The URBAN 2000 experiments were conducted in the complex urban and topographical terrain in Salt Lake City, Utah, in stable nighttime conditions. Unexpected plume dispersion often arose because of the interaction of complex terrain and mountain–valley flow dynamics, drainage flows, synoptic influences, and urban canopy effects, all within a nocturnal boundary layer. It was found that plume dispersion was strongly influenced by topography, that dispersion can be significantly different than what might be expected based upon the available wind data, and that it is problematic to rely on any one urban-area wind measurement to predict or anticipate dispersion. Small-scale flows can be very important in dispersion, and their interaction with the larger-scale flow field needs to be carefully considered. Some of the anomalies observed include extremely slow dispersion, complicated recirculation dispersion patterns in which plume transport was in directions opposed to the measured winds, and flow decoupling. Some of the plume dispersion anomalies could only be attributed to small-scale winds that were not resolved by the existing meteorological monitoring network. The results shown will make clear the difficulties in modeling or planning for emergency response to toxic releases in a nocturnal urban boundary layer within complex terrain.
An understanding of plume dispersion in urban environments adequate for the reliable prediction of plume behavior remains elusive. These environments are characterized by their complexity with urban canopies (Oke 1988; Grimmond and Oke 2002) featuring irregularly differential heights, widths, geometries, building aspect ratios, and possibly irregular street grid layout. Additional complications arise when the effects of thermal regime and terrain (e.g., topographic variation, land–sea breezes) are included. Add in mesoscale and synoptic variations, with continually varying wind speed and wind direction in the approach flow, and the uncertainty with regard to plume release location in a real emergency, and the scale of the problem becomes apparent. All of these factors will influence the flow field and plume dispersion pattern. Overall reviews of the effort to achieve this understanding have been presented by Britter and Hanna (2003) and Vardoulakis et al. (2003).
Research has focused on understanding the urban boundary layer, acquiring datasets that can be used for developing emergency response guidelines, and testing predictive urban plume dispersion models. Dispersion models attempt to describe very complicated physical situations and it is essential that they be evaluated against real-world data. Tracer datasets from field studies are a very important aspect of developing response protocols and guiding model development for complex urban environments.
Many tracer field studies have been conducted in urban settings (e.g., McElroy 1969; DePaul and Sheih 1985; Huang et al. 2000; Rotach et al. 2004; Venkatram et al. 2004; Cooke et al. 2000; Allwine et al. 1992; Thistle et al. 1995). A notable tracer field study was conducted in Oklahoma City in July 2003 [Joint Urban 2003 (JU03)] and was significant for its comprehensive scope and highly integrated multidisciplinary effort (Allwine et al. 2004; Allwine and Flaherty 2006; Clawson et al. 2005). This included a major program of meteorological measurements for understanding mean and turbulent flow conditions in the urban boundary layer, a major program of tracer concentration measurements for tracking the movement and dispersion of a pollutant in this environment, and a modeling effort designed to improve the ability to predict the movement and dispersion of toxic plumes in urban environments using the extensive meteorological and tracer concentration database generated. Both daytime and nighttime experiments were conducted. Numerous papers reporting on urban wind and turbulence measurements (Nelson et al. 2007a, b; Brown et al. 2004; Calhoun et al. 2006; Klein and Clark 2007; Ramamurthy et al. 2007; Lundquist and Chan 2007; Hanna et al. 2007), tracer dispersion (Doran et al. 2007; Flaherty et al. 2007a), and dispersion modeling efforts (e.g., Hendricks et al. 2007; Chan and Leach 2007; Flaherty et al. 2007b) have originated from this field study. While JU03 was highly relevant to the study of urban dispersion, it is still representative of an urban environment that features a canopy that is much less complex than that found in major metropolitan areas, such as New York City.
Another recent field study, designated MID05, was conducted during daytime conditions in Midtown Manhattan during August 2005 under the auspices of the Department of Homeland Security’s Urban Dispersion Program (Allwine and Flaherty 2007). It was a major effort designed to study the atmospheric transport and dispersion of pollutants in a complex, tall, dense urban core with large areal extent.
Urban dispersion is often difficult enough to characterize and understand in daytime, convective conditions because of the effects of the urban canopy alone. When the complications of nighttime conditions, terrain variations, and light winds in a stable boundary layer are added to urban canopy effects, the situation is much more physically complex and the difficulty of making reliable predictions is significantly magnified. The issue of nocturnal plume dispersion in very complex terrain combined with urban effects was not addressed in JU03 or MID05 and it is still not well understood.
Vertical Transport and Mixing (VTMX) (Doran et al. 2002) and URBAN 2000 (Allwine et al. 2002) were nocturnal experiments conducted in the Salt Lake Valley in October 2000. The combined effort was designed to generate tracer and meteorological datasets for developing pollutant dispersion models for use at range of scales in complex terrain at night. VTMX was a larger-scale perfluorocarbon tracer (PFT) and meteorological study that examined the complex wind field and tracer transport mechanisms across the Salt Lake Valley as a whole. URBAN 2000 was spatially nested within the VTMX experiment and focused on how a sulfur hexafluoride (SF6) tracer released from a site in the downtown area moved through the urban core and out into the neighboring suburbs. As such, it featured both urban canopy and topographic effects.
The URBAN 2000 effort used historical wind rose data together with the best available meteorological forecasts to select the tracer release periods for the individual experiments. By design the tracer was to be released at a site in the downtown area and transported by southeasterly (typically drainage flow) winds toward samplers mostly arrayed across the central downtown area and in the suburbs to the west and north. A few samplers were also located to the south and east of the release to confirm that the tracer was all dispersing in accordance with prediction. Plume dispersion commonly occurred as expected with drainage flows transporting the plume toward the west and northwest. However, sampling to the south and east of the release site did identify some unexpected dispersion events. It is these events that will be the emphasis of this paper.
While it is difficult to concisely portray the many nuances of plume dispersion with respect to the wind field observed in the complex setting of URBAN 2000, a representative dispersion example in typical drainage flow conditions will be shown. Then by way of contrast, cases were selected to illustrate some of the dispersion anomalies that were realized in URBAN 2000 and will be examined in detail. These more atypical cases portray some of the effects of mountain valley topography, synoptic influences, flow recirculation, flow decoupling, and the counterintuitive movement of plumes in directions different than those indicated by wind observations made on the available meteorological network. It is these more atypical scenarios that might be of particular importance in planning emergency response to the release of toxic plumes.
2. Experimental description and analytical methods
The URBAN 2000 experiment was designed to study plume dispersion in the complex urban terrain of Salt Lake City, further complicated by the effects of mountain valley topography. Six sets of SF6 tracer gas release experiments [intensive observational periods (IOP)] were conducted during URBAN 2000. Continuous releases of the SF6 tracer were made from both a point and a 30-m line source located on the nominally upwind side of the central downtown area. Concurrent and postrelease measurements of tracer concentrations were made at a gridded array of samplers across downtown Salt Lake City, at an array of samplers within one block of the release, at selected rooftop locations, and across a downwind array of samplers deployed along arcs in the nearby suburbs. Only results from line source releases are reported below. Table 1 summarizes the VTMX–URBAN 2000 IOPs.
All experiments were conducted between the hours of 2200 and 0700 mountain daylight time (MDT). Thus almost all of the measurements were made at night and, in general, within a stable boundary layer. Possible exceptions include measurements made in the dawn transition period, high wind periods, or within the urban core where urban heat island effects combined with the mechanical generation of turbulence by winds interacting with the urban canopy would have produced neutral to slightly unstable conditions (e.g., Fast et al. 2006; Britter and Hanna 2003; Hanna et al. 2007; Klein and Clark 2007; Lundquist and Chan 2007). The Field Research Division of the Air Resources Laboratory of the National Oceanographic and Atmospheric Administration (NOAA/ARLFRD) executed the SF6 tracer release and the majority of the tracer measurements for the six IOPs over the course of the study. The Lawrence Livermore National Laboratory (LLNL) measured SF6 concentrations on an array of samplers within one block of the release. The results from the LLNL samplers will not be discussed. Each IOP consisted of three separate 1-h periods during which the inert SF6 tracer was continuously released. The tracer was measured downwind on the sampler array during the 1-h release period and then for an additional hour following the end of each release. At the end of each 2-h period the next release would begin.
A complete description of the tracer measurements made by ARLFRD in URBAN 2000 can be found in Clawson et al. (2004). Two tracer concentration datasets were collected. The first were time-averaged concentration tracer measurements collected by ARLFRD Programmable Integrating Gas Samplers (PIGS). Thirty-six of these bag samplers were deployed at street corners on a grid covering the downtown area. Twenty-eight of these street corner units used a 30-min sampling average and were used in the analysis. Most of these fixed time-integrated samples were collected at half-hour intervals covering the 6-h period that composed each IOP. A total of 24 PIGS were deployed at midblock sites between the corner sites and were programmed to sample for 15-min periods. Two 15-min periods were averaged to match the half-hour sampling period of the street corner PIGS. This provided a total of 12 midblock samplers per release period. Four PIGS sampling at 30-min intervals were deployed on rooftops in the downtown area. Finally, there were 36 PIGS deployed along arcs at 2-, 4-, and 6-km distances from the release site in the suburbs north and west of the downtown area. Four of the street-level PIGS were for quality control purposes only. Four additional PIGS were deployed on 2- and 4-km arcs southeast of the downtown area using a 1-h sample interval. The two PIGS on the southeast 2-km arc were included in the analysis. All PIGS on the arcs and downtown streets were deployed 3 m above ground level (AGL). The collected PIGS samples were then analyzed in the ARLFRD Automated Tracer Gas Analysis System (ATGAS) facility. The rigorous quality control protocols used assured that the dispersion anomalies reported below were not artifacts of sample handling or analytical procedures. Further details on the ARLFRD SF6 tracer gas measurement instrumentation and quality control procedures can be found in Clawson et al. (2005).
The second concentration dataset was measured by four mobile and two quasi-stationary, continuous, real-time SF6 tracer gas analyzers (TGAs). These TGAs were deployed at distances of up to 6 km from the release site to identify in real time where the tracer plume was being transported. Data from the TGAs were acquired at a rate of 2 Hz. GPS coordinates of the van-mounted TGAs were continuously recorded.
The ARLFRD TGAs and ATGAS use electron capture detection (ECD). Method limits of detection (MLOD) for the TGAs, defined as the lowest concentration that can be determined to be statistically different from zero, typically extend down to a few tens of parts per trillion by volume (pptv). MLOD for the PIGS sampling system in URBAN 2000 was 14 pptv.
Most of the wind data acquired for these experiments and reported here were collected at a network of 11 meteorological stations located throughout the Salt Lake Valley. These were operated by the Pacific Northwest National Laboratory (PNNL) as part of the concurrent VTMX experiment. Measurements at these stations were made at a height of 3 m above the local surface (ground or building top) that varied from zero to 121 m (i.e., 3–124 m AGL). Four of these 11 stations were in the downtown area. With the exception of the Latter Day Saints temple (LDS) station at 121-m elevation, all other PNNL stations in the downtown area ranged from 4 to 20 m (measurement height 7–23 m AGL). The base level of the PNNL stations outside the downtown area ranged from 0 to 11 m. An additional station, operated by ARLFRD, was located in the suburbs 5 km southwest of downtown (Clawson and Crescenti 2002). Two additional sonic anemometers operated by ARLFRD were collocated in a parking lot 86 m from the release site at 6.9 and 9.81 m AGL. Finally, a set of six sonic anemometers were deployed by Los Alamos National Laboratory (LANL) in the vicinity of the release site at heights ranging from 3.5 to 46.3 m AGL (Streit et al. 2001). Five of these were deployed on rooftops near the release site, 39, 48, 69, 103, and 471 m away, and the remaining one was deployed at the surface 18 m from the release site. Data from the ARLFRD anemometer placed at 9.8 m were used to represent the wind measurements from the vicinity of the release site in the figures below.
The overall VTMX experiment also included radar wind profilers, RASS, sodar, Doppler lidar, rawinsonde, and tethered balloon measurements (Doran et al. 2002).
Concentration footprint-wind vector maps will be used to illustrate several features of plume dispersion observed during URBAN 2000. These maps show color-coded symbols for street corner (circle), midblock (diamond), rooftop (square), and arc (triangle) PIGS. All the concentration results are half-hour averages except for the two PIGS to the southeast. If real-time TGA mobile traverse concentration data were available at a desired time and location, these data were used to confirm the PIGS results, check for anomalous plume behavior, and/or define the plume’s location where it appeared that the plume might have dispersed to areas not covered by the PIGS sampling array. The thick colored lines represent selected mobile TGA traverses executed at some time during the half-hour period indicated on the map. Wind measurements were reported every 5 min and averaged to half-hour periods to be consistent with the PIGS tracer concentration data. Mean vector wind speeds are proportional to the wind scale included in the maps. Wind speeds less than 0.4 m s−1 are shown by a “+” symbol. An “X” indicates the release site.
a. Typical drainage flow conditions
A representative example of plume dispersion in typical nocturnal “well-developed drainage flow” conditions (Table 1) is shown in Figs. 1 –3 (IOP4, release 1). The overall wind field (Fig. 1) exhibited generally light winds with flows toward the axis of the valley from the surrounding high terrain and then down the valley toward the north. The wind field shown by the four PNNL stations plotted in the downtown area was typical for these stations in this situation. The winds observed at the station east of the main downtown area, 7 m AGL, were typically very light and easterly. Winds at the station to the south, 11 m AGL, were usually very light and most commonly easterly or northeasterly. The wind at the station to the west of downtown, 23 m AGL, was usually stronger and tended to be more southerly. This station tended to be more strongly correlated in wind speed and direction with other stations lying in the more open topography of the valley than with its immediate neighbor stations in the downtown area. The northernmost wind observation shown, immediately north of the downtown sampling array, was located at 124 m AGL (LDS station). Wind observations at this height tended to be very poorly correlated with nearby stations. Winds at the downtown ARLFRD station were generally very light but usually consistent with the PNNL downtown stations to the south and east. On the maps, this station is often difficult to distinguish because of the close proximity to the release site (immediately west) and low wind speeds.
In the first half-hour after the tracer release began (0100–0130 MDT), the tracer plume had moved less than 1 km (Fig. 2). The track of the plume was toward the west or west-southwest. The speed and direction of plume movement were consistent with the weak winds observed at the downtown area stations to the east and south and there was a suggestion of topographic influence. No significant tracer concentrations were observed at rooftop level.
During the second half-hour after the release began (0130–0200 MDT), the plume had reached the suburban arcs and the plume began tracking toward the northwest, perhaps aided by the wind direction shift and increase in wind speed observed atop the 121-m LDS building in this particular episode. Although the track of the plume was coincident with the winds aloft at the LDS station in this instance, this was frequently not the case. Significant tracer concentrations were measured at two of the three rooftop samplers although the concentrations were never as high as at street level.
The available TGA traverses corroborated the PIGS results.
The pattern illustrated in Figs. 2 and 3 is a variation on a common theme seen in URBAN 2000 drainage flow conditions with an initially west or southwest plume track followed by a turning toward the northwest. In some instances, the tracer concentrations at samplers on the northwest corner of the grid were low and the plume mainly exited the downtown area toward the arcs along the west and southwest edges of the grid. In other cases, the plume moved more directly toward the northwest initially. In general, the arc samplers most impacted by the plumes were those lying immediately to the northwest of downtown along the mountain front. In some cases the plume reached the more southern arc samplers first before turning toward the northwest. The general picture was one of weaker drainage flows transporting the plume out of the downtown area toward the west, southwest, or sometimes northwest. The plumes then tended to turn toward the northwest in the more open terrain of the valley where southerly winds prevailed in drainage situations.
b. Drainage flow conditions with very slow dispersion
During the second release of IOP2 on 7 October, a flow pattern was observed that was characterized by very slow dispersion (Figs. 4 –6). The overall wind pattern associated with this dispersion pattern was only subtly different from that described in the previous section for drainage flows in the downtown area (Fig. 1). The winds were poorly organized and the wind speeds were low. The mean half-hour wind speeds at the ARLFRD station near the release site ranged from a high of 0.42 m s−1 to as little as 0.06 m s−1. However, there were locally strong downslope winds exiting from some of the canyons to the south of the downtown area, in particular, Parley’s Canyon (not shown, but see example from a few hours later; Fig. 7). These locally strong canyon winds were due to cold air from east of the Continental Divide advancing westward toward the Salt Lake Valley at this time (Doran et al. 2002). The striking feature here is that dispersion of the tracer plume was limited to a few hundred meters from the release site over the course of 2 h, even failing to reach the edge of the downtown sampling grid in any significant way. The slightly elevated tracer concentrations observed at some of the arc samplers at the start of the illustrated release period (0300–0330 MDT) were a relic from the prior release from 0100 to 0200 MDT.
c. Strong downslope winds
By the third release of IOP2, consistently strong easterly downslope winds had developed, especially near some of the canyon gaps (e.g., Parley’s Canyon; Fig. 7). In the downtown area, the typical drainage flow pattern was replaced by what appeared to be a horizontal clockwise recirculation zone. While the overall wind field exhibited stronger easterly winds along the eastern edge of the valley and south-southeast winds in the valley itself, the winds in the downtown area tended to be from the north and west. A feature worth noting here is the fact that downtown sits in a bend in the Wasatch Front defined by a westward topographic flexure that encompasses Capitol Hill and Ensign Peak. This prominent ridge or buttress lies immediately north of the downtown area. It is conjectured that this topographic feature influenced the wind field. In this particular scenario, drainage flows transitioned into frontal boundary downslope flows and generated a recirculating eddy and upslope flows in the downtown area. In addition, this topographic feature might have played a role in what appears to be a delay or dampening of increased wind speeds in the downtown area as the overall wind speeds associated with the downslope flows began to increase.
This affected the evolution of the tracer concentration field in a way that might be expected (Figs. 8 –10). The tracer plume failed to reach the western or northern boundary of the downtown sampling grid. Instead, the plume was mostly driven toward the south and east. The contrast in the downtown area wind field between the normal drainage pattern and the recirculation pattern is also apparent in these figures. The available TGA traverse data corroborated the PIGS data. One intriguing anomaly worth noting was the high concentrations measured by a mobile TGA southwest of downtown as shown in Fig. 9.
d. Anomalous dispersion
A fairly typical drainage flow pattern was in place during the first release of IOP7, on 18 October, with light east and northeast winds along the eastern flank of the valley near the Wasatch Front and light southerly winds in the center of the valley (Fig. 11). The interesting feature of this release period is the very poor correlation between the details of the observed wind and tracer concentration fields. During the first half-hour (Fig. 12) only the winds measured near the release site at the ARLFRD station were in agreement with the observed westerly, downslope, drainage flow plume dispersion pattern. During the second half-hour of release, the tracer plume continued to follow the typical drainage flow pattern. The main tracer plume reached the edge of the downtown sampling grid and turned toward the northwest (Fig. 13). The available TGA data again corroborated the plume dispersion pattern given by the PIGS. What is curious is that this northwesterly plume dispersion was in apparent opposition to any of the winds measured at stations around the downtown area. Winds at the stations to the south and east were typical of drainage conditions and were out of the east or east-northeast. The other two stations showed northeast and southwest winds. In any case, the observed winds were inconsistent with the plume tracking toward the northwest (i.e., no southeast winds).
Another curious feature of this release was the broad, diffuse, relatively uniform concentration field observed at the outer arc samplers. The plume was still relatively narrow at the first arc but by the second and third arcs it had broadened greatly, resulting in low-level concentrations up to 100 pptv over a large area. To further complicate this picture, in the first half-hour after the release had ended (Fig. 14), east winds were being measured at all four stations surrounding the downtown area. Nevertheless, high concentrations were still being measured in the vicinity of the release site, even after the bulk of the tracer appeared to have passed through the sampling grid in the first hour, and some notably elevated concentrations were detected at “upwind” samplers. In particular, high concentrations were measured by a mobile TGA over a half-kilometer to the east of the release site. Finally, in the next half-hour, the concentration field measured at the arc samplers was almost completely homogeneous with some higher concentrations even being reported at some higher elevation samplers in the foothills to the north of downtown (Fig. 15). The highest concentrations in the grid sampler array lay on the eastern, upwind edge of the grid. Mobile TGA traverses between 0230 and 0300 MDT also still detected significant tracer concentrations more than a kilometer east of the downtown release site.
Figure 16 sheds some light on the origin of the tracer anomalies in Figs. 14, 15. The winds observed at the ARLFRD station and LANL stations [(F) and (L), respectively, in Fig. 16] near the release site were mostly easterly throughout IOP7 with the exception of a period encompassing 0200–0230 MDT, when they were westerly. The half-hour average wind speed at the ARLFRD station was 0.59 m s−1 during this half-hour. This temporary, localized shift to light westerly winds in the vicinity of the release site was not detected by the surrounding downtown area PNNL stations (P). Thus it appears as if part of the explanation for the tracer distribution pattern after 0200 MDT was that there was a local, small-scale surface wind that was acting to transport the tracer to the east and upslope, which was in opposition to all other locally measured winds. The greater variability observed in the wind direction between lower and higher heights (Figs. 16a,b, respectively) suggests possible urban canopy effects even though the localized westerly winds were detected at all the stations.
Some conjectures are suggested by this experiment. First, plume movement having little or no correlation with the observed winds implies that dispersion in typical drainage flow conditions is driven mostly by local, small-scale flows in a relatively shallow layer, strongly influenced by the local topography (Figs. 12, 13). Second, there was an obvious persistence of high tracer concentrations in the release area. This suggests that the tracer might have been trapped within poorly ventilated areas in the urban core, taken up by buildings and later off-gassed, and/or that the tracer was transported in some manner off the grid to the east and was later transported back toward the release site. The potential implication of the latter is that some sort of horizontal and/or vertical recirculation occurred. While rooftop concentrations were low with the lone exception of a measurement from 0200 to 0230 MDT (Fig. 14), the somewhat higher concentrations in the arc samplers north of downtown in Fig. 15 suggest the presence of some vertical mixing and transport. Third, the broad, uniform dispersion pattern in the outer arcs would tend to support the notion that simple gradient diffusion, in the absence of a well-defined controlling wind field, might have played a role in the tracer dispersion.
Finally, the postrelease results from 0200 to 0300 MDT (Figs. 14 –16) reinforce the idea that dispersion in these conditions is driven by local flows on scales that can only be measured by a dense meteorological monitoring grid. Furthermore, these weak local winds can sometimes override the effects of topography. The origin of the localized westerly winds measured by the ARLFRD and LANL stations near the release from 0200 to 0230 MDT is a matter of conjecture given the data available. One idea is that these arise from the interaction of near-surface outflow from City Creek Canyon with the urban canopy in the downtown area. City Creek lies in the distinct north–south-aligned canyon that emerges from the Ensign Peak area just north of downtown. Flows exiting the canyon would project southward into the downtown area but probably not be detected by the station at 124 m AGL (P124). It has been found that urban canopies induce channeled flow in street canyons oriented approximately perpendicular to the approach flow with the direction of channeling governed by small changes in wind direction (Klein and Clark 2007; Nelson et al. 2007a; Dobre et al. 2005). If outflow from City Creek Canyon had a slightly westerly component, it could have induced westerly winds in the downtown area as it was redirected by crosswind-oriented (east–west) street canyons. Alternately, the interaction of other flows with the heterogeneous Salt Lake City urban canopy could have induced down/updrafts that resulted in zones of convergence and divergence and the generation of local winds (Nelson et al. 2007a; Hosker 1987).
e. General observations
As noted earlier, the winds observed at the LDS station (P124) were often poorly correlated with neighboring stations. This suggests that flows aloft were commonly decoupled from flows closer to the surface (<20-m height) and that these elevated wind observations were a poor guide for anticipating or predicting plume dispersion for the downtown area. There is some suggestion, however, that when the LDS station wind direction was from the south or southeast and synoptic influences were stronger, then wind speeds were faster and flow coupling did occur (e.g., IOPs 9 and 10; Table 1). The best correlation between the track of the tracer plume and wind directions measured at the LDS and other stations was observed in this situation. In this situation the plume tended to track toward the northwest, along the mountain front, or sometimes even north into the higher elevations. This was not at all typical but notably elevated tracer concentrations were observed at some of the elevated arc samplers in this situation. Otherwise, encroachment of the tracer plume into the higher elevations to the north was generally limited and suggested limited vertical mixing and the importance of topography in controlling dispersion.
Street-level concentrations were always greater than concentrations measured at rooftop samplers and the differences were often large. Tracer concentrations at rooftop samplers were often low or negligible. Nevertheless, it was not unusual to measure significant tracer concentrations (e.g., up to 5825, 1868, and 3185 pptv in Figs. 3, 10 and 14, respectively).
Finally, there is evidence that dispersion in directions inconsistent with the observed winds was a common occurrence in typical drainage flow conditions, part of which can be seen in the figures presented. In Fig. 2, the middle sampler of the easternmost downtown grid samplers measured 293 versus 0 pptv for the samplers to the north and south of it. This sampler was clearly upwind of the release site based on the available wind observations. The feature is more evident in Fig. 3 with 3348 pptv at the upwind sampler on the southeast corner of the downtown grid. Later releases for IOP4 were even more distinct in this regard but were not selected for illustration since other aspects of the dispersion were more obscured. For example, interpretation of a release could be complicated by the persistence of relic tracer in the sampling area from earlier releases.
In Fig. 12, while the main plume track was toward the west, one of the easternmost grid samplers measured an isolated high tracer concentration of 3288 pptv. Given the wind vectors shown, it is not clear if this sampler was really upwind or not. However, a similar dispersion pattern exists in Fig. 13 that is inconsistent with the wind vector pattern, as previously discussed. Then, in the hour after the release (Figs. 14, 15), the highest concentrations were observed at the eastern end of the sampling grid, even in the presence of a distinctly easterly observed wind pattern. Some possible explanations for the evolution of this concentration pattern over the 2-h period were listed in section 3d.
The mobile TGAs also found evidence of complicated dispersion. One example of this was described in section 3d, especially Figs. 14, 15. Another example is shown in Fig. 9 from IOP2. It can be seen that the plume moved primarily toward the southeast at this time. This is approximately consistent with the wind vectors observed to the north and west of the downtown area. However, the winds observed at the stations to the south and east of downtown were easterly and in opposition to the observed plume track.
Doran et al. (2002), Darby et al. (2006), and Fast et al. (2006) have grouped the overall Salt Lake Valley nocturnal dispersion patterns into two broad regimes. One regime occurred during periods with weak surface pressure gradients and synoptic forcing (Table 1; “well-developed drainage flows”). It was characterized by thermally forced downcanyon and downslope drainage flows along the Wasatch toward the axis of the valley with well-developed low-level downvalley jets flowing from south to north. The downvalley jets were part of a diurnal cycle with northerly winds up the valley during the day and southerly winds down the valley at night. Vertical mixing occurred locally, associated with zones of convergence between slope, basin, and canyon flows, or due to turbulent mixing generated by wind shear across the top of the stable layer (Pinto et al. 2006; Banta et al. 2004; Fast and Darby 2004). The drainage flows were highly stratified below 25–50 m and largely decoupled from the winds above (Monti et al. 2002). While IOP4 has been classified overall as a modulated drainage circulation (Fast et al. 2006), the particular time period encompassed by Figs. 2, 3 was generally representative of the first regime wind field and dispersion pattern.
The second regime was driven by synoptic or mesoscale influences with well-organized, basinwide, large-scale winds and strong pressure gradients. It was characterized by higher wind speeds, enhanced vertical mixing, and suppression of thermally forced drainage flows. Plume transport was well coupled with the large-scale wind field and plume spread was less than in the first regime.
The IOP2 experiment represented a hybrid situation between the two regimes with drainage flows in the downtown area modulated by strong easterly canyon and downslope winds driven by high pressure to the east. The easterly winds increased in strength as time passed in the IOP. By the second release (Figs. 4 –6) the combination of relic drainage flows in the downtown area and the convergence of valley flows and synoptically driven canyon flows had set up a situation where dispersion in the downtown area was essentially shut down. The wind field in the downtown area was very light and poorly organized and was essentially in transition to forming a horizontal eddy with a circulation counter to the canyon and valley flows. By the third release (Figs. 8 –10) the horizontal eddy, located over the northeastern part of the valley (the downtown area), was fully developed. The tracer was then advected toward the south and east along the Wasatch Front.
The time period encompassed by Figs. 12 –15 (IOP7) was also generally representative of the first regime wind field and dispersion pattern. Again, this was an intriguing release period 1) because the dispersion trends were largely independent of the observed winds in the downtown area and 2) because of the persistence of the tracer signal in the release area and upwind of the release, after the release was off. The first point attests to the argument of Darby et al. (2006) that small-scale flows, often unresolved by the available meteorological grid, account for much of the dispersion in stable conditions in complex terrain. The second point 1) suggests that the tracer might have been trapped within buildings and poorly ventilated areas downtown and later released, 2) offers further support of the importance of small-scale flow features in nocturnal conditions in complex terrain, and/or 3) suggests that some sort of recirculation was occurring to account for the presence of the tracer detected by the TGAs well upwind of the release site.
The larger-scale VTMX study documented cases in which there was sufficient vertical mixing for the PFTs to escape above the stable, low–wind speed surface layer into a zone with higher wind speeds (Fast et al. 2006). If the wind direction aloft was consistent with the surface winds the tracer transport was accelerated in the same direction and then could potentially mix back to the surface at a point more distant than expected based on the surface winds. However, if the wind direction aloft was contrary to the drainage flow circulations, the winds aloft could shear off the plume and potentially advect it back upwind of the release site where it could mix back to the surface. Doppler lidar profiles for IOP7 show southerly winds aloft throughout the experiment (Darby et al. 2006), however, so it is difficult to understand how this mechanism would explain the presence of the tracer to the east of downtown well after the release was off. The more likely explanations for the anomalies associated with the first release period of IOP7 are tracer trapping and small-scale flows. The small-scale flows might have originated from surface-level outflow from City Creek Canyon and/or higher-level flows interacting with the heterogeneous urban canopy to create channeling and/or divergence–convergence in the local wind field. A hint of these small-scale flows was given by the observations at the ARLFRD and LANL stations, something that was completely missed by the rest of the meteorological grid.
Flow decoupling merits further comment. It has been established that nocturnal decoupling develops in open terrain above the stable boundary layer (Stull 1988) and there was direct evidence for this in the URBAN 2000 results. It was pointed out that the winds measured aloft at the LDS station were often in disagreement with those measured closer to the surface. For that matter, there was often disagreement even between stations at lower heights. Furthermore, examples were given showing how the plume often dispersed in directions inconsistent with the nearby measured winds, most notably IOP7 (Figs. 11 –15), and that an increase in wind speed promoted correlation between wind direction and plume transport direction.
Flow decoupling is probably characteristic of situations when the factors of weak synoptic forcing, a stable nocturnal boundary layer, and complex terrain are combined. In other studies featuring conditions comparable to the drainage flow regime in URBAN 2000, flow decoupling was also observed (Chen et al. 1999; Gudiksen 1984). The degree of interaction between the drainage flows (slope and valley, canyon, localized channeling) and flows aloft was related to how effectively topography shielded the drainage flows. The results of Mahrt et al. (2001) and Soler et al. (2002) further emphasize the importance of understanding the role of flow decoupling on dispersion in complex terrain. In terrain characterized by much less topographical variability and complexity than URBAN 2000, it was found that drainage flows a few meters in depth were present in shallow gullies. They were flowing in opposition to winds up to 10 m s−1. The decoupling sometimes occurred at heights less than standard observation heights of 3–10 m. When the influence of the urban canopy on plume dispersion is considered along with observations such as these, the challenges posed in trying to understand dispersion in urban complex terrain become apparent.
The question might be asked as to how frequently dispersion events such as those depicted in sections 3b, 3c, and 3d might occur. The data would suggest that situations such as in sections 3b and 3c would probably occur whenever first regime conditions are present in the Salt Lake Valley and synoptic conditions are such that strong canyon outflow winds are generated. The characteristic pattern for this is very slow and/or upslope dispersion in the downtown area. Transport modeling studies found that the 3b/3c scenario was associated with greater persistence of a larger fraction of pollutant mass released within the valley (Fast and Darby 2004). There was also a tendency for higher concentrations of the pollutant to be found in the northeast part of the domain (i.e., in the downtown area), in agreement with the slow dispersion reported here.
The upwind dispersion event (section 3d) might be more common than expected. If it is classified as a well-developed drainage flow, then it was observed 1 out of 7 times in which tracer data are available (Table 1). If it is categorized as a modified drainage flow due to later onset of synoptic influences during the IOP, but without the strong downslope winds, then it occurred 1 in 6 times for which data are available. Upwind dispersion was also detected in a VTMX perfluorocarbon tracer result (Fast et al. 2006). This was attributed to “temporal and spatial variations in the winds. . . . for a short period of time.” It should also be pointed out that the TGA efforts were focused on where the plume was expected. The TGA traverses that located the anomalous upwind plume (section 3d) were made because they were unable to locate the plume where it was anticipated. Thus, it is possible that more such events might have been observed if more TGA resources had been allocated to the task of looking for the unexpected. It is also interesting to note that in simulations of the averaged vertical velocities over the valley, IOP7 most closely resembled IOPs 2 and 3 (Fast and Darby 2004, their Fig. 13). This would suggest a possible role for canyon outflow in IOP7.
The significance of the chances of realizing scenarios such as in section 3d should be made clear. While there are no universally adopted emergency response protocols, the available guidelines would fail to account for a situation like this and people could be needlessly endangered (Brown and Streit 1998; Allwine et al. 2007). Given the poorly organized wind fields in the 3b/3c section scenarios, it is also a distinct possibility that errors in response could occur in these somewhat better-behaved situations if an unfortunate choice of which wind data to use was made.
Attempts have been made to model the tracer concentration fields observed during URBAN 2000 (Hanna et al. 2003; Gowhardhan et al. 2006; Chang et al. 2005). What should be clear by now is that accurate modeling is very challenging in stable, light wind conditions in complex terrain. Flow decoupling, extremely slow dispersion, apparent tracer trapping and rerelease, and plume transport in directions opposed to the observed winds are some of the challenges. The problem posed is being able to develop a model that can predict plume transport in a direction completely opposite to the observed wind field. In the absence of a model with this capability it would be necessary to have a dense meteorological grid that is able to resolve the small-scale flow field and/or tracer studies that can identify the types of dispersion anomalies that might occur in a given set of urban complex terrain conditions.
The choice of running dispersion experiments in nocturnal conditions in Salt Lake City would lead, at first glance, to a relatively predictable dispersion pattern dominated by drainage flows. In fact, a somewhat characteristic pattern of dispersion was usually observed. Tracer released in the downtown area usually dispersed toward the west and northwest, consistent with topography, the drainage flow pattern, and prediction. However, the URBAN 2000 dataset also brought to light exceptions when the effects of mountain valley terrain, drainage flows, stable nighttime conditions, urban canopy, and synoptic events sometimes combined to produce the unexpected. This dataset offered an opportunity to realize what can happen when these factors interact in a complex way. Some of the atypical or unexpected dispersion patterns that were observed during URBAN 2000 include
dispersion independent of and in opposition to almost all the wind direction observations made on the available meteorological grid [local flows on small scales below what could be resolved by anything but a dense meteorological grid were major drivers of dispersion in the stable, low-wind conditions (e.g., Figs. 11 –15)], and
very slow plume decay in the urban core near the release site, possibly due to trapping of the plume in buildings or poorly ventilated areas and then later rereleasing to the atmosphere.
In the nocturnal conditions and complex urban and topographical terrain characterizing URBAN 2000 it should be borne in mind that plume dispersion could be significantly different than what might be expected based upon the available wind data. It is problematic to rely on any one urban area wind measurement to predict or anticipate dispersion. Small-scale flows can be very important in dispersion and the potential effects of the larger-scale, synoptically driven flow field and its interactions with the smaller scales need to be carefully considered. Even when plume dispersion proceeds mostly as expected, the possibility for surprises still exists. Future dispersion experiments in complex urban/topographical terrain, especially in stable conditions, should include plans to systematically evaluate the possibility of dispersion in directions different than that suggested by the available wind data. Future model development should be guided by paired, finer-scale meteorological and tracer measurements that could potentially resolve the effects of the small-scale flows observed in URBAN 2000.
We acknowledge the efforts of the many people who contributed to the execution of the URBAN 2000 study and made possible the analyses contained in this manuscript. In particular we thank Debbie Lacroix, Dr. Tami Grimmett, Neil Hukari, Brad Reese, Dr. Jeff French, Randy Johnson, the late Dr. Timothy Crawford, Wayne Hooker, and Tom Strong from ARLFRD who helped with instrument preparation, data collection, chemical analyses, and report preparation. We also thank Jim Bowers of the U.S. Army Dugway Proving Ground, Dr. Joe Shinn of the Lawrence Livermore National Laboratory, and Dr. Gerald Streit and their staffs for their cooperation and assistance in making this project a success. We also acknowledge the assistance of local and state officials including Cindy Clark (Utah Automated Geographic Reference Center), Nick Kryger (Public Works Department of Salt Lake City), and Mark Miller (Salt Lake County Surveyor’s Office). This project was supported by the National Oceanic and Atmospheric Administration, by the U.S. Dept. of Energy’s Chemical and Biological National Security Program under Interagency Agreement DE-A101-01NN20120, and the Defense Threat Reduction Agency (DTRA) of the Department of Defense (DOD). The DOD Military Interdepartmental Purchase Request number was MIPR5KDPG87101.
Corresponding author address: Dennis Finn, NOAA/ARLFRD, 1750 Foote Drive, Idaho Falls, ID 83402. Email: email@example.com