mesoscale aspects of the urban heat island around new york ... · (renou, 1862; landsberg, 1981)....

Theor. Appl. Climatol. 75, 29–42 (2003)DOI 10.1007/s00704-002-0724-2

1 Department of Earth and Atmospheric Sciences, City College of New York, New York, NY, USA2 Computer Sciences Department, Physical, Environmental, and Computer Sciences Department,

Medgar Evers College, Brooklyn, NY, USA3 Weather Access Inc., Highland Park, NJ, USA4 Sussex County Weather Network, Wantage, NJ, USA5 Weather Access Inc., Chatham, NJ, USA6 South Jersey Resource Conservation and Development Council, Hammontown, NJ, USA7 Department of Geography, Rutgers University, New Brunswick, NJ, USA

Mesoscale aspects of the Urban Heat Islandaround New York City

S. D. Gedzelman1, S. Austin2, R. Cermak3, N. Stefano4, S. Partridge5,S. Quesenberry6, and D. A. Robinson7

With 19 Figures

Received August 16, 2001; revised October 6, 2002; accepted November 20, 2002Published online March 17, 2003 # Springer-Verlag 2003

Summary

A mesoscale analysis of the Urban Heat Island (UHI) ofNew York City (NYC) is performed using a mesoscalenetwork of weather stations. In all seasons the UHIswitches on rapidly in late afternoon and shuts down evenmore rapidly shortly after dawn. It averages about 4 �C insummer and autumn and 3 �C in winter and spring. It islargest on nights with clear skies, low relative humiditythrough much of the troposphere, and weak northwestwinds, when it may exceed 8 �C. The synoptic meteoro-logical situation associated with the largest UHI occursroughly two to three nights after cold front passages.

During spring and summer, sea breezes commonly reduceand delay the UHI and displace it about 10 km to the west.Backdoor cold fronts, which occur most frequently in springand early summer, reduce or even reverse the UHI, as coldair from the water to the northeast keeps NYC colder thanthe western suburbs. Cases documenting the sensitivity andrapidity of changes of the UHI to changes in parameters suchas cloud cover, ceiling, and wind speed and direction arepresented.

1. Introduction

Ever since Luke Howard discovered that Londonwas warmer than the surrounding countryside,

studies of the Urban Heat Island (UHI) havetended to emphasize its climatological aspects(Renou, 1862; Landsberg, 1981). The urbanenergy budget has usually been evaluated withoutincluding the troublesome effects of advection bythe wind (Oke, 1995). To predict the magnitude ofthe UHI under particular weather conditions,regression equations have been developed thatrelate the UHI to synoptic meteorological param-eters such as cloud cover or wind speed (Eliasson,1996; Runnalls and Oke, 2000) or to generalizedsynoptic weather regimes (Morris and Simmonds,2000). The major conclusions that emerge fromsuch studies are as follows: the UHI is most pro-nounced on calm, dry, clear nights near the centerof anticyclones. Under such conditions heat isretained in the city while a pronounced nocturnalinversion forms in the countryside as the groundradiates heat rapidly to space. Conversely, theUHI is reduced on windy, humid, overcast nightswith precipitation, often under cyclonic condi-tions, because the nocturnal inversion cannotform, and urban and rural heat budgets are similar.

In an attempt to produce a more complete pic-ture of the UHI, increased attention has beenfocused on the impact of horizontal advectionand on varying weather conditions. Thus, theUHI is displaced by the prevailing wind (Oke,1995). The energy balance is affected not onlyby the prevailing wind but also by the circulationset up by the temperature difference between cityand surrounding countryside (Haeger-Eugenssonand Holmer, 1999). The impact of mesoscalephenomena such as drainage winds (Hertig,1995) and the sea breeze (Yoshikado, 1990,1992) on the UHI are now examined in dataand modeling studies.

Many studies that attempt to produce moredetailed pictures of the UHI have been conductedas field experiments for limited time periods(Zois, 1986; Montávez et al., 2000). In such stud-ies data from a relatively small number of resi-dent weather stations are enhanced by finer-scalerecords obtained from instrumented vehicles per-forming transects of the region during the periodof the field experiment. General conclusionsmust be based on fortuitous cases.

The recent establishment of mesoscaleweather networks (Brock et al., 1995) has madeit possible to obtain pictures of the mesoscalemeteorological aspects of the UHI on a routinebasis. In the mid 1990’s a mesoscale network ofweather stations in the Greater MetropolitanArea around New York City (NYC) was estab-lished. In this article the data from this network isused to provide a mesoscale picture of the UHI inand around NYC.

The coastal setting of NYC imparts a strongmesoscale component to the UHI. NYC islocated on several islands and one peninsula ofthe mainland along the East Coast of the UnitedStates. Long Island extends to the east of NYCand is bounded on the south by the AtlanticOcean and on the north by Long Island Sound.New Jersey, situated to the west and mostly southof NYC, is bounded on the east by the AtlanticOcean. Along the coastal strip, appreciable land-sea temperature contrasts exist in all seasons andimpart a strong local signal to a range of meteor-ological phenomena. During winter storms, therain-snow line often lies directly over NYC andsnow cover increases almost abruptly to thenorthwest. Sea breezes that cool NYC more thanthe inland suburbs are common on warm spring

and summer days, while the ocean frequentlywarms coastal locations on cold nights inautumn, winter, and early spring. On manyspring days, ‘backdoor cold fronts’ (Richwien,1980) followed by cold northeast winds that arechanneled along the coast by the hills and moun-tains to the north and west overwhelm the UHIby chilling NYC more than the western suburbs.

2. Data sources and methodology

The primary data source for this study consists ofhourly surface weather observations for the two-year period, 1997–98 from a network of 50volunteer and about 25 National Weather Service(NWS) stations in the New York – New JerseyMetropolitan area that included several oceanbuoys (see Fig. 1). Surface meteorological dataat NWS stations includes cloud cover and height,visibility, weather, pressure, temperature, dewpoint temperature, wind speed and direction,and precipitation. Volunteer stations do notinclude visibility or cloud measurements. Thebuoys give water temperature but do not includevisibility, cloud cover or precipitation. The sur-face weather data is augmented by several other

Fig. 1. Weather stations and topography of the GreaterNew York City Metropolitan Area. Circles indicate NWSstations, squares, private stations. The four stations markedby concentric circles and concentric squares are used as theurban and rural stations respectively. The contour intervalis 100 m and the analysis omits fine scale features. Thesickle-shaped ridge about 10 km west of NYC representthe Watchung Mountains

30 S. D. Gedzelman et al.

sources including soundings, mainly from theNWS station, OKX at Brookhaven, NY (100 kmeast of NYC), satellite and NEXRAD radarimagery, and analyses and forecasts of variousparameters from NWS models.

Stations are several kilometers apart at best, sothat only the mesoscale aspects of the UHI areinvestigated here. The magnitude of the UHI isalmost certainly understated because traversesthrough the urban heartland were not included(Runnalls and Oke, 2000) and because nostations are located in the paved heart of theurban center. For example, station NYC islocated atop a small hill in Central Park,Manhattan roughly 300 m from the nearest gridof streets and buildings. Many of the volunteerstations, particularly in northern New Jersey, arelocated on the grounds of private homes and donot conform to international standards. Localtopography plays a significant role in tempera-ture variations, with differences of 3 �C or morecommonly occurring in less than 100 m of reliefon clear nights. Therefore, using internationalstandards to register variations of surface tem-peratures in and around cities has been recog-nized as a significant problem (Voogt and Oke,1997).

Given these problems regarding the reliabilityof the setting of stations and the accuracy of thedata, hourly surface weather maps of temperatureand wind were constructed and inspected toscreen data from for obvious errors and inconsis-tencies. Inconsistent or blatantly erroneous datawere deleted and stations whose data was biasedwere excluded. The final product was a consis-tent set of hourly surface weather maps for theyear 1998.

A major component of this research was toidentify illustrative or extreme examples of theUHI for case study. These were identified usingtwo techniques. First, the hourly surface analyseswere color-coded for temperature and animatedfor each day in 1998. In this way a host of phe-nomena including the UHI, sea breezes, frontalpassages, and squall line motions could be visual-ized in an unmistakable manner. Second, meteor-ograms and scatter diagrams relating the UHI tostandard parameters such as wind speed anddirection, cloud cover, and the strength of therural nocturnal inversion, were prepared to iden-tify anomalous cases. Each of these cases was

then analyzed in greater detail to identify causa-tive factors.

Climatological analyses included the two-yearperiod, 1997–98. In the climatological analyses,seasons were defined as three-month periodsusing the meteorological standard that the period,DJF (December, January and February) consti-tuted winter, MAM, spring, JJA, summer, andSON represented fall.

3. Climatological featuresof the Urban Heat Island

The UHI is a nocturnal phenomenon in all sea-sons. A surface plot of temperature at midnightduring Autumn (Sep–Nov) across the GreaterNew York Metropolitan area (Fig. 2) shows thehighest values occur in New York City whilemuch lower values occur in the northwest sub-urbs. Nocturnal temperature also drops substan-tially below urban values in the rural easternportions of Long Island.

The magnitude of the urban-rural temperaturedifference, �TU�R, was defined using the aver-age temperature of four urban stations (NYC,LGA, EWR, and TEB) minus the average tem-perature at four of the reliable and representativeinland rural stations (BLM, CHA, OKR, andSSX). The rural stations are centered 50 km westof the urban stations and further inland. Thisintroduces a bias whenever east–west tem-perature gradients existed, such as during frontal

Fig. 2. Map of mean temperature (�C) at midnight duringmeteorological autumn (Sep–Nov) across the region

Mesoscale aspects of the Urban Heat Island around New York City 31

passages and sea breezes. This bias was sub-tracted when appropriate (e.g. as in Fig. 6). After22 July 1998 station FOK (40.85 N, 72.83 Wand 95 km east of the urban stations) – beganreporting hourly weather data, which showed thatsimilar nocturnal cooling occurred east of NYCas well.

�TU�R has an average magnitude of 3�C in

winter and spring, and 4 �C in summer andautumn, when wind speeds are smaller on aver-age (Fig. 3). It increases rapidly from almost zeroin mid afternoon to an almost constant value allnight, as Oke and Maxwell (1975) and Hage(1972) have found. In the morning, �TU�Rdecreases even more sharply than it increases in

the evening, because the nocturnal inversiontends to steepen more than it deepens over thecourse of the night. This differs from the morerapid evening cooling found around St. Louisduring METROMEX (Hilberg, 1978). Even nearthe winter solstice, the rising sun can produceextremely rapid temperature increases in thecountryside because it only has to heat a veryshallow layer of air. In fact, 2-hour increases of�TU�R exceeding 6

�C occurred 22 times duringthe fall and winter months of 1998. This neverhappens during late spring and summer when thenights are shorter and clear, dry and, almost calmnights near the center of polar high pressure areasare least likely. Two-hour decreases of �TU�Rexceeding 6 �C occurred only 3 times in 1998,but each of these cases was associated with thepassage of a cold front or squall line during midafternoon, and not by nocturnal cooling.

For 1998, the maximum nightly value of�TU�R correlated positively (0.37) with themagnitude of the 1200 UTC surface temperatureinversion at OKX (�TINV), much as Ludwig(1970) found for a number of cities. However,considerable scatter exists for several reasons.First, the sun rises early enough in the monthsnear the summer solstice to reduce or destroythe nocturnal inversion by sounding time. Sec-ond, surface inversions also occur during seabreezes or frontal situations when �TU�R tendsto be small or negative. To reduce the impact ofthese cases, Fig. 4 shows �TU�R vs �TINV onlyon Year days 1–120 and 240–265 when theinversion height was less than 300 m. Under

Fig. 3. Hourly values of urban – rural temperature differ-ence, �TU�R, for each of the four meteorological seasons(Winter¼Dec–Feb)

Fig. 4. Relation between the magnitude of thesurface temperature inversion at OKX at 1200UTC and �TU�R for Year days 1–120 and240–365 on nights when the depth of the noc-turnal inversion was less than 300 m. Circledpoint near lower right is the night of 27–28Oct


these restricted conditions, the correlationbetween the nightly maximum �TU�R and themagnitude of the surface inversion increases to0.694.

Figure 4 reveals anomalous cases, such as thenight of 27–28 October (circled point), when�TU�R was only 1.8

�C despite a large�TINV¼ 9.2 �C. On this night, high pressure overNew England produced clear skies over NYCand OKX while the northwestern suburbs werecovered by a low overcast that kept temperaturesin the western suburbs relatively high. Removalof this single case raised the correlation between�TU�R and �TINV to 0.728.

Cloud cover and height, surface humidity, andwind speed, have the expected large impacts onthe magnitude of the UHI (Semonin, 1981).Under clear skies, �TU�R averages 4.5

�C.When the sky is overcast the average �TU�Rfalls to 2.3 �C, and large values never occur(Fig. 5). �TU�R also increases as the height ofthe cloud base increases. Opaque clouds radiatelike black bodies according to their absolute tem-perature. Therefore, as the height of cloud baseincreases less heat is radiated downward, assum-ing a positive lapse rate above the nocturnalboundary layer. The correlation between theheight of cloud base as determined from thesounding at OKX and the maximum �TU�Ron any night is 0.54. (Clear soundings areassigned a cloud base at 12 km.) �TU�R alsocorrelates significantly (0.49) with the depressionof the dew point temperature (T�Td) at the

surface since dry air radiates less heat down tothe ground. Average �TU�R increases from2.1 �C when T� Td14 �C.

Wind reduces the magnitude of the UHI in twoways – by mixing cold air at the surface withwarmer air aloft and by ventilating the city. Atmidnight, �TU�R averages 4.8

�C when windspeed is less than 5 kt but only 2.1 �C when windspeed is above 15 kt (Table 1a). Wind speed var-ies with the synoptic setting and is typically larg-er during cyclonic weather or in the hoursfollowing a cold front passage. As winds slowand high pressure with clear skies and drier airapproaches, �TU�R increases to a maximumvalue for any synoptic weather conditionbetween two and three nights after a cold frontpasses (Fig. 6). Because the rural stations arelocated an average of 50 km west of the urbanstations, the bias in the �TU�R due to the synop-tic scale temperature field was removed fromFig. 6 by subtracting the east–west temperaturedifferences at the 850 hPa level.

Wind direction has a profound influence on themagnitude and location of New York City’surban heat island because of its coastal setting.This makes it informative to group �TU�R bywind quadrants (Table 1b) and season. Duringspring and summer, �TU�R is significantly larg-er when the wind blows from either of the twowesterly quadrants, i.e. from the land. The largestaverage �TU�R occurs with winds from thenorthwest quadrant since skies tend to be clearestand have the lowest relative humidity. When thewind has an easterly component, it comes off thecold waters and tends to cool the city more thanstations located further inland. The smallest aver-age �TU�R at midnight occurs when the wind ofthe previous afternoon blows from the southeastquadrant, the direction of the sea breeze in thearea. Antecedent winds provide the highest cor-relation with �TU�R because the cooling caused

Fig. 5. Histogram of �TU�R for clear and overcastconditions

Table 1a. Magnitude of the UHI as a function of wind speedat JFK

Wind speed (kt) �TU�R (�C)

10 2.8>15 2.3>20 2.1


by the sea breeze tends to persist for severalhours after the wind direction has changed.

Even though the sea breeze has a major impacton temperatures, cooling the coastal strip andNYC more than the outlying rural areas to thenorth and west, its impact on �TU�R often dis-appears shortly after midnight. The sea breezedoes, however, delay the UHI for several hours.Figure 7 shows that on nights with a strong seabreeze, the �TU�R starts somewhat later andincreases more slowly, but both ultimately reachthe same magnitude around midnight. The cri-terion used to define a day with a strong seabreeze is that at 1600 local time, temperature atNYC must be at least 2.2 �C warmer than at JFK,and the southerly component of the wind at JFKmust exceed that at NYC by 5 kt.

During the early hours of the night, when thelingering effects of the sea breeze are still felt,the center of the heat island is often displaced tothe New Jersey Meadowlands, about 10 km westof Manhattan. Cooling by the sea breezediminishes with distance inland from the coastat a rate that varies with wind direction and speedand the depth of the chilled layer. Because of the

prevailing southerly winds during late spring andsummer, the sea breeze often penetrates furtherinland and is associated with a smaller tempera-ture gradient on the south-facing shores of LongIsland, NYC, and Connecticut than on the eastfacing shore of New Jersey.

Nocturnal cooling, sea breezes and backdoorcold fronts all produce surface temperature inver-sions. Nocturnal radiation inversions tend to bequite shallow (

is shallow enough to allow rapid warming andproduce a large temperature gradient inland fromthe coast in most cases during spring and summerdays. More elevated inversions are associatedwith backdoor cold fronts, enabling the cold airto penetrate much further inland.

Backdoor cold fronts are anomalous situationsin which cold air invades the NYC MetropolitanArea from the northeast. They are most commonin late winter and spring, and tend to produce thesmallest UHI in the Greater New York Metropol-itan Area. During intense backdoor cold fronts,�TU�R may even be negative. At such times,chilling, damp NE winds, from polar high pres-sure areas centered over southeastern Canada,blow down the length of Long Island Sound andinto NYC. The Appalachian Mountains to thewest simultaneously block the advance of warmerair from the southwest and dam the surface layerof cold air against the eastern slopes. This boththickens the cold air mass and deflects it downthe Eastern Seaboard as far south as the Carolinas(Richwien, 1980). As a result, the cold air layerassociated with pronounced backdoor cold frontsis often more than 1500 m deep. Backdoor situa-tions occur much less often than the sea breeze sothat some years have few good cases. In 1998,there were only a few relatively weak backdoorcold fronts. These seldom passed far south ofNYC, and had a mean inversion height on 804 mfor the 0000 UTC sounding at OKX.

The inland penetration of cold air followingbackdoor cold fronts is often increased by a blan-ket of low clouds and fog that is thickenedfurther by frictional convergence along the coast.At such times temperatures on the foggy coastcan be 25 �C colder than inland temperatures inthe sunny western and northern suburbs.

4. Meteorological features of the UHI

To illustrate how meteorological factors affectthe UHI we analyze exemplary cases of theUHI for a number of different weather situationsin 1998 revealed by the mesoscale network ofstations shown in Fig. 1.

4.1 Classical UHI

A classical case of a large UHI occurred on thenight of 9–10 December 1998, two days after

record warmth in the Northeast United States.The meteorogram (Fig. 8) shows that �TU�Rbegan to increase rapidly about two hours beforesunset and approached its maximum value of8 �C by 2000 EST. Thereafter, �TU�R remainedalmost constant until dawn, after which it rapidlydecreased and disappeared by 1000 EST.

The synoptic situation provided optimal con-ditions for a large UHI. On the previous night, alow pressure area moved up the coast and skieswere mostly overcast. Because no precipitationoccurred the ground remained dry. After thelow passed by to the northeast, sinking air pre-vailed and high pressure approached from thewest. Skies cleared by dawn on the 9th and windsbacked to northwest. Wind speed diminished to5 kt or less by late afternoon, everywhere exceptat LGA, where winds remained brisk until almostmidnight. The sounding at OKX for 1200 UTCof 10 December contains a sharp nocturnal inver-sion only 70 m deep, and dry, clear air to the topof the troposphere (Fig. 9). However, the appear-ance of a broken layer of high clouds shortlyafter dawn on the 10th slowed the morningwarming and the disappearance of the UHI.

The �TU�R for this case had one of the larg-est magnitudes of 1998. Most cases of largeUHI occur under the same general synoptic con-ditions, namely clear skies with gentle northwestwinds and dry air from ground up to the tropo-pause. Such conditions typically occur about twoto three nights after the passage of a cold front,when high pressure is centered a relatively shortdistance to the west of NYC. As is typical ofsuch situations, a map of the 3-h temperaturechange (from 1500 to 1800 EST on 09 Decem-ber) shows that the suburbs cool much morequickly than the urban centers in the early even-ing hours (Fig. 10). Once the rural nocturnal

Fig. 8. Meteorogram of �TU�R and wind at JFK for theclassical UHI of 9–10 December 1998. Dashed lines indi-cate times of sunrise and sunset. Length of the wind shaftsis proportional to wind speed


inversion is established, city and countrysidecool at almost the same rate. Cooling rates arealso highly influenced by topography and henceare more localized than the contour analysis sug-gests. For example, early evening cooling ratesare almost always anomalously small at HighPoint in extreme northwest New Jersey (elevation465 m) since this sloped, elevated site invariablyreduces the nocturnal inversion.

The map of surface temperatures at 2200 ESTon the 9th shows the large UHI centered overNYC (Fig. 11). A smaller and less pronouncedUHI is also apparent over Philadelphia. On thisnight, UHI was compounded by the proximity ofa very warm ocean. The map shows a striking

temperature contrast between land and water.Air at any station with a trajectory that passedover water was distinctly warmer than inlandlocations. LBI (Long Beach Island, NJ), whichwas roughly 6 �C warmer than the mainland tothe west, is a case in point. LBI is located on abarrier island and separated from the mainlandby Barnegat Bay, which at that point is 7 kmwide. The elevated temperature at LBI resultslargely from the turbulent transfer of sensibleheat from the warmer water surface to thethin layer of nocturnally cooled air passed overBarnegat Bay. Using the Austauch equation, the

Fig. 9. Vertical sounding of temperature anddew point temperature at OKX for 1200 UTCof 10 Dec 1998 with a sharp nocturnal inver-sion 70 m above the ground, and dry, clear airabove

Fig. 10. Surface map of the 3-hour temperature change(from 1500 to 1800 EST on 09 Dec 1998) with much morerapid early evening cooling in the rural suburbs than in theurban centers

Fig. 11. Surface weather map of temperature and wind at2200 EST on 9 Dec 1998 with a pronounced UHI centeredover NYC. Wind direction and speed are shown for selectedstations. Wind speed is proportional to the length of theshaft


sensible heat flux, F, for air as it first reaches theBay is roughly,

F ¼ C�cðTSST � TAIRÞv � 32 W m�2

where the sea surface and air temperaturesare respectively TSST¼ 10 �C, TAIR¼ 2 �C,C� (10)� 3 is the Austauch coefficient, �¼1.28 kg m� 3 is the density of the air, c¼1004 J kg� 1 K� 1 is the specific heat capacityof air at constant pressure and, v¼ 3 m s� 1 isthe wind speed. Since the nocturnal inversion atOKX was only 70 m (9 hPa) thick, the calculatedtemperature rise in crossing Barnegat Bay is8 �C, assuming the heating decreased linearly(a) with height to the top of the nocturnal inver-sion and (b) with distance from the mainland toLBI. This is more than enough to produce theobserved temperature at LBI.

4.2 Quick response UHI

Hour-by-hour comparison of rural and urban sta-tions reveals that during the night, �TU�R canrespond rapidly to changes of larger scalemeteorological parameters such as wind speedand cloud cover. Increased surface wind speedcauses mixing that warms the countryside byreducing the shallow nocturnal inversion at thesame time it cools the city through ventilationand advection. Onset of a cloud cover dramati-cally slows or even reverses cooling in the coun-tryside once a marked nocturnal inversion hasdeveloped while it has much less impact onurban temperatures.

A good case of a quick response UHI occurredon the night of 4–5 September 1998. �TU�Rincreased to 5.7 �C by 2300 EST under mostlyclear skies and gentle winds. Shortly before mid-night a broken deck of altocumulus (base2700 m) passed over the region. At the sametime, wind speed at both EWR and LGAincreased from near calm to about 10 kt. Inresponse to these changes, �TU�R decreasedby 1.8 �C as mean urban temperature fell andrural temperature simultaneously rose between2300 and 2400 EDT (Fig. 12).

Once the sun rises following clear, calmnights, �TU�R disappears rapidly because thenocturnal inversion in the countryside tends tobe quite shallow and steep. The largest 2-hourdecrease of �TU�R in 1998 (9.8

�C) took place

between 0600 and 0800 EST on the morning of25 October 1998. While the rural stationswarmed by an average of 10.8 �C, the urban sta-tions only warmed by 1.0 �C (Fig. 13). The OKXsounding at this time (Fig. 14) had a steep,shallow inversion (�TINV¼ 9.8 �C, �z¼ 82 m)topped by clear, dry air with the dew pointdepression exceeding 20 �C through most of thetroposphere. Simple heat balance calculationsshow that such an inversion will be erased in 2hours with a mean solar irradiance of only70 W m� 2.

Fig. 12. Surface map of temperature change showing urbancooling and rural warming from 2300 to 2400 EDT on 04Sept 1998. At the time, wind speed increased and a band ofaltocumulus crossed the region

Fig. 13. Surface map two hour temperature change from0600 to 0800 EST 25 Oct 1998 with much more rapidmorning warming in the rural areas than in New York City


4.3 UHI during a strong sea breeze

Prevailing southerly winds during the warm sea-son cause the sea breeze to penetrate furtherinland over the south facing shores of Long Islandand Connecticut than over the east facing shore ofNew Jersey. Because NYC extends into LongIsland, the sea breeze cools it more than the urban-ized counties of New Jersey immediately to thewest. During the peak summer months, the cool-ing effect of the sea breeze diminishes graduallyover the course of the night. This slows theincrease of �TU�R and displaces the center ofthe UHI about 10 km to the west for the first fewhours of the night so that it is often centered nearthe New Jersey Meadowlands. On most summernights the center of the UHI gradually shifts backtoward New York City shortly after midnight.

During spring, the sea breeze does not occur asfrequently as during summer because it is oftensuppressed by strong westerly or northerlysynoptic scale winds and polar air outbreaks.Nevertheless, the most dramatic sea breezesoccur on warm spring days because the largestland-sea temperature contrasts occur at this timeof year. From 27 March to 03 April 1998, unsea-sonably warm air invaded the Eastern Seaboardunder mostly south-southwest winds. Tempera-tures a short distance inland reached values ashigh as 30 �C while sea surface temperaturesouth of Long Island was only 5.5 �C. This estab-lished favorable conditions for extremely strongsea breezes. A comparison of soundings fromOKX and ALB (Albany, NY) at 0000 UTC on29 March (Fig. 15) shows that the sea breeze was

Fig. 14. Sounding of temperature and dewpoint temperature at OKX for 1200 UTC 25Oct 1998 with an intense but shallow noctur-nal inversion

Fig. 15. Comparison of soundings at OKXand ALB for 0000 29 Mar 1998 with warmair aloft at both locations but a sea breezeinversion only at OKX


associated with a pronounced inversion. The seabreeze inversion averaged about 400 m deep dur-ing the period from 27 March to 03 April anddisappeared only briefly when strong west windssuppressed the sea breeze at 0000 on 30 March.

On the afternoon of the 29th and again on the31st, the westerly wind component was strongenough to suppress the sea breeze and NYCrecorded its all time high temperature recordfor March. At the same time, however, the seabreeze reached JFK and cooled it significantly.On the days that the westerly component of thewind weakened, strong sea breezes cooled NYCwhile record high temperatures were recorded inthe western suburbs.

On the night of 28–29 March 1998, the seabreeze not only delayed the UHI and shifted itto the west, it also kept NYC colder than thewestern suburbs until 2 h before dawn (Fig. 16).

Fig. 16. Meteorogram for 28–29 March 1998. A strong seabreeze delayed UHI and reduced �TU�R

Fig. 17. Surface map for (a) 1500 EST 28 Mar 1998 with asharp sea breeze front and (b) 2100 EST 28 Mar 1998 withUHI displaced westward

Fig. 18. Surface map showing temperatures at (a) 1000EST 07 Jan 1998 with an advancing backdoor cold airand (b) 0300 EST 08 Jan 1998 with the backdoor frontretreating as a warm front


At that time the south wind at JFK weakened andthe winds inland turned north for a few hours.South-southwest winds limited the westwardpenetration of the sea breeze over Coastal NewJersey and led to the formation of a distinct seabreeze front a few miles inland that extendednorthward across the New Jersey Meadowlands(Fig. 17a). The temperature gradient associatedwith the sea breeze front sharpened just west ofNYC as the wind veered to almost due west atseveral stations just south of the WatchungMountains. This deflection of the winds in thisarea was observed on several other days whensynpotic scale winds were south-southwest andthe atmospheric boundary layer was staticallystable.

Cold air penetrated much further inland overLong Island and coastal Connecticut where thesoutherly sea breeze was superimposed on theprevailing south-southwest winds. At 2100 ESTthe highest temperatures were still centered westof NYC but Manhattan warmed somewhat whenthe wind veered south-southwest (Fig. 17b).

4.4 UHI following passage of a backdoorcold front

NYC tends to be much colder than the westernsuburbs under northeast winds and overcast skieswhen the ocean is relatively cold. Such condi-tions occur most often in spring and early sum-mer. However, a warm spell in early January1998 that produced a record-breaking ice stormin Northern New England, Quebec and Ontario

also provided the setting for an unusually earlyseason backdoor cold front in New York City. Onthe morning of 7 January, cold air from easternNew England crossed the New York City Metro-politan Area (Fig. 18a). This cold air massreached central New Jersey and eastern Pennsyl-vania. It remained entrenched over the Metropo-litan Area until the early morning hours of 08January, when warm southerly winds reinvadedthe region (Fig. 18b).

Because the front was located a short distancesouth and west of NYC, the cold air massremained relatively shallow over the Metropoli-tan Area with an inversion height of 600 m atOKX (Fig. 19). However, because the sun is sofeeble in early January, and because skiesremained mostly overcast, the air remained coldnorth of the front. As a result, NYC remaineddistinctly colder than the suburbs to the south-west, and so, �TU�R remained negative throughmuch of the time.

5. Summary and conclusions

A mesoscale analysis of the climatological andmeteorological aspects of the Urban Heat Island(UHI) around New York City (NYC) was pre-sented using a mesoscale network of weather sta-tions as the primary data source. On average,UHI increases rapidly in the late afternoon,remains almost constant during the night andthen decreases quickly after dawn. The urban –rural temperature difference, �TU�R is greateston clear nights with low humidity throughout the

Fig. 19. Sounding for 0000 UTC 08 Jan1998 with an inversion and low overcast as-sociated with a backdoor cold front. Windswere northeast below the inversion andsouthwest above it


troposphere and gentle northwest winds. Thistends to occur two to three nights after a coldfront passage. �TU�R is strongly reduced bysea breezes and by backdoor cold fronts, whichdrive cold air into the City during spring andsummer from the southeast and northeast respec-tively.

Case studies showing the impact of particularmeteorological situations on the UHI were pre-sented. A classic large magnitude UHI occurredon the night of 9–10 December near the center ofa high pressure area under clear skies with gentlenorthwest winds. In the early evening hours tem-peratures in the countryside dropped rapidly untilthey fell almost 10 �C colder than NYC. There-after, city and country cooled at the same rateuntil dawn, when the rural areas warmed rapidly.Rapid warming was due to the shallowness of thenocturnal surface inversion (82 m deep).

The second case involved a brief period ofcloud cover with increased wind speed in anotherwise clear, calm night (25 October 1998).During this time, �TU�R decreased in response.Rural temperatures rose as faster winds increasedvertical mixing in and above the nocturnal inver-sion and downward radiation increased becauseof the cloud layer. At the same time, temperaturedropped in NYC as faster winds increased venti-lation and advection from the colder suburbs.

In the third case several days of unusuallystrong southerly sea breezes at the end of March1998 during a record breaking heat wave causedadvection of cold air over Long Island and theeastern portions of NYC. A sharp sea breezefront formed a short distance to the west of theNew Jersey Coastline and extended northwardalong a line about 10 km west of NYC. The tem-perature gradient was enhanced just west ofNYC, where winds veered to skirt the WatchungMountains. The deflection of the winds by themodest topography will be examined further ina future study. The net result of the sea breezewas to displace the core of the UHI to the westduring the early part of the night and suppress itin NYC until after midnight.

The final case involved an unusual, midwinterbackdoor cold front on 7–8 January 1998. Amidwinter heat wave was interrupted when coldair invaded the region from the northeast. Thebackdoor cold front only passed a short distancesouth and west of NYC, where it stayed for less

than 24 hours. This rendered Long Island andNYC much colder than the western suburbs andcompletely squelched the UHI.

Our future studies have two goals. The meso-scale weather network has been augmented toinclude more stations inside New York City. Thiswill make it possible to obtain a finer scale viewof the UHI. We will also employ the mesoscaleweather model MM5 to assess the role of thevarious geographical and meteorological featuresrevealed by the mesonet and that contribute tothe UHI.

Acknowledgments

We thank the many dedicated weather aficionados who,often at their own expense and by their own sustainedeffort, have installed and maintained weather stations andarchived the data. The reviewers’ incisive comments helpedthe focus of the paper. This research was sponsored byNASA Grant #NCC 5–98 and NOAA CREST Grant.

References

Bornstein RD (1968) Observations of the urban heat islandeffect. J Appl Meteor 7: 575–582

Brock FV, Crawford KC, Elliott RL, Cuperus GW, StadlerSJ, Johnson HL, Eilts MD (1995) The Oklahoma Meso-net: A technical overview. J Atmos Oceanic Technol 12:5–19

Haeger-Eugensson M, Holmer B (1999) Advection causedby the urban heat island circulation as a regulatingfactor on the nocturnal urban heat island. Int J Climatol19: 975–988

Hage KD (1972) Nocturnal temperatures in EdmontonAlberta. J Appl Meteor 11: 123–129

Hertig JA (1995) Analysis of meteorological data and mainproblems related to the determination of building expo-sure. In: Cermak JE, Davenport AG, Plate EJ, Viegas DX(eds) Wind climate in cities. NATO ASI Series E: AppliedSciences – Vol. 277. Boston: Kluwer Academic Publish-ers, pp 153–182

Hilberg SD (1978) Diurnal temperature and moisture cycles.Summary of METROMEX, Vol. 2, Causes of Precipita-tion Anomalies. Illinois State Water Survey Bull 63,Urbana, 30–36

Howard L (1833) Climate of London deduced from Meteoro-logical Observations. London: Harvey and Darton

Kirkpatrick JS, Shulman MD (1987) A statistical evaluationof the New York City-Northern New Jersey urban heatisland effect on summer season daily minimum tempera-ture. Nat Wea Dig 12: 12

Landsberg HE (1981) The Urban Climate. InternationalGeophysical Series 28. NY: Academic Press, 275 pp

Ludwig FL (1970) Urban temperature fields. Urban Cli-mates, WMO Technical Note, No. 108: 80–112

Montávez JP, Rodrı́guez A, Jim�eenez JI (2000) A study of theurban heat island of Granada. Int J Climatol 20: 899–911


Morris C, Simmonds I (2000) Associations between varyingmagnitudes of the urban heat island and the synopticclimatology in Melbourne, Australia. Int J Climatol 20:1931–1954

Oke TR, Maxwell GB (1975) Urban heat island dynamics inMontr�eeal and Vancouver. Atmos Envrion 9: 191–200

Oke TR (1995) The heat island of the urban boundarylayer: characteristics, causes and effects. In: CermakJE, Davenport AG, Plate EJ, Viegas DX (eds)Wind Climate in Cities. NATO ASI Series E: AppliedSciences – Vol. 277, Boston: Kluwer AcademicPublishers, pp 81–108

Renou E (1862) Difference de temperature entre Paris etChoisyi-Roi. Societ�ee M�eet�eeorologique de France Annuaire10: 105–109

Richwien BA (1980) The damming effect of the SouthernAppalachians. Nat Wea Dig 5: 2–12

Runnalls KE, Oke TR (2000) Dynamics and controls of thenear-surface heat island of Vancouver, British Columbia.Physical Geography 21: 283–304

Semonin RG (1981) Surface weather conditions. In:Chagnon SA (ed) METROMEX: A Review and Sum-mary. Meteorological Monograph 18. American Meteor-ological Society, Boston (181 pp), 17–40

Voogt JA, Oke TR (1997) Complete urban surface tempera-tures. J Appl Meteor 36: 1117–1132

Yoshikado H (1990) Vertical structure of the sea breezepenetrating through a large urban complex. J Appl Meteor29: 878–891

Yoshikado H (1992) Numerical study of the daytime urbaneffect and its interaction with the sea breeze. J ApplMeteor 31: 1146–1164

Zois CNA, Sandoval AR (1985) Observations of the New-ark, New Jersey nocturnal heat island. Bull New JerseyAcad Sci 30: 1–16

Authors’ addresses: S. D. Gedzelman (e-mail: [email protected]), Department of Earth and Atmo-spheric Sciences, City College of New York, New York,NY 10031, USA; S. Austin, Physical, Environmental, andComputer Sciences Department, Medgar Evers College,Brooklyn, NY 11225, USA; R. Cermak, Weather AccessInc., 22 Felton Avenue, Highland Park, NJ 08904, USA;N. Stefano, Sussex County Weather Network, 890 GreenvilleRoad, Wantage, NJ 07461, USA; S. Partridge, WeatherAccess Inc., 55 Fairmount Ave., Chatham, NJ 07928,USA; S. Quesenberry, South Jersey Resource Conserva-tion and Development Council, 251 Belloeview Avenue,Hammontown, NJ 08037, USA; D. A. Robinson, Depart-ment of Geography, Rutgers University, New Brunswick,NJ 08901-8551, USA.

42 S. D. Gedzelman et al.: Mesoscale aspects of the Urban Heat Island around New York City

mesoscale aspects of the urban heat island around new york ... · (renou, 1862; landsberg, 1981)....

Documents