A better understanding of the interaction between the built environment and the atmosphere is required to more effectively manage urban airsheds. This paper reports an analysis of data from an atmospheric measurement campaign in Oklahoma City, Oklahoma, during the summer of 2003 that shows wind flow patterns, turbulence, and thermal effects in the downtown area. Experimental measurements within a street canyon yielded airflow patterns, stability conditions, and turbulence properties as a function of the incoming wind direction and time of the day. Air and surface temperatures at two different sites, one within the downtown urban canyon and the other in a nearby park, were measured. A study of the stability conditions within the urban canyon during the campaign indicates that dynamically stable conditions did not occur within the canyon. This provides evidence that the built environment can strongly influence the thermal characteristics in cities. Mean flow patterns close to the street level are analyzed for two different ranges of incoming wind directions and are compared with those obtained from a previous field experiment featuring idealized building configurations. This paper presents an approach allowing the estimation of wind direction in an urban canyon, given inflow conditions, that shows good agreement with wind patterns in the Oklahoma City street canyon. Turbulence statistics were calculated and normalized using different velocity scales to investigate the efficacy of the latter in specifying turbulence levels in urban canopies. The dependence of turbulence quantities on incoming wind direction and time of the day was investigated.
Central business districts (CBDs) are at the heart of urban areas and are characterized by an array of buildings of different shapes and sizes concentrated in an area of ∼(1–10) km2 In modern planned cities, such as Phoenix, Arizona, the buildings are regularly placed, separated by streets that usually run east–west and north–south, but in older cities the building/road networks are more circuitous and webbed, and thus are difficult to investigate within a general framework. With heightened awareness on urban security (e.g., the need to determine best emergency evacuation routes in case of a chemical or biological release) and because of concerns on pedestrian safety, indoor and outdoor air quality, human comfort, and urban energy conservation, interest in flow patterns within street canyons of CBDs has increased during the past decade. Intense field (e.g., Urban-2000, Allwine et al. 2002; Madison Square Garden-2005, Hanna et al. 2004) and laboratory (e.g., Macdonald 2000; Kastner-Klein et al. 2004; Princevac et al. 2010) observations on urban canopies have been made, and nested modeling systems with grid sizes down to 1 m (Fernando et al. 2001, 2010; Baik et al. 2003) have been developed with some success. The complexity of flow configurations involved, however, causes CBD flows to be largely site dependent, thus limiting the application of simple models and formulations to predict CBD flows and necessitating the use of high-resolution computational models for flow predictions (Calhoun et al. 2004). The development of such models requires fundamental understanding of the interaction between ambient and street canyon flows, which is perhaps best obtained using laboratory and numerical investigations based on idealized geometries. Yet, these models have to be evaluated against field data taken in real field canyons at large Reynolds numbers, which also can provide further fundamental insights on flow complexities. The latter approach is central to this paper, with use of the former to interpret results. In particular, the results obtained during the Joint Urban 2003 (JU2003) experiment are analyzed within the framework of past process studies conducted with idealized geometries, which are reviewed in section 2. JU2003 was conducted in Oklahoma City during 29 June to 30 July 2003 to study flow and dispersion in urban areas, as a part of which a simple, yet representative, street canyon was densely instrumented as discussed in section 3. The results presented herein were obtained using flow and turbulence sensors placed in this street canyon by a number of groups and they are presented and interpreted in section 4. Salient findings of the overall study are described in section 5.
2. Building blocks of urban canopies
As mentioned, given the complexity of building canopies, the flows through them are highly complex and site dependent, and detailed numerical simulations (e.g., Tseng et al. 2006) or scaled physical models (Kastner-Klein et al. 2004) are necessary to predict the flow structure. The approach flow is distorted by the canopy, and flow patterns within the canopy are governed by the building morphology, the approach flow direction θ, and velocity U of the approach flow. Simple flows in idealized morphologies can display essential features of complex flows in real canopies and have been extensively studied. Such studies of flows through idealized canopies provide useful guidance for interpreting, and sometimes estimating, flow through building canopies. Scaled down models of real and hypothetical cities have also been used in fundamental studies on the influence of building mosaics (Terjung and O’Rourke 1980). Detailed reviews on these are given by Hosker (1980, 1982, 1984, 1987), Hunt et al. (1978), Meroney (1982), Wilson and Britter (1982), Woo et al. (1977), and Oke (1988). In the context of field data, most of the work has been on flow above roof canopies (in the roughness sublayer), and a comprehensive review in this context is given in Roth (2000). Below, we shall briefly review several basic flow configurations with the aim of guiding the interpretation of JU2003 urban canyon data.
The well-known configurations of flow over a regular array of identical buildings are shown in Fig. 1. Figure 1i shows a simple configuration, where the along-wind building separation and building dimension are g and w, respectively, the crosswind dimension (breadth) is b and the crosswind gap is s. As the flow passes through the canyon, the communication with the background flow depends on several relevant aspect ratios; for example, λbs = b/s, λbw = b/w, and λhg = h/g. If λbs < 1 and λhg ≪ 1, the buildings are far enough apart that respective wakes of elements are approximately independent and corresponding analysis can be applied (Counihan et al. 1974). Typically in CBDs and where complicating factors such as courtyards and open gateways are present, separation distances are smaller (λbs ≫ 1, λhg ≫ 1). For the case of long buildings (b ≫ s), which is the two-dimensional (2D) case, with flow perpendicular to the street canyon in between the buildings, the flow has been studied extensively. The following flow regimes have been identified: 1) isolated roughness regime (little interaction between individual building wakes) occurs when the sum of frontal separation region (xF ≈ h) and rear (xR ≈ 1.5h) are smaller than the building separation; that is, g > 2.5h as shown in Fig. 1ii(a) (Hosker 1987); 2) at smaller separations (1.4 < g/h < 2.4), the wake of one building interferes with the one downstream [Fig. 1ii(b)]; and 3) at still smaller separations, flow skims over the street canyon, generating stable vortices, with the number of vortices depending on g/h [Fig. 2ii(c)]. These three general flow regimes have been confirmed by a number of studies (Counihan 1971; Hussain and Lee 1980; Raupach et al. 1980). Typical cities have g/h < 1.5 and thus belong to the skimming flow regime.
A significant feature of all canopy configurations is the strong shear layer at the building top, which oscillates with an instability frequency. In the case of skimming flow, vortices so emanating impinge on the windward face of the downwind building [Fig. 1ii(c)] and generate vertical shear layer oscillations with a frequency ω = (g/w)(n − 1/4), where n is an integer (Blevins 1979). The flow at the rooftop, therefore, is unsteady, with a buffeting frequency of ω.
An extension of the above case is that with unequal building heights, as shown in Fig. 2. Here the flow patterns depend on the windward (h1) and downwind building (h2) building heights as well as the separation g. Figure 2 illustrates how the height scale ratios h1/h2 and aspect ratio h1/g affect the flow and the number of vortices (Chan et al. 2003; Xiaomin et al. 2005).
When the wind speeds are small, the thermal effects come into play; buildings tend to be warmer than the surrounding air (say by Δθ), and depending on the intensity of heating either the approach wind or the thermal convection dominate (Kim and Baik 1999, 2001; Fernando et al. 2010). The magnitude of Δθ is dependent on many factors; including the anthropogenic heat flux in the canyon, sky view factor, and the albedo of the building surface (Sakakibara 1996). In addition to the dominance of thermal convection, at low wind speeds, the stability of in-canyon vortices also reduces, thus enhancing the relative turbulence intensity with respect to the mean winds (DePaul and Shah 1986). Sophisticated urban canopy parameterizations have been recently implemented in mesoscale models (DuPont et al. 2004), and such inclusions have indeed improved model predictions for densely populated urban canopies (Park and Fernando 2006).
When the flow becomes 3D, the effects of vertical side edges of the buildings become important, as shown in Fig. 3 for small and large w/h ratios and for the case of a sparse canopy. For taller buildings, h/b > 1, the recirculation bubble behind the building is overwhelmed by the side separation layer, and the resulting intense turbulence can destroy the separation bubble (Lawson et al. 1988). When the buildings are closer, the wakes and sidewall shear layers interact, and a set of parameters become important. For example, for intermediate values of b/s, the flow transitions in the simple case of Fig. 1 become independent on the building breadth and separation g. Edge effects of individual buildings become important in this case, with jetting flow through the gaps between the buildings (Fig. 4a), and the ensuing vortex dynamics is complex. For the case of similar building heights, the flow transition between the regimes is also dependent on multiple dimensionless parameters. For example, according to Hosker (1979), for a plan area of density λp ≈ 17%, the transition from isolated roughness to wake interference regime occurs when the separation g*/h = 1 + 1.4(b/h)0.25 and from wake interference to skimming flow occurs when g**/h = 0.15(b/h) + 1.25 for b/h ≤ 2 and g** ≈ 1.55 for b/h ≥ 2. The recirculation bubble aspect ratio xR/h also is dependent on b/h in this case (Wilson and Britter 1982; Hussain and Lee 1980). Figure 5 shows a general diagram on the influence of three-dimensional effects on flow regime transitions.
The nature of the flow also strongly depends on the angle of incidence (Fig. 4b) and relative height of the adjoining buildings (Fig. 4c). In the 3D case of uneven building heights, the flow becomes complex (Fig. 6), depending on the h1/h2 (Britter and Hunt 1979). A stagnation region occurs on the taller building, thus separating the flow above and below; the downwash is dominant only when the building separation is small enough (g < 2.4h1) to cause the shear layer to impinge on the downwind building (Fig. 6a), and the resulting intense flow causes building wakes to interact strongly and produce intense fluctuations.
If the building separation is large enough, say g/h > 3, then the flow behind the upwind building is much the same as that of an isolated building (Fig. 6b), but with corner vortices playing a role. Britter and Hunt (1979) classified the building clusters as those showing weak interactions (analysis can be performed as if buildings exert a small perturbation to the general flow field, as in Counihan et al. 1974) and those showing strong interactions (where vortex generation and stretching are important). Beranek (1979) introduced a speed amplification factor γ = 6 m s−1/U to determine the effects of vorticity generation and jetting, with low values of γ (γ < 1.5) associated with strong jetting and turbulence generation due to vortex stretching. The opposite case occurs when shorter buildings are located sufficiently close to, but downwind of, taller buildings. Wakes of tall buildings in this case overwhelm the entire wake structure and the intense turbulence can destroy the entire recirculation region of the shorter buildings (Hunt and Carruthers 2004).
Figure 7 shows flow arriving at an approach angle of ϕ (to the normal of the long side of the building). For the case b > 8h and large separations (g ≫ w), a spiral flow is generated behind the building, creating an along-canyon wind (Gandemer 1976). When the buildings are closer, a helical circulation is created for |ϕ| < 60° (Johnson et al. 1973) as in Fig. 7b, and for larger angles a streaming flow is generated (Nicholson 1975; Hunt and Carruthers 2004).
Flow arriving at a normal incident angle ϕ toward a regular array of buildings is shown in Fig. 8. As pointed by Hunt and Carruthers (2004), for a linear (tandem) array, somewhat independent wakes can be seen for smaller ϕ (cosϕ > 2w/g), with weak along-canyon helical circulation (cf. Fig. 7b). At larger ϕ, building wakes merge, resulting in a strong along-canyon flow (velocity Vs) superimposed by swirling motion in the perpendicular direction that extends up to the top of the building. Soulhac (2000) suggested VS/UH ≈ 0.3 for |ϕ| < 60°, noting that this ratio is sensitive to Vs at the upwind end of the canyon (e.g., presence of crossroads). Hunt and Carruthers (2004), however, estimate 0.5 < VS/UH < 1 for wide streets. When |ϕ| < 30°, along-canyon flow takes a greater distance to develop. In contrast, for the case of staggered arrays relative to the wind direction, the wake structure tends to disappear and the mean flow between the buildings consists largely of unidirectional bifurcating flows.
3. Field experiment setup and relevant previous work
During the JU2003 field campaign, measurements were obtained by a number of research groups at a variety of different locations around the urban area, including in the central business district (see Allwine et al. 2002 for broader overview of the experiment). The focus of this paper is on a building canyon within the CBD, known as the Park Avenue. A three-dimensional view of CBD is shown in Fig. 9, and a plan view of the canyon and the building configuration, including heights, is depicted in Fig. 10. The Park Avenue and the immediate proximity were densely instrumented, for example, see Figs. 11a,b, which show meteorological towers deployed by the Arizona State University (ASU) and University of Utah (UU) groups. The instrumentation deployed by the ASU team consisted of four 3D ultrasonic anemometers, radiation sensors, infrared temperature sensors, thermistors, a soil heat flux plate, a soil water content sensor, and a Doppler lidar. The instruments were located at three different sites, referred to as the energy budget station, the sonic tower, and the lidar location.
The ASU energy budget station (Fig. 11c) was located near the intersection of N Walker Avenue and NW 11th Street, 1.3 km northwest from the Park Avenue street canyon. The ASU energy budget station was placed outside of the CBD on a grassy surface surrounded by low-level buildings and some vegetation. This was a semirural site, and measurements obtained therein were used to elicit “urban” effects exhibited by the CBD street canyon. This tower was instrumented with the following: Kipp and Zonen net radiometer at 9.2 m above ground level (AGL), two cup anemometers at 8.9 and 1.5 m AGL, two thermistors at 8.3 and 1.1 m AGL, an IR thermometer, an upward facing pyranometer, and a downward facing pyrgeometer at 3.5 m AGL, a 3D sonic anemometer (Campbell Scientific), a Krypton hydrometer at 2.5 m AGL, a soil heat flux plate (6.5 cm below the ground level), and six thermistors (2 × 2 cm2, 3, 4, 5, and 8 cm below the ground level). A soil water content reflectometer was added on 13 July. Data from the net radiometer, cup anemometers, thermistors, pyranometer, pyrgeometer, and soil heat flux plate were stored as 5-min averages. Data from the IR thermometer, sonic anemometer, Krypton hydrometer, and soil water content reflectometer were stored as 1-min averages. These instruments provided detailed information on the thermal properties of the air and soil as well as radiation measurements. The ASU lidar was located approximately four kilometers to the south-southeast from the CBD, on the southwest corner of 25th Street and Amin Drive. It operated during intensive observing periods (IOPs) and nonintensive observing periods and provided wind information on the approach flow to the city (Calhoun et al. 2006).
The sonic tower was located in the CBD, approximately 22 m from the northeastern edge of the east–west-oriented Park Avenue, delimited by Broadway Avenue at the east end and Robinson Avenue at the west end. Distances from the northern curb edge and the building to the northern side of the canyon were 1.7 and 7.8 m, respectively. This section of Park Avenue is approximately 157 m long. The tower was equipped with three sonic anemometers (ASU1, ASU2, and ASU3) at heights 2.5, 5, and 8.5 m AGL with a sampling rate of 10 Hz. The IR sensor (sampling rate 8 Hz) was mounted on the same tower to measure the street surface temperature. Dataloggers and laptop computers stored raw data from sonic anemometers (three velocity components and temperature), as well as 1-min averages and rms values for surface IR temperature. The sonic at 5-m level and the IR thermometer were operated continuously during the whole campaign. During the first part of the campaign, the two sonics at other levels were operational only during IOPs. However, starting from IOP 6, all three sonics operated continuously.
The ASU sonic tower was one of the six meteorological towers located on the street level of this section of Park Avenue (see Fig. 10). A number of sonics were also located on the rooftops and on tripods at ground level during IOPs. In the Park Avenue canyon, buildings on the southern side are fairly uniform in height (50 m), except for a ∼120-m-tall building at the western end. Buildings on the northern side are also approximately the same height except for a small group of shorter buildings, a taller ∼107-m building in the western end, and a narrow passage near the middle of the eastern half. Since the width of the street canyon is close to 25 m, canyon height to width ratio is approximately h/g ≈ 2, corresponding to the skimming flow regime for winds approaching normal to the canyon (section 2).
Turbulence data were also collected in the Park Avenue canyon by the University of Utah (using a 3D sonic tower) and by Los Alamos National Laboratory (using a rooftop sonic hanging over northeast edge of the canyon at 47.5 m). The UU 10-m tower was located at the street level, at the southwestern part of the urban canyon, approximately 8 m from the southern side of buildings (Fig. 10). It consisted of five 3D Campbell Scientific sonic anemometers (UU1–UU5) operating at 10 Hz. The sonics were mounted at 3.19, 4.19, 5.04, 7.24, and 9.84 m above the ground level (see Nelson et al. 2007a and Ramamurthy et al. 2007 for more details). In addition to those from ASU and UU towers, data from eight Los Alamos National Laboratory (LANL) 2D sonic anemometers (employed only during IOPs) as well as instruments from Oklahoma University (OU) tower and two Defense Science and Technology Laboratory (DSTL) towers were used. Although Fig. 10 shows two OU towers, only the data from northern one were available to us. Four of the LANL sonics were located close to the eastern edge and the other four close to the western edge of the canyon at approximately 2.1 m AGL to better capture horizontal vortices close to the canyon edges. They were placed in such manner to form two rows (northern and southern) and four columns (see Fig. 10). The northern row of 2D sonics was between 0.7 and 1 m far from the buildings on the northern edge of the canyon, while the southern row was approximately 23 m from the northern row of buildings. The distances from the eastern edge of the canyon were 7.3, 12.6, 145, and 149.5 m for four columns of 2D sonics, and their operational frequency was 0.5 Hz. The location of the northern OU tower used in this study was 8.3 m from the northern row of canyon buildings and 90 m west from the eastern edge of the Park Avenue section (see Nelson et al. 2007a; Klein and Clark 2007). This tower had five R.M. Young sonic anemometers mounted at heights of 1.5, 2.96, 5.97, 9.91, and 15.08 m AGL and were operating with a sampling frequency of 10 Hz. The northern DSTL tower was located 8.4 m south and 118 m west from northeast corner of the Park Avenue section, while the southern DSTL tower was 15 m south and 22 m west from the same corner. Each DSTL tower had three Gill 3D sonic anemometers placed at 3, 5, and 10 m AGL on the northern tower and 3.5, 5, 6.5 m AGL on the southern DSTL tower. The sampling frequency from the DSTL instruments was 10 Hz but the data provided from these two towers were reported at frequency of 1 Hz.
Also used in the analysis below are data from a sodar [Pacific Northwest National Laboratory (PNNL)], a tethersonde tower (UU), and sonic measurements from a tower (Indiana University [IU]). The PNNL sodar was located approximately 2.4 km southwest from Park Avenue, at the southern edge of the Wheeler Park. It measured all three wind components as well as rms wind velocities between 30 and 910 m AGL. This enabled us to study how flow patterns inside the urban canyon depend on the incoming flow direction. The IU sonic anemometer was located at a height of approximately 80 m, 5 km south from CBD, and provided information about the mean flow and turbulence properties above the city. Most of the above instruments operated continuously during the whole campaign period and provided useful data on conditions inside the urban canyon, at rooftop level, and for the incoming wind conditions above the city.
Nelson et al. (2007a) analyzed flow patterns within the Park Avenue for selected time periods and incoming wind directions. They showed strong dependence of mean flow patterns and turbulence intensities on wind direction and suggested that idealized canyon studies are more applicable to European-type cities characterized by less variability in building heights. The along-canyon vortex typical of studies within idealized geometries was not detected. Detailed analysis of measured velocity spectra, cospectra, and quadrant analyses are given in Nelson et al. (2007b). Ramamurthy et al. (2007) investigated the influence of upwind stability conditions on turbulence statistics within the Park Avenue. They observed that momentum-related turbulence statistics is insensitive to upstream atmospheric stability, while the potential temperature and kinematic heat flux showed more sensitivity. The work by Klein and Clark (2007) focused on analysis of measurements from two towers in the central part of the Park Avenue canyon. They observed that small changes in incoming wind direction can cause drastic changes in flow within the canyon. Significant along-canyon flow exists even for winds perpendicular to the canyon. Two tall buildings within the canyon play an important role by enhancing downward mixing of higher momentum fluid from above. When wind speed at 80-m level was used for data normalization only minor stability effects were observed, while use of wind speed at 250 above ground showed strong influence of stability on in-canyon flow properties (e.g., turbulence).
a. Thermal effects in the urban core
Different thermal properties of surface materials in urban areas significantly influence the properties and stability conditions of flow inside the urban canyon. This is a consequence of increased heat capacities of dominant urban surfaces, such as asphalt and concrete, compared to vegetated surfaces in rural areas. The location of the energy budget station is not ideally representative of a rural site since it was located inside the city, but site characteristics such as surface properties (grassy area) and the significant distance from closest buildings make it a good approximation of a rural site. Although the distance between the two sites is only 1.3 km, their completely different land use characteristics created a significant temperature difference, clearly indicating the existence of an urban heat island.
Surface temperatures (5-min averages) measured at two sites, the Park Avenue street canyon site and the energy tower site (open grassy field), are presented in Fig. 12a for a 5-day period (190 to 195 day of year; 10 to 15 July). A sharp increase of temperature above the open grassy field begins to occur (∼0830 local times several hours after the sunrise (∼0630 LT). Morning heating in the urban canyon is slower because of shading effects of the buildings, and its surface temperature starts its fastest growth around 1030 LT, when solar energy begins to penetrate the urban canyon. During the daytime, the surface temperature in the urban canyon is slightly lower, again because of the shading effects of buildings and trees located on the sidewalk. Temperature begins to drop rapidly around 1600 LT at both sites, the sunset being at ∼2045 LT. After approximately one hour, however, the temperature drop in the urban canyon slows down, though it continues to cool steadily throughout the night until warming takes place in the morning. In contrast, the surface temperature in the open field drops rapidly (∼9°C h−1) until it reaches a state with a very low cooling rate (∼0.17°C h−1) that lasts until morning. Surface temperature difference (Park Avenue minus open field) for the same time period is shown in Fig. 12b. The largest difference of about −15°C is reached in morning hours (around 1030 LT) when the grassy field surface warms up rapidly relative to the urban core. The large temperature difference between the Park Avenue and the grassy field (∼10°C) also occurs in the evening hours (2030 LT) when the surface temperature at the grassy field begins to drop quickly.
Different surface properties and local site characteristics have an impact on the air temperature as well. Figure 13 shows the air temperature difference between the two sites at 8.5 m AGL. The daytime temperature inside the canyon is lower because of shading effects of buildings, while during the night the temperature is higher. The latter is expected because of higher urban surface temperature, additional anthropogenic heat fluxes (e.g., heat from buildings), and the trapping of radiation between tall buildings. Radiation emitted inside the canyon is reflected, reabsorbed, and reemitted by tall buildings. Although the results should be viewed with caution because of large measurement error (0.5°C) of the thermistor used at the open grassy field, the trend is obvious and consistent for the period of 11 days shown here.
Stability conditions can be broadly divided into three classes based on the parameter z/L (Rotach 1995), where z is the height above the ground and L is the Monin–Obukhov length scale: near neutral (|z/L| ≤ 0.05), weakly unstable (−0.5 < z/L < −0.05), and strongly unstable (z/L < −0.5). The Monin–Obukhov scale is defined as
where u*I is the friction velocity within the inertial sublayer, k is the von Kármán constant, β is the thermal expansion coefficient, g is the gravitational acceleration, and w′ and T ′ are the fluctuations of vertical velocity and temperature, respectively.
A 3-day time series of the parameter z/L is shown in Fig. 14, where the local value of L and two different levels on the UU tower are used. At the 9.8-m level, the parameter values are mostly within a range corresponding to near-neutral conditions with relatively short periods of weakly unstable conditions. Measurements by sonic closer to the ground, at 3.2 m, indicate mostly weakly and strongly unstable conditions with shorter periods of neutral conditions. This indicates that close to the ground there is a layer where stability conditions are almost always unstable during the period of observations. This is further demonstrated by plotting the profiles of stability parameter z/L versus nondimensional height z/H (Fig. 15) for different parts of the day (two hour periods) averaged over the whole period of the experiment. Note that the lower three sonic anemometers measure weakly unstable conditions both during the day and night. During the daytime, all sonic anemometers indicate weakly unstable conditions, with the lowest sonic closer to the transition between weakly and strongly unstable conditions and the upper sonics approaching (increasingly with height) stability parameters corresponding to neutral stability. During the night, profiles are shifted toward neutral conditions. The three lowest sonic anemometers (3.19, 4.19 and 5.04 m AGL) still measure weakly unstable values of the stability parameter, although absolute values are much lower, while the highest sonic (at 9.84 m AGL) indicates near-neutral conditions. The fourth sonic (at 7.24 m AGL) on the tower indicates that stability conditions are at the border between weakly unstable and near neutral.
Temperature profiles obtained with the UU tethersonde system (15-min averages) indicate near-neutral conditions inside the canyon (see Fig. 16). Profiles presented were taken during the daytime (1600) and nighttime (0300). During daytime, the temperature slightly decreases with height while during the night it increases, but these changes are small. During the day, the temperature difference between the lowest and the highest sonde is approximately 0.5 K, while during the night it is even smaller, less than 0.2 K. These observations show that stable conditions did not occur inside the Park Avenue canyon during the period studied here.
b. Flow patterns
The winds within the canyon respond to upstream and local conditions and morphology, and hence it is difficult to synthesize a general interpretation for the entire dataset. Nevertheless, the results can be interpreted on a broad framework using knowledge gained from previous studies of idealized buildings described in section 2. Since winds arrive mainly from the south, detailed analysis of only the following cases were considered: (i) the wind angle is between 130° and 180° (southeasterly), and (ii) the wind angle is between 180° and 215° (southwesterly). The upstream wind direction used here was measured at 250-m level using the PNNL sodar.
The dashed lines in Fig. 17a show the upstream wind direction and its interception with the building cluster for the case of 130°–180°. Shown at the end of the dashed lines are the corresponding wind patterns pertinent to the idealized cases discussed in section 2, assuming that the flow behaves similar to an idealized equivalent having a similar cross section (i.e., neglecting the along-canyon variation). If the flow is dominated by disparate velocity patterns induced by two close building features, the resultant flow direction can be derived by the vector sum of individual velocities inferred from the idealized cases. The measurement locations are marked by A, B, C, etc., and wind directions inferred for these locations are plotted. Corresponding measurements are shown in Fig. 17b and agreement is generally good. Only the flow direction can be inferred by this method, however, because no explicit parameterizations exist for velocity magnitude even for the simplest possible idealized cases.
Since for the above case the incoming winds are southeasterly to southerly close to the ground, air enters the canyon from eastern side, causing a large vortex in the horizontal plane at the eastern edge of the canyon. This feature is obvious from the flow measurements, where sonics on the northern side of eastern edge (A and B) show easterly winds while instruments on the southern side (M and N) measure westerly flows. Flow above the midwestern half of the southern row of buildings (48 m) impinges on the taller building on the opposite side (63 m), causing the flow to split toward the east and west, respectively. For this reason, G and F measure easterly winds, and D measures westerly to northwesterly winds. Sonic J is expected to measure northerly winds based on the inset (smaller building upstream of a larger downstream), as a result of the smaller vortex near the ground as shown in the inset. On the southern side toward the middle of the canyon, southerly winds can be expected (K and L), based on the inset depicting a taller building upstream of a shorter building. Sonics K and L are closer to the center of the street than 2D sonics, and they measure flow due to middle vortex shown in the corresponding inset. Since the building on the western edge of southern row is taller (123 m) than the one next to it (48 m), wind flowing over the shorter one interacts with the taller building to create a corner vortex, thus causing westward flow at H and I. The passage on the north side of the canyon channels the flow that converges from both east and west (westerly winds at D and easterly at B). Measured wind directions (Fig. 17b) indicate general agreement with flow patterns inferred by combining what is expected from idealized building configuration studies.
Similar analysis was performed when the incoming winds were in the range 180°–215° (Fig. 18). In this case, because of the presence of a westerly approach wind component, air enters the canyon from the western side, forming a vortex in the horizontal plane. This agrees well with measurements: G and F measure mainly westerly winds while H and I on the southern side measure easterly winds. A horizontal-plane vortex forms on the eastern edge of the canyon as well, and the flow channels through the narrow passage on the northern side, as in the previous case. This causes opposite flow directions at D and B. An analysis similar to that of Fig. 17, where insets indicate simple cases of two buildings, was used to estimate wind direction in the middle of the canyon on the southern side. Expected wind directions are southwesterly, which is in general agreement with the observations. A notable deviation could be seen at the L location, where the inferred patterns are southwesterly, whereas measurements are southeasterly–northeasterly with high variability.
c. Turbulence measurements
As mentioned, the distribution of turbulence quantities within a street canyon is crucial for many practical applications, such as the dispersion of contaminants. Turbulence intensities are of particular interest as they are robust quantities and can be independent of the wind direction. It has been argued that turbulence intensities can be independent of wind speed. Available measurements above the urban canopy layer in the roughness sublayer (RSL) have been reviewed by Rotach (1995) and Roth (2000), which indicate that turbulence intensities, scaled with the local friction velocity u*(z) = [(u′w′)2 + (υ′w′)2]1/4 tend to be constant. Another plausible scale is the friction velocity u*I based on the Reynolds stresses within the inertial sublayer (>2HB). Britter and Hanna (2003) suggested σu = 2.4u*I, συ = 1.91u*I, σw = 1.27u*I (also see Roth 2000), but alternative proportionality constants have also been reported; for example, Rotach (1995) suggested 1.7, 1.5, and 1.0, respectively. Fernando et al. (2010) explained the disparities as due to differing morphological parameters and stressed the importance of wind direction and thermal forcing. The results may also depend on the stability of the flow, which has been classified according to the z′/L values (Rotach 1995), where z′ = z − d, d is displacement height, and L is the Monin–Obukhov scale [see Eq. (1)].
Locally the turbulence is heavily dependent on factors such as the approach wind speed and direction and the building morphology. One may agree that the scaling velocity u*I is not the best candidate here, given that it is determined by inertial sublayer contributed by the interaction of many street canyons flows (i.e., an ensemble effect of city blocks). On the other hand, local u*(z) is highly spatially variable within the canyon and usually becomes very small under a certain height. Given these considerations, u*I has been used as a representative velocity. In cases in which u*I drops below the in-canopy convective velocity because of thermal effects,
(where hb is the building height, α is the thermal expansion coefficient, ρ0 is a reference density, and cp is the specific heat at constant pressure) can be used and u*I should be replaced by w*. Note that Hs is the heat flux within the canyon, but lacking reliable data for anthropogenic heat flux, Hs was approximated by the sensible heat flux. Previous studies of Snyder [quoted in Britter and Hanna (2003) and conducted with λf = 0.027] and Macdonald (2000; λf = 0.0625) show that σu/u*I ≈ 1.6, συ/u*I ≈ 1.4, σw/u*I ≈ 1, where the u component is along the canyon direction. Based on JU2003 data, Hanna et al. (2006) proposed for sonics based at z1 = 8 m height in the Park Avenue canyon that σh = 3.63u*(z1) and σw = 1.56u*(z1), where σh2 = σu2 + συ2. Similarly, for the Madison Square Garden experiment, they found σh = 5.63u*(z1) and σw = 1.54u*(z1).
More recently, attempts have been made to develop an in-canopy velocity scale uc. Bentham and Britter (2003) proposed, by projecting the obstacle drag on the surface area to obtain an equivalent friction velocity u*B, an in-canopy velocity of
where u*B = and τwB is the equivalent stress on the floor based on the total force FB acting on built elements, FB = τwBAfloor. Note that the measurement of τwB is not trivial. On the contrary, Zajic (2006) proposed, based on a bulk momentum balance, a canopy velocity scale of
where λfs is the frontal solidity (Fernando et al. 2010), CD is the representative drag coefficient of building elements, u*a is the friction velocity of the approach flow, and z0a is the roughness length of the upstream terrain.
Assuming that u*I based on the IU tower located ∼5 km from the city downtown gives a good indicator of approach friction velocity u*a for southerly winds, it was used to normalize the measured turbulence quantities. Also note that this upstream value can be used as u*a in evaluating uCZ. Figure 19 gives a comparison of normalized (by u*I) rms quantities, for UU, ASU, and LANL sites. Although the height of the LANL sonic anemometer used here is close to the building height hB and thus significantly higher than the other sonics, it is included in plots (star symbol) for comparisons. The average height of the building cluster was assumed as hB = 50 m. In calculating rms values, 15-min averaging was used, and the normalized rms values were again averaged over the entire observational period. The standard deviation of the latter is also shown as error bars, which is quite substantial.
In all cases, turbulence levels at the middle sonic (5 m AGL) at the ASU tower are lower than those measured by other sonics of ASU and UU towers, which could be attributed to the presence of a tree canopy in the immediate vicinity of the tower at this level. Maximum values of rms velocities (of all three components) were measured by the LANL sonic located close to the building top, which is not surprising because of the strong influence of the separating shear layer at the building height.
The highest sonic of the ASU tower measured higher turbulence levels than the lowest one, which is probably due to shear caused by the tree canopy located underneath the sonic. Results from the UU tower showed almost a constant rms horizontal velocity with height, although there was a slow increase of the vertical rms. In almost all cases, turbulence levels of the UU tower were higher than those measured at the ASU tower.
Figure 20 shows a similar plot to Fig. 19 except using the max(u*I, w*) for velocity scaling (i.e., u*I for u*I > w* and w* for u*I < w*). Conclusions are the same as in the previous case, and values are also similar to those in the previous plot. Figures 21 and 22 show nocturnal (2300–0500 LT) and daytime (1000–1900 LT) normalized rms velocities using u*I and max(u*I, w*), respectively, as velocity scales. Note that the nocturnal normalized values are higher (about 50%), which is a consequence of similar stability conditions inside the canyon during day and night (in the range −0.5 < z/L < 0 measured at different levels at UU tower). However, at the upstream IU tower location, the u*I values drop substantially at night because of stable stratification. During both the day and night, the LANL sonic close to the building top measured the highest values, while ASU sonics measured the lowest.
Figure 23 shows the total rms distribution, normalized by uCB [Eq. (3)]. In calculating uCB, a local canopy value of u* was used, and we opted to use average u* from the lowest sonics of UU and ASU towers. Values plotted represent the average for the whole period of the experiment. Table 1 summarizes values of all three rms velocities normalized using three different velocity scales for all incoming directions and over the whole period of the experiment. Since the IU sonic was located 5 km south from the CBD area, it can be assumed to measure upstream winds (as winds were typically southerly). Owing to the complexity of CBD geometry, frontal area density (λf) depends on the incoming wind direction, and according to Brown et al. (2004), it can change in the range 0.25–2.0. If we use CD = 1 and hb/z0a = 20, uCZ should be in the range ucz ≈ (0.3 − 2.5)u*I. Since the midrange of possible values of uCZ is reasonably close to u*I, which was already used for normalization, for the case of southerly winds, normalization with u*I can be used as a measure of performance of the velocity scale uCZ.
Finally, it is useful to investigate whether the rms turbulence velocities are sensitive to the wind direction, given the observation that a small shift of wind patterns can have a significant change of the circulation patterns, as shown in Figs. 17 and 18. To this end, the rms velocities (15-min averaged) calculated for the two cases, 130 < θ < 180 and 180 < θ < 230, were averaged over the entire experiment and the results are shown in Figs. 24 and 25 for UU and ASU sonics. When u*I is used for normalization (Fig. 24), lateral and vertical velocity rms’s are almost the same for ASU and UU towers, while LANL measures higher values for southeasterly (130 < θ < 180) winds. Along-canyon rms velocity measured close to the street level (ASU and UU towers) is very different for the two cases (wind directions ranges) analyzed, while the values measured close to the top are closer. Close to the street level, along-canyon rms velocity is higher for southwesterly (180 < θ < 230) winds, especially at the UU tower. When the uCB is used as a velocity scale (Fig. 25), all three rms velocities differ for two different incoming wind directions, with higher values registering for southwesterly winds.
The central business district is an important part of every urban population center, and there has been an increase in number of studies on flow and contaminant dispersion through building clusters that characterize CBDs. For this reason, part of the Joint Urban 2003 field experiment was focused on a detailed study of flow patterns and turbulence inside a section of Park Avenue in the Oklahoma City CBD. A large number of instruments provided information about the wind speed and air–surface temperatures on the street level and at building height. The ASU team also had another tower in the park just north of the CBD that enabled us to compare conditions inside the urban core to those outside. It was shown that stable stratification did not develop inside the Park Avenue during the period of observations. This is due to the higher heat capacity of the urban surface relative to the rural one, anthropogenic flux, and the trapping of radiation by the buildings in the CBD. This also causes differences in surface and air temperatures. Temperatures were higher at the semirural site during the day, while during the night temperatures in the Park Avenue street canyon were higher.
During the campaign period, the dominant large-scale winds were from the south when IOPs were conducted. During observational periods, the Park Avenue section was densely instrumented, as additional sonics were deployed during these periods. Flow patterns were studied in detail for winds coming from south-southeast and south-southwest. Flow patterns showed high sensitivity to large-scale wind direction changes, but some general conclusions can be made. Close to the canyon edges large vortices form in the horizontal plane and flow tends to channel through an opening on the northern row of buildings. Comparison with flow patterns predicted based on previous studies in simplified configurations showed good agreement with measurements. It appears that the use of canonical flow configurations as buildings blocks to construct more complicated multibuilding configurations works well for this case, and this methodology ought to be tested for other urban areas as information becomes available.
Turbulence statistics were calculated and analyzed using different available velocity scales for normalization. Analysis was conducted in order to observe how turbulence intensities depend on incoming wind direction as well as on the time period of the day. As expected, measurements near the top of the buildings showed the highest levels of turbulence, while on the street level turbulence intensities depended on location within the canyon and the immediate surroundings. Dependence of rms velocities on the incoming wind direction and time of the day was also observed.
A better understanding of the mean flow, turbulence, and thermal characteristics within urban areas is paramount for developing parameterizations for models dealing with contaminant dispersion, energy usage, and human comfort in densely build urban areas. The rapid expansion of cities creates larger areas of increased drag, roughness (due to buildings), and thermal forcing (due to anthropogenic heat flux, radiation trapping, and high heat capacity built elements) that can significantly impact the flow patterns on regional scale. For this reason it is important to provide urban models with a most appropriate description of thermal conditions and turbulence within urban regions as a function of synoptic meteorology so that the influence of larger flow patterns on urban scales can be predicted through downscaling. Another important goal of our work was to investigate the applicability of flow studies conducted in idealized geometries and in controlled laboratory conditions to real world situations. This would enable first-order predictions of flow through building canopies that enable planning and interpretation of field experiments and delineate nonlinear phenomena that arise because of strong interactions between processes that are intrinsic to canonical building configurations.
We thank NSF (ATM, CMG) and Army Research Office for supporting the analysis.
Corresponding author address: D. Zajic, Center for Environmental Fluid Dynamics, Department of Mechanical and Aerospace Engineering, Arizona State University, Tempe, AZ 85287-9809. Email: firstname.lastname@example.org