Anthropogenic heat island at Barrow, Alaska, during winter: 2001–2005



[1] The village of Barrow (71°N latitude) is the largest native community in the Arctic, with a population of approximately 4500 people. Situated on the coast of the Arctic Ocean in northernmost Alaska, the area is entirely underlain by permafrost. Although most supplies must be imported, Barrow relies on local natural gas fields to meet all energy requirements for building heat and electrical power generation. This energy eventually dissipates into the atmosphere, and can be detected as a pronounced urban heat island (UHI) in winter. Since 2001, a 150 km2 area in and around Barrow has been monitored using ∼70 data loggers recording air temperature at hourly intervals. The mean daily temperature of the urban and rural areas is calculated using a representative sample of core sites, and the UHI magnitude (MUHI) is calculated as the difference in the group averages. The MUHI is most pronounced in winter months (December–March), with temperatures in the urban area averaging 2°C warmer than in the surrounding tundra and occasionally exceeding 6°C. The MUHI is maximized under cold and calm conditions, and decreases with wind speed and warmer temperatures. It is strongly and directly correlated to natural gas utilization on a monthly basis. Integrated over the home heating season, there is an 8% reduction in freezing degree days in the village. It is unlikely that anthropogenic heat contributes to the forward shift in the snow meltout date that has been observed near Barrow over the past 60 years.

1. Introduction

[2] The clearest, least ambiguous signal of unintentional climate alteration by humans is provided by the urban heat island (UHI). Since the time of Howard's [1820] seminal work on the climate of London, it has been recognized that urban areas are noticeably warmer than the surrounding hinterland. Within the next few years, the planet will likely reach a milestone: for the first time in history, half of the global population will reside in a setting defined as “urban” [United Nations, 2005]. The potential consequences on human health and well-being, especially in the rapidly growing tropical and subtropical cities, is receiving considerable attention from urban climatologists using a suite of observational, modeling and mitigation tools [e.g., Rosenfeld et al., 1998; O'Neill et al., 2003; World Health Organization, 2003]. Less attention has been focused on urban heat-island effects in high-latitude regions, which have potential to increase downward heat flow and impact cryogenic processes.

[3] There are several interrelated causes invoked to explain warmer temperatures in urban settings [Oke, 1995; Arnfield, 2003]. Urbanization typically entails the conversion of a vegetated, permeable surface to dense, nonpermeable, low-albedo, heat-absorbing materials such as concrete, asphalt and brick. Further, the original surface is replaced with vertical and horizontal surfaces that absorb and retain heat energy. The net effect is to increase the net radiation flux by enhanced absorption of incident solar and thermal radiation, and to increase the sensible heat storage [Oke, 1987]. In addition, the resulting complex three-dimensional urban geometry has a much rougher surface, which disrupts wind patterns and serves to trap energy within the urban canopy, the layer between the ground surface and roof tops. Since solar energy dominates the net radiation budget, the UHI in midlatitude cities is best developed in summer [Klysik and Fortuniak, 1999; Philandras et al., 1999; Morris et al., 2001].

[4] Impermeable surfaces, combined with municipal drainage systems to remove surface water, effectively reduce surface moisture and therefore limit evaporative cooling. Combined with reduced total turbulent transport, a larger proportion of absorbed radiant energy is available as sensible heat to warm the air and substrate. These effects can be amplified by the aerosol or gaseous pollutants often concentrated in the urban setting. Although the impact on urban temperature is most obvious, other aspects of the urban climate are impacted including cloudiness, precipitation patterns and amounts, and the local wind regime [Landsberg, 1981].

[5] Given the geographic complexity of the urban mosaic, it is difficult to assess the role of the anthropogenic heat flux density (QA). This component of the surface energy balance equation results from the conversion of stored chemical energy in fossil fuels or biomass to heat energy. Grimmond and Oke [1991] defined three components of QA in a suburban setting, such that

equation image

where QAV is sensible heat generated by vehicles and similar mobile sources, QAH is energy released from homes and other buildings requiring internal temperature regulation (including stationary power generation), and QAM is energy contributed by animal metabolism. Although electrical power consumption and natural gas use, as it varies over time and space, can be used to estimate QAH if data are made available by the local power company [Grimmond, 1992], it is very difficult to assess the spatial and temporal impact of mobile sources (QAV). This is particularly true of large cities with major transportation arteries.

[6] The village of Barrow in northern Alaska is anomalous in this regard. A traditional whaling site near Point Barrow, the village is the largest native community in the Arctic with a population of about 4500, of which 59% is Inupiat Eskimo. Nearly all of the energy needed to support modern life styles is derived from local natural gas fields. There are no all-weather roads connecting Barrow to the Alaska's primary road network, so perishables and necessities are delivered daily by aircraft. Aviation fuel is stockpiled at the airport, and gasoline is available at a single gas station. Processed fuels and bulky goods are delivered by barge in August, when coastal shipping lanes are temporarily open. Vehicle traffic is sparse and restricted to the region within and around the village, and power-dependent industry is largely absent. Thus QAV is relatively small.

[7] “Typical” urban structures and complex urban morphology are not present at Barrow. Concrete and asphalt are supplanted by wood and synthetic materials. Like other high-latitude locations, there is a period of continuous darkness in winter during which the solar component of the surface energy budget is absent or negligible. Latent heat transfer is minimal because water remains in the frozen state [Benson et al., 1983; Magee et al., 1999; Steinecke, 1999]. Under these conditions, anthropogenic energy should be the dominant cause of an existing winter heat island. Moreover, because QAV and QAM are negligible, QA is defined by QAH. Records of monthly natural gas production and use are available for the village, making it possible to approximate QA at this temporal scale.

[8] There are few cities on Earth where the anthropogenic heat input can be so clearly isolated and quantified. Although midlatitude cities typically experience the strongest heat island in summer, and tropical cities tend also to respond to wet-dry seasons [Kumar et al., 2001; Tereshchenko and Filonov, 2001], the UHI at Barrow should be strongest in winter and should result exclusively from anthropogenic heat derived from local fossil fuel sources [Benson et al., 1983; Hinkel et al., 2003].

[9] Evaluating the heat island at Barrow is not an esoteric exercise. Data published by Stone et al. [2002] and updated by Hinzman et al. [2005] indicate that melting of the snow cover is occurring earlier in the season. Since 1941, the snow melt date has advanced ten days on the open tundra near Barrow, and the effect is more pronounced in the village. Stone et al. [2002] attributes this forward shift to regional synoptic circulation changes, although Dutton and Endres [1991] suggest that local factors may predominate.

[10] Barrow is situated in the continuous permafrost zone. A layer of near-surface soil, known as the active layer, thaws during the short summer and refreezes each winter. Extending the snow-free season by inducing the disappearance of high-albedo snow effectively extends the period during which the ground can thaw and promotes thickening of the active layer, which could destabilize the upper permafrost. Because near-surface permafrost is supersaturated with respect to ice [Sellmann et al., 1975], thawing would result in ground subsidence. The potential impact on roads, buildings, runways, pipelines and related infrastructure could be substantial and costly [e.g., Nelson et al., 2001].

[11] The objectives of this study are threefold: to (1) determine the magnitude of the Barrow UHI as it varies with time and over space; (2) to assess the influence of wind velocity on UHI magnitude, and (3) to estimate the integrated impact of anthropogenic heat on UHI magnitude, snow cover stability, and melt out date.

2. Study Area

[12] The coastal village of Barrow (71.3°N, 156.5°W) is located on the Chukchi Sea about 15 km southwest of Point Barrow (Figure 1). The mean annual temperature is −12.0°C, and ranges from −26.6°C in February to 4.7°C in July [National Climate Data Center, 2003a]. Only June, July and August have positive mean monthly temperatures. The sun remains below the horizon for the period 18 November to 23 January, and is continually above the horizon from 10 May until 2 August. Most of the 11 cm of precipitation falls as rain in summer. The snow cover is typically in place by September or October, and averages 50 cm in depth on the open tundra. Maximum active-layer thickness occurs in August, and averages about 35 cm [Nelson et al., 1998; Hinkel and Nelson, 2003].

Figure 1.

Landsat-7 ETM+ image of the 150 km2 Barrow Urban Heat Island Study (BUHIS) network showing site locations. Sites symbolized with squares are rural core sites; the six urban core sites are near the village center to the east of the “6.” The solid circle labeled “W” is the Barrow NWS station.

[13] Offshore sea ice is usually stable by October. In the Chukchi Sea it is often 1 m thick by December and thickens throughout the winter. Narrow openings in the ice cover, known as leads, occasionally develop in winter when wind conditions are favorable, but these tend to be episodic and of short duration. Ice in the shallow, protected Elson Lagoon remains intact through the winter. Spring breakup is highly variable, and occurs between April and June.

[14] The Barrow region is underlain continuously by permafrost to depths of about 600 m. Relief is confined to within 20 m of sea level and the land surface is covered nearly continuously by tundra vegetation. The largest structure in Barrow is the airport runway, which is visible in Figure 1, south of the village center. Housing density is relatively high near the village center and extends as a narrow band northeastward along the coast for several km. Overall, an area of about 4 km2 is covered with gravel pads and buildings [Klene, 2005]. Only a few buildings are multistoried, and nearly all structures are built atop wood pilings driven into the underlying permafrost. This creates a 1- to 2-m crawl space beneath the building permitting air circulation, and limits potentially destabilizing heat flow into the ground. Buildings are significant surface roughness elements in this flat terrain, but they are not too dissimilar from the ice pressure ridges that develop offshore in winter. Wood and siding are the primary building materials; brick, concrete and asphalt are largely absent.

[15] The limited road network is constructed of gravel berms about 1–2 m high. These are plowed in winter, resulting in an artificial topography that has a significant influence on snow drift patterns and snow depth.

3. Methodology

[16] In June 2001, approximately 50 data loggers were installed across the study area. The devices are custom Onset Computer HoboPro® two-channel data loggers with accuracy of ±0.2°C and precision of 0.02°C at the freezing point. The units were mounted on an instrument mast, with one thermistor inside a six-plate radiation shield located 1.8 m above the ground surface. The second thermistor was positioned 5 cm below the surface to measure near-surface ground temperature. Data loggers were synchronized and recorded at hourly intervals.

[17] About half the measurement sites were established in urban settings, mostly in the back yards of local residents. In some of the urban settings it was not feasible to deploy an instrument mast, so radiation shields and data loggers were attached to utility poles. The remaining sites were located in a rural setting to form an approximate grid pattern over the 150 km2 study area (Figure 1). In August 2001, additional data loggers were deployed such that a network of 68 sites was operational by September 2001. The data loggers were serviced each year in August, at which time the depth of thaw was also measured. Attrition rate was less than 10% per year, and was due primarily to instrument malfunction and animal activity. At the end of the 4-year study, about 70% of the loggers had a data record spanning the entire duration. The hourly data for each site were processed, and the daily mean temperature and temperature range calculated.

[18] The Barrow Urban Heat Island Study (BUHIS) network has unusually high instrument density, and was designed specifically to avoid many of the problems inherent in networks producing spatial climate data sets [Daly, 2006]. The low relief of the Barrow Peninsula and the compact nature of the study area facilitate use of straightforward interpolation techniques for BUHIS data [Hinkel et al., 2004; Klene, 2005]. A simple kriging algorithm [Golden Software, 2002] was used to interpolate station values over the entire area and produce temperature maps.

[19] The magnitude of the heat island is determined by comparing rural to urban temperatures over a specified time period. In a study such as this involving numerous instruments, it is first necessary to identify those sites to be used for the comparison. As an initial step, the mean temperature was calculated at each site for the period 1 Nov to 30 April using hourly data. This time interval was selected because the ground is snow covered throughout the duration, and both lakes and the ocean typically have a thick, stable ice cover that limit the effects of thermal advection on the temperature signal. The results are shown in Figure 2, which demonstrates interannual variability of several degrees over the four winter periods.

Figure 2.

Mean temperature of individual sites for 1 November to 30 April for each of the four winter periods, calculated using hourly data. Trace displacement along y axis reflects interannual variation with winter 2003–2004 about 3°C colder than winter 2002–03.

[20] The intent is to identify the warmest and coldest sites. As a target, six sites were selected (about 10% of the operational data loggers) that consistently recorded the warmest temperatures; these were U5, 13, 39, 42, 43, and 51. In a similar manner, the six coldest sites were selected. The mean and variance of each site within the group were compared to ensure statistical similarity; this resulted in eliminating one of the cold sites. The coldest sites with a continuous record throughout the duration were U15, 16, 18, 24 and 59.

[21] These sites were then mapped (Figure 1). The warmest sites were concentrated in the central urban area while the coldest were located in the undeveloped tundra away from large water bodies, indicating the existence of a winter urban heat island. To calculate the magnitude of the urban heat island (MUHI), the average temperature of all core sites for the urban (equation image) and rural (equation image) groups were used, such that

equation image

In this way, estimates utilize group averages rather than individual sites [Hinkel et al., 2003]. Unless otherwise noted, all comparisons between the rural and urban settings are based on these group averages of representative core sites.

4. Results

[22] The spatial and temporal patterns of the MUHI are analyzed at varying scales to identify controlling factors. These are discussed separately below.

4.1. Temporal Patterns of Daily MUHI

[23] The initial analytical step entailed comparing rural and urban core sites on a daily basis over the period of record (1 September 2001 to 31 August 2005). Results are displayed in Figure 3, which shows daily MUHI values as solid circles. Positive values indicate that the urban area is warmer than the rural sites. During winter, some days have a MUHI exceeding 6°C. To evaluate seasonal patterns, a weighted moving average was applied and is shown as the solid trace. The annual sinusoidal curves demonstrate the strengthening of the MUHI during autumn and winter, with a peak in January or February, followed by a decline in spring. As such, the cycles are nearly out of phase with the annual temperature cycles; peak MUHI values are associated with minimum air temperatures, as documented by Hinkel et al. [2003]. Further, the MUHI tends to be negative for several months in summer when air temperatures are warmest. This is largely a response to a slight maritime effect resulting from seasonal westerly winds [Hinkel et al., 2004].

Figure 3.

Time series of daily MUHI calculated using core sites averaged for period 1 September 2001 to 31 August 2005; ticks at approximate 1-month intervals. The solid trace was generated using 5-day weighted average algorithm. Numerical values represent mean MUHI calculated over the winter period, defined as 1 December to 31 March for each year.

[24] When calculating average MUHI, the time period must be stipulated. For example, if comparing daily means as in the previous example, MUHI is defined as MUHID; comparisons of monthly values are symbolized as MUHIM. Here we are interested in seasonal magnitudes and define the winter period as 1 Dec to 31 March because it consistently encompasses the coldest time of the year. Thus the MUHIW is reported in Figure 3 for each of the four winter (w) periods. Although there is considerable interannual variation, the average MUHIW is about 2°C. Specifically, the 485 MUHID values range from −0.05 to 8.37°C, with mean of 2.0°C and median of 1.5°C. In 25% of the cases, the MUHID is less than 0.7°C. A similar percentage is greater than 3.1°C, and 12% of the time the MUHID exceeds 4°C.

4.2. Wind Regime and Daily MUHI

[25] The heat island intensity is related to cloud cover, and is best developed under cloudless conditions. Clouds reduce solar forcing during the day, and limit net longwave radiation losses [Oke, 1987, 1998]. Given the same sky coverage, thick low-lying stratus clouds are more effective than thin cirrus clouds [Oke, 1987]. To quantify the impact of clouds requires measurements of sky coverage (oktas as measured by ceilometer and satellites), ceiling height and cloud type [Oke, 1998], but these observations are not available in the Local Climatological Data daily summaries for Barrow. However, given the cold winter air temperatures and thick sea ice cover, the specific humidity is low and cirriform clouds tend to be thin. Further, solar forcing is absent or limited throughout much of the winter period. Thus the impact of clouds is unknown but should be minimal.

[26] Previous studies have also indicated that the MUHI is strongly dependant on wind velocity [Oke and Maxwell, 1975; Yap, 1975; Figuerola and Mazzeo, 1998; Morris et al., 2001; Hinkel et al., 2003]. Calm conditions, defined as <1.3 m s−1 by Oke and Maxwell [1975], promote development of strong urban-rural temperature gradients, while windy conditions cause mechanical mixing and turbulence and thus reduce the MUHI. Wind rose diagrams for each winter period are shown as Figure 4 using data from the National Weather Service at Barrow [National Climate Data Center, 2003b]. Mean daily wind velocity (m s−1) averaged 5.5 m s−1 over the four winter periods (N = 485). Winter winds are predominately and consistently from the eastern sector. Although not shown, wind rose diagrams for the summer periods (I June to 31 August) are similar but flip between two opposite modes. They are either from the ENE or WSW, and the dominant direction can vary markedly each year. Westerly winds across open water in summer are associated with slightly negative MUHID values [Hinkel et al., 2004].

Figure 4.

Wind rose diagrams for each winter period (1 December to 31 March) using daily data from the Barrow National Weather Service.

[27] The impact of mean daily wind velocity on winter MUHID is shown in Figure 5. Here the median, quartiles, and extreme values for all winter observations (N = 485) are plotted using wind velocity bins of 1.5 m s−1. Most daily MUHI values range from 1.5° to 4.0°C when wind velocity is less than 3 m s−1; the median value hovers around 3°C. Wind velocity of 3–6 m s−1, which encompasses mean and median wind speed, are associated with a median MUHID of 2.0°–2.5°C. MUHID decreases steadily with higher winds, falling to <0.5°C when velocity exceeds 9 m s−1.

Figure 5.

Impact of mean daily wind velocity on daily MUHI values. Wind velocity bins have a width of 1.5 m s−1. Median, quartile, and extreme values are shown for each bin. Descriptive statistics are for the four winter periods (1 December to 31 March).

[28] It is not uncommon for the daily MUHI to reach 6°C when wind velocity is less than 6 m s−1. The anomalous maximum MUHID value of 8.4°C occurred on 3 December 2003, and is discussed below.

4.3. Spatial Patterns of the Daily MUHI

[29] Using mean daily temperature from all sites, the spatial patterns can be mapped across the study area. Four days have been selected to illustrate patterns as they vary with air temperature, wind velocity, and wind direction.

[30] Figure 6a shows the temperature field on 4 March 2004, and is representative of a statistically “average” winter day in Barrow. The mean daily temperature is −22°C, the wind is from the ENE at 5.2 m s−1, and the MUHID is 2.7°C. On the maps, wind regime characteristics are shown in the upper right corner. Daily summary statistics are indicated in parentheses. The range is the difference in mean daily temperature between the warmest and coldest sites in the study area; on this date it was 3.1°C. In all cases, the range exceeds the MUHID since the extremes in the data set are used instead of group averages. Maximum temperatures occur in the urban core, with warmer conditions extending along the populated portions of the western coast, and minimum temperatures in the unsettled interior. Because the sea ice cover was intact and winds were easterly during this period, advection cannot be the causal mechanism. This is a typical geographic pattern in winter that closely approximates the population distribution and results from anthropogenic heat.

Figure 6.

Mean daily temperature field on (a) “average” winter day in Barrow (4 March 2004); (b) calm day (17 December 2001); (c) windy day (9 March 2005); and (d) day with offshore open lead (3 December 2003). Wind direction and velocity, and descriptive statistics, are shown above map. Dots show site locations. Note difference in temperature scales.

[31] This pattern occurs in more extreme fashion on cold calm days. Figure 6b illustrates the daily temperature field on 17 December 2001. The mean daily temperature was −31°C and the 1.5 m s−1 wind was from the north. The MUHID was 5.9°C, with a temperature range across the area of 8.6°C. This strong gradient developed without diurnal forcing, as the sun remained well below the horizon on this date.

[32] At the other end of the spectrum, windy days eliminate spatial gradients, as shown in Figure 6c. Winds on 9 March 2005 were from the ENE at 13.0 m s−1. Mean daily temperature was −23°C and the MUHID 0.2°C; there was only a 0.9°C temperature difference across the study area. This pattern is typical for wind velocities exceeding 7.5 m s−1, which occur on about 20% of the winter days at Barrow. Although no heat island was detected in the temperature field, heat loss from buildings is likely to be at a maximum during windy periods owing to leakage of warm interior air and turbulent sensible heat transfer across all building surfaces including the floor of elevated buildings.

[33] The maximum MUHID value over the period of record was 8.4°C. This anomalous value occurred on 3 December 2003, when the mean daily temperature was −26°C (Figure 6d). Winds from the west at 4.2 m s−1 opened a lead in the Chukchi ice pack near shore (Barrow National Weather Service, unpublished data, 2004). The impact of advection is apparent along the entire western coast; not just the populated coastal regions. Isotherms parallel the shoreline and a large range of temperature (9.6°C) occurred across the study area. The same condition existed on the previous day, with westerly winds at 2.5 m s−1; the MUHID was 6.3°C and the geographic pattern was nearly identical. These events are not common during winter months, but must be identified and separated from UHI effects.

[34] Under calm conditions, the UHI is strongly developed in the urbanized core of Barrow, as reflected in the mean daily temperature fields. Research conducted in midlatitude and tropical cities demonstrates that hourly values of the MUHI typically peak in the late evening and early morning hours [Unwin, 1980; Ripley et al., 1996; Jauregui, 1997]. At these locations, daytime temperatures are similar in rural and urban settings, but nocturnal cooling is reduced in the city. Absorbed solar energy is retained in building and road materials, and is only gradually released as thermal radiation and sensible heat in the complex urban morphology. The net impact is to shift the daily minimum and mean temperature upward while reducing the temperature range.

[35] At Barrow, it is difficult to statistically summarize daily patterns owing to weak or nonexistent diurnal forcing in winter, making the notion of “day” versus “night” meaningless. To determine if there is a daily pattern to the MUHI, hourly data from one urban and one rural site were compared during several time periods. Hinkel et al. [2003] demonstrated attenuated nighttime cooling at Barrow in April, during a period of strong diurnal forcing; this pattern was also observed in Fairbanks in March by Magee et al. [1999]. However, during the period of complete darkness in January, hourly data show no diurnal forcing effects. Instead, the urban site was consistently warmer and the urban-rural contrast tended to increase with colder temperatures.

4.4. Anthropogenic Heat and the MUHI

[36] Barrow is highly unusual because nearly all of its energy needs are provided by local natural gas fields. The Barrow Utilities and Electrical Cooperative, Inc., compiles gas production and sales data, which are available as monthly values. These were correlated with monthly MUHI values calculated using the core groups; regression coefficients are given in Figure 7. The Pearson Product Moment coefficient of correlation for the 4-year period is 0.94, indicating a very high correspondence between the magnitude of the heat-island effect and energy use in winter.

Figure 7.

Linear best-fit regression between monthly MUHI and natural gas production/use for the 4-year period. Solid circles represent monthly winter (December–March) values. Solid squares represent summer months (June–August). Transition months are indicated by open triangles.

[37] About half of the natural gas produced locally is used to generate electricity, and the remainder is used directly to heat living space and water, for cooking, and to operate appliances. During summer, 2.0–2.5 × 106 m3 of gas are consumed by the populace each month for electrical lighting, cooking, and appliances. In winter about twice this amount is used monthly, and this largely reflects the impact of residential, commercial, and municipal space heating demands. This sensible heat ultimately escapes to the atmosphere, and is detectable as the winter UHI.

4.5. Seasonal Impact of the UHI on Freezing Degree Days

[38] One way to assess the temporally integrated impact of the warmer urban core is to compute the daily freezing degree days (DDF) at each site. Freezing degree days were calculated as the difference between the site-specific mean daily value and the ice point. The DDF were accumulated for a specified duration to yield a time-temperature integral (ADDF) with units of °C-days. These site-specific sums were interpolated from the irregular pattern of logger locations using a kriging algorithm [Golden Software, 2002] and mapped to quantify the overall impact of the UHI.

[39] This exercise could be performed for the core winter period (1 December to 31 March). It is more appropriate, however, to make the assessment over the entire period of subfreezing temperatures, when energy is being used for space heating purposes. This is roughly the 9-month period 1 September to 31 May. As shown in Figure 7, these months coincide with MUHIM values exceeding zero.

[40] Figure 8 maps ADDF over the 9-month winter periods. Although ADDF magnitudes differ annually, the geographic pattern is consistent, with fewer ADDF near the urbanized core area and along the populated west coast. Maximum ADDF values are found in the interior, away from the village and the coast. In general, annual patterns are quite similar and are clearly related to the effects of urbanization.

Figure 8.

Accumulated freezing degree days (DDF) over the home heating period (1 September to 31 May) for each winter. Avg Tr is the average temperature at the rural core sites; DDFu reflects the reduction in accumulated DDF at the urban core sites relative to the rural core sites.

[41] Summary data are shown in parentheses above each map. Owing to logger loss, the available sites vary from year to year. Thus interannual comparisons are made using the rural and urban core group averages. Average temperature, as measured at the rural core sites over the 9-month period, is reported as Avg Tr. To compare the integrated impact, the average ADDF of the rural and urban core sites were calculated. The overall reduction of ADDF at the urban sites is expressed as a percentage relative to the rural average. These values range from −6.4 to −9.1%; on average, the urban area has 8.1% fewer ADDF than the rural sites.

[42] This exercise is affected somewhat by conditions that vary from year to year. For example, in some years the snow cover may not be in place by 1 September, or the sea ice cover may not be stable at the same time of year. However, the same exercise was performed for the winter core period (1 December to 31 March). Although ADDF magnitudes were necessarily reduced, the geographical patterns are very similar to those shown in Figure 8. Moreover, the percentage reduction in ADDF in the urban area was nearly identical to those representing the 9-month period, which supports the methodology and the validity of the results.

5. Discussion

[43] Unlike midlatitude cities, which tend to experience the strongest heat island in summer, at Barrow the UHI develops in winter. Furthermore, the heat island at Barrow is largely due to the release of anthropogenic heat associated with maintaining the internal temperature of buildings (QAH). Building density, geometry, and construction materials are unlikely to have much impact, and other anthropogenic sources (QAV and QAM) are minimal. Mapping of the daily temperature field reflects the influence of the underlying pattern of residential settlement and energy use.

[44] Integrated over the heating season, there is a significant and consistent reduction of freezing degree days in the urban core area. There are several reasons, however, to conclude that the anthropogenic heat does not hasten spring snow melt. First, the UHI is only weakly developed during the spring meltout period. The monthly observed MUHI in May, for example, averaged 0.6°C, with a range of 0.4 to 0.7°C over the 4-year period. The June average was −0.1°C, and ranged from −0.3 to 0.1°C. A large amount of energy is required to cause snow melt (0.33 MJ kg−1) or sublimation (2.83 MJ kg−1 at 0°C). The contribution of anthropogenic heat during the snow melt period is minor relative to the insolation load at this time of year, given that the sun is continuously above the horizon after 10 May.

[45] Second, it is possible that less snow accumulates on the ground owing to the indirect effects of urbanization. The road network at Barrow is well maintained by the continual plowing made necessary in this windy location. This accentuates local snow drifting, which has become more pervasive as residential development continues, since new and existing structures contribute to surface roughness and drifting [e.g., Outcalt and Goodwin, 1980]. In the recent past, it was not uncommon for home entranceways to be buried beneath drifts several times each winter; this was especially true for houses on the eastern fringe of the village. The response has been the construction of a series of 4-m-high snow fences upwind of the village; one extends for more than 2 km [Hinkel and Hurd, 2006]. The net effect of these activities is to reduce the amount of snow delivered to the village, while the increasing number of obstruction-induced spatial variability in snow thickness within the village.

[46] Another indirect effect of urbanization is the generation and dispersal of dust. Snow cover is visibly darker in the urban area, especially downwind of gravel roads. Studies are ongoing at Barrow to determine if the impact on snow albedo is sufficient to induce earlier snow melt [e.g., Auerbach et al., 1997].

[47] The geographic patterns of the UHI reflect the residential settlement pattern. Although buildings are typically well insulated, sensible heat must ultimately escape and warm the atmosphere. Heat is probably dispersed downwind of the village over the Chukchi Sea, perhaps as an ascending thermal plume. Loggers monitor temperature at a standard height of 1.8 m. However, given that buildings are elevated 1–2 m on pilings, it is unlikely that the sensors are detecting the majority of the heat that escapes from buildings via cracks, gaps, entrance ways and windows. Efforts to explore the vertical temperature structure at heights above the urban canopy were not possible owing to Federal Aviation Administration restrictions associated with the nearby airport.

6. Conclusions

[48] Analysis of the 4-year temperature record from 68 sites covering the 150 km2 BUHIS study area yields several conclusions.

[49] 1. The MUHID at Barrow demonstrates a strong seasonal pattern, with a well-developed UHI in winter. Slightly negative magnitudes in summer reflect a maritime influence caused by episodic westerly winds.

[50] 2. In winter, the magnitude of the UHI increases with decreasing air temperature and reaches a peak in January-February. There is considerable interannual variability.

[51] 3. On the basis of rural and urban group averages for the period 1 December to 31 March of four winters, the urban area is ∼2°C warmer than the rural area. It is not uncommon for the MUHID to exceed 4°C.

[52] 4. The strongest heat islands are associated with calm days. When wind velocity is less than 3 m s−1, the median daily MUHI value is ∼3.0°C. Average daily wind of 3–6 m s−1 are associated with MUHID values of 2.0–2.5°C, whereas windy days (>9 m s−1) have daily MUHIs less than 0.5°C.

[53] 5. Warmest temperatures are consistently observed in the urban core area. However, the magnitude and spatial pattern of the UHI is strongly dependent on daily weather conditions. These include wind velocity and direction, mean daily temperature, and the influence of episodic events such as the opening of near-shore leads in the ice pack. On very windy days, the temperature field across the study area is uniform.

[54] 6. Monthly natural gas production/use is strongly and directly correlated with the monthly MUHI. Large winter MUHI values reflect the impact of anthropogenic sensible heat derived from local natural gas fields.

[55] 7. Using spatial averages integrated over the winter seasons of 1 September to 31 May, accumulated freezing degree days are reduced about 8% in the urban core compared to the hinterland. However, it is unlikely that anthropogenic heat contributes significantly to snow melt because the heat island is very weak or nonexistent during the snow melt period, and is negligent relative to the solar radiation load experienced at this high-latitude site.

[56] The availability of relatively inexpensive and reliable data loggers makes it possible to collect temperature time series across a large region. This permits mapping of the temperature patterns, which helps to identify and interpret operative processes. Furthermore, groups of sites can be used to define core regions, eliminating the reliance on single sites to represent “urban” or “rural” settings. This provides a conservative method of estimating the UHI impact.

[57] The network described here is expensive and time consuming to maintain. Most of the instruments have been removed from the field; only those sites that form the representative rural and urban core groups remain operational. The core groups will provide the data needed to continue basic monitoring of the UMH magnitude and the temporal patterns. Although it will not be possible to map the temperature fields in subsequent winters, this study demonstrates that the geographic patterns are coherent and internally consistent.


[58] This work was supported by the National Science Foundation under grants OPP-9529783, 9732051, and 0094769 to K. M. H. and OPP-0095088 and 0352958 to F. E. N. Any opinions, findings, conclusions, or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation. We are grateful for support from the Barrow Arctic Science Consortium, the Ukpeagvik Inupiat Corporation, Dorothy Savok at the Barrow Utilities and Electrical Cooperative, Inc., the North Slope Borough, and the many Barrow residents who allowed us to use their back yards and parking lots. R. Beck, J. Bell, B. Jones, A. Klene, and C. Tweedie assisted in this study, and H. Eicken provided data on ice freezeback and breakup. Our thanks go to three anonymous reviewers for their helpful comments.