TMCnet News

On the Drag and Heat of Washington, D.C., and New York City [Journal of Applied Meteorology and Climatology]
[July 16, 2014]

On the Drag and Heat of Washington, D.C., and New York City [Journal of Applied Meteorology and Climatology]


(Journal of Applied Meteorology and Climatology Via Acquire Media NewsEdge) ABSTRACT Data from a network of micrometeorological instruments, mostly mounted 10 m above the roofs of 12 buildings in Washington, D.C., are used to derive average values and spatial differences of the normalized local friction velocity ... (with u being the wind speed reported at the same height as the covariance is measured, w being the vertical wind component, primes indicating deviations, and the overbar indicating averaging). The analysis is extended through consideration of two additional sites in New York City, New York. The ratio u*/u is found to depend on wind direction for all locations. Averaged values of u*/u appear to be best associated with the standard deviation of local building heights, with little evidence of a dependence on any other of the modern building-morphology indices. Temperature covariance data show a large effect of nearby activities, with the consequences of air-conditioning systems being obvious (especially at night) in some situations. The Washington data show that older buildings, built largely of native limestone, show the greatest effects of air-conditioning systems. The assumption that the nighttime surface boundary layer is stable is likely to be most often incorrect for both Washington and New York City-the sensible heat flux resulting from heating and cooling of building work spaces most often appears to dominate.



(ProQuest: ... denotes formulae omitted.) 1. Introduction "DCNet" is a research program initiated in the aftermath of the terrorist attack of 11 September 2001, when it was recognized that the ability to describe or forecast dispersion in urban areas was poor despite recent short-term but intensive experimental activity re- lated to the meteorological behavior of cities and the dispersion of chemicals within them. The number of numerical models available and the different ways in which they accessed data were widely seen as evidence of unacceptable disarray, aggravated by the endemic lack of relevant real-time data and of systems to access such data should they exist (see OFCM 2002). Logistic difficulties in erecting tall towers compound the prob- lems encountered. To step past this constraint, rooftop towers have been contemplated, regardless of the fact that interpretation of their data could be a problem because they would necessarily violate some widely accepted micrometeorological constraints. One of the goals of the DCNet program considered here is to learn how to make use of such rooftop data. As recently as 2008, the Government Accounting Office reported (in response to a congressional request) that "[e]valuations and field testing have shown an unpredictable range of uncertainty in urban dispersion models' analyses" (GAO 2008, p. 44). To obtain the best model results, "ready availability of building-top winds is essential" (GAO 2008, p. 47).

Somewhat in anticipation of later examinations of urban dispersion capabilities, such as that led by the Government Accounting Office (GAO 2008), DCNet was initiated in 2002 with a small network of micrometeoro- logical towers set up on the roofs of selected buildings across the national capital region of Washington, D.C., and extending into neighboring areas. In subsequent years, the network expanded as awareness of the magni- tude and consequences of spatial variability in urban areas grew. DCNet was designed to be a long-term study, in contrast to the short-term intensive studies conducted in the last few decades-in the United States, there have been major short-term studies in, for example, Salt Lake City, Utah; Oklahoma City, Oklahoma; and New York City, New York (see Allwine et al. 2002; Allwine and Flaherty 2007; Hanna et al. 2003, 2007). These intensive studies have revealed the complexity of the urban envi- ronment in considerable detail. For example, it is now appreciated that details of buildings and street orientation can be controlling factors in the movement of pollutants, and methods to address such matters as the distribution and shape of upwind buildings have been proposed (e.g., Grimmond and Oke 1999; Theurer 1999; Burian et al. 2008; Ching et al. 2009). These complexities lead to a suite of difficulties complicating the provision of forecasts relevant to where most people are actually exposed- at the lowest level of the urban boundary layer and well below the tops of buildings.


Figure 1 shows the network of DCNet sites as of the end of 2010. DCNet employs three-dimensional sonic- anemometer systems at a nominal height of 10 m above large buildings (e.g., Fig. 2) and located to minimize pos- sible effects of roof edges and nearby structures. Data from the sonic anemometers are accessed at 10 Hz by local data- acquisition systems that compute all averages, variances, and covariances over 15-min periods. Every 15 min, computed results are transmitted via cellular modem to a central archive at the Atmospheric Turbulence and Dif- fusion Division of the National Oceanic and Atmospheric Administration (NOAA) Air Resources Laboratory (ARL) in Oak Ridge, Tennessee. All archived data are available, upon prior arrangement, via the Internet. There are no supporting wind measurements at different heights, as would be needed to quantify the roughness length z0 and the displacement height d directly.

Several considerations led to the selection of Washington, D.C., as a focal area: 1) the area was the site of an extensive study of dispersion in the 1980s, lasting more than 1 year (Draxler 1987a,b), 2) the downtown area is well defined, with building heights not exceeding about 27 m (as required by the Washington Buildings Act of 1910), and 3) the U.S. capital region is a proven target of terrorists, and results of the DCNet program might therefore be rapidly migrated to operational applications.

In recognition of the importance of the central business district of New York City, two additional sites were deployed, operating as remote members of the DCNet array: one in the immediate vicinity of Times Square (TSQ) and the other adjacent to the Houston subway station near the southern tip of Manhattan Island (EML).

The analysis presented here is intended to introduce the DCNet database to the urban-research community. The analyses here focus on variations of surface rough- ness with space and season and on sensible heat fluxes at night. In the first part of the analysis to follow, the em- phasis will be on the normalized local friction velocity u*/u = (2u0 w0 )1/2/u, where u is the wind speed at the height of measurement of the covariance. In the second part of the analysis, examination of the sensible heat exchange will make direct use of the temperature T and covariance w0T 0 evaluations yielded by the same sonic- anemometer systems.

2. Background considerations Analysis of the data considered here must be done cautiously, because DCNet stations violate standard micrometeorological practices regarding fetch unifor- mity and height above obstacles. Regardless of such problems, the Monin-Obukhov construct L appears to provide a reasonable basis for analysis.

Measurements of covariances can be made at any lo- cation. The challenges are to relate these covariances to eddy fluxes, to identify the upstream areas of relevance (the footprint), and to quantify the corresponding sur- face properties in a way that improves understanding and permits extrapolation to other situations. The re- gion of influence for the turbulence measurement foot- print is expected to be site specific (e.g., Schmid 2002; Leclerc et al. 2003). In the current case, the daytime footprint related to sensible heat is likely to be different from that for the momentum flux, because the former is controlled by buoyancy whereas the latter is not. In both cases, however, the greater the height of measurement is, the more likely it is to derive a spatially representative upwind value.

Over conventional micrometeorological sites (char- acterized by spatial homogeneity and time stationarity), the roughness length is derived from measurements of wind profiles and the momentum flux by using the neutral relationship ... (1) where u is the wind speed, u* is the local friction ve- locity, and z is the height of measurement above the ground. Here, the von Kármán constant k is taken to be 0.4. In numerical models, it is common to take d and z0 to be constants that characterize a grid cell. Because of subgrid-scale complexity, it is not immediately clear how to derive appropriate gridcell averages of eddy fluxes for an urban area, regardless of the size of the grid. This is not only because the constraints of spatial homogeneity and time stationarity are usually violated but also be- cause obtaining a "representative" set of measurements is demanding. The variable that is most affected by surface structures is the covariance u0 w0 (because sur- face obstructions strongly influence both u0 and w0 ). The usual experimental approach is to employ a strategically located tall tower, with wind speed gradients measured and with sonic anemometry high enough to minimize the influences of specific surface features. Any tall-tower installation is constrained by the "rules of thumb" generated in past micrometeorological field studies, however. These legacy guidelines require that mea- surements should be made sufficiently above surface obstructions that the influence of individual roughness elements is minimized but not so high as to be outside the surface "constant flux" layer (extending to 7%-10% of the relevant mixed layer). The definition of "suffi- cient" varies among researchers, and actual illustrative data for an urban area are largely lacking. These guidelines have been refined by detailed modeling [e.g., Raupach (1989) for the case of a densely vegetated surface], supporting the conclusion that an urban situa- tion constitutes a demanding application for both ex- perimentation and simulation.

In this regard, note that TSQ in New York City and the Silver Spring, Maryland, (SSG) site in the Washington suburbs are the main violators of the above guidelines among the sites now considered. Data from an additional elevated installation (CSPN) at a height of 87 m in the suburbs of Washington are presently excluded.

3. Data selection and analysis Table 1 lists details of the DCNet stations considered here. In the following discussion related to surface roughness, all available data collected during calendar year of 2008 have been used. Later, in considering heat exchange, data obtained over several additional years will be used. To eliminate questionable data, the fol- lowing constraints have been imposed.

1) For each 15-min period, complete data records are required.

2) To ensure optimal anemometer performance, cases for which the average wind speed is less than 1 m s21 have been excluded. (This exclusion is based on field experience with the devices of concern. All in situ anemometers suffer from performance defects at low wind speeds. Experienced field experimenters will have different criteria to apply that are based on their own tests of the instrumentation that they actually use).

3) Periods for which rainfall was reported are excluded.

4) After coordinate rotation, the average angle of attack of the wind relative to the horizontal must be less than 108.

The last criterion eliminates occasions that are egre- giously affected by upwind obstructions and the result- ing flow distortion. Every site of the DCNet program has unique characteristics that require consideration. For example, the NAS site is located on the roof of the National Academy of Science, on Constitution Avenue near the Vietnam Memorial. To the north is the complex of the U.S. Department of State. To the south is the grassland of the National Mall. To the east and west are buildings and gardens similar to the surroundings of the National Academy itself. It is clear that some conse- quences of upwind obstacles must be expected. As a first- order correction for the consequences of streamline deformation, this analysis will make use of coordinate rotation (see, e.g., Wesely et al. 1972). Such coordinate rotation is commonly applied to correct for sensor- induced (and other) deformations of local streamlines away from the true reference plane, across which there is no net exchange of dry air, or to derive orthogonal wind components from measurements made using a different coordinate system. Table 1 lists evaluations of coeffi- cients of variation (CoVs, defined as the ratios of stan- dard deviations to average values) of u*/u,computedfor near-neutral conditions (jLj . 1000 m) that satisfy the data-quality constraints given above. Two values of the CoV are given, one using raw data [CoV(1)] and the other using data after coordinate rotation [CoV(2)]. A lower value of the CoV means that the standard deviation in the derived friction velocity is reduced, even after al- lowance for the dependence of u* on the wind speed. The coefficients of variation show that coordinate rotation does indeed improve the dataset, with the resulting values of the CoV being reduced for all sites and by a substantial amount for some. The wind analysis here will make use of rotated datasets, arranged to yield stress evaluations along the vector wind with no average crosswind or ver- tical components.

Figure 3 shows the results of a first-cut inspection of the data, using results obtained at the American Geophysical Union (AGU) site. For the present, the low wind speed criterion is relaxed; all data are used for Fig. 3a,which shows how the covariance u0 w0 =2u2* varies with the square of the longitudinal wind velocity u2.Amore conventional plot might be of u* against u;however,the present intent is to look at the entire dataset with rec- ognition that site imperfections (and other factors) will give rise to positive and nonphysical values of u0 w0 and that analysis should not discount these measurements. Because of the large number of measurements, Fig. 3a presents every fiftieth record of the overall time se- quence, without any additional sorting. It is the high wind speed data shown in Fig. 3a that are most indicative of neutral conditions, and these data are widely scattered. Figure 3b shows how the near-neutral subset of the AGU dataset behaves. For this purpose, only those data for which the sensible heat flux is near zero have been se- lected (jw0 T 0j , 0.001 K m s21). All such data also satis- fying the other constraints above are shown. The scatter is reduced considerably. The lines drawn are the results of a regression analysis yielding the properties shown in the diagrams. Even though R2 is improved by the data sorting of near-neutral runs, the slopes of the two lines are nearly identical. Furthermore, the slope of a regression con- strained to pass through the origin is 20.0117 6 0.0030, as compared with values of 20.0122 for the near-neutral data of Fig. 3b and 20.0118 for the data of Fig. 3a,with similar standard errors. These values are determinations of the drag coefficient Cd prevailing at the times of mea- surement. The corresponding average friction coefficient Cf (5u*/u)is0.11.fNote that the drag coefficient used here is as appears in recent micrometeorological conven- tion [Cd = (u*/u)2], whereas elsewhere the drag coefficient is defined with a multiplicative coefficient of 0.5.} Analyses like that illustrated by Fig. 3 have no al- lowance for any variation with wind direction. More- over, they ignore the known criterion by which stability is measured (Z/L, where Z is the height above the rel- evant zero plane displacement d). In the lack of an ob- jective mechanism to define d, data have been selected to satisfy the criteria above but with the near-neutral constraint being that jLj . 1000 m. Figure 4 shows the resulting dependence of u*/u on wind direction, overlaid on a depiction of the location-in this case the NAS site. The effects of the large structure to the north (the U.S. Department of State building) are immediately obvious. With winds from the north, few sets of observations survive the data-selection criteria. Most sites suffer from similar directional favoritism.

Figure 5 is a similar illustration, for the installation at Howard University (HU). In this case there is also a considerable variability with wind direction. Close in- spection reveals some variation in u*/u that could cor- respond to street channeling, with somewhat lower values when the mean wind is aligned with the local streets. There is no apparent effect attributable to the presence of a lake to the northeast, about 0.5 km distant. To the northeast is also the main administration build- ing, a large structure located on a knoll. The region due east of the measurement location is a student mall, largely grassed. This corresponds to the minimum in u*/u seen in the diagram.

The top panel of Fig. 6 shows the results obtained if all near-neutral data collected within the District of Columbia are averaged, according to wind direction. The bottom panel presents averages for a subset of the D.C. sites-six stations within the central business district (CBD) of Washington (AGU,DOC,DOE,NEA,NAS,andWMC). The overall mean u*/u for the larger dataset (Fig. 6, top) is 0.143, and for the CBD it is 0.151. Averages and 61 standard-error bounds are indicated along the right-hand ordinates of the two panels. Directional variation remains similar, regardless of the CBD siting constraint. There ap- pears to be little basis for relating variations of the average results to street orientations: the street grid of Washington is largely aligned with the north-south/east-west Cartesian system, although with an overlay of angled boulevards. Both panels of Fig. 6 indicate, for example, that winds from the south yield a minimum value of u*/u, whereas winds from the opposite direction yield a maximum.

For the CBD dataset of Fig. 6, the average height of deployment of sensors is about 35 m above ground level. No assumption is made here about the zero plane dis- placement. This matter is complicated since some of the buildings are sufficiently large that they probably con- stitute the "relevant fetch" in their own right. For ex- ample, it is not clear whether the DOC observations relate to a height of measurement best taken to be 10 m above the local rooftop or 40 m above the level of the surrounding streets.

In Table 2, values are given of the mean value of u*/u (computed geometrically, i.e., by averaging the loga- rithms of the raw data) for each location of the Wash- ington DCNet, using only those observations that satisfy the criteria presented above (including the constraint jLj . 1000 m) and presented as averages in 308 wind direction sectors centered on 308,608, and so on. (Similar data were already summarized in Table 1, where they constituted support for the use of coordinate rotation. The data of Table 1 were not constrained to near neu- tral.) Two of the sites identified in Table 1 are excluded from Table 2: ARB is a subcanopy forest site (the U.S. National Arboretum in Washington), and APH is a ref- erence site well outside the D.C. area (at Fort A. P. Hill in Virginia). Some of the data given in Table 2 could be interpreted as indicating a possible influence of local street orientations (e.g., AGU, HU, and WTOP, all of which show a slight minimum in u*/u for winds from the east, along the adjacent street). This dependence is not seen at all sites.

Burian et al. (2008) present tabulations of building characteristics for a number of cities, including the central business districts of Washington, D.C., and New York City. Their data provide the opportunity to ex- amine the influence of the surroundings on eddy fluxes as are reported here. To this end, Table 3 gives a sum- mary of the surroundings of the six sites within the CBD of Washington. The properties listed are derived from the National Building Statistics Database, version 2 (Burian et al. 2008). Also listed are the overall average values of u*/u, quantified as the grand averages of the values given in Table 2. The overall average value of the normalized friction velocity is 0.159 (cf. 0.151 quoted above as the corresponding geometric mean) for the central Washington area. The value obtained for the EML site in New York City should not be considered as representative of the New York metropolis, since the site in question is a substantial distance from the area of the major buildings. The TSQ location, which is within the area of the tallest structures, yields data that are too scattered to be interpreted with confidence.

For the Washington CBD sites identified in Table 3, the most significant association is between u*/u and the reported standard deviation of the local building height, with correlation coefficient squared R2 = 0.77 (statisti- cally significant at the 95% level). Further examination, using data on the upwind building plan-area fraction [lP, also derived from the listings of Burian et al. (2008)] has proved to be inconclusive. For some sites, statistically significant relationships (at the 90%-95% confidence level) between u*/u and lP can be found, but not for all (or even for a majority) of the locations. The outcome of such examinations of the data depends on how the up- wind area of influence is determined-for example, the width of the sector over which average building mor- phologies are estimated and the radius of the sector.

The matter of upwind "footprints" has been the subject of substantial research, and many detailed models have been developed accordingly (see Schmid 2002). These models typically relate to good experimental sites and are often based on the findings of wind-tunnel studies. The case presented here differs, and hence the models de- veloped elsewhere are seen as indicative rather than ex- planatory. In this case, the local exposure of the sensors could be a dominant factor obscuring the role of the up- wind landscape, and additional uncertainty could result from the role of the pervasive atmospheric instability. (In this regard, discussion of the sensible heat flux regimes is in a later part of this paper.) It is concluded that attri- bution of surface-roughness properties to upwind con- figurations of buildings is a task that will require extensive study, as is well evident in the quantity of related literature already available. The conclusion found here regarding the relative significance of the standard deviation of local building heights (among other properties) is in agreement with the results of large-eddy simulations reported by Kanda et al. (2013), however. For a summary of other related findings, the reader is referred to Britter and Hanna (2003); there are many later examples, but there are insufficient DCNet data to explore this aspect in detail.

Inspection of the data reveals some patterns that ap- pear to be indicative of seasonal changes in the sur- face [of a kind similar to those reported by Gallo et al. (1993)]. Figure 7 presents evidence of some seasonality for two locations: ARB, where the instrumentation is mounted 10 m above ground level in a clearing largely surrounded by trees, and NAS, also largely surrounded by trees except to the north, for which direction there are few surviving data because of the presence of the large State Department structure (see Fig. 4). The lines shown are drawn by eye as an interpretation of the ob- servations that correlates well with the leafing of the surrounding deciduous trees in the early spring and the fall of leaves in the autumn. The differences apparent in the two diagrams are small but are statistically significant.

For NAS, the two lines as drawn correspond to u*/u = 0.185460.0029(samplesizeN5137)and0.161560*.0017 (N = 107) for the leafed and leafless periods, respectively. The corresponding results for ARB are 0.1695 6 0.0014 (N = 923) and 0.1455 6 0.0015 (N = 1053). No other site yields results as clear cut as the cases illustrated in Fig. 7, but no other site is as affected by trees. The association of the change in surface roughness with the leafing of trees is appealing but cannot be proved with these data. The two sites in New York City yield (as expected) measurements that are considerably more scattered than for Washington; the building height constraint im- posed in Washington is absent in New York City. As has already been emphasized, the TSQ site is too high to warrant application of conventional micrometeorologi- cal methods. On the other hand, the EML dataset per- tains to a location that is surrounded by structures of similar height, much as for the sites of the Washington CBD. The EML data yield a best estimate of the spatial average u*/u = 0.113.

4. Heat fluxes The heat island of Washington is a well-accepted feature of the area, having been the focus of research for many decades (e.g., Woollum 1964; Kim 1992) and having recently been examined using data from rooftop sensors (Hicks et al. 2010). The heat island is the result of many factors, all associated with the activities of the human population. In the case of New York City, the heat-island issue has also been the subject of consider- able research (e.g., Bornstein 1968; Kirkpatrick and Shulman 1987; Gaffin et al. 2008), culminating in studies of possible mitigation strategies (Rosenzweig et al. 2006). From the perspective of dispersion modeling, Hanna et al. (2011) summarize the variation of sensible heat fluxes across various urban areas.

In the analysis of roughness presented above, only 1 year of data was used so as to minimize variability in- troduced by year-to-year changes in buildings and their surroundings. In the following, a longer period of available data has been used, starting in 2004 and ending in 2010. Figure 8 shows monthly average diurnal cycles of local w0T 0 for the SSG location in the northern suburbs of Washington, D.C. Data are separated into the four cal- endar seasons (winter is January, February and March and so on), as described in the caption. Figure 9 parallels Fig. 8, but for the TSQ location in New York City. The New York City data differ from their Washington coun- terparts. For the elevated TSQ station (125 m), the av- erage midwinter nighttime value of w0T 0 is in the range from 0.10 to 0.15 K m s21, substantially greater than the maximum average value for the Washington area. More- over, for TSQ there are few months for which the average nighttime w0T 0 is negative. The EML location (in lower Manhattan, near the former site of the World Trade Center) displays a winter maximum similar to that for DOC, but once again there is no month with a negative average value of w0T 0. This is similar to what has been reported elsewhere. For example, Grimmond and Oke (2002) derive an anthropogenically imposed sensible heat fluxofabout100Wm22('0.08Kms21)forthedowntown area of Nagoya, Japan. Moreover, the result is sup- ported by the results of recent modeling studies (e.g., Giovannini et al. 2013; Pigeon et al. 2008).

Two notable features of Fig. 9 are 1) that the entire average diurnal cycles are elevated relative to their SSG counterparts and 2) that the early morning peak in w0T 0 seen in Fig. 8 is also seen in Fig. 9 for the same time period as for SSG. Figures 8 and 9 display features that are common among many locations: * For the Washington sites, nighttime sensible heat fluxes are typically close to zero; the negative average heat fluxes expected on the basis of classical micrometeorology are observed for only some months and some sites. For many locations, however, the average heat fluxes remain strongly positive throughout the entire diurnal cycle.

* On average, the months of November-March show short-term increases in w0T 0 in the hours immediately before dawn. A possible cause is the time-dependent ramping up of heating systems in the colder months, in advance of the start of the working day.

Detailed consideration of the daytime w0T 0 data would require information on such properties as net radiation, measurements of which are lacking. Consid- eration of the nighttime data is informative, however. Figure 10 presents a multiyear sequence of monthly averages of nighttime (from 2200 to 0500 LT) w0T 0 measurements for the U.S. Department of Commerce (DOC) site in the Federal Triangle area of central Washington. Also shown are corresponding average air temperatures. The dominant feature of Fig. 10 is the cyclic nature of the nocturnal heat flux, anticorrelated with air temperature. Some sites yield stronger signals of this kind, whereas other sites indicate negligible cor- relation. For the Washington case, the sites with the strongest negative correlation are the ones located in the older part of the city, where large limestone buildings dominate. These buildings were constructed more than 50 years ago, well before the modern emphasis on the need for energy conservation. In the temperature ex- tremes of winter and summer, there is a greater need for air conditioning of such buildings, the consequences of which appear to be evident in many of the datasets generated by DCNet. The dependence on building construction (among other local characteristics) be- comes most evident in a comparison between DOC and DOE. Whereas the DOC data show the consistently highest values of nocturnal w0T 0, the nearby DOE moni- toring station (the more modern Forrestal Building, about 1 km away) yields a comparatively constrained set of w0 T 0 values, with smaller excursions that could be attributed to summer cooling and winter heating.

Figure 11 plots the monthly-averaged nighttime values of local w0 T 0 against the measured air temperature for a different selection of stations, again intended to illus- trate the spread of behaviors. The regression lines shown are based on data points that are emphasized-the filled dots, generally corresponding to temperatures of less than about 188C. At higher temperatures, there is some evidence of a reversed relationship for some locations, but this is not a common feature.

Table 4 gives details of the regressions derived from plots such as are shown in Fig. 11 but for all of the lo- cations considered here. It is obvious that the variation among the datasets is extreme, with some showing results that are compatible with the more-heating-in-colder- weather syndrome and others showing negligible effects of this kind. Earlier, it was proposed that the present considerations might best be considered as a sampling of the Washington area. On this basis, it appears that the nighttime characteristics of the surface boundary layer might best be described using a relationship of the form assumed in consideration of Fig. 11: ... (2) with representative values of a (K m s21) and b (m s21) derived from the listing presented in Table 4: a [asymptotically =] 0.021 ± 0 : 006 and b [asymptotically =] -0.0013 ± 0.0003.

It is relevant to emphasize that these remarks apply to the relationships among averages and that any particular occasion might well depart from any expectations that are based on the current analysis. Regardless of this consideration, it is clear that these results suggest benefit from comparison with power-consumption data in the vicinities of the sites (see, e.g., Kato and Yamaguchi 2005), but such a study has not yet been attempted.

The variability evident in the sensible heat averages illustrated in Fig. 10 has been emphasized by Gaffin et al. (2008) for the specific case of New York City. They conclude that (for residential areas, at least) these var- iations correlate with changes in ambient temperature, much as is apparent in datasets presented here.

5. Spatial average fluxes The site-to-site variability evident in much of the preceding discussion emphasizes that no single DCNet observing system should be expected to yield data that are representative of an area like that of a mesoscale model grid cell. The matter is usually addressed by el- evating the instrumentation until the upwind footprint is sufficiently large. In central city locations, tall towers are usually difficult to arrange. Here, the logistical con- straints have been addressed by using many smaller towers and by viewing the problem as one of sampling.

Other models with smaller grid cells might find benefit in local data like those analyzed here. In such cases, the need to interpolate between observation points could prove to be a limitation unless the model in question makes full use of surface parameterization schemes [e.g., those proposed by Kanda et al. (2013) and Voogt and Grimmond (2000) for the mechanical and thermal attributes, respectively].

Figure 12 shows four examples of the diurnal cycle of w0T 0, spatially averaged from all rooftop sites opera- tional throughout calendar year 2007 and within the District of Columbia (i.e., AGU, DOC, DOE, EMA, HU,NAS,NEA,NRL, RFK, WMC,WTOP,andSSG). Average cycles are shown for the months of January, February, July, and August. Inspection shows that even in the average there is evidence of the wintertime predawn heating evident in the SSG data of Fig. 8 and the TSQ data of Fig. 9 (for the months of November-March).

Figure 13 combines spatial average surface roughness information with the sensible heat information dis- cussed immediately above. To construct Fig. 13, the original 15-min data have been averaged into hourly spatial averages. Thus, there are 31 values plotted for each hour (one for each day of the month). Values of the hourly average u0 w0 covariances and wind speeds have been used to determine the hourly values of the nor- malized local friction velocity u*/u. In this procedure, there is no allowance of (or correction for) the effects of stability. One particularly interesting observation from the resulting evaluations is that there is not a great variation of the scatter with time through the day.

So far, only two turbulence properties have been considered: w0T 0 and u*/u. Figure 14 shows how a number of familiar micrometeorological parameters vary with location across the array of stations now considered. The least variable property is the horizontal turbulence shape ratio s(y)/s(u). An assumption that this was about 0.82 would seem a good approximation regardless of the location. The corresponding ratio s(w)/s(u) is similarly tightly constrained, at about 0.55. Neither of these varies greatly from conventional mi- crometeorological expectations. The ratio s(w)/u* is also compatible with expectations-typically about 1.5. In comparison, the standard deviation of the wind di- rection s(u) appears to vary greatly from location to location, as also do the normalized local friction velocity u*/u and the u-w correlation coefficient. The overall conclusion to be drawn is that the wind speed and wind direction are the variables most affected by siting, with the structure of eddies being less influenced. In con- structing Fig. 14, analysis has been confined to near- neutral situations (as previously defined), and hence there is minimal confounding influence of stability.

6. Conclusions As anticipated, the surface roughness characteristics of Washington, D.C., depend on wind direction for all locations, but such directional variations tend to average out as a spatial average is constructed. For the down- town area of Washington, average values of u*/u range from 0.13 to 0.19 (see Table 3), with a best overall esti- mate of the (spatially averaged) normalized friction velocity of about 0.158. No corresponding result can be derived from the New York dataset, because the only site with low-enough scatter to warrant examination is located in a nonrepresentative area. The data available for Washington and New York City do not permit de- termination of optimal values of average displacement heights or roughness lengths.

Several of the Washington datasets show signs of "rectification" according to the orientation of adjacent streets. Lower values of u*/u are associated with wind directions along the streets rather than across them.

There is some evidence for a seasonality in the values of u*/u, corresponding to the leafing of deciduous trees in early spring. This small effect is most obvious in the data from the National Academy of Sciences location, on a tree-lined boulevard adjacent to the National Mall and near the Vietnam Memorial.

Discussion of the heat-island effect usually draws at- tention to the many causative factors, such as the urban changes in albedo and vegetation, but the data in this paper support the conclusion of Makar et al. (2006) that an important factor is the direct generation of heat by building climate controls and other human activity: the older the buildings are, the more striking is this effect. The present expectation that the nocturnal heat-island effect is indeed greater for the more massive structures of New York than for the height-constrained buildings of Washington appears to be in agreement with the re- sults of Hicks et al. (2010), on the basis of Washington and New York rooftop temperature data.

Across Washington, D.C., spatial average nighttime sensible heat fluxes are typically close to zero. For some months, the averages at night remain positive. The months of November-March show short-term increases in the w0T 0 average in the hours immediately before dawn. This increase is possibly due to the time-dependent ramping up of heating systems in winter, in advance of the start of the working day. For some sites (e.g., those closer to the downtown areas) the average heat fluxes remain strongly positive throughout the entire diurnal cycle; the vari- ability with time and space is such that an assumption of always-prevailing instability might be in significant error, however. For some of the sites, especially within the central business areas of Washington, there is a clear re- lationship between the nighttime sensible heat flux and the outside air temperature (with greater values of w0 T 0 corresponding to colder weather), presumably a conse- quence of air conditioning required to elevate the in- ternal temperature of buildings within the eddy flux footprint. This is most striking for the site above the Department of Commerce building in Washington-an old structure that occupies a complete city block.

Acknowledgments. This work was carried out as a contribution to the NOAA Dispersion Program, with support from the Air Resources Laboratory.

REFERENCES Allwine, K. J., and J. E. Flaherty, 2007: Urban dispersion program overview and MID05 field study summary. Pacific Northwest National Laboratory Rep. PNNL-16696, 63 pp. [Available on- line at http://www.pnl.gov/main/publications/external/technical_ reports/PNNL-16696.pdf.] _____, J. H. Shinn, G. E. Strait, K. L. Clawson, and M. Brown, 2002: Overview of URBAN 2000. Bull. Amer. Meteor. Soc., 83, 521-536, doi:10.1175/1520-0477(2002)083,0521:OOUAMF.2.3.CO;2.

Bornstein, R. D., 1968: Observations of the urban heat island effect in New York City. J. Appl. Meteor., 7, 575-582, doi:10.1175/ 1520-0450(1968)007,0575:OOTUHI.2.0.CO;2.

Britter, R. E., and S. R. Hanna, 2003: Flow and dispersion in urban areas. Annu. Rev. Fluid Mech., 35, 469-496, doi:10.1146/ annurev.fluid.35.101101.161147.

Burian, S., N. Augustus, I. Jeyachandran, and M. Brown, 2008: National building statistics database, version 2. Los Alamos National Laboratory Rep. LA-UR-08-1921, 77 pp. [Available online at http://glasslab.engr.ccny.cuny.edu/u/brianvh/UHI/ NBSD2_FinalReport.pdf.] Ching, J., and Coauthors, 2009: National urban database and access portal tool. Bull. Amer. Meteor. Soc., 90, 1157-1168, doi:10.1175/ 2009BAMS2675.1.

Draxler, R. R., 1987a: One year of tracer dispersion measurements over Washington, DC. Atmos. Environ., 21, 69-77, doi:10.1016/ 0004-6981(87)90272-1.

_____, 1987b: Accuracy of various diffusion and stability schemes over Washington, DC. Atmos. Environ., 21, 491-499, doi:10.1016/0004-6981(87)90032-1.

Gaffin, S. R., and Coauthors, 2008: Variations in New York City's urban heat island strength over time and space. Theor. Appl. Climatol., 94, 1-11, doi:10.1007/s00704-007-0368-3.

Gallo, K. P., A. L. McNab, T. R. Karl, J. F. Brown, J. J. Hood, and J. D. Tarpley, 1993: The use of NOAA AVHRR data for assess- ment of the urban heat island effect. J. Appl. Meteor., 32, 899-908, doi:10.1175/1520-0450(1993)032,0899:TUONAD.2.0.CO;2.

GAO, 2008: First responders' ability to detect and model hazard- ous releases in urban areas is significantly limited. Govern- ment Accounting Office Rep. GAO-08-180, 74 pp. [Available online at http://www.gao.gov/new.items/d08180.pdf.] Giovannini, L., D. Zardi, and M. de Franceschi, 2013: Character- ization of the thermal structure inside an urban canyon: Field measurements and validation of a simple model. J. Appl. Meteor., 52, 64-81, doi:10.1175/JAMC-D-12-06.1.

Grimmond, C. S. B., and T. R. Oke, 1999: Aerodynamic properties of urban areas derived from analysis of surface form. J. Appl. Meteor., 38, 1262-1292, doi:10.1175/1520-0450(1999)038,1262: APOUAD.2.0.CO;2.

_____, and _____, 2002: Turbulent heat fluxes in urban areas: Ob- servations and a local-scale urban meteorological parameter- ization scheme (LUMPS). J. Appl. Meteor., 41, 792-810, doi:10.1175/1520-0450(2002)041,0792:THFIUA.2.0.CO;2.

Hanna, S. R., R. Britter, and P. Franzese, 2003: A baseline urban dispersion model evaluated with Salt Lake City and Los Angeles tracer data. Atmos. Environ., 37, 5069-5082, doi:10.1016/ j.atmosenv.2003.08.014.

_____, J. White, and Y. Zhou, 2007: Observed winds, turbulence, and dispersion in built-up downtown areas in Oklahoma City and Manhattan. Bound.-Layer Meteor., 125, 441-468, doi:10.1007/ s10546-007-9197-2.

_____, E. Marciotto, and R. Britter, 2011: Urban energy fluxes in built-up downtown areas and variations across the urban area, for use in dispersion models. J. Appl. Meteor. Climatol., 50, 1341-1353, doi:10.1175/2011JAMC2555.1.

Hicks, B. B., W. Callahan, and M. Hoekzema, 2010: On the heat islands of Washington, DC, and New York City, NY. Bound.- Layer Meteor., 135, 291-300, doi:10.1007/s10546-010-9468-1.

Kanda, M., A. Inagaki, T. Miyamoto, M. Gryschka, and S. Raasch, 2013: A new aerodynamic parametrization for real urban surfaces. Bound.-Layer Meteor., 148, 357-377, doi:10.1007/ s10546-013-9818-x.

Kato, S., and Y. Yamaguchi, 2005: Analysis of urban heat island effect using ASTER and ETM1 data: Separation of anthro- pogenic heat discharge and natural heat radiation from sen- sible heat flux. Remote Sens. Environ., 99, 44-54, doi:10.1016/ j.rse.2005.04.026.

Kim, H. H., 1992: Urban heat island. Int. J. Remote Sens., 13, 2319- 2336, doi:10.1080/01431169208904271.

Kirkpatrick, J. S., and M. D. Shulman, 1987: A statistical evaluation of the New York City-northern New Jersey urban heat island effect on summer season daily minimum temperatures. Natl. Wea. Dig., 12, 12-15. [Available online at http://nwas.org/digest/ papers/1987/Vol12-Issue1-Feb1987/Pg12-Kirkpatrick.pdf.] Leclerc, M. Y., A. Karipot, T. Prabha, G. Allwine, B. Lamb, and H. L. Gholz, 2003: Impact of non-local advection on flux footprints over a tall forest canopy: A tracer flux experiment. Agric. For. Meteor., 115, 19-30, doi:10.1016/S0168-1923(02)00168-5.

Makar, P. A., S. Gravel, V. Chirkov, K. B. Strawbridge, F. Froude, J. Arnold, and J. Brook, 2006: Heat flux, urban properties, and regional weather. Atmos. Environ., 40, 2750-2766, doi:10.1016/ j.atmosenv.2005.11.061.

OFCM, 2002: Proceedings of the Workshop on Effective Emer- gency Response. Office of the Federal Coordinator for Me- teorological Services and Supporting Research Rep., 47 pp. [Available online at http://www.ofcm.gov/atd2/proceedings/ pdf/eer_proceedings.pdf.] Pigeon, G., M. A. Moscicki, J. A. Voogt, and V. Masson, 2008: Simulation of fall and winter surface energy balance over a dense urban area using the TEB scheme. Meteor. Atmos. Phys., 102, 159-171, doi:10.1007/s00703-008-0320-9.

Raupach, M. R., 1989: Applying Lagrangian fluid mechanics to infer scalar source distributions from concentration profiles in plant canopies. Agric. For. Meteor., 47, 85-108, doi:10.1016/ 0168-1923(89)90089-0.

Rosenzweig, C., W. D. Solecki, and R. B. Slosberg, 2006: Mitigating New York City's heat island with urban forestry, living roofs, and light surfaces. New York City Regional Heat Island Initiative Rep. NYSERDA 6681, 133 pp. [Available online at http://www.fs.fed. us/ccrc/topics/urban-forests/docs/NYSERDA_heat_island.pdf.] Schmid, H. P., 2002: Footprint modeling for vegetation atmosphere exchange studies: A review and perspective. Agric. For. Me- teor., 113, 159-183, doi:10.1016/S0168-1923(02)00107-7.

Theurer, W., 1999: Typical building arrangements for urban air pol- lution modeling. Atmos. Environ., 33, 4057-4066, doi:10.1016/ S1352-2310(99)00147-8.

Voogt, J. A., and T. R. Oke, 1997: Complete urban surface temperatures. J. Appl. Meteor., 36, 1117-1132, doi:10.1175/ 1520-0450(1997)036,1117:CUST.2.0.CO;2.

_____, and C. S. B. Grimmond, 2000: Modeling surface sensible heat flux using surface radiative temperatures in a simple urban area. J. Appl. Meteor., 39, 1679-1699, doi:10.1175/1520-0450-39.10.1679.

Wesely, M. L., C. B. Tanner, and G. W. Thurtell, 1972: An im- proved pressure-sphere anemometer. Bound.-Layer Meteor., 2, 275-283, doi:10.1007/BF02184769.

Woollum, C. A.,1964: Notes froma study ofthe microclimatology of the Washington, D. C. area for winter and spring seasons. Weatherwise, 17, 262-271, doi:10.1080/00431672.1964.9927020.

BRUCE B. HICKS Metcorps, Norris, Tennessee WILLIAM R. PENDERGRASS III NOAA/ARL /Atmospheric Turbulence and Diffusion Division, Oak Ridge, Tennessee CHRISTOPH A. VOGEL NOAA/ARL/Atmospheric Turbulence and Diffusion Division, and Oak Ridge Associated Universities, Oak Ridge, Tennessee RICHARD S. ARTZ NOAA/Center for Weather and Climate Prediction/Air Resources Laboratory, College Park, Maryland (Manuscript received 17 April 2013, in final form 3 December 2013) Corresponding author address: Bruce Hicks, Metcorps, P.O. Box 1510, Norris, TN 37828.

E-mail: [email protected] DOI: 10.1175/JAMC-D-13-0154.1 (c) 2014 American Meteorological Society

[ Back To TMCnet.com's Homepage ]