1. Introduction
The heterogeneity of the urban surface presents a challenge to the development of accurate meteorological and air-quality models. The concept of a street canyon (Nunez and Oke 1977), a roadway flanked by buildings, is central to the conception and development of most physical and theoretical attempts to model urban areas at scales from local to mesoscale (Arnfield 2003; Masson 2006). In urban areas devoid of vegetation, this simplification enables the surface to be represented by two land-cover types: roof and street canyon. They have very different radiation and surface energy balances (Harman and Belcher 2006), which locally depend on building materials, geometry, and orientation. As a result, the fraction of the surface covered by street canyons and roofs plays an important role in determining the turbulent sensible heat flux QH from a particular urban area and hence the turbulent characteristics of the urban atmosphere (Harman and Belcher 2006).
Mesoscale urban meteorological models often derive their mean boundary layer turbulent flux input from an average of the surface fluxes from street-canyon and roof surfaces (Masson 2000; Martilli 2002). Successful validation of these models using blended data from the inertial sublayer (ISL) of the urban boundary layer (UBL) further supports this approach: see, for example, Masson et al. (2002). Using a “top down” method, observational studies also imply that in summer clear-sky conditions the turbulent characteristics of the UBL are driven primarily by QH fluxes from the roof surfaces during the day and the release of stored heat from the street canyons during the night (Christen and Vogt 2004; Grimmond et al. 2004; Salmond et al. 2005; Offerle et al. 2006, 2007).
Difficulties associated with the measurement of representative fluxes from individual street roofs and canyons limit the ability to evaluate this hypothesis from a “bottom up” perspective. Close to the surface, in the roughness sublayer (RSL) that includes the urban canopy layer below roof level, turbulent transfer of energy and mass is highly variable in space and is fully three-dimensional. Here, the exchange of momentum, energy, and mass is not restricted to turbulence and may include the mean flow around obstacles (“dispersive fluxes”; Raupach and Shaw 1982). In real cities it is difficult to establish whether flow is responding to a single structure (roof or canyon) or whether it is blended. Hence, Monin–Obukhov similarity (MOS) theory has limited applicability close to the surface and care must be taken to interpret the results in an appropriate framework (Roth 2000). All of this makes it experimentally challenging to estimate the relative contribution of individual surface types to the energy balance at the local (neighborhood) scale in a city.
Given these challenges, it is not surprising that so far there are few examples of field campaigns specifically designed to examine energy partitioning close to the surface or the spatial variability of the fluxes in the urban RSL. Such studies are resource intensive, typically using multiple single-point eddy-covariance (EC) sensors to observe local-scale variability due to differences in solar heating, wind direction, wind speed, building materials, and canyon geometry (Nunez and Oke 1977; Schmid et al. 1991; Eliasson et al. 2006; Offerle et al. 2007). Early studies of simultaneous turbulent exchanges in canyons and above their adjacent rooftops demonstrate the spatial complexity of the urban RSL (e.g., Yoshida et al. 1991; Rotach 1995). Subsequent studies demonstrate differences with height and wind direction relative to canyon orientation (Christen et al. 2009) and between different sides of a single street canyon (Offerle et al. 2007).
Microscale advective effects in the RSL often confound attempts to measure fluxes from a single surface component. For example, measurements of fluxes within street canyons may be affected by the characteristics of the surrounding roof surfaces. During daytime, warm air from the hot roofs may be entrained into the canyon by coherent structures, whereas at night cool air may slump off roofs into the street canyon. This makes it difficult to determine a mean flux for a standard unit street canyon. In a similar way, the continually changing characteristics of the “flapping shear layer” close to roofs (Louka et al. 2000) makes it hard to obtain representative measurements of roof fluxes using point measurements.
Scintillometers offer the ability to make path-averaged measurements of turbulent fluxes of heat and momentum. This is a potentially attractive approach to obtaining spatially representative data. Scintillometry has been shown to be effective in the measurement of turbulent sensible heat fluxes in cities. Small-aperture scintillometers, which measure fluctuations in the refractive-index parameter caused by small turbulent eddies, are well suited for use close to the surface (e.g., de Bruin et al. 2002). They have been used successfully near the top of the RSL and in the ISL of cities (Kanda et al. 2002) and even across the top of street canyons (Roth et al. 2006; Nadeau et al. 2009). Standard scintillometer theory relies on MOS theory, which is only valid in the ISL, but modified forms of the necessary MOS equations have been developed for the RSL (Rotach 1993; Roth 1993; Roth and Oke 1993). Roth et al. (2006) show that by using locally modified MOS equations scintillometers can be effectively used to isolate fluxes generated by individual urban surface-cover types.
The aim of this paper is
to assess diurnal variations of, and differences between, QH fluxes from roofs and street canyons,
to determine whether it is possible to provide a bottom-up approach to the calculation of ISL QH in an urban area by weighting the roof and canyon fluxes within the source area and combining them, and
to discuss limitations associated with such an approach.
2. Experimental considerations
a. Experimental design
1) Experimental setup
The field campaign took place in Basel, Switzerland, as part of the Basel Urban Boundary Layer Experiment (BUBBLE), a project to investigate boundary layer structure over a European city (Rotach et al. 2005). The site for this study (Basel—Sperrstrasse, 47.566°N, 7.597°E) is representative of the urban core of the city [LCZ2 in the classification of Stewart (2011)]. It is located in an area of relatively uniform land use: street canyons bounded by mainly 3–4-story residential buildings that form semienclosed quadrangles with courtyards and gardens in their interior. It is an area with little vegetation. The site is chosen because, in plan view, roofs and canyons dominate the land cover at the neighborhood scale. Therefore, the observational aims of the study are implemented by instrumenting a street canyon and its adjacent roofs (Fig. 1). The intention is to measure separately the turbulent sensible heat fluxes to and from the roofs and the top of the canyon and to see if their net value, when weighted by their respective land-cover fractions in the source area and then summed, is equivalent to that measured by a sensor located in the ISL above the same area.
(a) Map and (b) photograph of the Sperrstrasse street canyon. The lines show the approximate location of the two scintillometer paths. Arrowhead indicates position and direction from which (b) and Fig. 5 were taken.
Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-078.1
(a) Map and (b) photograph of the Sperrstrasse street canyon. The lines show the approximate location of the two scintillometer paths. Arrowhead indicates position and direction from which (b) and Fig. 5 were taken.
Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-078.1
(a) Map and (b) photograph of the Sperrstrasse street canyon. The lines show the approximate location of the two scintillometer paths. Arrowhead indicates position and direction from which (b) and Fig. 5 were taken.
Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-078.1
2) Site
An instrument tower was erected in Sperrstrasse, a canyon (mean width of 13 m and local canyon height-to-width ratio of ~1) considered to be representative of those in the neighborhood (Fig. 1). The tower was located near midcanyon length, between street intersections (Fig. 1). The open-lattice tower reached up to 32 m, slightly greater than 2 times the local average building height zH. It was instrumented with sonic anemometry at six levels. From prior results at this site, it has been shown that the sensor at 31.7 m (2.2z/zH) is located within the ISL (Christen 2005). Local-scale fluxes from the roofs along one side of the canyon and across its top were measured separately, using scintillometers.
Morphometric properties of the area within a 250-m radius of the tower are zH = 14.6 m, standard deviation of building height σzH = 6.9 m, plan aspect ratio (λp; plan area of roughness elements relative to total plan area) = 0.54, mean frontal aspect ratio (λf; frontal area of roughness elements “seen” by the oncoming wind relative to total surface area) = 0.37, and complete aspect ratio (λc; total surface of roughness elements exposed to airflow relative to total surface area) = 1.92 (Rotach et al. 2005; Roth et al. 2006). The plan areas covered by surfaces within a 200-m radius of the tower are 56.6% for buildings (range by wind sector 47.9%–67.7%) and 43.4% for ground.
b. Methods
1) Sonic anemometry
The tower was instrumented with sonic anemometers mounted at six heights above street level: two within the canyon (3.6 and 11.3 m; both Model USA-1 from METEK GmbH, Elmshorn, Germany), one near mean roof level (14.7 m; Model R2O from Gill Instruments, Ltd., Lymington, United Kingdom), and three above roof level (17.9 m: Gill Model R2O; 22.4 m: Gill Model R2A; 31.7 m; Gill Model HS). Raw data were recorded at 20 Hz. The coordinates were rotated with a single rotation into the mean horizontal wind, and linear detrending was performed to calculate (co)spectra, fluxes, and other turbulence statistics averaged over 30 min. The signals were corrected for humidity and density effects (Schotanus et al. 1983; Webb et al. 1980). Data were excluded from analysis if they came from wind directions with flow interference by the tower or by other sensors, if the QH flux density was very low (0–5 W m−2), or if the average wind speeds were too low (<0.1 m s−1). The sonic anemometer mounted at approximately 32 m is hereinafter referred to as the ISL, or EC system, respectively.
The turbulent source areas for the EC system calculated using the analytical source-area model of Kormann and Meixner (2001) are given in Fig. 2. The model was run for each 30-min interval, and ensemble-average footprints were calculated by summing individual runs over a 2 km × 2 km raster with 2-m resolution, following the procedure of Chen et al. (2009). Source areas were summed separately for daytime (1000–1600 LT) and nighttime (2200–0400 LT) cases. From the analysis by Christen (2005), calculations were conducted with the following settings: z′ = z − zd = 31.7 − 10.8 = 20.9 m, z0 = 2.05 m, and zd = 10.8 m. Values of συ, w′T′, u*, and T were taken directly from sonic measurements at 31.7 m for each 30-min step.
Ensemble-average flux source area of the EC system (31.7 m) for all (a) day (1000–1600 LT) and (b) nighttime (2200–0400 LT) cases during the entire IOP. Maps are the spatial average of n = 522 (day) and n = 419 (night) individual source areas over 30 min. Read as follows: the area enclosed by an isoline contributes the given percentage to the measured signal (not flux) at the tower top.
Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-078.1
Ensemble-average flux source area of the EC system (31.7 m) for all (a) day (1000–1600 LT) and (b) nighttime (2200–0400 LT) cases during the entire IOP. Maps are the spatial average of n = 522 (day) and n = 419 (night) individual source areas over 30 min. Read as follows: the area enclosed by an isoline contributes the given percentage to the measured signal (not flux) at the tower top.
Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-078.1
Ensemble-average flux source area of the EC system (31.7 m) for all (a) day (1000–1600 LT) and (b) nighttime (2200–0400 LT) cases during the entire IOP. Maps are the spatial average of n = 522 (day) and n = 419 (night) individual source areas over 30 min. Read as follows: the area enclosed by an isoline contributes the given percentage to the measured signal (not flux) at the tower top.
Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-078.1
For analysis here, roof and canyon-top fluxes are combined according to the plan fraction of the source area each occupies. Details of land cover by direction around the tower are given in Table 2 of Roth et al. (2006). It is assumed that the ~10-m-high buildings found in some courtyards display heat flux characteristics that are similar to those of roofs on the canyon buildings. To test the impact of variability in land cover by wind sector, the analysis is repeated using weighted averages specific to each wind direction. The difference in calculated fluxes between those calculated using the mean land cover and those that use weighted averages specific to each wind direction was <8%. Nevertheless, the difference was deemed sufficient to retain the latter estimates.
2) Scintillometry
Two small-aperture scintillometers (Model SLS20, Scintech AG, Rottenburg, Germany) were placed in close proximity to the tower at Sperrstrasse. One was mounted with its optical path oriented diagonally across the top of the canyon, just below roof level, to give the canyon flux (Figs. 1 and 2). The pathlength was 116 m, and its height was 13.5 m above street level (z/zH = 0.92). The second scintillometer was located ~3–5 m above the irregular roofs along the north side of the Sperrstrasse canyon (Figs. 1 and 2). Its optical path was 171 m long and 19.3 m above street level (z/zH = 1.32).
Turbulent statistics for the scintillometers were calculated initially using a 1-min averaging time and were then averaged over 30 min to correspond in length to the sonic-anemometer measurements. Prior to deployment at the present site, the scintillometers were compared during a warm and clear afternoon at a rural site covered with short grass. The sensors were installed at 1.2 m above the surface with the two paths (pathlength = 150 m) separated horizontally by about 0.7 m. Differences between individual 1-min values were generally smaller than 10%, and agreement was within 5% for 30-min averages, which was the averaging period for the observations in this study. Periods during which the diagnostic software of the scintillometer reported errors of >20% in the 1-min statistics were culled from the dataset. Such data were usually associated with rain events, during which the scintillometer does not work properly. Application of MOS theory is sensitive to accurate determination of the zero-plane displacement height. This is problematic in the RSL because several approaches are valid but provide different answers. Following the sensitivity analysis in Roth et al. (2006), here we use z′ = 3.4 and 4.3 m for the canyon and roof scintillometers, respectively.
The instrument used to obtain the thermal images was a FLIR Systems, Inc., SC500 thermal scanner. The images shown are brightness temperatures, uncorrected for atmospheric or emissivity effects. The instrument was located on the top of a tall (~67 m) apartment building located to the southwest of the main Sperrstrasse canyon site with a fixed view toward the northeast. This view provides a good overview of roof surfaces and the most thermally active canyon wall.
c. Field campaign
Data were collected from 26 June to 12 July 2002 during the BUBBLE summer intensive observation period (IOP). Weather conditions were variable during the study period. Hourly air temperature at 31.7 m varied between 9.9° and 29.8°C (mean 18.8°C). Light winds prevailed throughout the IOP (mean speed of 1.9 m s−1 at 17 m above zH). Average daily shortwave irradiance at the top of the urban canopy was 21.1 MJ m−2 day−1, and average daily total net radiation was 15.8 MJ m−2 day−1.
Three days were cloud free (5, 8, and 12 July 2002), with a peak incoming solar radiation of ~920 W m−2. On those days, wind direction followed a regular diurnal reversal consistent with a well-known thermally driven local circulation associated with the Rhine Valley (Kaufmann and Weber 1996). This gives airflow from the northwest in the boundary layer during daytime, reversing to southeast at night. The pattern was most marked in anticyclonic weather but was present throughout the IOP (Fig. 3).
Time series of wind direction at 31.7 m for the IOP (2–13 Jul 2002).
Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-078.1
Time series of wind direction at 31.7 m for the IOP (2–13 Jul 2002).
Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-078.1
Time series of wind direction at 31.7 m for the IOP (2–13 Jul 2002).
Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-078.1
3. Results and discussion
a. Observed roof and canyon heat fluxes
There are systematic differences in hourly QH from the roof and canyon scintillometer measurements due to the differing geometric configuration and material properties of the two surface components. When QH is small, individual canyon values are slightly larger than those from the roofs, but at large magnitudes the scatter increases considerably and QH over the roofs is the greater (Fig. 4a). The ensemble-averaged values confirm, as expected, that the larger values are measured in the afternoon and that the smaller ones are observed at night (Fig. 4b). Given the purpose and design of roofs, their thermal behavior, with relatively large values by day but relatively small values at night, is as expected. Roof cladding (e.g., tiles, metal sheeting, or membranes) is designed to intercept but not of itself retain heat. Further, the aim is to maintain a reasonably stable thermal climate inside the structure by eliminating large heat gain or loss. This is achieved by insulating the roof either through the use of very-low-conduction material or often by an attic air space between the cladding and the living space.
(a) Scatterplot of 30-min sensible heat fluxes observed by the roof and canyon scintillometers during the IOP. (b) Ensemble-averaged sensible heat flux for the roof and canyon-top scintillometer paths.
Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-078.1
(a) Scatterplot of 30-min sensible heat fluxes observed by the roof and canyon scintillometers during the IOP. (b) Ensemble-averaged sensible heat flux for the roof and canyon-top scintillometer paths.
Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-078.1
(a) Scatterplot of 30-min sensible heat fluxes observed by the roof and canyon scintillometers during the IOP. (b) Ensemble-averaged sensible heat flux for the roof and canyon-top scintillometer paths.
Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-078.1
The roofs on the southeast side of the canyon are angled and covered in ceramic tiles. This is the same for about one-half of the roofs on the northwest side, with the others being flat and covered with gravel on top of a protective membrane. The materials have relatively low shortwave reflectivity (good solar absorptivity), which, combined with their excellent solar exposure, insulation, and lack of permeability giving dryness, generates high roof surface temperatures on sunny days (Fig. 5a). Those roofs sloped to the southeast show brightness temperatures of over 50°C, with maximum values in excess of 60°C. Flat roofs are cooler by 5°–7°C. Roof facets sloped to the northwest average approximately 40°C from the limited surfaces that can be assessed. Further, these hot surfaces are fully open to gusty airflow above the city, which creates almost ideal conditions to support large but variable turbulent heat losses. During the IOP the maximum individual 30-min flux registered by the roof scintillometer was 400 W m−2.
Thermal image of Sperrstrasse canyon and adjacent courtyards for (a) daytime at 1430 central European time (CET) and (b) nighttime at 2300 CET 12 Jul 2002 to show temperature variations that are due to surface materials and facet position periods. The tower is visible as a thin dark (“cold”) line. Viewing direction is to the northeast (see Fig. 1a). No correction has been made for surface emissivity; hence metals appear anomalously “cold” (tower, skylight flashing, and metal roofs).
Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-078.1
Thermal image of Sperrstrasse canyon and adjacent courtyards for (a) daytime at 1430 central European time (CET) and (b) nighttime at 2300 CET 12 Jul 2002 to show temperature variations that are due to surface materials and facet position periods. The tower is visible as a thin dark (“cold”) line. Viewing direction is to the northeast (see Fig. 1a). No correction has been made for surface emissivity; hence metals appear anomalously “cold” (tower, skylight flashing, and metal roofs).
Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-078.1
Thermal image of Sperrstrasse canyon and adjacent courtyards for (a) daytime at 1430 central European time (CET) and (b) nighttime at 2300 CET 12 Jul 2002 to show temperature variations that are due to surface materials and facet position periods. The tower is visible as a thin dark (“cold”) line. Viewing direction is to the northeast (see Fig. 1a). No correction has been made for surface emissivity; hence metals appear anomalously “cold” (tower, skylight flashing, and metal roofs).
Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-078.1
In comparison, the thermal climate and heat flux exchange of the canyon are more conservative, as a result of several factors. Prime among these is the canyon morphometry, including vertical surfaces. A canyon takes the solar input entering across its top and distributes it over a much larger surface area of walls and floor (road). Further, the thermal properties of those surfaces allow them to store daytime heat relatively readily and retain it for later release. The net outcome is that surface temperatures of canyon surfaces are lower than on the roofs by day (Fig. 5a). Southeast-facing walls are approximately 35°C at this time while northwest walls are very close to the canyon air temperature. The most-shaded road surfaces are slightly warmer than the northwest walls and are about 2°–3°C above air temperature, whereas the most sunlit portions of the road are up to 20°C warmer than the canyon air and are 10°–15°C warmer than the sunlit southeast wall. Further, given the shelter found inside the canyon, its wind speeds are subdued and the peak daytime heat fluxes measured by the canyon scintillometer are only about 200 W m−2, about one-half times the roof value.
At night, turbulent heat fluxes from both the roof and the canyon are relatively small and mostly positive. Although it is not possible to determine the direction of the flux from the scintillometer data alone, measurements from the EC system, located near the center of both scintillometer paths, also showed positive values through the night at both locations (see Fig. 8 in Roth et al. 2006). Although this does not ensure that fluxes are positive along the entire path nor that they remain positive for each 1-min averaging period, it is reasonable to assume that they were generally positive throughout the night. Nocturnal turbulent heat fluxes from the canyon are persistently higher after the roof values have dropped to their minimum (Fig. 4b). It is well known that the urban canopy layer remains warm after sunset (see Fig. 5b). The canyon floor is warmer than the average wall temperature (23°C) by approximately 3°C, and the walls in turn are warmer than the roofs by 4°C in this image. Wall temperatures are nearly identical. The canyon warmth is due to the relatively slow release of heat from storage related to the thermal properties and restricted sky view for radiative cooling plus anthropogenic heat emissions from traffic and buildings. The relative warmth of the canyon air promotes intermittent venting to the UBL. In the IOP, nocturnal heat fluxes from both surfaces increase from day to day, which suggests that regional-scale advection of cool air may have been present.
Differences between the roof and canyon QH fluxes are most marked in clear-sky conditions, for which daytime fluxes are consistently higher than on cloudy days (Fig. 6). The difference is greatest for the roofs. Their peak values show an increase of ~120 W m−2 (~100%) as compared with ~50 W m−2 (~50%) for the street canyon. There is also a corresponding net increase in the ISL QH flux of ~100 W m−2 on cloud-free days. As solar input increases, so does the temporal variability of QH observed over the roofs (Fig. 4a). On cloudy days the difference between the daytime peak QH for the roofs and canyon is smaller.
Ensemble-average sensible heat fluxes observed by the canyon and roof scintillometers and the ISL sonic anemometer on (a) sunny days (n = 3) and (b) cloudy days (n = 6) during the IOP.
Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-078.1
Ensemble-average sensible heat fluxes observed by the canyon and roof scintillometers and the ISL sonic anemometer on (a) sunny days (n = 3) and (b) cloudy days (n = 6) during the IOP.
Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-078.1
Ensemble-average sensible heat fluxes observed by the canyon and roof scintillometers and the ISL sonic anemometer on (a) sunny days (n = 3) and (b) cloudy days (n = 6) during the IOP.
Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-078.1
If the assumption that the urban surface can be broken down into two units (roof and canyon) is correct, then the combined fluxes from the roof and canyon-top scintillometers, weighted by the relative abundance of those land covers in the surrounding area, should be approximately equal to the flux measured in an ISL, since that should be a spatially averaged value for the neighborhood. The result of that comparison in this study (Fig. 7) indicates a significant flux underestimation by day and a slight overestimation at night, relative to those measured in the ISL.
Area-weighted, ensemble-average sensible heat flux calculated from the roof and canyon scintillometers and the sensible heat flux from the EC system at 31.7 m, during the IOP.
Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-078.1
Area-weighted, ensemble-average sensible heat flux calculated from the roof and canyon scintillometers and the sensible heat flux from the EC system at 31.7 m, during the IOP.
Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-078.1
Area-weighted, ensemble-average sensible heat flux calculated from the roof and canyon scintillometers and the sensible heat flux from the EC system at 31.7 m, during the IOP.
Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-078.1
Before concluding that this result invalidates the hypothesis that the ISL flux can be obtained additively from area-weighted component-tile fluxes, we suggest it is prudent to investigate whether there are sources of error that might account for this discrepancy. Several possible sources are present in this study: 1) instrumental error, 2) fundamental differences in the source areas sampled by the two techniques, 3) failure to sample representative street-canyon and roof surfaces effectively, 4) effects of local flow patterns, and 5) limitations of the methods to fully capture transfer of energy in the RSL. These are examined in the next section.
b. Discussion of potential errors in the comparison
1) Instrumental error
Past studies that considered intercomparison of high-quality eddy-covariance instruments collocated over the same, relatively homogeneous source area have found differences in average QH fluxes of between ~5% (Mauder et al. 2006) and 8% (Loescher et al. 2005). This is thought to be due to flow distortion by the EC system, different path-averaging or internal-flow-correction algorithms, and differences in the postprocessing. Further, the existence of large organized structures embedded within the turbulent flow (e.g., Roth and Oke 1995) makes single-point eddy-covariance measurements for a particular averaging period not necessarily representative for the spatial average, even over “ideal” (flat and homogeneous) surfaces (Kanda et al. 2004, 2006).
Because of spatial averaging, intercomparisons of QH fluxes from scintillometers over homogeneous terrain show smaller errors on the order of, for example, 3%–5% (Kleissl et al. 2009). If one assumes the worst-case scenario in which the above errors are additive, an 8%–13% difference between the calculated and measured neighborhood-scale fluxes could be accounted for by unavoidable instrumentation issues. The maximum daytime differences in this study between the aggregated scintillometer and eddy-covariance data are larger than 50% and therefore cannot be explained solely by instrumental errors.
2) Differences in the source areas sampled by the two techniques
The models used to estimate the source areas have been developed for homogeneous, two-dimensional surfaces. They do not consider the verticality of the scalar source and sink distributions found in urban areas, which possess three-dimensional morphometry. Exact estimation of the actual source area in an urban context is uncertain using present methods. More significant is that differences in the size and shape of the source area of a scintillometer relative to those of an EC sensor render direct comparison of the two techniques in a city extremely difficult.
The cumulative surface area contributing to 50% of the EC signal at the present site is 4.6 and 7.5 ha by day and night, respectively (Fig. 2). Daytime experiences a wider range of wind directions and stronger lateral mixing, which together with the predominantly northwest winds ensure a relatively uniform urban surface in the source area. At night the source area is biased toward the southeast where some large commercial and exhibition buildings are located. Surface properties and the source-area-averaged building plan area (λb = 44.2% and 43.8% for day and night, respectively) are relatively uniform in the EC cumulative source area. If only buildings that are >10 m (i.e., without courtyard infill) are considered, then λb>10m = 32.4% (30.8%) for day (night). The average building height including courtyard infill is zH = 13.9 (14.4) m during day (night).
The dominant source area of the scintillometer is smaller (primarily because of the lower height of the sensor) and is restricted to within 10–200 m of the instrument (Roth et al. 2006). Within a circle of 200 m around the center of the scintillometer paths, zH is 13.3 m (roof scintillometer) and 13.6 m (canyon-top scintillometer), respectively. The building plan area in the scintillometer source area is 10% higher than that for the EC system, λb = 55.3% (roof) and λb = 53.3% (canyon top) and without courtyard infill λb>10m = 38.4% (roof) and λb>10m = 35.7% (canyon top), respectively.
These estimates of the ensemble-averaged (cumulative) source areas explain some of the bias between the two techniques. For individual runs the contrast is amplified by the heterogeneity of the surface, which introduces additional (unbiased) uncertainties due to the differential representation of the actual surface sampled by the respective sensors, as mentioned below.
3) Failure to sample representative street-canyon and roof surfaces effectively
This study area is relatively homogeneous by urban standards, but the source areas of the two systems might still be either too small or not representative of the neighborhood-scale mix of canyons and roofs. Spatially separated EC measurements in an urban area of Vancouver, Canada, taken at about 3 times the mean building height show that QH can vary by 25%–40% over horizontal scales of 102–103 meters, with increasing variability as the source area becomes smaller (Schmid et al. 1991). Within the EC source area here, there is considerable variation of urban form (street geometry, roof geometry, intersections, and courtyards—some of which are vegetated and others include single-story buildings) and facet materials (asphalt, concrete, tile roofs, gravel roofs, soil, and vegetation), but their influence on turbulence characteristics is likely to be blended by the height of the ISL.
In comparison, the surface morphometry and materials within the source areas of the canyon and roof scintillometers are much more limited. The Sperrstrasse canyon is oriented at 67°–247°. Although measurements from the site are likely to be characteristic of other northeast–southwest canyons within the source area, those that are orthogonal to it (i.e., northwest–southeast) and intersections are not sampled, nor are courtyards. Further, the path of the rooftop scintillometer traverses both flat (~50%) and northwest–southeast-sloping (pitched) roof surfaces (~50%), and the dominant source area is relatively small, as noted in section 3b(2). Different combinations of roof type and pitch orientation than these may be present within the source area of the EC sensor. At different solar zenith angles, this could cause preferred heating patterns that may affect the magnitude of the fluxes.
(i) Roof scintillometer
Recent studies suggest that QH fluxes from roofs dominate urban surface–atmosphere heat exchanges by day (Christen and Vogt 2004; Grimmond et al. 2004; Offerle et al. 2006, 2007). This, combined with the large plan area occupied by roofs within the present source area (~35% >10-m buildings plus 10% courtyard building infill) and the strength of QH fluxes from roof surfaces (section 3a), suggests that if the bottom-up approach is to work then estimation of the roof component must be accurate.
Given the magnitude and dominance of the roof fluxes, we might expect their mean daytime values to be similar to, or higher than, those observed by the EC system in the ISL. That is not the case, however; mean roof QH fluxes are consistently lower than those from the EC approach (Fig. 6). Analysis of the roof QH fluxes by wind direction shows that when winds are from the southeast the peak magnitude is more than double that observed when the winds are from the northwest (Fig. 8a).
Ensemble-average sensible heat flux observed when airflow was from the southeast and northwest for the (a) roof scintillometer, (b) canyon scintillometer, and (c) EC system during the IOP.
Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-078.1
Ensemble-average sensible heat flux observed when airflow was from the southeast and northwest for the (a) roof scintillometer, (b) canyon scintillometer, and (c) EC system during the IOP.
Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-078.1
Ensemble-average sensible heat flux observed when airflow was from the southeast and northwest for the (a) roof scintillometer, (b) canyon scintillometer, and (c) EC system during the IOP.
Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-078.1
Several effects may underlie the lack of agreement and sensitivity to wind direction. First, it may be partly due to the failure to sample other roof pitches and orientations, as outlined above. Second, it is possible that the roof scintillometer may be registering QH fluxes from more than roof surfaces. It may be picking up those emerging from the adjacent canyon and courtyards that are located within its source area. Previous studies suggest that, close to rooftops, the flow is very complex and a flapping shear layer may exist (Louka et al. 2000). This shear layer is thought to play an important role in determining intermittent turbulent exchange between the canyon and the UBL above. Third, the roof scintillometer is centered along the roofline of the northwest side of the canyon (see Fig. 1). Hence, when flow is from the southeast it passes over heated roof surfaces prior to reaching the scintillometer path. There is a corresponding increase in the QH flux of 100%–150% (Fig. 8a). Conversely, much smaller fluxes are observed with northwest flow, which is across the largely northwest-facing pitched roofs that are shaded and relatively cool (Fig. 5a), prior to reaching the roof scintillometer path. This suggests that local differences in roof orientation may be important to setting QH fluxes from roofs. Fourth, the geometry and surface cover of adjacent or nearby courtyards are likely to affect the measured QH flux. The courtyard on the northwest side of the canyon has a relatively high H/W (~0.6), which will restrict solar access, and it is partially vegetated, whereas the courtyard on the southeast side is more open (H/W ~ 0.15), which enhances its receipt of solar radiation. Further, it is mostly filled with commercial 1-story structures covered with metal roofs. It follows that greater QH is likely to originate from the courtyard on the southeast side when compared with that on the northwest side. Both courtyards are in the source area of the roof scintillometer.
(ii) Canyon scintillometer
The diagonal nature of the path across the canyon top might be expected to ensure that measurements from the canyon scintillometer are representative regardless of the asymmetry of solar radiation receipt within the canyon. Roofs of the buildings that flank the canyon differ between the two sides, however, as does their height. This may affect the rate of exchange between the canyon and the UBL above (Barlow and Belcher 2002). For example, on the northwest side of the canyon, roof surfaces are typically 50% pitched and 50% flat, whereas those on the southeast side are all pitched.
Although less obvious than for the roof case, during the daytime there is a small systematic difference between canyon-top fluxes associated with the northwest and southeast; higher values are associated with southeast flow (Fig. 8b). Differences might be related to the extra roughness (TKE) generated by the greater proportion of pitched roofs on the southeast side. Further, with southeast flow, air first passes over the relatively warm southeast courtyard and then the flow encounters the row of hot southeast-facing pitched roofs (Fig. 5a), prior to crossing the canyon.
The scope of this study does not make it possible to determine the impact of only sampling a single street canyon or rooftop on the estimation of neighborhood-scale fluxes. Because both southeast and northwest wind flows were observed by day during the IOP, however, it is possible to investigate (but not to isolate) the potential impact on QH of source areas characterized by different heating patterns. In this case, the southeast-facing pitched roofs are heated for most of the daytime, together with the flat roofs. On the other hand, the northwest roof surfaces receive considerably less direct solar radiation and are cooler (Fig. 5a). As shown later (Fig. 8), the scintillometer results do show considerable variability with wind direction, which is largely absent in the EC measurements (although long-term statistics do show some dependence on wind direction; Christen et al. 2009). This suggests that undersampling of roof and canyon surfaces may be important in the interpretation of our results.
4) Effects of local flow patterns
The local-scale variability in urban structure, materials, radiative heating, and the microscale advection of TKE and sensible heat that depends on wind direction clearly affects the scintillometry observations in the RSL. Since the bottom-up approach uses these fluxes, we ought to see if light can be shed on the discrepancy, seen in Fig. 7, between the two ways to arrive at a neighborhood-scale flux of turbulent sensible heat.
The results show that for southeast winds the agreement between the two approaches is relatively good (Fig. 9a). By day the bottom-up scheme slightly underestimates fluxes, especially the large fluxes around midday. At night this approach overestimates the ISL values, but the flux differences are small (<25 W m−2). This agreement might seem to demonstrate that the weighted-tile approach shows promise. That view is not, however, supported by the equivalent results with northwest winds (Fig. 9b). The bottom-up scheme substantially underestimates fluxes relative to the directly measured ISL flux. It is of course a fact that neither method is considered to be a standard able to yield the true value; there is no standard against which turbulent fluxes can be calibrated. There is little doubt, however, that results from an EC sensor mounted at >2zH above an extensive and relatively homogeneous site are regarded as being the best available at present.
Comparison of ensemble-average sensible heat flux from the bottom-up area-weighted scheme when airflow is from the (a) southeast and (b) northwest. Results are averaged over all cloud conditions during the IOP.
Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-078.1
Comparison of ensemble-average sensible heat flux from the bottom-up area-weighted scheme when airflow is from the (a) southeast and (b) northwest. Results are averaged over all cloud conditions during the IOP.
Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-078.1
Comparison of ensemble-average sensible heat flux from the bottom-up area-weighted scheme when airflow is from the (a) southeast and (b) northwest. Results are averaged over all cloud conditions during the IOP.
Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-078.1
The remarkably clear effect of wind direction evident in Fig. 9, at what was initially considered to be a relatively homogeneous site, begs for further explanation or comment. It is the larger daytime flux estimates that vary with wind direction; nocturnal differences between the bottom-up and ISL values are relatively small. The canyon-top fluxes may or may not be the correct magnitude, but they approximately agree with each other for both airflow directions (Fig. 8b). On the other hand, the much larger roof fluxes show clear directional dependence by day (Fig. 8a). The ISL flux is largely unaffected by the direction of the above-roof flow (Fig. 8c). So with respect to wind direction, it seems that the daytime roof fluxes are largely responsible for the differences evident in bottom-up estimates of the neighborhood flux.
Several factors could be responsible for the differences between Figs. 8a and 8b and between Figs. 9a and 9b, especially during the daytime (Fig. 10a). The complexity can be simplified by considering some underlying systematic controls, on heating, flow, and heat fluxes at this site, as sketched in Figs. 10b and 10c:
The northeast–southwest orientation of the canyon means that the sun preferentially heats southeast-facing walls and sloping roofs rather than their equivalent northwest-facing facets that are largely in shade or are illuminated for only short periods. The combination of this fixed thermal condition with winds that can approach from either side of the canyon means that systematic directional differences in QH are to be expected.
The roof scintillometer path runs along only one side of the canyon (the northwest). Hence, it samples only one set of roofs—those with a mixture of both flat and sloping aspect and that are located downstream from the more vegetated courtyard in northwest flow. The roofs on the other side (the southeast) of the canyon are different—all sloping and placed downstream from the less-vegetated courtyard in southeast flow. Christen (2005) shows that there are systematic differences in the effects on the canyon vortex related to the direction of flow over the two sets of roofs [discussed further in section 3b(5)]. So potentially the roof and courtyard differences between the two sides could contribute systematic directional effects to the tile-average QH. The diagonal scintillometer path across the canyon means it is less likely to be subject to such systematic differences, although the roof flow effects may be relevant.
Illustrative sketch of probable differences in surface heating and the airflow patterns. (a) The site cross section with instrument locations. (b) Daytime northwest flow. (c) Daytime southeast flow.
Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-078.1
Illustrative sketch of probable differences in surface heating and the airflow patterns. (a) The site cross section with instrument locations. (b) Daytime northwest flow. (c) Daytime southeast flow.
Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-078.1
Illustrative sketch of probable differences in surface heating and the airflow patterns. (a) The site cross section with instrument locations. (b) Daytime northwest flow. (c) Daytime southeast flow.
Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-078.1
In hypothetically tracing the path of air as it approaches the roof scintillometer during northwest flow in the daytime (Fig. 10b), we note that it first flows over the courtyard gardens (cool), then up a shaded (cool) northwest-facing wall, and then across the heated flat roof but that its internal boundary layer heat plume probably lies beneath the height of the scintillometer path (see Fig. 1b). So, lacking local buoyant lift or extra heat, the flux measured by the roof scintillometer is probably slightly suppressed. This contrasts with the case of southeast flow (Fig. 10c) wherein air traverses the warm heated metal roofs in the southeast courtyard before encountering the heated southeast-facing walls and roof facets of the southeast side of the canyon. The gabled roof form on that side enhances turbulent activity in the flow just as it crosses the canyon. On the other side, it picks up heat rising from the hot canyon wall and carries it over the heated flat roof. This is likely to include a buoyant plume of warm air from the canyon wall and so it is more likely to affect (boost) the roof flux than when flow is from the opposite direction.
5) Limitations of the methods to fully capture transfer of energy in the RSL
(i) Inappropriate assumption of local turbulence closure
Standard scintillometer theory relies on MOS relationships, which in turn assume equilibrium between local production and dissipation of turbulence. In the RSL here, however, strong horizontal advection and vertical transport of TKE have been observed (Christen et al. 2009). Locally scaled nondimensional turbulence relationships therefore vary spatially, even at a given height, which explains why location-specific similarity functions are different than those derived in the homogeneous ISL (Roth et al. 2006). Given the extreme spatial variability of turbulence and its dependence on wind direction above the canyon and rooftops, the empirically derived similarity functions used in this study, which are representative of general conditions during the IOP, may not fully capture the actual variability present along the scintillometer path for all wind directions and cases. For example, winds from the southeast show higher-than-expected ϕε values (when compared with those from the ISL over a flat and homogeneous surface), suggesting that more TKE is locally produced by shear than is dissipated. A possible explanation is horizontal or vertical advection of turbulence away from the tower, which might be located in the region of strong shear caused by the pitched roofs to the southeast of the tower. In contrast, winds from the northwest show lower-than-expected values for ϕe at the tower location, indicating that more TKE is dissipated locally than is produced locally (Christen et al. 2009).
(ii) Variability of sensible heat flux with height in the RSL
Previous studies based on point measurements unsurprisingly demonstrate that the vertical profile of turbulent sensible heat flux is not constant with height in the urban RSL (Rotach 1991; Roth 1991; Feigenwinter et al. 1999). Such vertical flux divergence could affect the energy balance in the air layers between the sensor pairings of this study, namely between the two scintillometer paths, or between either scintillometer path and the EC system. Because this could contribute to the differences we see between these systems, they are considered here. Vertical divergence could be due to
vertical divergence of the net radiation flux density within the air volume,
vertical divergence caused by sensible heat fluxes to/from active surfaces that may be present in the layer,
vertical divergence of the dispersive fluxes (i.e., transport arising from the spatial inhomogeneity in mean vertical flow stemming from motion around individual obstacles), or
a change of state of the water vapor in the layer as a result of evaporation or condensation associated with cloud (fog) dissipation or formation, respectively.
Figure 11 is a measured profile of turbulent sensible heat flux density in the RSL normalized by the value in the ISL at the study site. It shows a maximum at z/zH = ~1.2 (i.e., just above roof level and about one-half of the way between the two scintillometer paths). Above this height, there is little vertical divergence, but below it a sharp decrease with height (i.e., strong divergence) is observed. This profile of turbulent sensible heat flux density gives the general context of flux observations in the RSL. It is not directly applicable to the scintillometer measurements because the results show point measurements of turbulent fluxes from all surfaces, not just the canyons or roofs like the scintillometers do. Because the canyons and roofs are the two dominant contributors to the total turbulent sensible heat flux, however, we make an untested assumption that their relative input to the total divergence remains constant with height. Given the divergence profile at the level of the canyon scintillometer, and the canyon weighting fraction, it seems that the canyons may contribute about 10% to the underestimate. This may be due to causes 1 and 2 in the list above, especially the irregular mix of built surfaces and atmosphere in the zone straddling the canyon-top layer.
Horizontally averaged vertical profiles of kinematic sensible heat flux for all wind directions (
Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-078.1
Horizontally averaged vertical profiles of kinematic sensible heat flux for all wind directions (
Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-078.1
Horizontally averaged vertical profiles of kinematic sensible heat flux for all wind directions (
Citation: Journal of Applied Meteorology and Climatology 51, 1; 10.1175/JAMC-D-11-078.1
In the ISL, under horizontally homogeneous conditions, the transfer of sensible heat is assumed to be entirely carried out by turbulent exchange of randomly distributed eddies of various sizes. In the RSL, however, the mean flow around objects can further contribute a neighborhood-scale transport of sensible heat, called the dispersive flux (e.g., Poggi and Katul 2008; Raupach and Shaw 1982). This is cause 3 in the above list and may make contributions to QH that are not fully registered by the instrumental setup in Sperrstrasse. For momentum transfer, dispersive fluxes are largest close to the top of the roughness elements, as shown in numerical simulations over cubic arrays by Coceal et al. (2006). Indirect estimates at the current site (using tower measurements) suggest that dispersive fluxes of momentum are relevant in the range 0.75 < z/zh < 1.25 (Christen et al. 2009), which is exactly where the scintillometers are located. It is unclear whether the results for momentum can be applied to sensible heat exchange, but dispersive fluxes could explain an underestimated QH measured by the scintillometers because part of the flux is not transported by turbulence and hence is not registered by the scintillometers. Christen (2005) shows that with northwest flow a canyon vortex develops in Sperrstrasse with a persistent pattern of vertical velocities. This could mean that a significant part of the sensible heat is transported by the mean vortex within the canyon (i.e., a dispersive flux). With flow from the southeast, probably due to the pitched roofs on that side of the canyon, no clear mean flow pattern evolves in the canyon (i.e., no vortex and little vertical mean motion) and hence any contributions from dispersive fluxes would be smaller.
4. Conclusions
It is concluded that daytime roof heat fluxes at the current site are subject to systematic controls that in combination lead to greater values in southeast flow than in northwest flow. Some controls are general and fixed (solar angle), whereas others are specific to the site (canyon morphometry, roof aspect, and courtyards) and experimental arrangement (placement of roof and canyon scintillometer paths) and yet others are due to environmental variability (wind). The results of the bottom-up versus ISL flux comparison are inconclusive. In some flow conditions, reasonable agreement is achieved, possibly even lending support to the notion of summing tiles to obtain area-average fluxes over a city. In other flow conditions there is less support for such an approach. If the heuristic arguments about the flux discrepancies, which depend on the controls listed above, have merit, then it seems that site inhomogeneity and incomplete spatial sampling may provide a basis for explanation. That would mean that weaknesses in the study design and logistics at the site presented here are responsible and the concept is neither proven nor disproven.
That conclusion might logically suggest that using a larger experimental array (e.g., with instruments on both roofs of a canyon and/or use of multiple canyons with different orientation and street intersections) might improve agreement between the bottom-up and ISL estimates of neighborhood fluxes. On the other hand, it is also possible that the site and measurement challenges are too great to allow the question posed in the title to be satisfactorily answered using field measurements in a real-world city. To move forward, it might be helpful to study idealized cases (e.g., scintillometry over model arrays) over which experimental control is more readily established.
Acknowledgments
This field project was made possible thanks to the loan of scintillometers by Drs. Roland Vogt (University of Basel) and Manabu Kanda (Tokyo Institute of Technology). Roland Vogt also provided scientific advice. The core BUBBLE project was funded by the Swiss Ministry of Education and Science (Grant C00.0068). Author MR acknowledges financial support from the National University of Singapore (R-109-000-037-112). Authors TO and JS were funded by the Canadian Foundation for Climate and Atmospheric Sciences (GR-022). The contributions by TO, JV, and AC were also supported by Discovery Grants from the Natural Sciences and Engineering Research Council of Canada.
REFERENCES
Arnfield, A. J., 2003: Two decades of urban climate research: A review of turbulence, exchanges of energy and water, and the urban heat island. Int. J. Climatol., 23, 1–26.
Barlow, J. F., and S. E. Belcher, 2002: A wind tunnel model for quantifying fluxes in the urban boundary layer. Bound.-Layer Meteor., 104, 131–150.
Chen, B. Z., T. A. Black, N. C. Coops, T. Hilker, J. A. Trofymow, and K. Morgenstern, 2009: Assessing tower flux footprint climatology and scaling between remotely sensed and eddy covariance measurements. Bound.-Layer Meteor., 130, 137–167.
Christen, A., 2005: Atmospheric turbulence and surface energy exchange in urban environments—Results from the Basel Urban Boundary Layer Experiment (BUBBLE). Rep. Stratus 11 (Ph.D. dissertation, University of Basel, Switzerland), 140 pp. [Available online at http://edoc.unibas.ch/228/1/DissB_7159.pdf.]
Christen, A., and R. Vogt, 2004: Energy and radiation balance of a central European city. Int. J. Climatol., 24, 1395–1422.
Christen, A., R. Vogt, and M. W. Rotach, 2009: The budget of turbulent kinetic energy in the urban roughness sublayer. Bound.-Layer Meteor., 131, 193–223.
Coceal, O., T. G. Thomas, I. P. Castro, and S. S. Belcher, 2006: Mean flow and turbulence statistics over groups of urban-like cubical obstacles. Bound.-Layer Meteor., 121, 491–519.
De Bruin, H. A. R., W. M. L. Meijninger, A. S. Smedman, and M. Magnusson, 2002: Displaced-beam small aperture scintillometer test. Part 1: The Wintex data set. Bound.-Layer Meteor., 105, 129–148.
Eliasson, I., B. Offerle, C. S. B. Grimmond, and S. Lindqvist, 2006: Wind fields and turbulence statistics in an urban street canyon. Atmos. Environ., 40, 1–16.
Feigenwinter, C., R. Vogt, and E. Parlow, 1999: Vertical structure of selected turbulence characteristics above an urban canopy. Theor. Appl. Climatol., 62, 51–63.
Grimmond, C. S. B., J. A. Salmond, T. R. Oke, B. Offerle, and A. Lemonsu, 2004: Flux and turbulence measurements at a densely built-up site in Marseille: Heat, mass (water and carbon dioxide), and momentum. J. Geophys. Res., 109, D24101, doi:10.1029/2004JD004936.
Harman, I. N., and S. E. Belcher, 2006: The surface energy balance and boundary layer over urban street canyons. Quart. J. Roy. Meteor. Soc., 132, 2749–2768.
Kanda, M., R. Moriwaki, M. Roth, and T. R. Oke, 2002: Area-averaged sensible heat flux and a new method to determine zero-plane displacement length over an urban surface using scintillometery. Bound.-Layer Meteor., 105, 177–193.
Kanda, M., A. Inagaki, M. O. Letzel, S. Raasch, and T. Watanabe, 2004: LES study of the energy imbalance problem with eddy covariance fluxes. Bound.-Layer Meteor., 110, 381–404.
Kanda, M., R. Moriwaki, and F. Kasamatsu, 2006: Spatial variability of both turbulent fluxes and temperature profiles in an urban roughness layer. Bound.-Layer Meteor., 121, 339–350.
Kaufmann, P., and O. Weber, 1996: Classification of mesoscale wind fields in the MISTRAL field experiment. J. Appl. Meteor., 35, 1963–1979.
Kleissl, J., C. J. Watts, J. C. Rodríguez, S. Naif, and E. R. Vivoni, 2009: Scintillometer Intercomparison Study—Continued. Bound.-Layer Meteor., 130, 437–443.
Kormann, R., and F. X. Meixner, 2001: An analytical footprint model for non-neutral stratification. Bound.-Layer Meteor., 99, 207–224.
Loescher, H., and Coauthors, 2005: Comparison of temperature and wind statistics in contrasting environments among different sonic anemometer–thermometers. Agric. For. Meteor., 133, 119–139.
Louka, P., S. E. Belcher, and R. G. Harrison, 2000: Coupling between air flow in streets and the well-developed boundary layer aloft. Atmos. Environ., 34, 2613–2621.
Martilli, A., 2002: Numerical study of urban impact on boundary layer structure: Sensitivity to wind speed, urban morphology, and rural soil moisture. J. Appl. Meteor., 41, 1247–1266.
Masson, V., 2000: A physically-based scheme for the urban energy budget in atmospheric models. Bound.-Layer Meteor., 94, 357–397.
Masson, V., 2006: Urban surface modeling and the meso-scale impact of cities. Theor. Appl. Climatol., 84, 35–45.
Masson, V., C. S. B. Grimmond, and T. R. Oke, 2002: Evaluation of the Town Energy Balance (TEB) scheme with direct measurements from dry districts in two cities. J. Appl. Meteor., 41, 1011–1026.
Mauder, M., C. Liebethal, M. Göckede, J.-P. Leps, F. Beyrich, and T. Foken, 2006: Processing and quality control of flux data during LITFASS-2003. Bound.-Layer Meteor., 121, 67–88.
Nadeau, D. F., and Coauthors, 2009: Estimation of urban sensible heat flux using a dense wireless network of observations. Environ. Fluid Mech., 9, 635–653.
Nunez, M., and T. R. Oke, 1977: The energy balance of an urban canyon. J. Appl. Meteor., 16, 11–19.
Offerle, B., C. S. B. Grimmond, K. Fortuniak, and W. Pawlak, 2006: Intra-urban differences of surface energy fluxes in a central European city. J. Appl. Meteor. Climatol., 45, 125–136.
Offerle, B., I. Eliasson, C. S. B. Grimmond, and B. Holmer, 2007: Surface heating in relation to air temperature, wind and turbulence in an urban street canyon. Bound.-Layer Meteor., 122, 273–292.
Poggi, D., and G. C. Katul, 2008: The effect of canopy roughness density on the constitutive components of the dispersive stresses. Exp. Fluids, 45, 111–121.
Raupach, M. R., and R. H. Shaw, 1982: Averaging procedures for flow within vegetation canopies. Bound.-Layer Meteor., 22, 79–90.
Rotach, M. W., 1991: Turbulence within and above an urban canopy. Dissertation 9439, ETH, Zurich, Switzerland, 240 pp.
Rotach, M. W., 1993: Turbulence close to a rough urban surface. Part II: Variances and gradients. Bound.-Layer Meteor., 66, 75–92.
Rotach, M. W., 1995: Profiles of turbulence statistics in and above an urban street canyon. Atmos. Environ., 29, 1473–1486.
Rotach, M. W., and Coauthors, 2005: BUBBLE—An urban boundary layer meteorology project. Theor. Appl. Climatol., 81, 231–261.
Roth, M., 1991: Turbulent transfer characteristics over a suburban surface. Ph.D. thesis, The University of British Columbia, Vancouver, BC, Canada, 292 pp.
Roth, M., 1993: Turbulent transfer relationships over an urban surface. II: Integral statistics. Quart. J. Roy. Meteor. Soc., 119, 1105–1120.
Roth, M., 2000: Review of atmospheric turbulence over cities. Quart. J. Roy. Meteor. Soc., 126, 941–990.
Roth, M., and T. R. Oke, 1993: Turbulent transfer relationships over an urban surface. I: Spectral characteristics. Quart. J. Roy. Meteor. Soc., 119, 1071–1104.
Roth, M., and T. R. Oke, 1995: Relative efficiencies of turbulent transfer of heat, mass, and momentum over a patchy urban surface. J. Atmos. Sci., 52, 1863–1874.
Roth, M., J. A. Salmond, and A. N. Satyanarayana, 2006: Methodological considerations regarding the measurement of turbulent fluxes in the urban roughness sublayer: The role of scintillometery. Bound.-Layer Meteor., 121, 351–375.
Salmond, J. A., T. R. Oke, C. S. B. Grimmond, S. Roberts, and B. Offerle, 2005: Venting of heat and carbon dioxide from urban canyons at night. J. Appl. Meteor., 44, 1180–1194.
Schmid, H. P., H. A. Cleugh, C. S. B. Grimmond, and T. R. Oke, 1991: Spatial variability of energy fluxes in suburban terrain. Bound.-Layer Meteor., 54, 249–276.
Schotanus, P., F. T. N. Nieuwstadt, and H. A. R. de Bruin, 1983: Temperature measurement with a sonic anemometer and its application to heat and moisture fluxes. Bound.-Layer Meteor., 26, 81–93.
Stewart, I. D., 2011: Advancing the study of urban heat islands through research synthesis and local climate classification. Ph.D. thesis, The University of British Columbia, Vancouver, BC, Canada, 351 pp.
Thiermann, V., and H. Grassl, 1992: The measurement of turbulent surface-layer fluxes by use of bichromatic scintillation. Bound.-Layer Meteor., 58, 367–389.
Webb, E., G. Pearman, and R. Leuning, 1980: Correction of flux measurement for density effects due to heat and water vapour transfer. Quart. J. Roy. Meteor. Soc., 106, 85–100.
Yoshida, A., K. Tominaga, and S. Watatani, 1991: Field measurements on energy balance of an urban canyon in the summer season. Energy Build., 15, 417–423.
