ArticlePDF Available

Modelling the ground heat flux of an urban area using remote sensing data


Abstract and Figures

During the Basel Urban Boundary Layer Experiment (BUBBLE) conducted in 2002, micrometeorological in-situ data were collected for different sites using a variety of instruments. This provides a unique data set for urban climate studies. Nevertheless, the spatial distribution of energy and heat fluxes can only be taken into account with remote sensing methods or numerical models. Therefore, multiple satellite images from different platforms (NOAA-AVHRR, MODIS and LANDSAT ETM+) were acquired, processed and analysed. In addition, a high resolution digital elevation model (DEM) and a 1 m resolution digital surface model (DSM) of a large part of the city of Basel was utilized. This paper focuses on the calculation and modelling of the ground (or storage) heat flux density using remotely sensed data combined with in-situ measurements using three different approaches. First, an empirical regression function was generated to estimate the storage heat flux from NDVI values second approach used the Objective Hysteresis Model (OHM) which is often used for in-situ measurements. The last method used information of the geometric parameters of urban street canyons, computed from the high resolution digital urban surface model. Modelled and measured data are found to be in agreement within ±30 Wm−2 and result in a coefficient of determination (R2) of 0.95.
Content may be subject to copyright.
Theor. Appl. Climatol. 90, 185–199 (2007)
DOI 10.1007/s00704-006-0279-8
Printed in The Netherlands
Department of Environmental Sciences, Institute of Meteorology, Climatology and Remote Sensing,
University of Basel, Basel, Switzerland
Modelling the ground heat flux of an urban area
using remote sensing data
G. Rigo and E. Parlow
With 6 Figures
Received February 6, 2006; revised August 31, 2006; accepted September 10, 2006
Published online February 21, 2007 # Springer-Verlag 2007
During the Basel Urban Boundary Layer Experiment
(BUBBLE) conducted in 2002, micrometeorological in-situ
data were collected for different sites using a variety of in-
struments. This provides a unique data set for urban climate
studies. Nevertheless, the spatial distribution of energy and
heat fluxes can only be taken into account with remote
sensing methods or numerical models. Therefore, multiple
satellite images from different platforms (NOAA-AVHRR,
MODIS and LANDSAT ETMþ) were acquired, processed
and analysed. In addition, a high resolution digital elevation
model (DEM) and a 1 m resolution digital surface model
(DSM) of a large part of the city of Basel was utilized.
This paper focuses on the calculation and modelling of the
ground (or storage) heat flux density using remotely sensed
data combined with in-situ measurements using three dif-
ferent approaches. First, an empirical regression function
was generated to estimate the storage heat flux from NDVI
values second approach used the Objective Hysteresis Model
(OHM) which is often used for in-situ measurements. The
last method used information of the geometric parameters
of urban street canyons, computed from the high resolution
digital urban surface model.
Modelled and measured data are found to be in agreement
within 30 Wm
and result in a coefcient of determina-
tion (R
1. Introduction
During summer 2002 field measurements of
the Basel Urban Boundary Layer Experiment
(BUBBLE) were conducted, covering a wide
range of micrometeorological techniques and
a unique set of urban climatological variables
(Rotach et al., 2005). A total of seven flux towers
were used to cover urban, suburban, and rural
surface conditions. Vertical profiles of radiation
and heat fluxes in a street canyon were measured
up to 38 m above ground, 22.5 times the height
of buildings. These data provide an excellent
opportunity to study the local heat flux densities
and their temporal modification during the inten-
sive operation period (IOP). The objective of this
experiment was to analyse the spatial distribution
over a large area of the city of Basel and its sur-
roundings. Satellite data from various platforms
were investigated (NOAA-AVHRR, MODIS and
LANDSAT ETMþ ) offering different temporal
and spatial resolutions. In contrast to rural sites,
the ground heat flux plays an important role for
studies in urban environments. Under rural con-
ditions the ground heat flux can be estimated to
be roughly 10% of the net radiation during day-
time conditions. In urban environments this value
increases to 3040% of the net radiation making
ground heat flux a major factor in urban heat
flux studies (Christen and Vogt, 2004). Maximum
values of up to 58% of net radiation are report-
ed from Mexico City (Oke et al., 1999). This
intensive storage of energy during the day in the
urban structure is a basic requirement for the de-
velopment of an urban heat island during night-
time. The ground heat flux during the night fully
compensates the negative net radiation of urban
areas and in addition, maintains a small sensible
heat flux directed into the atmosphere maintain-
ing urban air temperatures several degrees higher
than those in rural regions (Parlow, 2003).
This paper concentrates on computing the
ground heat flux density from remotely sensed
data. Different approaches have been used in the
past to calculate the ground heat flux density. In
most cases it has been computed for rural areas
by deriving an empirical function to estimate
ground heat flux by using NDVI values from sat-
ellite data (Kustas and Daughtry, 1990) or by
using NDVI values as input for more complex nu-
merical models (Baastiansen et al., 1997; Kustas
et al., 2004). A completely different approach
uses urban street canyon geometry elements
such as complete aspect ratio, sky-view-factor or
width-to-height-ratio of streets (Voogt and Oke,
1997) to parametrise the complex surface condi-
tions for radiation and heat fluxes. Grimmond
and Oke (1991, 1999) introduced the Objective
Hysteresis Model (OHM) to compute the ground
heat flux of an urban surface based on micro-
meteorological measurements including the hys-
teresis effect between ground heat flux and net
radiation. In this paper the OHM approach is
applied for a spatial analysis using satellite de-
rived data sets.
2. Study area and data
The city of Basel is located in NW Switzerland
in Central Europe at the southern end of the
Upper Rhine Valley, at the confluence of the
French, German and Swiss borders. The city has
about 200,000 inhabitants, with a total of 350,000
in the immediate vicinity. In 2002, the BUBBLE
project (Rotach et al., 2005) collected data from
eight micrometeorological sites set up in and
around the city, these data however, only rep-
resent isolated measurements. The locations of
these sites, Sperrstrasse (U1), Spalenring (U2),
Messe (U3) Allschwil (S1), Lange Erlen (R3),
Village-Neuf (R2) and Grenzach (R1), are shown
in Fig. 1. All the ‘U’ sites are urban, the ‘‘S’’ are
suburban and ‘‘R’’ are rural sites. The urban sites
at Sperrstrasse and Spalenring were in typical
urban street canyons, and the Allschwil site
was situated in a suburban residential area with
planted gardens. The Grenzach and Lange Erlen
rural sites were in grassland regions whereas the
Village-Neuf site was located over agricultural
fields. For an extended overview of the microme-
teorological instrumentation of these sites see
Christen and Vogt (2004) and Rotach et al. (2005).
To form an idea of the spatial distribution of
heat fluxes during the BUBBLE IOP, a set of
Fig. 1. Map of the measure-
ment sites during BUBBLE-IOP
based on a land use classifica-
tion derived from an ASTER=
LANDSAT ETMþ fused dataset
from June 12, 2001. Geographic
grid is UTM, Zone 32N, WGS-84
186 G. Rigo and E. Parlow
different satellite data from LANDSAT ETMþ,
MODIS, NOAA-AVHRR and ASTER were used.
Satellite overpasses integrated in this study are
reported in Table 1.
Surface geometry and urban morphometry
were integrated in the analysis through the offi-
cial digital elevation model (DEM) with a 25 m
resolution from Swiss Topo. In addition, a digital
city surface model with a 1 m resolution from the
local urban planning authorities in Basel, offer-
ing information on building heights and roof ge-
ometry was used. A land use classification (Fig. 1)
has been calculated from a fused data set from
ASTER and LANDSAT ETMþ (both overpasses
on June 12, 2001) with a 30 minute time differ-
ence between the two overpasses.
3. Methods
3.1 Radiation and energy balance
Net radiation data are needed to describe the heat
fluxes of an urban or rural surface. This is the sum
of all incoming and outgoing radiation fluxes and
thus a key factor for the energy available for heat
fluxes. If net radiation is positive, as is mostly
the case during the day, energy can be transferred
into turbulent heat fluxes (sensible and latent)
and=or into the ground heat flux. If net radiation
is negative, as is usual at night, it has to be com-
pensated for by the heat fluxes. The net radiation
equation can take the form:
þ Q
þ Q
¼ Q
net all wave radiation
shortwave upward radiation
shortwave downward solar radiation
longwave upward terrestrial radiation
longwave downward atmospheric radiation.
The heat flux balance can then be described by
the following equation:
þ Q
þ Q
þ Q
þ Q
¼ 0 ð2Þ
ground or storage heat flux density
latent heat flux density
convective sensible heat flux density
anthropogenic heat flux density.
The anthropogenic heat flux density Q
is very
small in European cities like Basel and typical-
ly ranges from 5 Wm
in suburban areas, to
20 Wm
in the city centre (Christen and Vogt,
2004). For this analysis the anthropogenic heat flux
is therefore not considered. Equation 2 is only valid
in an idealized environment, where advection and
the storage in the layer between the observation
and surface can be neglected. The radiation values
and energy flux densities were taken by sonic an-
emometers from the highest available measure-
ment point on the flux towers, 32 m above street
level, (for further information see: http:==pages.
htm). In addition, the satellite data have a spatial
resolution of 30 m, which renders it impractical for
modelling the ground heat flux inside a canyon.
The computation of net radiation was perform-
ed in several stages. Solar irradiance (Q
) (short-
wave downward radiation) was modelled with
SWIM (Shortwave Irradiance Model) (Parlow,
1996b), with a digital elevation model (DEM)
as input data. SWIM calculates the total short-
wave downward radiation Q
for a clear sky
day depending on geographic location, slope, ele-
vation and aspect (e.g., different standard atmo-
spheres for summer and winter conditions, and as
a function of latitude as well as altitudinal effects
for adjustment of percentages of diffuse and di-
rect shortwave downward radiation).
Table 1. Analysed satellite data
Satellite Time
Date Additional
10:10 08.07.2002
AVHRR 14 06:47 08.07.2002
AVHRR 16 day 11:10 08.07.2002
AVHRR 16 night 2:13 08.07.2002
MODIS day 11:20 08.07.2002
MODIS night 22:20 08.07.2002
AVHRR 15 7:23 25.06.2002
MODIS 11:40 25.06.2002
AVHRR 16 12:45 25.06.2002
AVHRR 16 12:34 26.06.2002
MODIS 10:40 26.06.2002
MODIS 21:50 26.06.2002
10:10 12.06.2001 Used for land
use classification
ASTER 10:40 12.06.2001 Used for land
use classification
Modelling the ground heat flux of an urban area 187
Model runs were computed for different dates
and times relating to the satellite overpasses used
in this study. Longwave upward radiation (Q
was calculated from the thermal infrared channels
of different satellite platforms after correcting
for atmospheric influences. Information regard-
ing the accurate assessment of thermal infrared
data analysis can be found in Rigo et al. (2006).
The albedo, which leads to shortwave upward
radiation (Q
), was derived using linear regres-
sion between in-situ measurements and a syn-
thetic panchromatic channel computed from the
visible and near infrared channels of LANDSAT
ETMþ, according to Parlow (1996a). Since long-
wave downward radiation (Q
) is uniform in a
region of the size in this study, it was assumed to
be the average of all in-situ measurements at
the time of a satellite overpass. From these data
fields the spatially distributed net radiation was
calculated. A comparison with in-situ measure-
ments of net radiation gave a RMS error of
25 Wm
, which is less than 5%.
3.2 In-situ data of ground heat flux density
At the rural sites, the ground heat flux density
was measured directly by means of four soil
thermistors and three heat flux plates inserted at a
depth of between 3 and 5 cm. The data are cor-
rected for flux density divergence in the soil layer
above the plates using measured soil tempera-
tures. The average daily value of the ground heat
flux density for rural areas is about 60 Wm
which corresponds to 1015% of the net radia-
tion (Christen and Vogt, 2004). However, even
heat flux plates are prone to measurement errors
which are associated with various problems, e.g.
poor plate contact with the substrate. Twine et al.
(2000) note a probable error of 15% in soil heat
flux measurements with heat flux plates even with
careful calibration. Instrument accuracy may also
cause an error of up to 20% according to Weber
(2006) who made some measurements in a bal-
last layer. Spatial variability of the ground heat
flux density in rural environments is low.
The Q
of an urban surface incorporates
all storage into artificial surfaces (streets, buil-
dings), into urban vegetation and into the ground
(Grimmond and Oke, 1999). Measuring the
ground heat flux over an urban surface is even
more difficult than over rural areas (Weber, 2006).
As discussed by Weber (2006) for urban, and
Twine et al. (2000) for rural environments,
ground heat flux is mostly estimated as residuum
from eddy-covariance measurements, because the
use of heat flux plates in the urban environment
is still at an experimental stage. The errors for
different methods of measuring the ground heat
flux are described by Weber (2006) who found
relative deviations of up to 31% for heat flux
plate measurements and 38% for the residual
term from the eddy-covariance method. Data in
this study were taken from sonic-anemometers
mounted above the urban canopy to simulate the
resolution of the satellite image. Since the high-
est spatial satellite data resolution available is
30 m no useful data were available from inside
the canyon. Another difficulty would be the urban
anisotropy (Voogt and Oke, 1997), which would
add some specific problems to high resolution
data. The instruments were mounted about 30 m
above ground level at U1 and U2 (average buil-
ding height was 14 m and at S1 about 15 m above
the ground with average building height of 7.5 m.
Data at U3 were derived from 2 m above the
surface of a multi-storey car park.
Most direct energy balance measurements of
flux densities over natural and agricultural sur-
faces show that Eq. 2 does not balance and
has a closure gap of approximately 20% of Q
(Bernhofer and Vogt, 2000; Oncley et al., 2005;
Wilson et al., 2002). For the measurement sites
R1 and R2, the average daytime closure gap was
17 and 18%, respectively, while during the night-
time it reached 30%. It is obvious that whenever
is determined as a residual term from the
eddy-covariance method (sites U1, U2, S1) there
is no closure gap. Therefore, any Q
as a residual term must be interpreted as an upper
limit. More information can also be found in
Christen and Vogt (2004). Both in-situ measure-
ment methods (heat flux plate and resid-
uum estimation) are therefore used and are the
best available methods, although the residuum
method is less accurate as described by Weber
3.3 Ground heat flux density from remotely
sensed data
For the calculation of the ground heat flux density,
three approaches were used and compared with
188 G. Rigo and E. Parlow
one hour averaged in-situ Q
measurements for
the different sites. It is important to note that the
parameters were derived from the in-situ sites as
averages over the whole IOP period and are not
specific values for July 8, 2002. In addition, the
applied factors are averages of the different sites
(see Table 2) and literature values.
The remotely sensed data are used to model
the ground heat flux Q
and are considered
independent due to the fact that the complete
aspect ratio
, the net all radiation Q
and the
NDVI were calculated without in-situ data, and
only from remote sensing datasets. The ground
heat fluxes were validated with the in-situ data
and were not fitted. The remotely sensed dataset
has also been used to validate the Q
because only the clear day average IOP Q
were used from the in-situ sites and not the whole
IOP dataset. The in-situ measured data were mea-
sured and not acquired using the approaches and
models presented below.
3.3.1 The complete aspect ratio approach
This approach is based on the calculation of the
complete aspect ratio (CAR
), which describes
the enlargement of the surface due to the 3-D
structure of the city, which almost doubles the
surface for storage of heat fluxes. CAR can be
derived from the high resolution (1 m) digital
surface model (DSM) which provides informa-
tion on building height, roof structure etc. As a
first step the sky-view-factor was calculated from
the surface model (see Fig. 2). As a second step,
the complete aspect ratio
was computed.
Following this the data were downsampled to a
30 m resolution for a more general overview and
to enable a better comparison with the other
methods and the in-situ data. As mentioned by
Christen and Vogt (2004), there is a correlation
and Q
, which can be de-
scribed as a hysteresis suggesting the following
f f
þ Q
where Q
is the ratio measured or mod-
elled over rural surfaces, and Q
is the
Table 2. The OHM parameters for the different land use
Parameters=Land use class a
Forest (Grimmond
and Oke, 1999)
0.11 0.11 12.3
Industrial (U3) 0.46 0.16 49
Medium S1 þ U1 þ U2) 0.42 0.27 36
Agricultural (R2) 0.21 0.34 25
R1 þ R3) 0.16 0.05 16
Fig. 2. Sky-View-Factor of the
City of Basel
Modelling the ground heat flux of an urban area 189
theoretical value asymptotically reached with in-
, set to a constant value of 0.45
(personal communication A. Christen). The rela-
tionship between complete aspect ratio
, net all
wave radiation Q
and storage heat flux Q
with in-situ measured values is shown in Fig. 3.
The factor f is used to describe the curvature
that is highly dependent on time of day due to the
diurnal hysteresis. It is set to values between 10
(morning) and 0 (evening) to best fit the observa-
tions. Since the LANDSAT ETMþ overpass was
between 10 and 11 a.m. UTC, f was set to a value
of 5 for best fit according to the in-situ derived
values. It is important to note that f was derived
from the in-situ averages during the whole IOP
and are not values specifically for July 8, 2002.
Small changes of f from 4.5 to 5.5 result in dif-
ferences in Q
of 5Wm
. The application
of f to the remotely sensed data to model the
ground heat flux density Q
is independent
due to the fact that the complete aspect ratio
was calculated without in-situ data from the
DSM dataset.
3.3.2 The NDVI approach
This approach is based on the assumption that
vegetation reduces the ground heat flux so that
with increasing biomass, ground heat flux density
decreases. The Normalized Difference Vegetation
Index (NDVI) can be treated as an indicator of
biomass density. Kustas and Daughtry (1990)
documented a linear relationship between the
ratio of Q
and the NDVI. In the Surface
Energy Balance Algorithm for Land (SEBAL)
(Bastiaanssen et al., 1997) NDVI was selected
to describe the general effect of vegetation on
surface heat fluxes. For the city of Basel, Parlow
(2003) used a NDVI approach to estimate spa-
tially distributed ground heat fluxes. In this case,
Parlow (2000) modified the Bastiaanssen et al.
(1997) equations based on rural field measure-
ments for use in an urban situation. The equa-
tions for urban and rural sites are thus:
¼ð0:3673 0:3914 NDVIÞQ
¼ð0:3673 0:3914 NDVIÞQ
ð0:8826 lnðQ
where Q
is the net all wave radiation, NDVI
the Normalized Difference Vegetation Index and
the shortwave net radiation (Q
). The
NDVI can be calculated from the LANDSAT
ETMþ image, and net radiation (Q
) and short-
wave net radiation Q
have been computed as
described above. Equations 4 and 5 were used for
urban and rural sites, respectively (see Fig. 6b).
3.3.3 The OHM approach
The Objective Hysteresis Model (OHM) was
introduced by Oke and Cleugh (1987) and
Grimmond and Oke (1999) for field measure-
ments. In this paper it has been applied to remote
sensing data for the first time. Oke and Cleugh
(1987) used a hysteresis-type equation to charac-
terize the ground heat flux as:
¼ a
þ a
þ a
where t is time [h] and the parameter a
cates the overall strength of the dependence of
the ground heat flux on net radiation. The param-
eter a
describes the degree and the direction
of the phase relations between Q
and Q
The parameter a
is an intercept term that indi-
cates the relative timing when Q
and Q
become negative. The hysteresis curve is shown
for the observed series (Fig. 4) as calculated by
Christen et al. (2003) for the BUBBLE sites for
the whole IOP.
Fig. 3. Intensity of the daytime storage heat flux density
in dependence of the complete aspect ratio
. The
data used was taken from the IOP period June 10 to July 10
2002. Error bars include 50% of all single 1 h-runs
190 G. Rigo and E. Parlow
The parameters a
, a
and a
were derived
from in-situ measurements for the specific land
use class of each site as averages (Christen and
Vogt, 2004) during the whole IOP, not specific
values for July 8, 2002. For the ‘Forest’ land
use class values were taken from Grimmond and
Oke (1999) as no BUBBLE measurements were
collected from inside a forest.
Instead of using the 2002 satellite data of
LANDSAT ETMþ, land use classes were calcu-
lated from a LANDSAT ETMþ and ASTER
fused image from June 12, 2001 (Fig. 1) by ap-
plying a maximum likelihood approach. The over-
all accuracy of the land use classification was
81.2% with a Kappa coefficient of 0.80. The sat-
ellite data from 2001 were selected because of the
superior spatial coverage and because the fused
ASTER-LANDSAT ETMþ data set had a higher
spatial resolution (15 m instead of 30 m) for the
initial land use classification.
Each set of parameters a
and a
can there-
fore, be applied to a specific land use class only
(urban, agricultural, forest etc). For this reason
the fourteen land use classes derived from satel-
lite data were reclassified to seven major land use
classes for which parameters were available: a)
water, b) densely built-up and industrial areas, c)
medium and sparsely built-up areas, d) forest,
e) grassland, f) agricultural fields and g) clouds
and not classified (see Table 2 and Fig. 1). The
parameters a
, a
and a
were therefore used with-
out specific fitting to the remotely sensed datasets.
For the application of Eq. 6 with remotely
sensed data, a minimum of two data layers repre-
senting net radiation Q
are necessary to calcu-
late Q
=t. Therefore, in a first step, modelled
net all wave radiation (Q
) values were com-
pared to in-situ measurements. The overall accu-
racy of the AVHRR 14, MODIS and LANDSAT
ETMþ data yielded the following mean differ-
ences from measured data (except the Messe site
(U3)): 12 Wm
for the MODIS daytime, 5 Wm
nighttime, 22 Wm
for AVHRR-14. AVHRR-16
data for July 8, (2:10 UTC) shows greater dif-
ferences at the Village-Neuf site, but good agree-
ment at the Messe site with a mean of 5 Wm
difference. One has to keep in mind that AVHRR
and MODIS only have a spatial resolution
of 1.1 km and 0.93 km, respectively, whereas
LANDSAT ETMþ has a 30 m spatial resolution
(in visible and near infrared bands; 60 m in the
thermal band).
4. Results
The in-situ measurements taken during the
BUBBLE IOP yielded ground heat flux density
measurements for the six sites previously men-
tioned (unfortunately the sites at Grenzach and
Gempen were not covered by the LANDSAT
ETMþ image). The raw data were integrated to
hourly averages. At some sites the ground heat
flux density was measured directly (Allschwil,
Grenzach, Village-Neuf), for the other sites it was
calculated as a residual term from eddy-covariance
measurements (Christen and Vogt, 2004; Webb
et al., 1980). The results were computed for July
8, 2002 and June 25 & 26, 2002, when multiple
satellite overpasses were available (see Table 1).
For the daytime overpasses (from 4 UTC to
9 UTC) net radiation was modelled using the
respective satellite data and SWIM (Short Wave
Irradiance Model). For the nighttime overpasses,
only the longwave atmospheric counter radiation
and longwave terrestrial emission had to be con-
sidered to calculate the net radiation according to
Eq. 1. Due to limited availability and known dif-
ficulties with in-situ measurements of the ground
heat flux density in urban environments (see also
Weber, 2006), hourly values for the respective
satellite overpass times for all clear-sky days dur-
ing the IOP were averaged for comparison with
Fig. 4. Mean diurnal hysteresis of the ground or storage
heat flux density Q
vs. the net radiation Q
at the rural
and urban sites (adapted from Christen, 2003)
Modelling the ground heat flux of an urban area 191
the results from the three different approaches
(see Fig. 5).
As can be seen, for urban site R1, the differ-
ences between the average and the values from
July 8 are quite small. This is due to the low daily
dynamic range of the ground heat flux in rural
environments, and also due to the relatively small
fluxes themselves. As for S1 or U2, the values
Fig. 5. Hourly average ground heat
flux density measurements at three se-
lected sites for the 8th of July together
with the IOP average values with a)
R1, b) S1 and c)U1
192 G. Rigo and E. Parlow
show much higher differences. The similarity
between the modelled and observed ground heat
flux values is far better for average values than
for daily values, which makes the average values
more realistic especially at U2 where there were,
unfortunately, no useful data available in the
morning hours due to instrument errors.
For the complete aspect ratio (CAR) only the
urban sites could be used for comparison. The
modelled values were compared with the respec-
tive in-situ observations for each measurement
site. For the MODIS daytime, LANDSAT and the
MODIS nighttime images more than one OHM
approach was used because more than one image
was available for the calculation (see Table 3).
In advance, the modelled net all wave radia-
tion (Q
) values were compared to the in-situ
measurements as described above to ensure the
accuracy of the input data for the OHM model.
The overall mean absolute difference (MAD) was
26 Wm
for the modelled Q
during the IOP
with an RMSE of 29 Wm
The comparison of the computed ground heat
flux densities with measured data revealed the
following results (Table 3) with (a, b, and c)
denoting the approach used. The values for R2
lated from the average value for the areas with
a complete aspect ratio of about 1 surrounding
the built up area (see Fig. 6a) and therefore
they are not included in the statistics. The coef-
ficient of determination (R
) for all OHM data is
0.96 with an RMSE of 16 Wm
which shows
the very high correlation between the modelled
and measured datasets. The individual MAD in
for each site and each calculation are
Table 3 shows that U1, R3 and S1 reveal the
best results, while U2 and U3 show the worst re-
sults. There is therefore no clear distinction be-
Table 3. Absolute differences (bias) between in-situ and the modelled ground heat flux densities at six sites (a) with CAR, b)
with NDVI and c) with OHM). Values in parentheses represent hypothetical rural differences for the CAR approach
Sites Mean absolute differences (MAD) between modelled and measured
ground heat fluxes in W=m
L Erlen
a) LANDSAT 08.07.02 4 19 103 (44) (15)
a) MODIS day 08.07.02 1 32 96 (36) (13)
b) LANDSAT 08.07.02 3 26 1 18 2 69 20
b) MODIS day 08.07.02 1 25 3 9 3 74 19
c) MODIS night=MODIS day 08.07.02 15 4 11 22 20 5 13
c) MODIS night=LANDSAT 08.07.02 1 11 2 4 14 1 5
c) MODIS night=AVHRR 16 day 08.07.02 1 12 5 5 16 1 7
c) MODIS night=AVHRR 14 08.07.02 4 7 0 3 6 1 3
c) MODIS night=AVHRR 16 night 6 5 2 7 1 3 4
c) MODIS day=LANDSAT 08.07.02 26 21 78 8 31 2 23
c) MODIS day=AHVRR 14 08.07.02 27 20 71 13 26 2 27
c) MODIS day=AVHRR 16 night 08.07.02 25 18 69 10 21 2 24
c) AVHRR 16 day=AVHRR 14 08.07.02 17 40 24 15 3 11 15
c) AVHRR 16 day=AVHRR 16 night 11 25 24 15 3 11 15
c) LANDSAT=AVHRR 14 08.07.02 5 17 8 10 5 3 8
c) LANDSAT=AVHRR 16 night 08.07.02 7 6 7 14 1 2 6
c) AVHRR 14=AVHRR 16 night 08.07.02 25 54 33 4 9 22 24
c) AVHRR 16=AVHRR 15 25.06.02 30 15 7 10 27 6 16
c) MODIS day=AVHRR 15 25.06.02 39 10 76 9 16 1 25
c) AVHRR16=MODIS day 25.06.02 15 28 19 36 44 9 25
c) AVHRR 16=MODIS day 26.06.02 14 37 7 40 37 12 25
c) MODIS night=AVHRR16 26.06.02 13 24 9 15 18 1 13
c) MODIS night=MODIS day 26.06.02 15 22 9 13 17 1 13
Site mean absolute difference 13 21 29 13 15 11 17
RMSE 11 12 34 10 13 21 17
Modelling the ground heat flux of an urban area 193
tween rural or urban sites but nevertheless the U2
and U3 sites are generally the least accurate.
When different overpasses are examined with
the OHM model it becomes clear that the runs
with ‘MODIS night’ on July 8, 2002 show very
high correlations together with the LANDSAT
ETMþ, irrespective of the satellite scene used
as a second image. In contrast, the ‘MODIS
day’ shows the worst results with all scenes,
even if they are used as the second scene as
occurred on June 25 and 26, 2002. The overall
MAD is 17 Wm
with an RMSE of 17 Wm
Fig. 6. Ground heat flux in
in the city of Basel. a)
Modelled with the complete as-
pect ratio approach, b) modelled
with the NDVI approach and
c) modelled with the OHM ap-
proach applied on AVHRR 16
and LANDSAT ETMþ data d)
night time ground heat flux from
194 G. Rigo and E. Parlow
for all the sites and all approaches used. If only
the OHM is taken into account, the values are
16 Wm
and 13 Wm
, respectively.
We can see in Fig. 6a, b, and c that the results
of the three different approaches show the same
patterns and similar values for the city and its
surroundings at about the same time of the day.
The surrounding data of the built up area in
Fig. 6a is treated as a rural field and corresponds
reasonably well with rural measured values of R3
Fig. 6 (continued)
Modelling the ground heat flux of an urban area 195
(see Table 3 and the results of a) for rural sites).
Since the surrounding area is not rural, but was
only treated as rural due to a lack of data, the
resulting complete aspect ratio was 1. Therefore,
the results are for information only and were not
included any further in the statistics.
Figure 6d shows that the nighttime ground
heat flux density Q
, with the OHM model ap-
proach, is positive. There is also a clear differ-
ence between urban and rural sites, but to a much
smaller extent than during the day (90 Wm
50 Wm
). The higher storage heat flux in the
historic downtown area (about 240 Wm
) and
around the harbour basin (industrial) is clearly
visible in Fig. 6b, c and d whereas the parks stand
out as ‘green’ spots with much lower Q
values (about 100 Wm
). Also, the airport in
the northwest corner of the image is clearly visi-
ble together with the forest areas which have the
lowest Q
of all land cover, 80 Wm
the day, and 30 Wm
at night.
5. Discussion
5.1 The CAR approach
The results of the CAR approach are somewhat
limited, because the available data only covers
the built-up areas of the city of Basel. The results
for the Sperrstrasse and Spalenring urban sites
are similar to the values obtained from the
other models (differences between 1 Wm
32 Wm
). However, the Messe site shows much
higher differences of 100 Wm
on average. Sev-
eral explanations are possible for this extremely
high difference. The Messe site was located on
the flat roof of a multi-storey car park. When the
calculation of the CAR from the sky view factor
was applied, the resolution was reduced to 30 m
instead of the original 1 m. Therefore, the roof
shows an aspect ratio of about 1, which qualifies
it as an open space. Since the aspect ratio for
areas with vegetation is taken as equal to that
for an open space, the result shows a ground heat
flux density typical for areas with vegetation
such as the Lange Erlen or Grenzach sites. This
explains the difference and also shows that on
open sites without vegetation the CAR-model is
not really useful, although the averaged values of
the surroundings come close to the rural mea-
sured values (see Table 3). In high density areas,
where we have a high CAR, the results can be
better compared. Another problem with the CAR
approach is the fact that although net radiation
can be modelled and measured with enough accu-
racy, the closure of the energy balance (Q
þ Q
¼ Q
) is simply not given for each hour
(see also Christen and Vogt (2005)). Therefore,
the net radiation at night can be lower than only
the storage heat flux Q
itself. Due to these
facts, the application of the CAR model to night-
time or morning data is difficult and makes valid
calculations possible only between 10 h and 16 h
CET. The dependency on the f parameter is low
as mentioned previously for midday values.
5.2 NDVI approach
For the NDVI approach, the spatial resolution of
the thermal data is not very important as is shown
in Table 3. The differences between the modelled
LANDSAT and MODIS-derived ground heat flux
densities are minimal with the exception of the
Lange Erlen site which shows far too high differ-
ences when compared to the other sites. Even the
Spalenring site is in good agreement, showing an
average difference of 25 Wm
. Why this hap-
pens is difficult to determine because the calcu-
lation with a MODIS derived Q
image showed
almost the same results. One explanation could
be that the NDVI value of R3 was different dur-
ing the overpass time than the average ground
heat flux would suggest, especially if it over-
estimates the in-situ measured values. The NDVI
also changes during the season and is sometimes
due to soil moisture influences. In general, the
NDVI approach is still the easiest to use of the
three approaches and requires the least data for
acceptable to good results. However, its biggest
drawback is the fact that it is limited to daytime
computation only.
5.3 OHM approach
The most promising of the three approaches is
the OHM approach with an R
of 0.95, an overall
RMSE of 13 Wm
, and a mean absolute differ-
ence (MAD) of 17 Wm
(even when the Messe
site is included). The rural sites perform better
than the urban ones but not by much. Neverthe-
less, the accuracy of in-situ measurements made
by heat flux plates is still higher than with eddy
196 G. Rigo and E. Parlow
covariance over the urban sites, as already men-
tioned. The OHM modelled values generally
overestimate daytime values and underestimate
nighttime values. The higher accuracy with the
nighttime imagery is important because the night-
time values of the storage heat flux density are
much lower than during the day. Therefore, a
better accuracy during nighttime seems reason-
able and minimizes the relative differences.
Most problems with the storage heat flux den-
sity at the Messe site arise with the OHM ap-
proach with MODIS daytime imagery resulting
in differences of more than 70 Wm
site is situated on a car park, the big differences
with the MODIS OHM approach can be ex-
plained with the far coarser spatial resolution
of 1 km, whereas smaller local extreme values
cannot be detected so accurately. But, because
the values for U1 and R2 are also much higher
than the average, this explanation is not suffi-
cient and it seems that the source of error is
the Q
for the MODIS overpass. This is sup-
ported by the results of Q
for the MODIS day-
time scene with a MAD of 44 Wm
modelled and measured values. This example
shows how important exact modelling of net
radiation Q
is for a successful application of
OHM to model the storage heat flux. For the
Lange Erlen site, it is remarkable that the differ-
ences are the lowest of all sites even with the
MODIS daytime overpass, but here the Q
values are also accurate enough.
Generally the more accurate modelling of the
rural sites can be attributed to the more homoge-
neous surfaces and also to more accurate mea-
surements made with heat flux plates. Although
the urban sites also yield reasonable results it
remains difficult to take into account all the spe-
cific parameters of urban surfaces (e.g. urban an-
isotropy (Voogt and Oke, 1997).
Another issue are the parameters a
, a
and a
They can either be derived from in-situ measure-
ments or from literature values e.g., Grimmond
and Oke (1999). It is more reliable to use the
specific values from BUBBLE because they are
more accurate than values taken from the litera-
ture. Nevertheless, for different cities without
measured in-situ derived parameters, data from
literature should be used which correspond as
closely as possible to the actual morphology and
structure of the city.
6. General discussion and conclusions
For all of the three approaches Complete Aspect
Ratio (CAR), NDVI or Objective Hysteresis
Model (OHM) the results correspond quite well
with the measured values. If we take into
account that it is not easy to measure the ground
heat flux with great accuracy in rural areas and
even less so in urban areas, the results can be
considered very good. Each of the approaches
has its limitations and advantages which make
them more or less suitable for the modelling of
the ground heat flux, but as shown in Table 3,
the results from all three approaches show sim-
ilar performance. Differences between rural and
urban areas are minimal, although rural areas
can be still modelled more accurately, espe-
cially with the OHM approach. A possible ex-
planation is in the method of storage heat flux
density measurement. Whereas over urban areas
it was determined as a residual, over rural areas
heat flux plates were used, which is much more
accurate than an assumed energy balance clo-
sure. On the other hand, the suburban site S1
also shows a high degree of accuracy. When
using a spatial model, we cannot ignore vege-
tated sectors in urban areas, therefore the model
performs well for urban and rural surfaces even
according to the uncertainty introduced by the
two different methods of in-situ measurement
of Q
Despite, the CAR approach being limited in
urban canyons, it can be easily derived from high
resolution digital surface data since only the
complete aspect ratio is needed. Of the three ap-
proaches presented, the CAR approach is supe-
rior to the OHM approach, which also uses a
hysteresis curve for the modelling. The biggest
source of error in the CAR approach is the de-
pendency on f and, as shown by Christen and
Vogt (2004), the difficulties for modelling the
nighttime storage heat flux. Further testing of
the CAR approach on other cities with 3-D mod-
els is advised and would be a second step in
establishing the transferability of the model.
The NDVI approach needs additional infor-
mation from remotely sensed imagery (Q
For the given overpass time between 10 h and
15 h local time (UTC þ 2withdaylightsaving
time in Central Europe) the NDVI approach is
rightly suitable, yields good results and is easy
Modelling the ground heat flux of an urban area 197
to use but should be used with data with a spatial
resolution of at least 60 m because NDVI from
AVHRR or MODIS (16 Day average) are less
accurate due to their lower spatial and=or tem-
poral resolution.
The results from all the approaches showed
good agreement with MAD around 17 Wm
(except at U3), particularly given the fact that
measurement of the ground heat flux density have
an operational error of more than 15%, depending
on the method used (heat flux plate or calculation
with eddy covariance from Sonic-anemometer
measurements) according to Weber (2006) or
Twine et al. (2000). Kustas et al. (1995), with their
two layer turbulent and therefore more compli-
cated model, showed a difference of 35 Wm
for the ground heat flux. Kustas et al. (2004),
with their two-source model approach, still show-
ed average differences between 21 Wm
32 Wm
with modelled LANDSAT TM and
LANDSAT ETMþ data for the ground heat flux
densities over croplands in Iowa.
The OHM approach is the most promising,
although we need at least two net radiation scenes
to be able to calculate Q
classification system to model the ground heat
fluxes. The parameters a
, a
and a
can be
derived either from in-situ measurements or from
values already published in the literature (see
Grimmond and Oke, 1999). Using the OHM
approach it is also possible to compute the night-
time ground heat flux densities depending on
satellite overpass times, if the correct parameters
are available from in-situ measured data. A large
in-situ network was available during BUBBLE,
however, the parameters can be used together
with other literature values for similar cities or
urban areas to model the ground heat flux with
satellite data. This is possible even if no energy
flux measurements are available. It is of course
a different matter to use literature values, but if
we have about the same urban parameters (e.g.
and similar surface materials), the
OHM model would probably yield rather accu-
rate results, although this would be another task
to be tested.
The facts mentioned above and the good accu-
racy of the OHM model favour it for further use
in modelling the storage heat flux in urban and
rural environments with satellite data in the spa-
tial domain.
The project BUBBLE-SARAH is funded by the Swiss
National Science Foundation Project No. 200020-105299=1.
Thanks go to MeteoSwiss# for providing radiosonde
data. The authors are also grateful for the two anonymous
reviewer’s comments on the manuscript.
Bastiaanssen WGM, Meneti M, Feddes RA, Holtslag AAM
(1997) A remote sensing surface energy balance algo-
rithm for land (SEBAL). Part 1: Formulation. SC-DLO,
Wageningen NL
Bernhofer C, Vogt R (2000) Energy balance closure gaps -a
methodical problem of eddy covariance measurements?
In: de Dear RJ, Kalma JD, Oke TR, Auliciems A (eds)
Biometeorology and urban climatology at the turn of the
millennium Selected papers from the conference ICB-
ICUC’99. WMO=TD No. 1026, pp 199203
Capparini F, Castelli F, Enthekabi D (2003) Mapping of
land-atmosphere heat fluxes and surface parameters with
remote sensing data. Bound-Layer Meteor 107: 605633
Christen A, Bernhofer C, Parlow E, Rotach MW, Vogt R
(2003) Partitioning of turbulent fluxes over different urban
surfaces. Proceedings of the 5th International Conference
on Urban Climatology (ICUC 5), Lodz, Poland, 285288
Christen A, Vogt R (2004) Energy and radiation balance of a
central European city. Int J Climatol 24: 13951421
Christen A, Vogt R (2005) Quality control and corrections of
the radiation measurements at the BUBBLE-surface sites.
Stratus Technical Report=Arbeitsbericht 003,18 pp, Insti-
tute of Meteorology, Climatology and Remote Sensing,
University of Basel (
Grimmond CSB, Oke TR (1991) An objective urban heat-
storage model and its comparison with other schemes.
Atmos Environ, Part B: Urban Atmosphere 25: 311326
Grimmond CBS, Oke TR (1999) Heat storage in urban areas:
local-scale observations and evaluation of a simple model.
J Appl Meteor 38: 922940
Kustas WP, Daughtry CST (1990) Estimation of the soil heat
flux=net radiation ratio from spectral data. Agric Forest
Meteorol 49: 205223
Kustas WP, Norman JM, Anderson MC, French AN (2003)
Estimating subpixel surface temperatures and energy
fluxes from the vegetation index-radiometric temperature
relationship. Remote Sensing Environ 85: 429440
Kustas WP, Li F, Jackson TJ, Prueger JH, MacPherson JI,
Wolde M (2004) Effects of remote sensing pixel resolu-
tion on modelled energy flux variability of croplands in
Iowa. Remote Sensing Environ 92: 535547
Norman JM, Kustas WP, Humes KS (1995) Source approach
for estimating soil and vegetation energy fluxes in ob-
servation of directional radiometric temperature. Agric
Forest Meteorol 77: 263293
Oke TR (1982) The energetic basis of the urban heat island.
Quart J Roy Meteor Soc 108: 124
Oke TR, Cleugh HA (1987) Urban heat storage derived as
energy budget residuals. Bound-Layer Meteor 39: 233–245
198 G. Rigo and E. Parlow
Oke TR, Spronken-Smith RA, Jauregui E, Grimmond CSB
(1999) The energy balance of central Mexico City during
the dry season. Atmos Environ 33: 39193930
Oncley SP, Foken T, Vogt R, Kohsiek W, deBruin H,
Bernhofer C, Christen A, van Gorsel E, Grantz D, Lehner
I, Liebethal C, Liu H, Mauder M, Pitacco A, Ribeiro L,
Weidinger T (2005) The energy balance experiment
EBEX 2000. Part I: Overview and energy balance.
Bound-Layer Meteor (submitted)
Parlow E (1996a) Net radiation in the REKLIP-area A
spatial approach using satellite data. In: Parlow E (ed)
Progress in environmental research and applications.
Rotterdam: Balkema, pp 429435
Parlow E (1996b) Correction of terrain controlled illumina-
tion effects in satellite data. In: Parlow E (ed) Progress in
environmental research and applications. Rotterdam:
Balkema, pp 139145
Parlow E (2000) Remotely sensed heat fluxes of urban areas.
In: de Dear RJ, Kalma JD, Oke TR, Auliciems A (eds)
Biometeorology and urban climatology at the turn of the
millennium Selected papers from the conference ICB-
ICUC’99. WMO=TD No. 1026, pp 523528
Parlow E (2003) The urban heat budget derived from
satellite data. Geographica Helvetica 2: 99111
Piringer M, Grimmond CBS, Joffre SM, Mestayer P,
Middleton DR, Rotach MW, Baklanov A, De Ridder K,
Ferreira J, Guilloteau E, Karpinen A, Martilli A, Masson
V, Tombrou M (2002) Investigating the surface energy
balance in urban areas – recent advances and future needs.
Water Air Soil Poll: Focus 2: 116
Rigo G, Parlow E, Oesch D (2006) Validation of satellite
observed thermal emission with in-situ measurements
over an urban surface. Remote Sensing Environ 104:
Rotach MW, Vogt R, Bernhofer C, Batchvarova E, Christen
A, Clappier A, Feddersen B, Gryning S-E, Martucci G,
Mayer H, Mitev V, Oke TR, Parlow E, Richer H, Roth M,
Roulet Y-A, Ruffieux D, Salmond JA, Schatzmann M,
Voogt JA (2005) BUBBLE an urban boundary layer
meteorology project. Theor Appl Climatol 81: 231261
Twine TE, Kustas WP, Norman JM, Cook DR, Houser PR,
Meyers TP, Prueger JH, Starks PJ, Wesely ML (2000)
Correcting eddy-covariance flux underestimates over a
grassland. Agric Forest Meteor 103: 279300
Voogt JA, Oke TR (1997) Complete surface temperatures.
J Appl Meteor 36: 11171132
Weber S (2006) Comparison of in-situ measured ground heat
fluxes within a heterogeneous urban ballast layer. Theor
Appl Climatol 83(14): 169179
Wilson K, Goldstein A, Falge E, Aubinet M, Baldocchi D,
Berbigier P, Bernhofer C, Ceulemans R, Doman H, Field
C, Grelle A, Ibrom A, Law BE, Kowalsk A, Meyers T,
Moncrieff J, Monson R, Oechel W, Tenhumen J, Valentini
R, Verma S (2002) Energy balance closure at FLUXNET
sites. Agric Forest Meteor 113: 223243
Authors’ address: Gergely Rigo (e-mail: Gergely.rigo@, Eberhard Parlow, Institute of Meteorology, Cli-
matology and Remote Sensing, Klingelbergstraße 27, 4056
Basel, Switzerland.
Modelling the ground heat flux of an urban area 199
... While the UCL positive air temperature anomalies are mostly found during the nocturnal period, the main cause for the UHI stems from the landuse/land cover (LULC) changes (and corresponding physical properties), as the urban artificial materials respond differently to the diurnal incoming solar radiation, compared to more natural/rural surfaces. As previously mentioned by several authors (e.g., Chrysoulakis et al., 2016Chrysoulakis et al., , 2018Feigenwinter et al., 2018;Grimmond, 2014a, 2014b;Lopes, 2003;Nadeau et al., 2009;Oke, 1988;Rigo and Parlow, 2007;Wang et al., 2010), during the day, impervious/artificial urban surfaces are responsible for a greater absolute storage heat flux component (compared to rural/natural surfaces); in addition, at the city-level, urban areas are typically poorly ventilated (due to greater roughness length from the built infrastructures), while evapotranspiration is reduced, due to lower vegetation cover; during the night, the diurnal storage heat flux is progressively released back into the atmosphere, reducing the nocturnal cooling rate, compared to rural sites (Oke, 1982(Oke, , 1988Oke et al., 2017aOke et al., , 2017b). An additional contribution to the UHI is the heat emitted due to anthropogenic activities, such as air conditioning systems and vehicles exhaust, although typically being a smaller contributor to the UHI intensity (Oke et al., 2017a(Oke et al., , 2017b. ...
... Despite the diurnal acquisition-related constraints, Landsat-based diurnal surface heat fluxes have been shown to provide greater insights on the culprit of the atmospheric UHI signalparticularly, the diurnal storage heat flux has been noted for being much greater in impervious/artificial urban surfaces during the mid-morning overpass, which partially contributes to the greater availability of nocturnal sensible heat in the atmosphere (Lopes, 2003;Oke, 1982Oke, , 1988Oke et al., 2017aOke et al., , 2017bParlow, 2003;Parlow et al., 2014;Rigo and Parlow, 2007). The diurnal latent hat flux is also affected by the absence of extensive green coverage and is reduced in urban areas, which negatively affects natural cooling processes. ...
... The diurnal latent hat flux is also affected by the absence of extensive green coverage and is reduced in urban areas, which negatively affects natural cooling processes. Hence, the spatial patterns of the diurnal heat flux components are known to explain the corresponding nocturnal atmospheric temperature anomaly (Lopes, 2003;Oke, 1982Oke, , 1988Oke et al., 2017aOke et al., , 2017bParlow, 2003;Parlow et al., 2014;Rigo and Parlow, 2007). ...
Southern European functional urban areas (FUAs) are increasingly subject to heatwave (HW) events, calling for anticipated climate adaptation measures. In the urban context, such adaptation strategies require a thorough understanding of the built-up response to the incoming solar radiation, i.e., the urban energy balance cycle and its implications for the Urban Heat Island (UHI) effect. Despite readily available, diurnal Land Surface Temperature (LST) data does not provide a meaningful picture of the UHI, in these midlatitudes FUAs. On the contrary, the mid-morning satellite overpass is characterized by the absence of a significant surface UHI (SUHI) signal, corresponding to the period of the day when the urban-rural air temperature difference is typically negative. Conversely, nocturnal high-resolution LST data is rarely available. In this study, an energy balance-based machine learning approach is explored, considering the Local Climate Zones (LCZ), to describe the daily cycle of the heat flux components and predict the nocturnal SUHI, during an HW event. While the urban and rural spatial outlines are not visible in the diurnal thermal image, they become apparent in the latent and storage heat flux maps – built-up infrastructures uptake heat during the day which is released back into the atmosphere, during the night, whereas vegetation land surfaces loose diurnal heat through evapotranspiration. For the LST prediction model, a random forest (RF) approach is implemented. RF results show that the model accurately predicts the LST, ensuring mean square errors inferior to 0.1 K. Both the latent and storage heat flux components, together with LCZ classification, are the most important explanatory variables for the nocturnal LST prediction, supporting the adoption of the energy balance approach. In future research, other locations and time-series data shall be trained and tested, providing an efficient local urban climate monitoring tool, where in-situ air temperature observations are not available.
... Unfortunately, heat storage in urban areas remains largely under-researched, likely due to the aforementioned reasons. Considering the scant availability of studies on the subject, fewer exist in the spatial domain by way of satellite remote sensing, and diurnal variability is merely nonexistent -all of which are major motivators for the present research (Kato and Yamaguchi 2007;Rigo and Parlow 2007;Liu et al. 2016;Weng et al. 2013). ...
... The residual method is perhaps the easiest method to employ, but also arguably carries the most error in that it lumps together all of the calculable variables in the surface energy balance and uses the amount leftover from the subtraction of the incoming radiation and the sensible and latent heats (along with any other incorporated fluxes such as advection, anthropogenic heat, etc.). The residual method, therefore, calculates the heat storage by the following balance (Offerle et al. 2005;Mirzaei and Haghighat 2010;Kato and Yamaguchi 2007;Rigo and Parlow 2007): ...
... This, leads to the issue of spatial resolution, as many towers are located very far from one another. And for the few studies that employ the OHM with satellite data, they use the net radiation directly and some of the land cover-based coefficients available in the literature (Rigo and Parlow 2007). ...
Full-text available
The energy exchanges at the Earth’s surface are responsible for many of the processes that govern weather, climate, human health, and energy use. This exchange, commonly known as the surface energy balance (SEB), determines the near-surface thermodynamic state by partitioning the available energy into surface fluxes. The net all-wave radiation is often the primary energy source, while the heat storage and sensible and latent heat fluxes account for the majority of energy distributed elsewhere. While the SEB of various natural environments (trees, crops, soils) has been well-observed and modeled, the urban surface energy balance remains elusive. This is due to the heterogeneity of urban land cover, where the surface cover is dominated by impervious materials (buildings, roads, and pavements) interspersed with vegetation and bare soil. The impervious materials differ in their hygro-thermal properties based on their inherent capacity to conduct and retain heat and moisture. Traditional observation techniques are unable to capture all the relevant processes in cities, and as a result, the urban surface energy budget remains mostly unknown. In this seminar, novel techniques that combine traditional boundary layer turbulence measurements and advanced remote sensing methods are presented as solutions to advance our understanding of urban surface energy balance. Here, new methodologies are developed that apply remote sensing-based algorithms to urban environments. The first topic uses satellite measurements to derive near-surface air temperature for urban areas- this has yielded a publication (DOI: 10.1016/j.rse.2019.111495). Next, a satellite-based algorithm that approximates the net all-wave radiation is presented, using machine learning and land cover information. Lastly, two novel methods for predicting the heat stored in cities are introduced (one of which resulted in a publication with DOI:10.1016/j.rse.2020.112125). Overall, this dissertation presents new knowledge and develops novel algorithms that expand and advance our understanding of urban thermodynamics, which impacts how we observe and model agricultural processes, human vulnerability to weather and climate, better predict energy use, and improve the sustainability of our cities.
... To compute the net radiation of the whole area, spatially distributed solar irradiance was calculated using a numerical model [22,23], and calibrated by local measurements at seven urban and rural flux towers; shortwave reflection was computed using a linear combination of Landsat visible bands, and atmospheric counter-radiation was assumed to be constant in that small area. For further information, I refer you to Rigo and Parlow [24]. The basic statement of Figure 2 is that areas with the highest LST have the lowest net radiation of Q*. ...
... In urban areas, storage heat flux into urban fabrics, asphalt, concrete, etc., is 4-5 times higher [25]. Urban net radiation values in Figure 2 (city center of Basel around 550 Wm −2 ) show storage heat flux gains between 200 and 350 Wm −2 during the satellite overpass, which have also been validated by measurements during BUBBLE [24]. This happens during the daytime when most LST satellite data are taken, and it can be considered akin to loading the "urban storage-heat-fluxbattery" during the daytime; this energy is then re-invested at night to keep urban air temperatures on a high level and manifest a nocturnal urban heat island. ...
Full-text available
This paper attempts to illustrate the complexity of thermal infrared (TIR) data analysis for urban heat island studies. While a certain shift regarding the use of correct scientific nomenclature (using the term “surface urban heat island”) could be observed, the literature is full of incorrect conclusions and results using erroneous terminology. This seems to be the result of the ease of such literature implicitly suggesting that “warm surfaces” result in “high air temperatures”, ultimately drawing conclusions for urban planning authorities. It seems that the UHI is easy to measure, easy to explain, easy to find, and easy to illustrate—simply take a TIR-image. Due to this apparent simplicity, many authors seem to jump into UHI studies without fully understanding the nature of the phenomenon as far as time and spatial scales, physical processes, and the numerous methodological pitfalls inherent to UHI studies are concerned. This paper attempts to point out some of the many pitfalls in UHI studies, beginning with a proper correction of longwave emission data, the consideration of the source area of a thermal signal in an urban system—which is predominantly at the roof level—demonstrating the physics and interactions of radiation and heat fluxes, especially in relation to the importance of urban storage heat flux, and ending with an examination of examples from the Basel study area in Switzerland. Attention is then turned to the analysis of spatially distributed net radiation in the day- and at nighttime as a minimum requirement for urban heat island studies. The integration of nocturnal TIR images is notably recommended, as satellite data and the UHI-phenomenon cover the same time period.
... The ideal method to assess UHI from surface thermal imagery is to either have enough daily samples (e.g., hourly) to compute a daily SUHI cycle, or to use that data to compute the flux components of the urban energy balance (UBE): heat storage, sensible heat, and latent heat components. The UBE approach has evolved (Oke 1982;Oke 1988;Stewart et al. 2014;Oke et al. 2017) and its application to remote sensing data has been explored by several urban climate research groups (Rigo et al. 2006;Rigo and Parlow 2007;Wicki et al. 2018;Chrysoulakis et al. 2018) with encouraging results. Nevertheless, the main challenge is the need for in situ monitoring to validate approaches. ...
... Nevertheless, the main challenge is the need for in situ monitoring to validate approaches. Flux towers are one method of validating satellite-based results but are difficult to access because they require substantial financial resources and appropriate site location (Rigo and Parlow 2007;Wicki and Parlow 2017;Wicki et al. 2018). In cases where flux towers have been constructed (e.g., UrbanFluxes project), there are also challenges in evaluating the accuracy of satellite-based results (i.e., each pixel covering several hundreds of square meters) given urban spatial heterogeneity. ...
Full-text available
Sensing and measuring meteorological and physiological parameters of humans, animals, and plants are necessary to understand the complex interactions that occur between atmospheric processes and the health of the living organisms. Advanced sensing technologies have provided both meteorological and biological data across increasingly vast spatial, spectral, temporal, and thematic scales. Information and communication technologies have reduced barriers to data dissemination, enabling the circulation of information across different jurisdictions and disciplines. Due to the advancement and rapid dissemination of these technologies, a review of the opportunities for sensing the health effects of weather and climate change is necessary. This paper provides such an overview by focusing on existing and emerging technologies and their opportunities and challenges for studying the health effects of weather and climate change on humans, animals, and plants.
... Many studies have computed ∆Q s using surface observations, while very few have done so from the satellite perspective. In urban areas, the number of heat storage-related studies involving satellite observations is even lower [23]. This is likely due to the lack of standardization for computing heat storage, and the simultaneous difficulty of representing accurate fluxes from remotely-sensed imagery. ...
... The heat storage is often estimated as a residual to the surface energy balance in Equation (1), with varying degrees of components omitted. For many applications, ∆Q s is simplified as [6,7,23]: ...
Full-text available
Heat storage, Qs, is quantified for 10 major U.S. cities using a method called the thermal variability scheme (TVS), which incorporates urban thermal mass parameters and the variability of land surface temperatures. The remotely sensed land surface temperature (LST) is retrieved from the GOES-16 satellite and is used in conjunction with high spatial resolution land cover and imperviousness classes. New York City is first used as a testing ground to compare the satellite derived heat storage model to two other methods: a surface energy balance (SEB) residual derived from numerical weather model fluxes, and a residual calculated from ground-based eddy covariance flux tower measurements. The satellite determination of Qs was found to fall between the residual method predicted by both the numerical weather model and the surface flux stations. The GOES-16 LST was then downscaled to 1-km using the WRF surface temperature output, which resulted in a higher spatial representation of storage heat in cities. The subsequent model was used to predict the total heat stored across 10 major urban areas across the contiguous United States for August 2019. The analysis presents a positive correlation between population density and heat storage, where higher density cities such as New York and Chicago have a higher capacity to store heat when compared to lower density cities such as Houston or Dallas. Application of the TVS ultimately has the potential to improve closure of the urban surface energy balance.
... Only slightly better results were reported when Landsat imagery was considered (with a significantly higher spatial resolution (60 m vs 1 km). In the BUBBLE experiment, when simultaneous direct and remote sensing observations were used, more attention was paid to the determination of the ground heat flux than to the atmospheric sensible heat flux (Rigo and Parlow, 2007). The methodologies used in the BUBBLE project were further improved during the URBANFLUXES project ). ...
Full-text available
This study explores the possibility of estimation of the sensible surface heat flux using satellite-derived surface temperature and road pavement temperature together with in-situ wind and air temperature measurements by the profile method. A 10-year series of data from the roadside weather observation network was used. This dataset contained wind (measured at 5.8–9.5 m above ground) and air temperature (measured at 2.6–4.8 m) together with road surface temperature. Another dataset consisted of 254 simultaneous MODIS observations. A high correlation (0.94) of the surface temperature measured by both methods was noted despite coarse pixel size. We considered satellite-derived surface temperature to determine the sensible heat flux by the profile method; these results were compared to the values obtained using road temperature measured by pavement-mounted sensors. While the overall correlation is relatively strong (0.70) and considerable systematic differences exist, the values of heat flux calculated at different locations show a high spatial coherence - either when using the in situ pavement temperature (correlation ranging from 0.84 to 0.94 for daytime and 0.63–0.84 for nighttime) or the satellite-derived temperature (correlation coefficient 0.72). In most cases, differences between the two flux estimates can be linked to local factors such as the land use structure.
... The LST observations at the overpass time of 13:30 and 01:30 local time were taken as proxies of the daytime and nighttime temperature. The data accuracy was shown to be better than 1 K for most cases, and the difference with in situ measurements was generally less than 5% in urban areas [75,76]. Pixels with water bodies, or those with no value or of bad quality, were masked out through quality control prior to evaluation. ...
Full-text available
Global urbanization significantly impacts the thermal environment in urban areas, yet urban heat island (UHI) and urban heat wave (UHW) studies at the mega–region scale have been rare, and the impact study of urbanization is still lacking. In this study, the MODIS land surface temperature (LST) product was used to depict the UHI and UHW in nine mega–regions globally between 2003 and 2020. The absolute and percentile–based UHW thresholds were adopted for both daily and three–day windows to analyze heat wave frequency, and UHW magnitude as well as frequency were compared with UHI variability. Results showed that a 10% increase in urban built-up density led to a 0.20 °C to 0.95 °C increase in LST, a 0.59% to 7.17% increase in hot day frequency, as well as a 0.08% to 0.95% increase in heat wave number. Meanwhile, a 1 °C increase in UHI intensity (the LST differences between the built-up and Non-built-up areas) led to a 2.04% to 92.15% increase in hot day frequency, where daytime LST exceeds 35 °C and nighttime LST exceeds 25 °C, as well as a 3.30% to 33.67% increase in heat wave number, which is defined as at least three consecutive days when daily maximum temperature exceeds the climatological threshold. In addition, the increasing rates of UHW magnitudes were much faster than the expansion rates of built-up areas. In the mega–regions of Boston, Tokyo, São Paulo, and Mexico City in particular, the increasing rates of UHW hotspot magnitudes were over 2 times larger than those of built-up areas. This indicated that the high temperature extremes, represented by the increase in UHW frequency and magnitudes, were concurrent with an increase in UHI under the context of climate change. This study may be beneficial for future research of the underlying physical mechanisms on urban heat environment at the mega–region scale.
... Q G is expressed as the heat storage capacity on surfaces including soil, roads, buildings, etc. The surface absorbs heat during the daytime (positive Q G ) and releases it at night (negative Q G ) (Rigo & Parlow, 2007). The surface includes the wall and the roof of the buildings in the SLUCM, so the heat storage capacity at the surface also increases, as ZR increased. ...
Full-text available
An urban heat island (UHI) is a well-known urban climatic feature; however, an opposite effect, i.e., an urban cool island (UCI), was recently observed at locations where high-rises were concentrated. Here, we analyzed the impact of two urbanization factors (anthropogenic heat flux (AH) and building height (ZR)), which could affect the formation of UHIs and UCIs, on urban precipitation over the Seoul Metropolitan area. AH caused an increasing precipitation in urban and downwind areas, especially during daytime. AH directly contributed to the sensible heat flux, resulting in surface warming associated with an UHI. Surface heat was transferred within the planetary boundary layer by turbulence, which increased the low-level instability, resulting in increased precipitation. In the experiments with the ZR, both surface temperature and precipitation increased (decreased) during the nighttime (daytime). The diurnal variation of ground heat flux intensified with an increasing ZR, inducing a smaller variation of sensible heat flux for surface energy balance. Changes in the ground heat flux could also explain occurrence of UCIs in cities with high-rises. Tall buildings reduced daytime surface temperature, thereby facilitating atmospheric stabilization, and reducing precipitation. Also, the surface friction increased with a higher ZR, resulting in enhanced convergence in the western part of urban areas where dominant westerlies first met land, which was favorable for increased precipitation. In summary, a study of variations in AH and ZR demonstrated how UHIs and UCIs significantly altered precipitation, respectively. In particular, an UCI alleviated an increment in thermo-driven daytime precipitation caused by an UHI.
... In contrast, significantly less latent heat is released by Noah03, especially in urban areas, releasing more sensible heat to the atmosphere and more ground heat flux to the soil. Of importance is that the above distributions of sensible heat flux and ground heat flux are highly consistent with the distribution of urban land, and previous studies also showed that they were greater in the urban area (Kato & Yamaguchi, 2005;Rigo & Parlow, 2007). Therefore, in this case, the simulation of surface heat flux by CLM LSM coupled in WRF model may have some defects, which needs to be further studied in the future. ...
Full-text available
A record-breaking extreme rainfall event with a maximum rainfall amount over 24 h of 524.1 mm occurred in Guangzhou, China, on May 06–07, 2017. To study the impact of land surface processes on this extreme rainfall, two 21-member convective-permitting ensemble forecasts over South China were performed based on two land surface models (LSMs), Noah and Community Land Model (CLM), and 21 forecasting members of the Global Ensemble Forecast System (GEFS). The results showed that, in general, members using the Noah LSM could better simulate urban heat islands (UHIs) and urban convection than members using the CLM LSM in this case. By investigating the ensemble member that most resembled the observations, it was found that the high temperature center in the urban area caused a thermal low in the early stage. As the southerly winds strengthened, the low-level convergence line continued moving northward and eventually triggered convection in the mountainous region. A sensitivity experiment showed that the impact of land surface heterogeneity on precipitation could be reflected on a finer scale, and heavy rainfall was very sensitive to the changes in small-scale land surface forces, including terrain and land use. Slight variations in small-scale land surface conditions caused great responses in the total precipitation, indicating that for the occurrence of such quasi-stationary extreme rainfall, a subtle balance between different atmospheric and land surface factors may be required.
Full-text available
In recent years, those who create strategies and policies for urban tourist destinations have been increasingly concerned with the greater or lesser capacity to enjoy public space. Furthermore, the growth of urban areas on a global scale has caused significant changes in the (micro)climate, due to the increase in impermeable surfaces, the anthropogenic heat generated by human activities and the change in air circulation. Taking into account the increasing demands of tourists and residents and the need to improve cities in the face of climate change, the option is to design new measures and action solutions. However, the lack of quality of the input data or their (total) absence, as well as their low spatial resolution, are common. The inadequacy of structures for sharing information is also noted, which significantly limits planning and adaptation actions. This investigation aims to identify the main methods of analysis to monitor the current ability to enjoy tourism based on the integration of objective and subjective domains; and contribute to the definition of action plans which seek to mitigate and adapt the tourism sector to climate change, in the medium and long-term. To assess the validity of these assumptions, the Porto Metropolitan Area, in general, and the municipality of Porto, in particular, were used as case studies. In this investigation, different methods of information and units of analysis were combined, based on a meso approach and local scale for: (i) the identification of critical areas, in an office analysis based essentially on Big Data (i.e., Flickr photographs, AirBnB accommodation and MODIS and LANDSAT satellite imagery); (ii) the assessment of the comfort level for enjoyment in critical areas with high tourist potential through field data collection; and (iii) the identification of prioritization actions and measures to maintain tourism attractiveness in view of climate change, in the medium and long-term. This research highlights the need for more detailed information, the weak interaction between stakeholders and the limitation of resources. Thus, considering that Porto is a destination with a good climate for tourism, and committed to mitigating the effects of climate change, the proposed methodological triangulation allows to outline some measures with predictable action in the short, medium and long-term. Finally, this study aims to make some contributions at national and international level, with the likelihood of the methodological approach adopted to be replicated in other geographical areas, taking into account the particularities of each territory under analysis.
Full-text available
The Urban Heat Budget Derived from Satellite Data The study of the interactions between urban surfaces and the urban boundary layer plays an important role in urban climatology, especially seen against the background of increasing urbanisation in most parts of the world. Measurements of radiation and heat fluxes suffer from the extreme heterogeneity of the urban landscape. It is therefore difficult to get accurate and representative measurements. To bridge the gap between accurate point measurements and their spatial representation, satellite data from Landsat-TM are used.
Conference Paper
Full-text available
Results from an experimental network of six eddy-covariance stations in and around the city of Basel (Switzerland) are presented. Three sites provide turbulent fluxes over dense urban surfaces (two towers, one parking lot). A suburban site and two rural reference sites with turbulent flux instrumentation complete the setup and allow simultaneous site to site comparison of the energy balance partitioning. The results support previous studies, that over dense urban surfaces also nighttime turbulent fluxes transport always energy away from the surface. Latent heat fluxes and bowen ratios are highly dependent on the urban vegetation fraction, while storage heat fluxes are coupled with buidling geometry.
Full-text available
The Basel UrBan Boundary Layer Experiment (BUBBLE) was a year-long experimental effort to investigate in detail the boundary layer structure in the City of Basel, Switzerland. At several sites over different surface types (urban, sub-urban and rural reference) towers up to at least twice the main obstacle height provided turbulence observations at many levels. In addition, a Wind Profiler and a Lidar near the city center were profiling the entire lower troposphere. During an intensive observation period (IOP) of one month duration, several sub-studies on street canyon energetics and satellite ground truth, as well as on urban turbulence and profiling (sodar, RASS, tethered balloon) were performed. Also tracer experiments with near-roof-level release and sampling were performed. In parallel to the experimental activities within BUBBLE, a meso-scale numerical atmospheric model, which contains a surface exchange parameterization, especially designed for urban areas was evaluated and further developed. Finally, the area of the full-scale tracer experiment which also contains several sites of other special projects during the IOP (street canyon energetics, satellite ground truth) is modeled using a very detailed physical scale-model in a wind tunnel. In the present paper details of all these activities are presented together with first results.
Full-text available
Recent advances in understanding of the surface energy balance of urban areas, based on both experimental investigations andnumerical models, are reviewed. Particular attention is directedto the outcome of a COST-715 Expert Meeting held in April 2000,as well as experiments initiated by that action. In addition, recentcomplete parameterisations of urban effects in meso-scalemodels are reviewed. Given that neither the surface energybalance, nor its components, normally are directly measuredat meteorological stations, nor are there guidelines for theset-up of representative meteorological stations in urbanareas, this paper also provides recommendations to closethese gaps.
The first measurements of the energy balance fluxes of a dry, densely built-up, central city site are presented. Direct observation of the net radiation, sensible and latent heat flux densities above roof-top in the old city district of Mexico City allow the heat storage flux density to be found by residual. The most important finding is that during daytime, when evaporation is very small (< 4% of net radiation), and therefore sensible heat uses dominate (Bowen ratio > 8), the uptake of heat by the buildings and substrate is so large (58%) that convective heating of the atmosphere is reduced to a smaller role than expected (38%), The nocturnal release of heat from storage is equal to or larger than the net radiation and sufficient to maintain an upward convective heat flux throughout most nights. It is important to see if this pattern is repeated at other central city, or dry urban sites, or whether it is only found in districts dominated by massive stone structures. These findings have implications for the height of the urban mixing layer and the magnitude of the urban heat island.
An observation program using ground and airborne thermal infrared radiometers is used to estimate the surface temperature of urban areas, taking into account the total active surface area. The authors call this the complete urban surface temperature. This temperature is not restricted by the viewing biases inherent in remote sensors used to estimate surface temperature over rough surfaces such as cities. Two methods to estimate the complete surface temperature are presented. Results for three different land-use areas in the city of Vancouver, British Columbia, Canada, show significant differences exist between the complete, nadir, and off-nadir airborne estimates of urban surface temperature during daytime. For the sites and times studied, the complete surface temperature is shown to agree with airborne off-nadir estimates of the apparent surface temperature of the most shaded walls. Some implications of using the complete surface temperature to estimate screen level air temperature and to calculate surface sensible heat flux are given.
A two-layer model of turbulent exchange that includes the view geometry associated with directional radiometric surface temperature is developed and evaluated by comparison of model predictions with field measurements. Required model inputs are directional brightness temperature and its angle of view, fractional vegetation cover or leaf area index, vegetation height and approximate leaf size, net radiation, and air temperature and wind speed. One advantage of the approach described in this paper is that directional brightness temperatures are considered so that the model should have wider applicability than single-layer models, and it opens the possibility of a simple solution if directional measurements are available from two substantially different view angles. Comparisons with several hundred measurements from two large-scale field experiments were performed. One study was conducted in a semiarid rangeland environment in Southern Arizona (Monsoon '90) while the other was conducted in a subhumid environment, namely the tall grass prairie in Eastern Kansas (FIFE). For the Monsoon '90 site, root-mean-square-differences (RMSD) between model predictions and measurements were between 35 and 60 W m−2 for soil, sensible and latent heat flux. With the FIFE site data RMSD values were between 50 and 60 W m−2. The larger scatter with the FIFE data was mainly caused by the model having difficulty reproducing the fluxes for the observation period with dormant vegetation. Considerations of the expected variability associated with flux measurements over complex surfaces suggests that model-derived fluxes were in acceptable agreement with the observations. However refinements in formulations of soil heat flux probably would improve agreement between model predictions and measurements.
Results from an experimental network of seven energy balance stations in and around a European city are presented. The network of micrometeorological stations was part of the Basel Urban Boundary Layer Experiment (BUBBLE) carried out in the city of Basel, Switzerland. Three urban sites provided turbulent flux densities and radiation data over dense urban surfaces. Together with a suburban site and three rural reference sites, this network allowed the simultaneous comparison of urban, suburban, and rural energy balance partitioning during one month of summertime measurements. The partitioning is analysed together with long-term data to evaluate the magnitude of the urban flux density modification, and to document characteristic values in their diurnal and yearly course. Simple empirical relations between flux densities and surface characteristics are presented. The energy balance partitioning is addressed separately for daytime and nocturnal situations. All four components of the surface radiation budget are analysed. Moreover, the vertical flux density divergences within the urban canopy layer are discussed. Copyright  2004 Royal Meteorological Society.