Urban wind field analysis from the Jack Rabbit II Special Sonic Anemometer Study

Abstract

The Jack Rabbit II Special Sonic Anemometer Study (JRII-S), a field project designed to examine the flow and turbulence within a systematically arranged mock-urban environment constructed from CONEX shipping containers, is described in detail. The study involved the deployment of 35 sonic anemometers at multiple heights and locations, including a 32 m tall, unobstructed tower located about 115 m outside the building array to document the approach wind flow characteristics. The purpose of this work was to describe the experimental design, analyze the sonic data, and report observed wind flow patterns within the urban canopy in comparison to the approaching boundary layer flow. We show that the flow within the building array follows a tendency towards one of three generalized flow regimes displaying channeling over a wide range of wind speeds, directions, and stabilities. Two or more sonic anemometers positioned only a few meters apart can have vastly different flow patterns that are dictated by the building structures. Within the building array, turbulence values represented by normalized vertical velocity variance (${\sigma }_w^2$) are at least two to three times greater than that in the approach flow. There is also little evidence that ${\sigma }_w^2$ measured at various heights or locations within the JRII array is a strong function of stability type in contrast to the approach flow. The results reinforce how urban areas create complicated wind patterns, channeling effects, and localized turbulence that can impact the dispersion of an effluent release. These findings can be used to inform the development of improved wind flow algorithms to better characterize pollutant dispersion in fast-response models.

1. Introduction

Dense and irregular building geometries within urban areas lead to complex turbulence, dispersion, and flow regimes that significantly influence urban pollutant concentrations. These situations present challenges in estimating effluent dispersal and associated human exposures in highly populated regions. Boundary layer wind profiles are altered by wake turbulence around buildings and channeling within street canyons (Belcher et al., 2013; Barlow and Coceal, 2009; Britter and Hanna, 2003). This results in reduced wind speeds, greater turbulent intensities and turbulent kinetic energy (TKE) in the lee of buildings, and a tendency toward neutral stability due to added mechanical turbulence (Arya, 2001; Briggs, 1973). A better understanding of the turbulent wind flows within urban areas is fundamental to the successful development of improved urban parameterizations in fast-running dispersion models (Lateb et al., 2016), such as those that are Gaussian or Lagrangian-based. In relation to accidental or intentional hazardous releases, these models help inform emergency preparation and response efforts by approximating the extent of a released plume and indicate where to evacuate, sample, decontaminate surfaces, and manage waste generated from the remediation process. Improved dispersion research can aid the US Environmental Protection Agency’s (EPA) emergency response mission in preparing for and responding to large-scale chemical, biological, radiological, and nuclear (CBRN) incidents. As such, the focus of the work presented in this paper is to analyze the wind flow patterns from the recent Jack Rabbit II (JRII) Field Study within a mock urban building array to inform wind flow and dispersion algorithms.

Dispersion may be fairly well represented as a Gaussian distribution when considering the average plume at citywide scales, but the complex flow around individual or small groups of buildings requires a more detailed description of the pollutant plume. As the airflow encounters an obstacle, such as a building with sharp edges, separation of the flow occurs due to the surface’s abrupt discontinuity. As the flow (and plume) is forced up and over the building, high turbulence intensity and reverse flow form a recirculating cavity zone on the lee side of the building (Monbureau et al., 2018), which may quickly bring contaminants to the street level and lead to downwind stagnation, especially under low wind speeds (DePaul and Sheih, 1986). This flow effect can entrain outside pollution and accumulate any contaminants emitted inside the cavity, leading to elevated concentrations due to low wind speeds, but high wind shear and turbulence intensity (Arya, 2001). The size of the cavity depends on the along-wind length (L_b), crosswind width (W_b), and height (H) of the building, aspect ratio of the street canyon (H/W_c, where W_c indicates the street canyon width), and the characteristics of the approach flow. Sufficiently wide street canyons (H/W_c < ~0.3) may retain much of the building-induced wake effects seen around isolated structures. However, the disturbed air might not have enough downwind distance to return to ambient conditions before encountering the next obstacle as buildings become more closely spaced (H/W_c ≈ 0.5), creating wake interference flow (Vardoulakis et al., 2003; Oke, 1988). Within narrow street canyons (H/W_c > 1), skimming flows over the tops of buildings often leave conditions at breathing level relatively stagnant (Grimmond and Oke, 1999).

Effluent dispersion within urban environments has been a popular research area for many decades through field and laboratory studies and mathematical modeling approaches. Meteorological wind tunnel experiments (e.g. Fuka et al., 2018; Garbero et al., 2010; Brixey et al., 2009; Heist et al., 2009) have helped develop a fundamental understanding of flow and turbulence within an array of urban structures and have supported the development of regulatory and operational dispersion models such as EPA’s AERMOD model (Cimorelli et al., 2005) and the Interagency Modeling and Atmospheric Assessment Center’s (IMAAC) model suite (Nasstrom et al., 2007). Laboratory experiments have also informed well-known “rules of thumb” (Brown and Streit, 1999) used by emergency responders to predict downwind dispersion for various source heights and simple building configurations.

Field experiments are also invaluable for understanding how urban structures channel wind flows, enhance turbulence, and affect atmospheric stability. In 2001, the Mock Urban Setting Test (MUST) was performed at the US Army’s Dugway Proving Ground (DPG) using an array of CONtainer Express (CONEX) shipping containers designed to represent a simplified, near full-scale urban area with neutrally buoyant tracer releases (Biltoft, 2001). While there were only five sonic anemometer towers to measure turbulence, the MUST experiment focused on the concentration fields by quantifying the degree to which the lateral and vertical plume spread was enhanced, and how the centerline concentration decreased compared to an open-terrain plume (Yee and Biltoft, 2004). Various tracer experiments have also been carried out within existing urban environments, such as the Join Urban 2003 (JU03) Tracer Field Tests in Oklahoma City (Clawson et al., 2003), London’s Dispersion of Air Pollutants and their Penetration into the Local Environment (DAPPLE) (Arnold et al., 2004), and EPA’s Brooklyn Traffic Real-Time Ambient Pollutant Penetration and Environmental Dispersion (B-TRAPPED) study (Hahn et al., 2009). Using tracers such as SF₆ and a network of sonic anemometers, these types of experiments inform dispersion on fine, neighborhood-scales (Carpentiri and Robins, 2015) and in street networks (Hertwig et al., 2018) while employing actual, complex urban morphologies.

More recently, a series of field studies called Jack Rabbit have simulated the accidental release and dispersion from toxic industrial chemicals (TICs). When the release involves a dense gas such as chlorine or ammonia in a built-up area, the dispersion of TICs can lead to elevated risks of exposure by adversely affecting human health. The 2010 Jack Rabbit I release trials were led by the Department of Homeland Security (DHS) to examine a chemical jet release and the associated dense gas behavior (Hanna et al., 2012; Fox and Storwold, 2011). In 2015, 2016, the DHS performed an expanded study called JRII at DPG (Gant et al., 2018) where a portion of the experiments were performed within a mock-urban environment designed from CONEX shipping containers. Large quantities of chlorine gas were released to study downwind dense gas concentrations for improving dispersion models and emergency response techniques (Nicholson et al., 2017). The JRII release trials and accompanying data have been documented within several works (Mazzola, 2020; Chang and Mazzola, 2020; Hanna, 2020; Nicholson et al., 2017; Gant et al., 2018; Spicer et al., 2019) and are fully described within the introductory paper of this special issue (Fox et al., 2020).

After the first phase of JRII, a network of sonic anemometers was deployed within the CONEX array to examine the localized effects of urban wind flow modification and test the parameterizations in existing urban flow and dispersion models. This secondary experiment, called the JRII Special Sonic Anemometer Study (JRII-S) was void of the chlorine gas release and was meant to collect flow and turbulence measurements around the CONEX obstacles. JRII-S was performed aside from the chlorine trials for two reasons. Firstly, the gas was released at such high concentrations (10,000–100,000 ppm) that the sensors would have been severely damaged. Also, the release jet’s momentum was so powerful that it caused a local modification to the boundary layer as it dispersed in all directions around the CONEXs out to distances of about 50 m.

JRII-S offers a valuable opportunity to study atmospheric turbulence within an array of building-like structures due to its extensive sensor network and study duration. This paper describes the JRII-S sonic anemometer study, data analysis methodologies, and observed wind flow and stability trends adjacent to the structures in comparison to the free-stream boundary layer flow. The purpose of this project was to identify various flow regimes and turbulence patterns seen within an idealized urban environment and then examine the flow patterns around the structures. Analyses from JRII-S may help us better understand complex flow problems with the implication that any neutrally buoyant release will be dispersed within the wind’s streamline flow.

2. Description of JRII-S

JRII was carried out at DPG, a US Army facility located in remote northwest Utah, west of the Great Salt Lake and adjacent to the Bonneville Salt Flats. DPG is a cool, semi-arid desert region (Köppen climate: BSk) with less than 40 cm of annual precipitation. The flat, packed salt playa surface of the field site has an overall roughness length of Z₀ ~ 0.5 mm and is considered to have a horizontally homogeneous boundary layer but is susceptible to mesoscale drainage flows from distant topographical features. The first phase of JRII involved the deployment of 80 CONEX shipping containers in 12 equally spaced rows (Fig. 1a) staggered laterally by various sizes. This arrangement was chosen to allow for the development of a recirculating cavity in the wake of obstacles. The rows were spaced 6.71 m from the back edge of the upwind CONEX to the front face of the next row, creating a wide street canyon aspect ratio (H/W_c ≈ 0.4) at the boundary between the isolated roughness and wake-interference flow regimes (Oke, 1988). The row spacing was very similar to the MUST field study, which reflects a flow regime typical for many U.S. and European urban areas (Biltoft, 2001). The CONEXs, numbered by row (from south to north) and position from west to east, varied in size by manufacturer but were generally of two main sizes, 6.1 and 12.2 m (20 and 40 ft) long with a cross section of 2.44 m (8 ft) wide by 2.44–2.59 m (8–8.5 ft) high (we define the average building height, H = 2.5 m). A few smaller containers in the front of the array were as short as 3.66–1.27 m long in the direction perpendicular to the CONEX centerline. In row 11, a tall “building” of two CONEXs wide stacked by three high and rotated 90° from the rest of the array (L_b = 2.42H, W_b = 1.96H, H = 3.1H), was arranged to simulate the effects of a tall building among an array of similar sized, uniform structures. Wind tunnel and computational fluid dynamics (CFD) modeling experiments have shown that this causes dramatic effects on street canyon and the freestream flow aloft (Heist et al., 2009). In addition, all containers had flat roofs except for two trailers with a 5° pitched roof (maximum height = 3.4 m) placed to the left of the tall building and used for indoor dispersion measurements during the chlorine release trials.

Fig. 1.

a) Full plan view of the 12 rows of 80 total CONEX containers with the focus area outlined by a red box. The array was oriented −14.5° from true north to account for the local climatological prevailing wind flow, b) Subset of the CONEX array and the sonic anemometer deployment locations in March 2016. Color dots indicate the sonic instrumentation heights and locations. The sonic anemometers and CONEXs were numbered in a grid fashion. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

A common methodology for characterizing building density has been proposed by Grimmond and Oke (1999), who described characteristics of groups of buildings for use in dispersion models for the estimation of average wind and turbulence profiles. The closeness of the buildings is important and is characterized by two dimensionless morphology parameters, λ_p and λ_f. The first, λ_p, is defined as the ratio of the total horizontal plan area of the buildings to the plan area of the ground that the set of buildings covers. The second, λ_f, is defined as the ratio of the total frontal area (which the wind encounters) of the buildings to the plan area that the set of buildings covers. Short, flat buildings would have λ_p/λ_f > 1. In a large city, the total plan area might be the whole city, a neighborhood, or a few blocks. Hanna (2020) reports that these JR II building array parameters are λ_p = λ_f = 0.18.

The chlorine release tank was positioned in the center of a 25 m diameter circular concrete pad laid between rows 3 and 5. This central release point was designated as the origin (0, 0) for the coordinate system for all JRII field tests. The entire building array was placed on a packed and graded gravel pad that was 120 m wide by 180 m long and approximately 1 m thick above the playa surface and rotated approximately −14.5° counterclockwise from true north. This was done to account for the climatological wind patterns at DPG such that the along-wind flow at around 8 a.m., when the chlorine release trials were planned, would intercept the long faces of the CONEXs most of the time. This effect would also cause the downwind rows to be directly within the wake of the upwind rows.

A plan view of the CONEX numbering and sonic anemometer deployment from March 2016 is shown in Fig. 1b. Thirty sonic anemometers were installed in two main clusters surrounding CONEX 9.4, which was 12.2 m long, and CONEX 11.4, the tall, stacked building structure. Locations denoted with black dots had mast-mounted sonics 1 m above the ground surface. Red dots indicate sonics on 1 m masts on top of a CONEX. Locations with green dots had three sonics mounted on a small tower at 1, 2, and 5 m heights, while orange dots represent four sonics on towers at 1, 2, 5, and 10 m above the ground. Sonics with “L” in their names indicate the sonic height level (i.e. L1 is the lowest 1 m sonic height on that tower). The sonic towers, as seen from the tall building vantagepoint, are shown in Fig. 2a. An additional 32 m meteorology tower was positioned approximately 115 m south of the CONEX array with sonics at 2, 4, 8, 16, and 32 m to provide a characterization of the unobstructed approach flow (Fig. 2b). The reader is referred to Fox et al. (2020) for a more detailed schematic of the full instrumentation and CONEX layout during the chlorine release trials.

Fig. 2.

a) Photograph of part of the CONEX mock-urban array and the sonic anemometer instrumentation towers looking southeast from the top of the stacked 2W_b × 3H tall CONEX building. Sonics were deployed on 1 m masts on the ground and on top of the CONEXs, as well as on 10 m towers at 1, 2, 5, and 10 m heights. A tall sonic anemometer tower can also be seen in the distance in the top right of a) to gather unobstructed flow. The tower is seen in b) with sonics at 2, 4, 8, 16, and 32 m. (Photo: DPG Meteorology Division).

3. Data and methodology

3.1. Sonic anemometers within the CONEX array

The sonic anemometers recorded data for a total of 49 days under a variety of atmospheric states. The JRII-S project occurred in two phases: from 2100 UTC October 11 to 1700 UTC November 2, 2015 and from 2100 UTC March 2 to 1600 UTC March 28, 2016. Most of the sonics were in nearly continuous operation, although some experienced gaps in data collection. Since many interruptions occurred in October 2015, our analysis was focused on the 24 complete diurnal cycles captured during March 2016. All sonics except S74, S75, S83, S101, and S111 had between 93 and 100% data completeness during the March 2016 collection period. Those five sonics gathered data between 40 and 43% of the time during the same period.

A total of 35 three-dimensional sonic anemometers (Model 81,000, R.M. Young Company, Traverse City, MI) recorded data at 10 Hz (10 samples per second) at an accuracy of ± 0.05 m/s and ± 2° for wind speeds below 30 m/s (R.M. Young Company, 2010). Data from the sonic anemometers were output to a datalogger and retrieved every three days. Mean values, heat fluxes, and turbulence statistics were produced from the raw 10 Hz data with an eddy-covariance software developed at DPG (Bowers et al., 1994). The software also provided quality assurance methods for spikes, stationarity, and missing data. Thirty-minute averages were selected for analysis to separate turbulence effects from the larger-scale mesoscale and synoptic changes in atmospheric conditions (Stull, 1988).

Since the centerline through the CONEX array was oriented −14.5° counterclockwise from true north, a modified coordinate system was introduced where the new north-south axis (0-180°) ran parallel to the short sides of the CONEX containers in the array (i.e. a formerly 165° wind direction would now be 180°) to simplify the description of local flow phenomenon. This conversion was applied to all the mean wind components and turbulence terms such as the fluxes, variances, and stresses. For the mean wind, + U, V, and W refer to wind components from the west, south, and upward, respectively. From this point forward, any wind directions reported in this paper, unless otherwise noted, reflect the new, transformed coordinate system in relation to the orientation of the CONEX array.

3.2. Unobstructed 32 m sonic tower

To characterize the approach boundary layer characteristics and atmospheric stability just above the desert surface, data from sonics on a tall, unobstructed tower were acquired for the same March 2–28, 2016 time period. The tower had sonics positioned at 2, 4, 8, 16, and 32 m heights. These data had not been previously processed by DPG. We analyzed the raw data into 30-min averages with the open-source QA/QC software package called TurblenzKnecht3 (“Turbulence Knight”, TK3; Mauder and Foken, 2015). TK3 was developed to process, quality control, despike, and average raw data from sonic anemometers (Mauder and Foken, 2004; Foken and Wichura, 1996). The data were first checked for missing data and then spikes (where values exceeding the median ± 7 times the standard deviation) were detected based on an algorithm from Mauder et al. (2013) and replaced with a missing data value. Thirty-minute averaging periods with less than 90% of data completeness were discarded. Covariances, mean wind components, standard deviations, and heat fluxes were summarized with a data quality flag (Foken, 1999). Stationarity was checked by comparing 5-min averages with the 30-min interval to be sure that the statistical parameters did not vary largely within the averaging time (Foken et al., 2004).

Since TK3 is a different software package than the one used by DPG, we tested using raw 10 Hz data within the building array by comparing the TK3 result with the QA/QC dataset provided by DPG. Results were nearly identical to the DPG-processed results for our test cases using three randomly chosen sonics at various dates, heights, and locations. The average percent difference for a few select variables including the mean vector wind speed, u-wind component variance (${\sigma }_u^2$), heat flux ($\overline{w^{’}{T}^{’}}$), U and V component eddy covariance ($\overline{u^{’}{v}^{’}}$), and friction velocity (u*) between DPG and TK3-processed variables for the S73 and S61L1 sonic on March 10 and 21 were generally less than 0.25%. One of the possible sources of difference was that DPG provided 30-min averages rounded to two or four decimal places. TK3 generated values to six decimal places. Most 30-min values were comparable to the hundredth or thousandths place. The error was slightly greater for the sonics positioned higher above the surface.

In processing the 32 m tower data, some spikes were detected, but only one 30-min period exceeded the program’s 10% spike threshold and was eliminated. A final quality control flag was assigned, which was a combination of the steady state flag for friction velocity (based on the eddy covariances or fluxes for the U and W and V and W components) and the highest flag value for the integral turbulence characteristic (ITC) for the separate U, W, and sonic temperature (T) components. Ten 30-min periods were eliminated, leading to 95–98% data completeness. The data were then rotated −14.5° to match the sonic data from the building array.

4. Results

4.1. Overview of general meteorological conditions during March 2016

The broad and flat desert valley in which JRII took place is flanked with distant and gently sloping mountains, resulting in drainage flows and vertical directional shear (Hanna, 2020). During the nighttime and daytime hours, a dominant south-southeasterly and north-northwesterly prevailing flow developed as the desert surface cooled and warmed, respectively. A wind rose of the unrotated wind distribution (Fig. 3) demonstrates the prevailing flow observed at the 2 m sonic on the unobstructed tower for all 30-min average observations. The wind distribution at the remaining sonic heights were similar. Most wind speeds fell within the range of 2–4 m/s with a mean wind speed of 3.62 m/s. Winds most frequently blew from the south-southeast, although the strongest winds originated from the west-northwest. Hanna (2020) noted that the lowest levels of the boundary layer were indicative of prevailing down-valley drainage flows from the south-southeast, while the overall synoptic flow above 100 m was from the west. Note that all wind directions reported in this paper refer to the direction that the wind is coming from.

Fig. 3.

Wind rose of the unrotated wind direction and speed distributions (m/s) from the 2 m sonic on the unobstructed 32 m tower. Prevailing winds were predominately from the south-southeast and northwest. The wind rose is based on 30-min average observations over the 26-day period in March 2016.

March 2016 was generally characterized by long stretches of low to moderate winds. The average 2 m sonic temperature was 7.8 °C with an average daily high of 14.5 °C and low of 2.9 °C. The daily maximum and minimum temperatures usually occurred 1–3 h before sunset and sunrise, respectively due to the arid, sparsely vegetated surface. The climatological average minimum and maximum temperature for March is 12.3 °C and −1.9 °C, respectively with 19.3 mm of average precipitation. From a synoptic perspective, the test area was influenced by three mid-latitude cyclones with strong cold fronts, gusty winds, and rapidly changing weather including rain and snow. The periods around 2100 UTC March 6, 1800 UTC March 14, and 0600 UTC March 22 are interpreted with caution, as there were large wind and velocity shifts due to frontal passages.

Atmospheric stability was classified using the 1/L (1/m) stability parameter from the 2 m sonic on the unobstructed tower, where L is the Obukhov length. The quantity was generated by the TK3 software using the friction velocity and heat flux at that sonic height. A histogram of 1/L is shown in Fig. 4. We defined three stability categories with units 1/m, where neutral is −0.02 ≤ 1/L ≤ 0.02, stable is 0.05 ≤ 1/L ≤ 0.25, and unstable is −0.25 ≤ 1/L ≤ −0.05. We chose to classify periods that were moderately to strongly stable or unstable while also eliminating the large positive and negative outliers. The gaps selected between neutral and stable and neutral and convective serves to more clearly define the stability categories. Our choice was informed by the Pasquill-Gifford stability class curves referenced in Randerson (1984) for the given DPG roughness length. Many periods were classified within the neutral range with a lesser frequency of both unstable and stable cases.

Fig. 4.

Histogram of the inverse of Obukhov length (1/L, 1/m) from the 2 m sonic on the unobstructed tower shaded by stability classification. Each bin represents a 0.01 increment in 1/L.

To analyze the prevailing wind’s effects on the mock-urban flow behavior, there is value in considering time periods when the flow was approximately perpendicular to the long face of the CONEX buildings to maximize the associated building-induced wake-effects. A tolerance of ± 15° in the rotated ambient approach flow was selected, approximate to northerly and southerly winds between 345 and 15° and 165–195°, respectively. Moderate wind speeds (U > 2.5 m/s) characteristic of more neutral boundary layer flow was desirable, as turbulence is relatively continuous vertically throughout the boundary layer compared to stable conditions. A list of these time periods is provided in Table A1 in Appendix A, as a guide for selecting periods for comparison to other studies in this special issue, including CFD-based building wake characterizations (Carissimo et al., 2020) and magnetic resonance velocimetry (MRV) in a scaled JRII model array (Owkes et al., 2020). The selection of time periods was informed by the 30-min averaged 8 m sonic data on the 32 m tower.

4.1.1. Meteorological analysis of a sample day

A time series meteogram of wind speed (U, m/s), wind direction (°), T (K), L (m), and $\overline{w^{’}{T}^{’}}$ (m K/s) for March 24 at the unobstructed tower is shown in Fig. 5 to demonstrate some of the diurnal meteorological patterns. This day is also simulated by Carissimo et al. (2020) since the wind flow was moderately strong and primarily from the south in relation to the CONEX array, which provides a good case to study building wakes in their CFD simulations. The synoptic pattern for March 24 was mostly dominated by clear skies and a broad area of (approximately 1025 hPa) high pressure building from interior portions of the Pacific Northwest in lee of an easterly advancing mid-latitude cyclone and trailing cold front east of the Rockies. At DPG, the UTC time difference was UTC-7 hours until March 13, 2016 and then UTC-6 hours under Mountain Standard Time (MST). Therefore, sunset for March 23 occurred at 0145 UTC, midnight was 0600 UTC, sunrise for March 24 was around 1330 UTC, and the peak heating of $\overline{w^{’}{T}^{’}}$ was around 2100 UTC (1500 MST). Sunset is evident as the temperature began to slowly fall a few hours after reaching its greatest value while the heat flux dropped to near zero by 0200 UTC as surface heating was cut off. As stability increased, L remained positive during the overnight hours and a surface-based inversion developed.

Fig. 5.

Meteogram of 30-min averaged wind speed (U, m/s), wind direction (°), sonic temperature (T, K), Obukhov length (L, m), and heat flux ($\overline{w^{’}{T}^{’}}$, m K/s) for March 24, 2016 from different levels of the 32 m unobstructed sonic tower.

As remaining effects from the weather system were departing during the overnight hours, northeasterly winds subsided by sunset and gently shifted from the north-northwest. A considerable wind speed difference with height of about 2–3 m/s between 2 m and 32 m was observed throughout the evening as the winds shifted to be from a more prevailing south-southeasterly flow and dominated during the daylight hours of March 24. The latter was due to downsloping winds as the valley’s surface cooled overnight. Sunrise is evident around 1330–1400 UTC as temperature and $\overline{w^{’}{T}^{’}}$ gradually started to rise and stability rapidly changed from stable to unstable. A missing data value for L at 1400 UTC is due to a spike where TK3 flagged the u* quantity. The atmosphere was likely transitioning through neutral stability during this half-hour period. The atmosphere remained unstable throughout the day as the wind speed and direction stayed steady, with some reduction by sunset. Less variability of wind speed with height was seen during the daylight hours.

4.1.2. CONEX array flow patterns on March 24, 2016

Fig. 6 shows a spatial depiction of 30-min average sonic wind vectors on March 24, 2016 within the mock-urban array. The horizontal pattern in Fig. 6a was measured at the 1 m height. Two half-hour periods from 1130 UTC (black) and 2130 UTC (red) show how a slight difference in direction and strength of southerly winds can change the channeling flow through the building array. During the overnight hours at 1130 UTC, the incoming 2 m wind speed was calmer (3.2 m/s) and oriented nearly southerly (185°), while the wind was 4.5 m/s from the south-southeast (155°) at 2130 UTC. The weaker winds at 1130 UTC directed flow around CONEX 9.4 and caused a slight flow reversal in the building’s cavity (the recirculating wake on the lee side of the CONEX). This feature is seen in sonics S31 and S41 in the vertical profile (Fig. 6b), especially closer to the surface. At 2130 UTC, the flow was forced to the west around the building in addition to reversal in flow in the building’s wake. The sonic on top of CONEX 9.4 (S23) and on tower S31 at 2 and 5 m heights show that the wind speed increased as it ascended over the top of the CONEX.

Fig. 6.

a) Spatial example of 30-min average sonic wind vectors at 1 m for 1130 UTC (black) and 2130 UTC (red) March 24, 2016 to show near-surface wind interactions within the CONEX array under slightly different approach flows, b) A vertical profile of the x = 0 centerline sonic anemometers. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Acceleration of the airflow up, over, and around the buildings is also seen from the sonics around the top of the tall stacked CONEX 11.4 building (S61L4, S73, and S82L4). The data from the S61 sonic show that as the wind approached the upwind face of the building, air was generally forced downward and outwards around the sides of the obstacle. Despite the approach flow direction, the flow was nearly calm at the lower sonics at S82 (1 and 5 m) in lee of the tall building. In these locations, TKE and turbulence intensity were high when the flow was weak or reversed. S101, which was 2.44 m from the downwind edge of CONEX 11.4, was still within the building’s wake, but S111 (twice the distance downwind at 4.88 m) experienced less of an effect of the tall building, at least for the 2130 UTC period. In general, the building array acted to slow the flow near the ground level and around some structures in relation to the unobstructed approach flow. Additional analysis for 1100 and 1330 UTC March 24 has also been shown through CFD modeling of building wakes (Carissimo et al., 2020) within this special issue.

4.2. Unobstructed tower vs. building array flow

The CONEX buildings strongly influenced the localized flow and often caused an abrupt flow redirection at some places within the mock urban array. A visual comparison of the average 30-min wind directions for all 1-m high sonics within the building array versus the 2 m sonics on the unobstructed tower is depicted in Fig. 7a and b for each cluster of sonics, one surrounding the tall building (CONEX 11.4) and the other around a long, low building (CONEX 9.4), respectively. The subplots are arranged to approximate their relative positions with respect to the CONEXs they surround. The data would fall along the 1:1 diagonal line if both the wind direction measured within the array and approach flow wind direction were similar. All data from the 26-day period in March 2016 are displayed on this plot regardless of flow direction or atmospheric stability.

Fig. 7a.

Rotated wind direction (in relation to the CONEX array) (°) for the sonic anemometers within the array vs. the 2 m unobstructed tower measurement. The figure shows the northernmost sonic cluster surrounding the ‘tall building’ CONEX 11.4 (see red box on inset plot). Each sonic is labeled prevailing, north-south, or east-west, fitting a flow regime that was influenced by its location, along with an R² value for the goodness of fit for that regime. Color bars differ by regime so that warmer colors indicate a better fit. Shading for the prevailing plots reflects the absolute value of the unobstructed wind direction subtracted from the sonic wind direction. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Fig. 7b.

Same as a) but showing the southernmost sonic cluster around CONEX 9.4 (see inset plot).

We have classified each sonic as following one of three flow “regimes” that approximate prevailing, north-south, or east-west flow and have indicated the classification following the sonic number in Fig. 7. For north-south, for example, flow at the sonic was forced nearly northerly or southerly most of the time regardless of the incoming wind direction or stability. The east-west regime had flow forced mainly easterly or westerly, while prevailing generally followed the unobstructed flow for at least half of the potential wind directions. The color bars adjacent to each subplot in Fig. 7a reflect the strength of the fit to the flow regime where warmer colors (reds) show data that fit the regime the best. Therefore, the east-west regime data is shaded red when the wind direction is aligned with the street canyon (i.e. 90° or 270°), but cooler colors (blues) when the flow is perpendicular (0° or 180°). North-south follows the same concept. Shading for the prevailing flow reflects the absolute value of the free-stream wind direction subtracted from the sonic wind direction to show the difference. Very low wind direction differences are shaded in warm colors and tend to fall along the 1:1 line or on the edges due to wind directions wrapping from 0° to 360°. A goodness of fit (R²) value is also reported next to each sonic plot to display how well the data fit that regime. It is important to note that these regimes held their relationships throughout a range of atmospheric stabilities, wind speeds, and wind directions. Even for southerly winds where the flow would have encountered eight rows of CONEX containers before intercepting the first sonic anemometer, these regimes are still observed.

4.2.1. East-west regime

The southernmost sonic (S11L1), which was positioned 1.22 m south of the middle of CONEX 9.4 (Fig 7b), followed a remarkably consistent flow regime of either easterly or westerly regardless of the unobstructed direction. Its flow was almost entirely dictated by the building forcing the flow either one way or another as it encountered the obstacle. If the free-stream flow was from about 30 to 180° (180°-345°), the local flow was forced mainly easterly (westerly). The only anomaly was when the unobstructed wind flow was coming directly from the northerly or southerly direction. This perpendicular flow in relation to the CONEX then forced the flow southerly as the wind loaded on the windward face of the building and caused air to reverse in direction. For this reason, the R² relationship was only 0.78. The rest of the sonics positioned along the centerline (except for S111) also followed the east-west regime. The flow for S31L1 (R² = 0.93), S41 (R² = 0.94), and S51 (R² = 0.84), which were centered laterally and positioned 1.22, 2.44, and 4.88 m north, respectively, from CONEX 9.4 followed a similar pattern as S11L1, with a few exceptions. When the incoming flow was directly from the south, the sonic flow varied from northeasterly or northwesterly, which can be explained by the wake induced by the building. The S51 data resembled that of S11L1 since S51 was 2.44 m upwind of the front face of CONEX 10.4. This sonic displayed a near reversal of flow when the approach wind was from the north.

For the specific wind direction of 180°, we now consider the expected wake dimensions for CONEX buildings 8.5 and 9.4. Earlier, we mentioned that JRII street canyon aspect ratio (H/W_c ≈ 0.4) is situated at the boundary between isolated roughness and wake-interference flow regimes (Oke, 1988). The H/W_c value defining the boundary decreases as L_b/H (the crosswind length of a typical building divided by its height) increases. That is, longer buildings require wider streets canyons to be considered isolated and avoid a wake-interference flow regime. According to formulas reported in Hosker (1984) and Schulman et al. (2000) for isolated buildings, the length of the recirculating cavity (L_R) downwind of CONEX 8.5 is expected to be 10.0 m and 6.8 m, respectively, while downwind of CONEX 9.4 (which was twice the length of CONEX 8.5) L_R ≈ 13.3 m and 9.9 m, respectively. W_c = 6.71 m for both street canyons. The manner in which the wind direction in the canyon transitions between 90° and 270° as the approach flow nears 180° (Fig 7b) suggests whether the sonic is within a recirculating region or not. Specifically, sonic S11L1, which was 5.5 m from the downwind edge of CONEX 8.5, has an observed wind direction of 180° when the approach flow nears 180°, suggesting that it is outside the cavity region of CONEX 8.5. While the above calculations indicate that the sonic should be within the cavity, the results in Fig. 7b show that the cavity is smaller due to wake interference. For sonics S31, S41, and S51, the wind direction transitions through 0° as the approach flow nears 180° enveloping the sonics in the recirculating cavity of the longer CONEX 9.4.

Three anemometers from the northernmost cluster of sonics around the tall building (CONEX 11.4; see Fig. 7a) had less of a clearly defined east-west flow pattern compared to what is seen in the southern cluster of sonics. Sonics S61L1 (R² = 0.78) and S82L1 (R² = 0.78) were 1.22 m south and north of the tall building, respectively, while S101 (R² = 0.94) was 2.44 m north. The recirculation length of the cavity associated with CONEX 11.4 is expected to be L_R = 4.8 m and 8.2 m for Hosker (1984) and Schulman et al. (2000), respectively. The shorter recirculation length for CONEX 11.4 compared to CONEX 8.5 and 9.5 is due to an increase in the along-wind building dimension (L_b) and reduction in crosswind width relative to H which leads to a decrease in cavity length (Hosker, 1984). The data in Fig. 7a confirm that S82L1 and S101 are within the recirculating building cavity when the free-stream flow was southerly.

4.2.2. Prevailing regime

The flow at some locations followed a more prevailing wind direction as it was relatively unaltered from the incoming flow. The best example of this was S111 (R² = 0.91) which was located far enough north of the tall building (4.88 m) to experience little impact of the building’s wake, except under direct southerly flow. The Hosker (1984) calculations of L_R above supports the conclusion that S111 may have been outside the recirculating wake of CONEX 11.4. S111 may have also been rather unaffected since no buildings were directly north, east, or west of its position. Similar prevailing effects were seen from S81 and S83, which were positioned 2.44 m from the sides and northernmost corners of the tall building. It could be argued that these sonics loosely fit either the prevailing or north-south flow regimes (for S81, R² = 0.86 vs 0.75; and S83, R² = 0.93 vs 0.80). The sonics were also influenced by the horseshoe vortices (Hosker, 1984) within the building’s wake but were likely centered far enough between the street canyons of rows 11 and 12 to experience prevailing flow. The relationship did not hold when the flow intercepted the tall building first and then downwashed at the sonic site (i.e. ESE for S81 and SSW for S83).

Four other sonics fitting the prevailing regime, S21 (R² = 0.78), S22 (R² = 0.81), S24 (R² = 0.71), and S25 (R² = 0.84), were sited 1.22 and 2.44 m on each short side of the long CONEX 9.4. S21 and S25 were generally in the middle of the street canyons between the adjacent buildings both laterally, northerly, and southerly (7.92 m from rows 8 and 10). The pairs of sonics on either side of the building followed an almost identical, but mirrored, prevailing flow relationship that was influenced by CONEX 9.4. When the unobstructed flow was within the range of 180°-360°, S22 followed the prevailing pattern since the flow did not intercept any portion of the nearby building within the street canyon. S24 held the same trend but from the range of 0-180°, since the building was on the opposite side. At S22 (and opposite for S24), if the flow came from the easterly direction, the sonic flow resembled a reversed westerly direction as the sonic was now positioned within the wake and cavity created by the short end of the building. However, as the incoming flow shifted to be more northeasterly (~45°) or southeasterly (~135°), flow at the sonic responded with an opposite range of wind directions from west-northwesterly to west-southwesterly. This can be due to influences of the now-distorted cavity region and subsequent flow separation. In a similar sense, wind tunnel and CFD studies have demonstrated that oblique mean wind incident on a long narrow building creates a lateral shift in any effluent entrained in the cavity (Monbureau et al., 2018). The effluent is thus moved laterally along the building and reemitted at the lee edge, causing what appears as a secondary emission source (Yang et al., 2020). Direct northerly or southerly unobstructed flow also showed considerable variation, as flow was directed outwards and around the building obstacle within horseshoe vortices on the lateral sides. S21 and S25 followed a similar pattern as the neighboring sonics around the northeasterly and southwesterly range, respectively, but were less susceptible to the previously mentioned features and more aligned with the prevailing flow.

4.2.3. North-south regime

The north-south regime was prevalent for the four sonics surrounding the long sides of CONEX 11.4 (S71, S72, S73, and S75) and was the result of channeling flow between the street canyons formed by CONEX 11.3 and 11.5. The structure created a narrow 3.66 m alley on its west side, resulting in a consistent north-south flow redirection for sonics S71 (R² = 0.94) and S72 (R² = 0.91), especially when unobstructed wind directions were about ±150° from the north or south. Specifically, incoming flow from about 90°-240° resulted in local flow to be consistently from the south (180°). Varying approach flow from about 270°-60° channeled the sonic flow northerly (~0°). Spread was seen when the unobstructed wind direction transitioned to easterly or westerly, resulting in an opposite flow at the sonic. As seen from some of the other sonics, this was likely due to downwash as the flow was forced up and over the building, creating a recirculating vortex on the lee side. A similar, but looser relationship was observed from the sonics on the east side (S74 (R² = 0.80) and S75 (R² = 0.67)) of the tall building because the street canyon was much wider (W_c = 6.71 m).

4.2.4. Flow characteristics

Several patterns were identified within the mean turbulence variables under each flow regime. Fig. 8 shows rotated 30-min averaged U, V, and W wind components, and $\sqrt{\mathit{TKE}}$ versus the 2 m unobstructed wind direction for select sonics of each flow regime type: prevailing (S111), north-south (S72), east-west (S11L1). The 2 m unobstructed measurements are also provided for comparison. The U, V, W components were normalized by the scalar wind speed (U) from the 2 m measurement on the 32 m sonic tower, which was calculated from the averaged U and V components after Merceret (1995). TKE was normalized by the local scalar wind speed at that sonic. Since the flow resulted in greater fluctuations at low wind speeds due to the nature of turbulence, wind speeds U < 2 m/s were shaded in blue for the U and V wind component and $\sqrt{\mathit{TKE}}$ plots to visualize the spread for nominal winds. Negative vertical velocity is shaded red in the W component plots.

Fig. 8.

Nondimensional U, V, W wind components and $\sqrt{\mathit{TKE}}$ versus the 2 m unobstructed sonic tower wind direction for select sonic anemometers of each flow regime type; prevailing (S111), north-south (S72), east-west (S11L1) and unobstructed (2 m tower). The U, V, W components are normalized by U from the 2 m height on the unobstructed sonic tower. Wind speeds U < 2 m/s are shaded in blue for the U, V, and $\sqrt{\mathit{TKE}}$ plots. Negative vertical velocity is shaded red in the W plots. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

The free-stream U and V component plots exhibit the expected sinusoidal and bell-shaped curves, respectively, where + U indicates wind blowing from west to east and + V is from south to north. Low positive W-component values are generally seen throughout the entire wind direction distribution at the unobstructed tower, indicating upward momentum flux. $\sqrt{\mathit{TKE}}$ values were generally low and homogeneous with wind direction at the 2 m sonic, with somewhat of a gap in observations from southeasterly or westerly flow since the wind direction rarely blew from these direction ranges (Fig. 3).

Both the prevailing sonic (S111) and east-west sonic (S11L1) showed a loose sinusoidal curve for the U component, however their amplitudes were stretched and compressed, respectively, relative to the unobstructed flow. This follows the fact that both a positive and negative U component distribution would indicate crosswind flow from the east or west. Together, both the U and V components show the most spread for the prevailing case since wind flowed freely around the unobstructed coordinate plane. The absence of V-component spread and a sinusoidal curve for the U component at S11L1 signifies flow forced towards the east or west, while limited U-component spread and a V-component bell shape for S72 indicated flow directed mainly towards the north or south. This observation is also reflected in the $\sqrt{\mathit{TKE}}$ plots where the lowest values were observed along the prevailing wind flow pattern. Outside of those regimes, $\sqrt{\mathit{TKE}}$ was high due to a greater building-generated shear.

There were some exceptions to these general observations. The north-south sonic (S72) showed a nearly flat distribution of the U component with some negative spread between 180 and 270° when the sonic encountered the wake of the tall building. This feature is evident in the W-component, as all values in this wind direction range were negative. Most of the positive vertical velocities for S72 occurred in the northwest and northeast directions as the flow encountered the lateral side of the tall building and was drawn upwards. S111, which was about 2H from the lee face of the tall building, also experienced a dip in V-component observations to zero where the apex of the bell curve was expected. Values of zero for both U, V, and (in general) W components indicate that the wind was mostly stagnant under southerly approach flow. Low to no wind velocity in the building wake also resulted in a spike in $\sqrt{\mathit{TKE}}$.

4.3. Effect of stability on turbulence

We examined the relationship between stability category and turbulence both within and upwind of the building array using the variance of vertical velocity as a surrogate for turbulence. Fig. 9 shows a box and whisker panel plot of the effect on vertical variance normalized by the average 2 m unobstructed wind speed squared (${\sigma }_w^2/{\mathbf{U}}^2$) for a) the five levels of the unobstructed tower, b) the four heights of the S61 sonic tower, and c) the 1 m sonics along the x = 0 centerline. All stabilities are based on the 2 mL on the unobstructed tower and numbers on the plot indicate the n-value (number of occurrences) for each stability class. Our focus in this stability analysis was on the southerly quadrant of wind directions (i.e. 135°-225°) which follows the prevailing wind flow.

Fig. 9.

Vertical velocity variance (${\sigma }_w^2$) normalized by the southerly quadrant of the 2 m unobstructed tower scalar wind speed (${\sigma }_w^2/{\mathbf{U}}^2$) for a) the five heights on the unobstructed 32 m tower, b) the four heights on sonic tower S61, and c) the 1 m sonics along the array centerline.

For the approach flow shown in Fig. 9a, our data is consistent with the findings of Panofsky et al. (1977) in that the mean and median values of ${\sigma }_w^2$ monotonically increase with height for the convective stability cases. Instability is indicative of greater values of vertical turbulence and buoyancy seen in the figure. Arya (1995) reports that standard deviation (σ) of wind components, when normalized by u*, are universal functions of the Monin-Obukhov stability parameter (z/L) in a convective environment, and the values increase with height. The neutral and stable cases show minimal ${\sigma }_w^2/{\mathbf{U}}^2$ variability and variation with height.

In contrast to the unobstructed approach flow, the flow within the building array exhibits an increase in turbulence for both sonics near the ground and throughout the canopy. Turbulence is at least two to three times greater in the building array compared to the free-stream flow measured at similar heights. It is interesting to note that ${\sigma }_w^2/{\mathbf{U}}^2$ shows little variation with height for the S61 sonic tower (Fig. 9b), possibly because these sonics were within the influence of the nearby 3H tall building even though the 5 and 10 m sonics were above the building canopy created by the shorter buildings. With a southerly flow, this would position the S61 tower just 1.22 m upwind of the tall CONEX building. There also appears to be no indication of a relationship between the free-stream stability and the turbulence values in the S61 profile. In contrast to the unobstructed tower, there is very little to distinguish among the three stability categories. Some higher values of ${\sigma }_w^2/{\mathbf{U}}^2$ may be attributable to the very light wind speeds associated with extreme stabilities rather than an increase in ${\sigma }_w^2$.

Fig 9c shows the results of the 1 m sonics along the array centerline within the building canopy. While there is considerable variation in the magnitude of ${\sigma }_w^2$ along the array centerline, the turbulence levels measured at any height or region are not a strong function of stability category in contrast to the approach flow measurements. A noteworthy observation is that the ${\sigma }_w^2$ values, especially at S82, are lowest immediately downwind of the buildings as the sonics are enveloped in the building wake. A considerable increase in ${\sigma }_w^2$ then occurs at S101 and S111 in the recirculating wake in lee of the tall building but note that the data completeness of S101 and S111 is much lower than the other upwind sonics. Wind tunnel measurements reported in Uehara et al. (2000) and large eddy simulations (LES) from Boppana et al. (2014) demonstrate stronger relationships between turbulence levels over a wide range of stabilities. A possible explanation for the lack of stability dependence of ${\sigma }_w^2$ in the JRII canopy measurements may be a stronger generation of mechanical turbulence produced in this irregular and staggered building array, in contrast to the more regular arrays of Uehara et al. (2000) and Boppana et al. (2014).

To compare the turbulence within the JRII array to measurements within actual cities, we note that Hanna et al. (2007) found average values of σ_w/U_local ~0.3 using sonic anemometer data from field campaigns within three large US cities (with U_local being the wind speed at sonics within the canopy). Hanna et al. (2007) further reports that wind speeds at the street level are about 1/3 of their unobstructed values (at locations such as airports), which suggests σ_w/U ~0.1 for those full-scale field studies. We report σ_w/U to be approximately 0.12 within the CONEX array, where U is the 2 m unobstructed wind speed. While Hanna et al. (2007) mentions that turbulence levels are often higher at H compared to street level, their results also support the postulation that overall turbulence levels are greater within the canopy due to meandering of wakes and mechanical turbulence generated by the structures, thereby reducing the overall effects of stability (Hanna and Chang, 1992).

5. Summary and discussion

JRII-S occurred during Fall 2015 and Spring 2016 at DPG to study the flow and turbulence within a systematically arranged mock-urban style environment constructed from 80 CONEX shipping containers. The JRII-S study occurred between the two phases of chlorine release trials with the goal of characterizing the flow between streets, intersections, and in wake of the buildings. Thirty sonic anemometers were sited at multiple heights and locations around CONEXs of various sizes, including one tall building approximately three times the height of the surrounding array. An additional 32 m unobstructed tower of five sonics was operated concurrently outside of the array to document the approach wind flow. The March 2016 dataset, with over 26 days of consistent data collection, was chosen for our analysis due to its completeness.

While the overall JRII project has been described within other papers (i.e. Fox et al., 2020; Gant et al., 2018; Nicholson et al., 2017), the sonic anemometry component (JRII-S) has not been discussed nor analyzed in detail until this work. Analysis from this study has helped inform components of other JRII-related projects within this special issue (Carissimo et al., 2020; Owkes et al., 2020). For example, Carissimo et al. (2020) uses our processed sonic datasets to inform boundary conditions and ideal time periods (Appendix A) for their CDF model simulations. This paper describes the JRII-S project in detail, documents the data analysis methodologies, and reports some of the observed flow patterns within the building array in comparison to the free-stream boundary layer flow.

We have provided a meso-to-microscale meteorological overview of the JRII site from March 2016 by analyzing data from the unobstructed sonic anemometer tower. Since the flat, arid site was situated away from topographical features, the month was characterized by long periods of classic diurnal variations indicative of a desert. Radiational cooling led to episodes of temperature inversions and stable boundary layers in the evening hours, while weak instability and a frequent occurrence of somewhat neutral flow was observed. Prevailing southerly wind, and to a lesser extent, northerly, aided to narrow our analysis when the incident flow intercepted the entire CONEX array.

We have also shown a comparison between the approach and local flow where the wide simulated street canyons (H/W_c ≈ 0.4) and staggered intersections often caused a complete flow redirection at some places within the mock urban array. We have classified the sonic locations to follow a generalized flow regime of either prevailing, north-south, or east-west in relation to the free-stream flow. Despite some caveats, our results show a tendency of channeling over a wide range of wind speeds, directions, and atmospheric stabilities. These relationships hold quite strongly and point to the fact that wind direction within the urban canopy is strongly connected to the proximity and morphology of the adjacent buildings and obstacles. Our analysis has shown that flow patterns at two sonics positioned less than 4 m horizontally from each other can show a great degree of variation which is dictated by the presence of the buildings, such as the flow patterns seen between sonics S72 and S81 or S101 and S111. In addition, a sonic as close as 3–5 m from an obstacle may show little difference in flow in comparison to the unobstructed wind pattern, as observed at sonics S25, S81 or S111. Although not totally unexpected, these findings reiterate just how complicated wind patterns, channeling, and turbulence are within urban areas and infers an associated difficulty in simulating pollutant dispersion. Our results also demonstrate how a small variation of incoming wind direction can significantly alter the streamline flow between buildings and streets. These challenges are further compounded when buildings of various heights, shapes, and sizes are present, as typical for any complex urban environment.

In addition, our analysis reinforces the fact that urban areas tend to neutralize atmospheric stability within the building canopy despite the characteristics of the approach boundary layer. In the approach flow, ${\sigma }_w^2$ values increase with height for the convective (unstable) cases but do not show much variability for neutral and stable periods. Within the building array and within the canopy, values of ${\sigma }_w^2$ are at least two to three times greater and show little variation with height. There is also little evidence of stability effects on turbulence in the mock urban array in contrast to the approach flow characteristics, regardless of the height or location within the array. In actual cities, urban heat island effects may further exacerbate stability differences than what is seen in our idealized mock-urban setup.

Findings from the JRII-S project may help us better understand complex urban flow with the implication that any neutrally buoyant release, whether it is a homeland security concern or a routine release, will be dispersed by the local flow. These mock-urban findings may elucidate flow and dispersion phenomena and support the development of improved algorithms within fast-response Gaussian dispersion models, such as EPA’s AERMOD regulatory model. Ongoing CFD modeling and concurrent studies within EPA’s Fluid Modeling Facility Meteorological Wind Tunnel are currently being pursued to complement the flow and dispersion study within the JRII mock-urban array under carefully controlled atmospheric conditions. The wind tunnel offers flexibility in changing building geometries, incoming wind directions, and tracer release locations and heights. Other wind tunnel and CFD studies such as Marucci and Carpentieri (2020) or Soulhac et al. (2013) provide a characterization of neutral to stable flow and the associated fate and transport of effluent plumes within urban areas and street canyons. The combination of wind tunnel, CFD, and field studies provides a solid basis for examining urban flow characteristics including turbulence, wind profiles, and street canyon flows with the ultimate goal of dispersion model improvement.

HIGHLIGHTS.

• Jack Rabbit II Special Sonic Anemometer Study is described in detail.

• Wind within the mock-urban building array is characterized by three flow regimes.

• Turbulence and mixing increase within the array while overall wind speed decreases.

• Flow at adjacent sonics varies strongly reinforcing urban dispersion complexity.

• Atmospheric stability effects are moderated within the mock-urban array.

Appendix A

Table A1:

Desirable time periods for those wishing to perform further analyses on the JRII-S database in which the prevailing wind flow was incident on the long, front faces of the CONEX buildings. A tolerance of ±15° from north and south as well as relatively moderate wind speeds (U > 2.5 m/s) characteristic of more neutral flow were identified.

Date	Time Range (UTC)
Northerly Flow:
March 4	0030–0530
March 8	0830–1100
March 13	0200–0430
March 14–15	2200–0100
March 19	0200–1300
March 20	0600–0930
March 22–23	1000-1200, 2100-0030
March 23	1700–1900
March 26	1600–2000
Southerly Flow:
March 3	0900–1100
March 5	1630–2330
March 6	0000–1300
March 7	1230–1830
March 10	0500-0700, 1200-0000
March 11	1000–1500
March 12	0100–0700
March 13	1300–2130
March 15	1430–1930
March 16	1330–1800
March 17	1330–1700
March 24	0730–2100
March 27	0600–1500