As a greenhouse gas, atmospheric water vapour (WV) plays a major role in the climate (Trenberth et al., 2005; Kay et al., 2009; Schmidt et al., 2010; Devasthale et al., 2011; Doyle et al., 2011; Solomon et al.2011). In the Arctic, downwelling longwave radiation, and therefore the climate, is particularly sensitive to the total column precipitable water (PW), and to the PW in near surface WV inversions (Curry et al., 1995). Doyle et al. (2011) found that WV intrusions, which measurably increase the total column PW in winter, increased the downwelling longwave radiation by about 15%. Ohmura (2001) showed that approximately 90% of the downwelling longwave radiation to the surface is emitted from a surface layer no deeper than 1000 m. At high latitudes, Zhang et al. (2001) found that downwelling atmospheric longwave radiation to snow surfaces increases logarithmically with increasing PW. The rate of increase is the greatest in the PW range from 0 to 5 mm. Zhang et al. (1996, 1997) also demonstrated that during snowmelt at high latitudes changes to the surface net radiation budget are primarily due to changes in downwelling longwave radiation. Variation in atmospheric water content (vapour, liquid and solid) is the most important factor controlling the variation in downwelling longwave radiation. Thus, it is critical to the onset and timing of snow melt.
In the Northern Hemisphere, the meridional distribution of atmospheric WV has a maximum in the Tropics, a secondary maximum in mid-latitudes, and a minimum in the Arctic. WV generally decreases with height with a maximum near the surface. Both the meridional distribution and the vertical structure of atmospheric WV are primarily controlled by the Clausius–Clapeyron relationship—the dependence of saturation vapour pressure on temperature. However, as a result of other factors such as seasonality, advection, storm tracks, convergence zones, variable surface evaporation/transpiration, and convection, WV exhibits a high degree of spatial and temporal variation (Trenberth et al., 2005).
In the Arctic, the total column PW is relatively low with a definite seasonal cycle. Serreze et al. (1995) analysed a comprehensive Arctic rawinsonde data set, and found that total column PW (surface 30 kPa) ranged from about 2.5 mm in January to 14.0 mm in July. For all months, approximately 95% of PW was in the surface to 50 kPa layer. Doyle et al. (2011) analysed 2007–2011 microwave radiometric profiles (MWRP) from Eureka (80°N, 86°W), and found that the meridional transport of moisture or WV intrusions increased the PW well above its climatological baseline value about 30% of the time in winter.
A near surface WV inversion (an increase of WV with height) is common in the Arctic (Devasthale et al., 2011). Moisture inversions have been attributed to the high frequency of temperature inversions, and the strong influence of meridional moisture transport on the WV profile. In winter, decoupling of the snow ice surface from the atmosphere, with the transported of WV peaking above a stable boundary layer, often creates a surface WV inversion (Curry, 1983). Deposition of WV on the surface may also contribute to the high frequency of winter WV inversions. Higher near surface temperatures in summer, and the associated increase in the depth and turbulent mixing of the boundary layer, particularly over land surfaces, means that surface/atmosphere decoupling is a rarer occurrence. These seasonal differences result in WV inversions occurring less frequently in summer than in winter.
Devasthale et al. (2011) have recently expanded our knowledge of the spatial distribution and vertical structure of atmospheric WV in the Arctic using data from an infrared sounder onboard NASA's Aqua satellite. They found the highest frequency of WV inversions during winter and the lowest during summer. Over the Arctic Ocean in winter, the mean frequency-of-occurrence was > 50%; high occurrence frequencies were also found over Siberia and the Canadian Archipelago. In summer, the mean occurrence frequency was less than 10%. For all seasons, WV inversion strengths (top–bottom) were most often between 0.3 and 0.7 × 10−3 kg m−3. Devasthale et al. (2011) also found that more than one inversion was often present in the near surface column. They calculated the contribution of the strongest WV inversion to the total column PW, and identified two apparent regimes. For relatively low values of PW, the PW in WV inversions often contributed up to 40%, and occasionally more, to the total column PW. When the total PW was relatively large, the inversion usually contributed less than 15%.
Gerding et al. (2004) detected WV inversions at around 1500 m at the Ny-Alesund research station (79°N), Svalbard Islands, Norway with LiDAR measurements. Treffeisen et al. (2007) provided statistics on WV inversions based on 15-years of humidity observations at the same site. While more frequent in winter, Tjernstrom et al. (2004) found WV inversions during summer using data from the 2001 Arctic Ocean Experiment. Data from the Arctic Summer Cloud Ocean Study (ASCOS) showed that summer WV inversions are typically shallower than winter inversions. Sedlar et al. (2011) reported frequent WV inversions associated with low-level cloud cover from pan-Arctic observation sites during all seasons.
The characterization of WV profiles over the Arctic remains limited in spite of its significance to the climate. Field observations of MWRP from the International Polar Year, Circumpolar Flaw Lead (CFL) System Study, mid-November 2007 through July 2008, and the 2009 ArcticNet cruise, August through mid-November, provided a unique opportunity to examine WV inversions, WV intrusions, and the annual cycle of total column PW over the southeastern Beaufort Sea–Amundsen Gulf region. This analysis should expand the knowledge of atmospheric thermodynamics over the western maritime Arctic, and contribute to improved climate modelling in this data sparse region.
In winter, most of the Beaufort Sea experiences consolidated pack ice whose usual anticyclonic motion is largely determined by the synoptic-scale winds (Thorndike and Colony, 1982). Amundsen Gulf, an easterly extension of the southern Beaufort Sea bounded to the north by Banks Island and to the south by Cape Bathurst, is generally covered by unconsolidated mobile pack ice with a vast network of inter-ice pan leads that generally persists from freeze-up in the fall to break-up in the spring. A recurrent flaw lead polynya can often be found between the land-fast ice, which generally forms along the north, south and eastern perimeters of the Gulf, and the mobile unconsolidated pack ice (Carmack and MacDonald, 2002; Barber and Hanesiak, 2004; Galley et al., 2008).
At the beginning of the 2007–2008 CFL period, the total sea-ice cover of Amundsen Gulf was around 97% with 3% leads, and it exhibited little variation from 7 January to 21 April 2008. The total sea-ice cover approached 100% from 28 April to 12 May. From May 19 onward, the total sea-ice cover declined rapidly to its July minimum of about 3% (Canadian Ice Service (CIS) digital ice charts http://www.ec.gc.ca/glaces-ice/default.asp?lang = En&n = D32C361E-1; Raddatz et al., 2012). At the start of the 2009 ArcticNet period, the total sea-ice cover of the Gulf was < 10% on 20 August. It dropped to 4% by 15 October before increasing rapidly to > 90% by 5 November. The total sea-ice cover was once again greater than 95% in December 2009 completing the annual cycle for the composite year (Figure 2). Compared to the average (1980–2004) year, the 2008 spring break-up began about 5 weeks early (Galley et al., 2008). The 2008 summer and 2009 summer total sea-ice covers did not mesh smoothly due to typical interannual variability during the summer season. In 2009, freeze-up began 2 weeks later, and ended 1 week later than the long-term mean.
Ware et al. (2003) noted that radiometric profiles are smoother than radiosonde soundings because the former observes a volume of air while the latter provides a point measurement. Cadeddu et al. (2002) determined that the vertical resolution of a microwave radiometric profiler was 125 m up to 400 m. Raddatz et al. (2011) concluded from the atmospheric boundary layer structure apparent in radiometric profiles, and from verification statistics that the vertical resolution may approach 100 m below 500 m as claimed by the manufacturer. The resolution became coarser with height above 500 m.
Throughout the 2007–2008 CFL and 2009 ArcticNet field campaigns, weather balloons carrying Vaisala RS92-SGPD radiosondes were launched from the Amundsen providing 68 radiosonde profiles that were used to validate the MWRPs. The Vaisala radiosondes took 45 min to 1 h for ascent. To account for the radiosonde ascent time, the MWRP data was averaged over the same hour. During all seasons, the root mean square (RMS) differences in WV density between the two instruments increased with height through the lowest 2000 m, and then decreased with height. The RMS differences for WV density (radiosonde, MWRP) in the lowest 2000 m averaged 0.25 × 10−3 kg m−3 during the winter (JFM) season, 0.32 × 10−3 kg m−3 for the spring (AMJ), 0.74 × 10−3 kg m−3 for the summer (JAS), and 0.37 × 10−3 kg m−3 for autumn (OND). Expressed as a percentage of the vertically averaged median WV density of the 0–2000 m layer, the RMS differences are 58, 11, 18 and 39%, respectfully, for winter to autumn. Radiosondes drift with the wind, so their profiles deviate from the zenith as viewed by a MWRP. This drift and inherent measurement errors of 10% for relative humidity (Miloshevich et al., 2006) make radiosondes less than ideal for verifying microwave radiometric profilers. During the CFL study, a problem with the software on the radiosonde base station caused frequent unrealistic drops in relative humidity. The WV profiles were corrected using linear interpolation. These factors, the relatively small number of comparable profiles, and the low WV densities in the colder seasons undoubtedly contributed to the fairly large differences between the radiosonde and MWRP.
The amount of atmospheric WV in a vertical column of unit cross-sectional area is termed PW (American Meteorological Society, 2000). PW can be determined for a column, z1 to z2, from WV density, profiles.
In practice, when WV has been measured at specific levels, Equation (1) is solved by multiplying each layer's mean WV density by the layer thickness, and then summing from the bottom to the top of the atmospheric column. The total column PW is commonly expresses as the depth of liquid water (mm) if the WV were condensed. PW (kg m−2) converts directly to mm as 1 kg of water spread over 1 m2 would produce a layer of water 1-mm deep given that the density of water is 1 × 103 kg m−3 (Oke, 1987).
From the composite year of hourly WV density profiles (November 2007 to July 2008 and August 2009 to November 2009) monthly median, 25 and 75 percentile PW values were calculated to describe the annual cycle of total column (0–10 000 m) PW, the percentage of PW below 5000 m, and in near surface WV inversions. Median rather than average values were used as a non-normal distribution was anticipated (Devasthale et al., 2011).
The time series of hourly PW values for each month was plotted to identify WV intrusions which were attributed to the meridional transport of moisture from more southerly latitudes. Doyle et al. (2011) found that, in winter, WV intrusions occur about 30% of the time. Thus, hourly PW amounts above the 75 percentile value for more than three consecutive hours were defined as WV intrusions, and the magnitude of the peak PW in WV intrusion was determined for each month.
Vertical profiles of the monthly median, 25 and 75 percentile WV density values were plotted at the levels observed by the microwave radiometric profiler. These plots were analysed to characterize the seasonality and strength of near surface WV inversions for the composite year. The hourly WV profiles for January and for July were also examined to provide further insight into near surface WV inversions during winter and summer.
4. Results and discussion
The monthly median total column (0–10 000 m) PW values (Table I) ranged from 1.4 mm for January to 16.8 mm for August, and the annual cycle (Figure 3) was similar to that reported by Serreze et al. (1995). The 25 and 75 percentile values (Figure 3) illustrate the cumulative frequency distribution of each month's PW. In all months, the 0–5000 m column held greater than 92% of the median total column PW, Table I.
Table I. Monthly median total precipitable water (PW mm) in the 0–10 000 m column, percentage of PW in the 0–5000 m column and in near surface inversions. Strength from the surface to the top of the highest inversion (gm m−3). Monthly peak PW in WV intrusions
Median PW (mm) in 10 000 m
% WV in 5000 m
% WV in inversions
Top of highest inversion (m)
Strength from surface to top of highest inversion (10−3 kg m−3)
Peak PW in WV intrusion(mm)
11/08 and 09
Winter and summer WV intrusions are illustrated by the time series plots for January and July 2008 (Figure 4). Peak PW in WV intrusions ranged from 3.4 mm for November to 37.3 mm for August (Table I). WV intrusions were found to measurably enhance the total column PW relative to the median total column PW (Table I). The seasonality of the WV intrusions over the southeastern Beaufort Sea-Amundsen Gulf region was consistent with that found by Doyle et al. (2011) for Eureka.
Plots of the monthly profiles of the median, 25 and 75 percentile WV density values at each microwave radiometer level revealed the presence of near surface WV inversions in the winter and early spring (January–April), and in the late fall (November and December, Table I). Winter and summer profiles are illustrated by the plots for January and July 2008 (Figure 5). When WV inversions were present, there were often two or three with the top of the highest inversion between 150 and 1400 m (Table I). The strength of the inversions, defined as the difference between the WV density at the top of the highest inversion and at the surface was 0.08 to 0.24 × 10−3 kg m−3 (Table I). Devasthale et al. (2011) also found multiple near surface WV inversions; however, the strengths of their inversions were generally greater than the inversion strengths found in our observations. This difference is likely due to differences in approach. Devasthale et al. (2011) calculated the strength of the strongest inversion; we calculated the inversion strength from the surface to the top of the highest inversion.
The PW in near surface WV inversions accounted for up to 41% of the median total column PW in winter when the median PW was 1–3 mm. In summer when the median PW was three to eight times larger, inversions were not evident in the median WV profiles (Table I). A plot of the percentage of PW in the near surface inversions versus the monthly median total column PW, with the percentage set to zero when inversions were absent, gave an L-shaped relationship (Figure 6). A similar relationship was found by Devasthale et al. (2011) who compare the WV in the strongest near surface inversion to the mean total column PW. This L-shaped relationship suggested that, over the seasonally varying sea-ice cover of the western maritime Arctic like elsewhere in the Arctic, near surface WV inversions are a common occurrence in winter, early spring and late autumn when the atmosphere is relatively cold and the median amount of PW is small. These inversions generally contain a large fraction of the total column PW. WV inversions are much less common in the warmer months when the total column PW is relatively large.
The hourly WV density profiles for January, a month with a very low median PW value (1.4 mm), and for July a month with a relatively high median PW value (14.5 mm) were examined to provide further insight into the L-shaped relationship between the monthly median total column PW and the percentage of the PW in the near surface inversions. WV density maxima in the lowest 2000 m of each profile were defined as tops of near surface WV inversions. That is, levels whose WV density was greater than the WV values of the two levels below and the two levels above where considered to be the tops of inversions (Uttal et al., 2002; Raddatz et al., 2011). Inversions with tops ≥ 1000 m were only classified as near surface inversions if they were the highest inversion in a series of inversions extending upward from the surface. One to three inversions were present in 90% of January's hourly profiles, but only in 24 % of July's profiles. This confirmed that WV inversions are common in winter, but occur much less frequently in summer. A plot of the percentage of the total column PW from the surface to the top of the highest inversion in each hourly profile, h, for January and July (Figure 7) revealed a grouping of data that suggested a family of curves. Each curve showed a decrease in the percentage of the total PW in near surface inversions as the total column PW increased. Distinct steps in h suggested that the percentage of the total PW in near surface inversions was related to the height of the highest inversion top. This hypothesis was tested and found to be true. A plot of hours with one or more near surface inversions revealed that the height of the top of the highest inversion explained about 90% of the variability in the percentage of PW in the inversions (Figure 8). Furthermore, the distribution of h differed between January and July (Table II). For January, over 50% of the hourly profiles had h values greater than 1200 m, while the top of the highest inversion extended to 1200 m for only about 8% of July's hourly profiles. In contrast, July's h values were less than or equal to 200 m for about 13% of the hourly profiles, while in January, h≤200 m for only 2% of the hourly profiles. Thus, the low PW month, January, had many more hours with one to three near surface inversions than the high PW month, July. In addition, in the low PW month the height of the top of the highest inversion was generally much higher than in the high PW month. Deeper and more frequent near surface WV inversions in winter than in summer is consistent with the results of others (Tjernstrom et al., 2004; Devasthale et al., 2011). These results support the validity of an L-shaped relationship between monthly median total column PW and the percentage of the PW in near surface inversions.
Table II. Heights of the tops of highest inversions (m) in the hourly WV profiles for January and July 2008 with WV expressed as a percentage of the total column PW
Highest inversion (h) top (m)
200 < h≤400
400 < h≤600
600 < h≤800
800 < h≤1000
1000 < h≤1200
1200 < h≤1400
h > 1400
A composite year of microwave radiometric hourly WV density profiles from November 2007 to July 2008 and August to November 2009 for the western maritime Arctic, a region with seasonally varying surface sea-ice cover, was analysed. The annual cycle of total column (0–10 000 m) PW was consistent with that obtained from a comprehensive set of Arctic radiosondes by Serreze et al. (1995). The monthly median PW values ranged from 1.4 mm for January to 16.8 mm for August with greater than 92% of the PW below 5000 m in all months.
Peak PW in WV intrusions ranged from 3.4 mm for November to 37.3 mm for August. These values are consistent with those of Doyle et al. (2011) for Eureka. The measurable increase in total column PW found in WV intrusions over the southeastern Beaufort Sea-Amundsen Gulf region suggested that WV intrusions can have an impact on the magnitude and variability of atmospheric downwelling longwave radiation. An L-shaped relationship was found between the monthly percentages of PW in near surface WV inversions and the median total column PW. This relationship is consistent with the pan-Arctic results reported by Devasthale et al. (2011). Near surface inversions were found in profiles of the median WV values for the winter and early spring months (January–April), and in the late autumn (November and December). PW in near surface WV inversions accounted for up to 41% of the monthly median total column PW in winter when the median PW was 1–3 mm. In the warmer months, when the median PW was three to eight times larger, inversions were not evident in the median WV profiles. This suggests that WV inversions may have a greater impact on the magnitude of atmospheric downwelling longwave radiation in the winter, early spring and late autumn, when the median amount of PW is low, than in summer when it is relatively high. This hypothesis was supported by an analysis of the hourly profiles for January (a low PW month), and July (a high PW month); 89% of the variability in the percentage of PW in near surface inversions was explained by the height of the highest inversion top which was generally much higher in January than in July. In addition, near surface WV inversions occurred with a much higher frequency in the low PW month than in the high PW month. Thus, atmospheric WV plays an important role in the cold season climate of the western maritime Arctic as downwelling longwave radiation is sensitive to both WV intrusions and to near surface WV inversions.
Near surface WV inversions also create a downward WV flux. Thus, WV inversions may also help to maintain the boundary layer at or near saturation with respect to ice. This may contribute to the climatologically high fractional cloud cover in the Arctic as near surface WV inversions are a common cold season occurrence (Curry, 1983; Andreas et al., 2002; Persson et al., 2002; Tjernstrom et al., 2004).
Due to the dependence of saturation vapour pressure on temperature, an increase in atmospheric moisture is expected to accompany future warming caused by increasing atmospheric concentrations of CO2. The greenhouse effect of increased atmospheric WV will further enhance climate warming (Bony et al., 2006). Changes in the atmospheric moisture budget are also likely to be accompanied by altered cloud cover which will further impact the radiation budget (Curry et al., 1996).
This study has characterization of the total column PW, WV intrusions, and near surface WV inversions for a composite year over the southeastern Beaufort Sea, Amundsen Gulf region. This information will expand the knowledge of atmospheric thermodynamics over the western maritime Arctic, and may contribute to improved climate modelling in this data sparse region.