H. C., Matvienko, G. G., Kasimov,
N. S., Guo, H., Bondur, V., Zilitinkevich, S., and Kulmala, M.: Pan-Eurasian
Experiment (PEEX): towards a holistic understanding of the feedbacks and
interactions in the land–atmosphere–ocean–society continuum in the northern
Eurasian region, Atmos. Chem. Phys., 16, 14421–14461, 10.5194/acp-16-14421-2016, 2016.Lareau, N. P. and Horel, J. D.: Dynamically Induced Displacements of a
Persistent Cold-Air Pool, Bound.-Lay. Meteorol., 154, 291–316,
10.1007/s10546-014-9968-5, 2015.Lareau, N. P., Crosman, E., Whiteman, C. D., Horel, J. D., Hoch, S. W.,
Brown, W. O. J., and Horst, T. W.: The Persistent Cold-Air Pool Study, B.
Am. Meteorol. Soc., 94, 51–63, 10.1175/BAMS-D-11-00255.1, 2013.Letzel, M. O., Krane, M., and Raasch, S.: High resolution urban large-eddy
simulation studies from street canyon to neighbourhood scale, Atmos.
Environ., 42, 8770–8784, 10.1016/j.atmosenv.2008.08.001, 2008.Letzel, M. O., Helmke, C., Ng, E., An, X., Lai, A., and Raasch, S.: LES case
study on pedestrian level ventilation in two neighbourhoods in Hong Kong,
Meteorol. Z., 21, 575–589, 10.1127/0941-2948/2012/0356, 2012.Lin, J.-S. and Hildemann, L. M.: Analytical solutions of the atmospheric
diffusion equation with multiple sources and height-dependent wind speed and
eddy diffusivities, Atmos. Environ., 30, 239–254, 10.1016/1352-2310(95)00287-9, 1996.Lo, J. C. F., Lau, A. K. H., Fung, J. C. H., and Chen, F.: Investigation of
enhanced cross-city transport and trapping of air pollutants by coastal and
urban land-sea breeze circulations, J. Geophys. Res., 111, D14104,
10.1029/2005JD006837, 2006.Lüpkes, C., Gryanik, V. M., Witha, B., Gryschka, M., Raasch, S., and
Gollnik, T.: Modeling convection over arctic leads with LES and a
non-eddy-resolving microscale model, J. Geophys. Res., 113, C09028,
10.1029/2007JC004099, 2008.Mahrt, L.: Stably Stratified Atmospheric Boundary Layers, Annu. Rev. Fluid
Mech., 46, 23–45, 10.1146/annurev-fluid-010313-141354, 2014.Maronga, B., Gryschka, M., Heinze, R., Hoffmann, F., Kanani-Sühring, F.,
Keck, M., Ketelsen, K., Letzel, M. O., Sühring, M., and Raasch, S.: The
Parallelized Large-Eddy Simulation Model (PALM) version 4.0 for atmospheric
and oceanic flows: model formulation, recent developments, and future perspectives,
Geosci. Model Dev., 8, 2515–2551, 10.5194/gmd-8-2515-2015, 2015.Mason, P. J. and Derbyshire, S. H.: Large-Eddy Simulation of the
stably-stratified atmospheric boundary layer, Bound.-Lay. Meteorol.,
53, 117–162, 10.1007/BF00122467, 1990.Moeng, C.-H. and Wyngaard, J. C.: Spectral analysis of large-eddy
simulations of the convective boundary layer, J. Atmos. Sci., 45,
3573–3587, 10.1175/1520-0469(1988)045<3573:SAOLES>2.0.CO;2, 1988.Mohamad, A. A. and Viskanta, R.: Flow and heat transfer in a lid-driven
cavity filled with a stably stratified fluid, Appl. Math. Model.,
19, 465–472, 10.1016/0307-904X(95)00030-N, 1995.Nakayama, H., Takemi, T., and Nagai, H.: Large-eddy simulation of urban
boundary-layer flows by generating turbulent inflows from mesoscale
meteorological simulations, Atmos. Sci. Lett., 13, 180–186, 10.1002/asl.377, 2012.Nordbo, A., Järvi, L., and Vesala, T.: Revised eddy covariance flux
calculation methodologies – effect on urban energy balance, Tellus B, 64,
1–20, 10.3402/tellusb.v64i0.18184, 2012.Norwegian Institute for Air Research: Luftkvaliteten i Norge,
available at: http://admin.luftkvalitet.info, last access: 30 September 2016.Norwegian Meteorological Institute: eKlima, the climate database of the
Norwegian Meteorological Institute, available at:
http://sharki.oslo.dnmi.no/portal/page?_pageid=73,39035,73_39101&_dad=portal&_schema=PORTAL,
last access: 30 September 2016.Park, S.-B., Baik, J.-J., and Lee, S.-H.: Impacts of Mesoscale Wind on
Turbulent Flow and Ventilation in a Densely Built-up Urban Area, J. Appl.
Meteorol. Clim., 54, 811–824, 10.1175/JAMC-D-14-0044.1, 2015a.Park, S.-B., Baik, J.-J., and Han, B.-S.: Large-eddy simulation of turbulent
flow in a densely built-up urban area, Environ. Fluid Mech., 15,
235–250, 10.1007/s10652-013-9306-3, 2015b.Patnaik, G. and Boris, J.: FAST3D-CT: an LES model for urban aerodynamics,
in: The Fifth International Symposium on Computational Wind Engineering (CWE2010),
Chapel Hill, North Carolina, USA, available at:
ftp://ftp.atdd.noaa.gov/pub/cwe2010/Files/Papers/394_Patnaik.pdf
(last access: 12 June 2017), 2010.Raasch, S. and Schröter, M.: PALM – A large-eddy simulation model
performing on massively parallel computers, Meteorol. Z., 10,
363–372, 10.1127/0941-2948/2001/0010-0363, 2001.Reeves, H. D. and Stensrud, D. J.: Synoptic-Scale Flow and Valley Cold Pool
Evolution in the Western United States, Weather Forecast., 24,
1625–1643, 10.1175/2009WAF2222234.1, 2009.Rimetz-Planchon, J., Perdrix, E., Sobanska, S., and Brémard, C.: PM10 air
quality variations in an urbanized and industrialized harbor, Atmos.
Environ., 42, 7274–7283, 10.1016/j.atmosenv.2008.07.005, 2008.Rotach, M. W., Calanca, P., Graziani, G., Gurtz, J., Steyn, D. G., Vogt, R.,
Andretta, M., Christen, A., Cieslik, S., Connolly, R., De Wekker, S. F. J.,
Galmarini, S., Kadygrov, E. N., Kadygrov, V., Miller, E., Neininger, B.,
Rucker, M., Van Gorsel, E., Weber, H., Weiss, A., and Zappa, M.: Turbulence
Structure and Exchange Processes in an Alpine Valley: The Riviera Project,
B. Am. Meteorol. Soc., 85, 1367–1385, 10.1175/BAMS-85-9-1367, 2004.Saiki, E. M., Moeng, C., and Sullivan, P. P.: Large Eddy Simulation of the
stably stratified planetary boundary layer, Bound.-Lay. Meteorol., 95,
1–30, 10.1023/A:1002428223156, 2000.Schalkwijk, J., Jonker, H. J. J., Siebesma, A. P., and Van Meijgaard, E.:
Weather Forecasting Using GPU-Based Large-Eddy Simulations, B. Am.
Meteorol. Soc., 96, 715–723, 10.1175/BAMS-D-14-00114.1, 2015.Schicker, I. and Seibert, P.: Simulation of the meteorological conditions
during a winter smog episode in the Inn Valley, Meteorol. Atmos. Phys.,
103, 211–222, 10.1007/s00703-008-0346-z, 2009.Sheridan, P. F., Vosper, S. B., and Brown, A. R.: Characteristics of cold
pools observed in narrow valleys and dependence on external conditions, Q.
J. Roy. Meteorol. Soc., 140, 715–728, 10.1002/qj.2159, 2014.
Steyn, D. G., DeWekker, S. F. J., Kossmann, M., and Martilli, A.: Boundary
Layers and Air Quality in Mountainous Terrain, in: Mountain Weather Research
and Forecasting, edited by: Chow, F. K., De Wekker, S. F. J., and Snyder, B.
J., Springer Netherlands, Dordrecht, 261–289, 2013.Tomas, J. M., Pourquie, M. J. B. M., and Jonker, H. J. J.: Stable
Stratification Effects on Flow and Pollutant Dispersion in Boundary Layers
Entering a Generic Urban Environment, Bound.-Lay. Meteorol., 159, 221–239,
10.1007/s10546-015-0124-7, 2016.van Heerwaarden, C. C., Mellado, J. P., and de Lozar, A.: Scaling Laws for
the Heterogeneously Heated Free Convective Boundary Layer, J. Atmos. Sci.,
71, 3975–4000, 10.1175/JAS-D-13-0383.1, 2014.Whiteman, C. D. and Doran, J. C.: The Relationship between Overlying
Synoptic-Scale Flows and Winds within a Valley, J. Appl. Meteorol., 32,
1669–1682, 10.1175/1520-0450(1993)032<1669:TRBOSS>2.0.CO;2, 1993.Wicker, L. J. and Skamarock, W. C.: Time-Splitting Methods for Elastic
Models Using Forward Time Schemes, Mon. Weather Rev., 130, 2088–2097,
10.1175/1520-0493(2002)130<2088:TSMFEM>2.0.CO;2, 2002.Wolf, T. and Esau, I.: A proxy for air quality hazards under present and
future climate conditions in Bergen, Norway, Urban Clim., 10, 801–814,
10.1016/j.uclim.2014.10.006, 2014.Wolf, T., Esau, I., and Reuder, J.: Analysis of the vertical temperature
structure in the Bergen valley, Norway, and its connection to pollution
episodes, J. Geophys. Res.-Atmos., 119, 10645–10662, 10.1002/2014JD022085, 2014.Wyszogrodzki, A. A., Miao, S., and Chen, F.: Evaluation of the coupling
between mesoscale-WRF and LES-EULAG models for simulating fine-scale urban
dispersion, Atmos. Res., 118, 324–345, 10.1016/j.atmosres.2012.07.023, 2012.Xie, Z. T., Hayden, P., and Wood, C. R.: Large-eddy simulation of
approaching-flow stratification on dispersion over arrays of buildings,
Atmos. Environ., 71, 64–74, 10.1016/j.atmosenv.2013.01.054, 2013.
Zängl, G.: The impact of upstream blocking, drainage flow and the
geotrophic pressure gradient on the persistence of cold-air pools, Q. J. Roy.
Meteorol. Soc., 129, 117–137, 10.1256/qj.02.99, 2003.Zhang, J. P., Zhu, T., Zhang, Q. H., Li, C. C., Shu, H. L., Ying, Y., Dai, Z.
P., Wang, X., Liu, X. Y., Liang, A. M., Shen, H. X., and Yi, B. Q.: The impact
of circulation patterns on regional transport pathways and air quality over
Beijing and its surroundings, Atmos. Chem. Phys., 12, 5031–5053, 10.5194/acp-12-5031-2012, 2012.Zhong, S., Whiteman, C. D., Bian, X., Shaw, W. J., and Hubbe, J. M.:
Meteorological Processes Affecting the Evolution of a Wintertime Cold Air
Pool in the Columbia Basin, Mon. Weather Rev., 129, 2600–2613,
10.1175/1520-0493(2001)129<2600:MPATEO>2.0.CO;2, 2001.
Zilitinkevich, B. S., Kulmala, M., and Esau, I.: Megacities – Refining
Models to Client Environment, WMO Bull., 64, 20–22, 2015.