Moisture flux changes and trends for the entire Arctic in2003–2011 derived from EOS Aqua data

Linette N. Boisvert,1 Thorsten Markus,2 and Timo Vihma3

Received 22 April 2013; revised 29 August 2013; accepted 24 September 2013.

[1] The Arctic sea ice acts as a barrier between the ocean and lower atmosphere, reducingthe exchange of heat and moisture. In recent years the ice pack has undergone manychanges, in particular a rapid reduction in sea ice extent and compactness in summer andautumn. This, along with modeling studies, would cause one to believe that the moistureflux would be increasing. We estimate the daily moisture flux from 2003 to 2011 usinggeophysical data from multiple sensors onboard NASA’s Aqua satellite, taking advantageof observations being collected at the same time and along the same track. Our findingsshow the moisture flux, averaged over the entire Arctic, has had large interannualvariations, with smallest fluxes in 2010, 2003, and 2004, and largest ones in 2007, 2008, and2005. Increases in air specific humidity tend to reduce the moisture flux, whereas thedecrease in sea ice cover tends to increase the flux. Statistically significant seasonaldecreasing trends are seen in December, January, and February because of the dominatingeffect of increase in 2 m air specific humidity increasing, reducing the surface-air specifichumidity difference by �0.0547 kg/kg in the Kara/Barents Seas, E. Greenland Sea, andBaffin Bay regions where there is some open water year round. Our results also show thatthe contribution of the sea ice zone to the total moisture flux (from the open ocean and seaice zone) has increased by 3.6% because the amount of open water within the sea ice zonehas increased by 4.3%.

Citation: Boisvert, L. N., T. Markus, and T. Vihma (2013), Moisture flux changes and trends for the entire Arctic in 2003–2011derived from EOS Aqua data, J. Geophys. Res. Oceans, 118, doi:10.1002/jgrc.20414.

1. Introduction

[2] The compact Arctic sea ice is a very efficient insula-tor between the surface and the lower atmosphere becauseit prevents heat and moisture exchanges from occurringbetween the relatively warm ocean surface and the cold airabove. But in recent years, the Arctic sea ice pack hasundergone many changes. Its areal extent has decreased ata rate of 4.1%/decade [Cavalieri and Parkinson, 2012], ithas become much thinner, losing 47% of the ice volumeduring the autumn since 2003 [Laxon et al., 2013], and ithas become predominantly first-year ice [Polyakov et al.,2012]. These changes cause the ice pack to become a lessefficient insulator and could cause increased exchanges ofheat and moisture between the ocean surface and atmos-phere. This is because the moisture flux (i.e., evaporation)

over the open water can be tens of times larger than over thesolid ice pack [Launiainen and Vihma, 1994]. Recent studiesshow increases in the amount of water vapor in the Arctic tro-posphere, though time periods, data, and methods vary [Deeet al., 2011; Screen and Simmonds, 2010a; Rinke et al.,2009; Serreze et al., 2012], and it is not clear if theseincreases are due to an increase in the evaporation in the Arc-tic or an increase in the atmospheric transport of moisturefrom lower latitudes to the Arctic. Increases in the air relativehumidity can warm up the lower atmosphere via the releaseof condensation heat in cloud formation and an increase in theair specific humidity causes warming because water vapor isa greenhouse gas. Further, low-level clouds effectively trapoutgoing long-wave radiation, increasing the net radiation atthe ice surface [Walsh and Chapman, 1998]. Thus, excessmoisture can increase ice ablation and will cause the ice packto be more vulnerable in the following years.

[3] As in our previous work, the moisture flux is definedas the vertical flux of water vapor due to atmospheric turbu-lent transport [Boisvert et al., 2012]. It is affected by thedifference between the saturation specific humidity corre-sponding to the surface temperature of the ocean or sea iceand the air specific humidity close to the surface, as well asby three factors affecting the intensity of the turbulentexchange: wind speed, surface roughness, and thermalstratification [Launiainen and Vihma, 1994].

1Department of Atmospheric and Oceanic Sciences, University ofMaryland, College Park, Maryland, USA.

2NASA Goddard Space Flight Center, Greenbelt, Maryland, USA.3Finnish Meteorological Institute, Helsinki, Finland.

Corresponding author: L. N. Boisvert, Department of Atmospheric andOceanic Sciences, University of Maryland, College Park, MD 20742,USA. (linette.n.boisvert@nasa.gov)

©2013. American Geophysical Union. All Rights Reserved.2169-9275/13/10.1002/jgrc.20414

1

JOURNAL OF GEOPHYSICAL RESEARCH: OCEANS, VOL. 118, 1–15, doi:10.1002/jgrc.20414, 2013

[4] Large-scale moisture fluxes are available in variousatmospheric model reanalysis, but these are known to havediscrepancies in the moisture variables [Cullather et al.,2000; Jakobson and Vihma, 2010; L€upkes et al., 2010;Jakobson et al., 2012]. Using satellite data on air moisturein order to calculate the moisture flux would allow an alter-native approach for large-scale moisture flux estimatesover the Arctic.

[5] In our previous work, Boisvert et al. [2012] com-puted the moisture flux for a series of polynya events fromthe North Water Polynya in order to test the accuracy ofthe air moisture observed from the NASA’s Earth Observ-ing System (EOS) Aqua satellite. They found that instru-ments onboard Aqua yielded sufficiently accurate data onnear-surface air moisture, allowing reliable calculation ofthe moisture flux. Aqua was chosen because it has multipleinstruments that collect a wide variety of geophysical datasimultaneously and along the same track. This is a veryuseful feature of Aqua, but it is often under-utilized instudying the Arctic.

[6] We have taken full advantage of this feature of Aquain order to calculate and study recent changes in the mois-ture flux over the entire Arctic from 2003 to 2011.

2. Data

[7] Data from NASA’s EOS Aqua satellite are used toestimate the daily moisture flux for the entire Arctic from 1January 2003 to 30 September 2011. Aqua was launchedon 4 May 2002 and continues to operate. Aqua carries sixEarth-observing instruments that collect a wide variety ofglobal data [Parkinson, 2003]. It has a near-polar low-Earthorbit with a period of 98.8 min and equatorial crossingtimes of 01:30 and 13:30 LT. Data used came fromNASA’s Atmospheric Infrared Sounder (AIRS) and the Ja-pan Aerospace Exploration Agency (JAXA)’s AdvancedMicrowave Scanning Radiometer for EOS (AMSR-E).

[8] AIRS is a cross-track scanner collecting data with a13.5 km spatial resolution in the horizontal and a 1 km re-solution in the vertical. It has 2378 infrared channels andfour visible/near infrared channels, which obtain highlyaccurate temperature and humidity profiles and many otherphysical products dealing with the Earth and its atmos-phere. From AIRS we use surface skin temperature of thesea ice, air temperature at 1000 hPa pressure level, and rel-ative humidity at 1000 hPa. We use geopotential heightsfrom AIRS in order to determine the actual heights of the1000 hPa level. All standard temperature and relative hu-midity products are level quantities, meaning that the val-ues are reported at fixed pressure levels [Olsen, 2011].These values are used in the calculation of the moistureflux and are Level 3 mean daily gridded products coveringa 24 h period for the ascending (equatorial crossing southto north at 13:30 LT) portion of the orbit [Aumann et al.,2003]. Both the temperature and relative humidity productsare level quantities, which mean that the values arereported at fixed pressure levels. This differs from layerquantities, which are also reported on the fixed pressurelevels, but represent the layer bounded by the level onwhich they are reported and the next height level (in alti-tude). Thus, the 1000 hPa temperatures and relative humid-ity are not vertical averages. In the Polar Regions, Aqua

makes multiple passes over the Arctic each day, allowingfor daily averages to be produced. These parameters aremapped onto a 1� � 1� global grid.

[9] AMRE-E is a conically scanning global passivemicrowave radiometer that measures microwaves in boththe horizontal and vertical directions and covers all of thechannels of scanning multi-channel microwave radiometer(SMMR) and special sensor microwave imager (SSM/I)combined, also having a higher resolution. It has 12 chan-nels, with horizontal and vertical polarization for each ofsix frequencies, and a spatial resolution ranging from 5.4 to56 km depending on frequency. On 4 October 2011AMSR-E stopped functioning after nine plus years of suc-cessful operations. Due to this malfunction, we stoppeddaily moisture flux estimates on 30 September 2011. Thesea ice concentrations are produced by the NASA Team(NT2) algorithm [Markus and Cavalieri, 2000]. It isdefined as the percentage of a pixel that is covered by seaice, and the concentration values are mapped onto a 25 kmby 25 km polar stereographic grid of the Arctic. AMSR-Eis also used to produce sea surface temperatures in the Arc-tic. These sea surface temperatures (SSTs) are defined as thetemperature of the top layer of water approximately 1 mmthick and are derived from brightness temperature data[Wentz and Meissner, 2000]. The SSTs are only valid in ice-free areas and further than 75 km off coastlines. The freezingpoint of seawater (�1.8�C) is used where the AMSR-ESSTs are not provided.

[10] Since none of the instruments onboard Aqua pro-duce accurate wind speeds, 10 m wind speeds wereobtained from ECMWF ERA-Interim reanalysis (http://data-portal.ecmwf.int/data/d/interim_daily/). The reanaly-sis combines a first-guess field (based on a 6 h forecast) aswell as in situ and remote sensing data into an assimilateddata set using the 4D-VAR method [Dee et al., 2011].Wind speed data are provided at 6 h time intervals with a0.73 by 0.73� spatial resolution.

[11] These data sets were all transposed onto a 25 km by25 km polar stereographic grid in order to simplify the cal-culations of the moisture flux. Calculations at 25 km resolu-tion indeed require interpolation of Aqua and wind speeddata. A horizontal resolution equal to that of the sea iceconcentration data is, however, essential for the moistureflux calculations, because the spatial variations of the mois-ture flux are mostly controlled by spatial variations of thesurface temperature (which depends above all on the stateof the surface: sea ice or open water). The air moisture andwind speed have weaker spatial gradients [Tisler et al.,2008].

3. Methodology

[12] The Monin-Obukhov similarity theory, which char-acterizes the vertical behavior of nondimensionalized meanflow and the turbulence properties in the surface layer ofthe atmosphere [Monin and Obukhov, 1954], is used to esti-mate the turbulent surface fluxes in the atmospheric bound-ary layer. An iterative method along with the Monin-Obukhov similarity theory [see Launiainen and Vihma,1990, for a complete description] allows for use of the airtemperature and relative humidity at 1000 hPa from AIRSto estimate the surface moisture flux.

BOISVERT ET AL.: MOISTURE FLUXES IN THE ARCTIC: 2003–2011

2

[13] Application of the theory also requires parameter-izations based on observations, and numerous studies havebeen carried out to improve the parameterizations of theroughness lengths [e.g., Dyer and Bradley, 1982;Högström, 1988; Garratt, 1992; Andreas, 2002] and thestability effects on the flux-profile relationships [e.g., Hol-stag and De Bruin, 1988; Beljaars and Holtslag, 1991;Dyer and Hicks, 1970 ; Paulson, 1970]. While these studieshave produced more accurate model results for turbulentsurface fluxes in low and midlatitudes, major challengesremain in high latitudes [Tjernström et al., 2005; Tastulaand Vihma, 2011].

[14] The flux-profile relationships depend on whether theboundary layer is stably or unstably stratified. For unstablestratification, the flux-profile relationships proposed byDyer and Hicks [1970], Paulson [1970], and Businger et al.[1971] are among the most widely used. For stable stratifi-cation, most of the flux-profile relationships are based onstudies of the nocturnal boundary layer over land; those byHoltslag and de Bruin [1988] are commonly used. Studiesby Forrer and Rotach [1997], Klipp and Mahrt [2004], andCheng and Brutsaert [2005] suggest that in very stable con-ditions, like those seen frequently over the Arctic sea ice,the turbulent fluxes decrease with increasing stratificationmore rapidly than according to the formulae by Holslagand De Bruin [1988]. Extensive in situ measurements weremade over the Arctic sea ice during the Surface HeatBudget of the Arctic Ocean Project (SHEBA) campaign in1997–1998, and Grachev et al. [2007] used these to derivehighly accurate flux-profile relationships for stable condi-tions over sea ice. Algorithms from Grachev et al. [2007]better fit the very stable boundary layer conditions in theArctic and we use these in our calculations.

[15] Accurate roughness lengths for the wind speed, hu-midity, and temperature profiles over the ice are required toestimate the transfer coefficients used to calculate the mois-ture flux. Andreas et al. [2010a, 2010b] used observationsfrom the SHEBA campaign to derive an algorithm for theroughness length over the Arctic in the winter when the iceis covered with compact, dry snow and in the summerwhen the sea surface is a mixture of leads and sea ice, thelatter being often covered by wet snow and melt ponds. Asthese are among the most accurate estimates made for theArctic sea ice in different seasons, these new roughnesslengths will be used in our model (an alternative wouldhave been L€upkes et al. [2012]).

[16] Since the air temperature and relative humidityobservations originate from a variable height (the 1000 hPapressure level), which differs from the height of the windspeed data (10 m) based on ERA-Interim, we have adoptedand further improved the method of Launiainen and Vihma[1990]. This bulk-aerodynamic method utilizes the Monin-Obukhov similarity theory and includes an iterative calcu-lation, which allows for variables to have different observa-tion heights above the surface. Taking into account thestability and roughness effects on the vertical profiles, theair temperature, relative humidity, and wind speed are allstratified onto a reference height where the calculations aremade.

[17] Our updated algorithm from Launiainen and Vihma[1990] includes the following changes, which improve theaccuracy of the moisture flux calculations in the Arctic.

When calculating the moisture flux, E, Andreas et al.[2010a] use a combination of E from the ice surface andthe fraction of melt ponds and leads in the ice pack.

E ¼ �CEzSr ciqs;i þ 1� cið Þqs;w

� �� qz

� �ð1Þ

where � is the air density, CEz is the bulk transfer coeffi-cient of evaporation, Sr is the effective wind speed, ci is theice concentration, qs,i is the specific humidity at the surfaceof the ice (i.e., the saturation specific humidity for the sur-face temperature), qs,w is (analogously) the specific humid-ity at the surface of the ocean, and qz is the air specifichumidity at 2 m. CEz in (1) is defined by the roughnesslengths and stability corrections for stable and unstableconditions:

CEz ¼1

ln z=z0ð Þ ��M z=Lð Þ½ � ln z=zq

� ���E z=Lð Þ

� � ð2Þ

where z is the measuring height, z0, zT, and zq are the rough-ness lengths for the wind speed, temperature, and watervapor. The Obukhov length, L, is a parameter characteriz-ing the dynamic, thermal, and buoyant processes of the sur-face layer [Obukhov, 1971]. �M, �H, and �E are theintegrated universal functions of momentum, sensible heat,and moisture, respectively, which depend on z and L, andquantify the effects of the surface layer stratification on theprofile gradients [Launiainen and Vihma, 1990].

[18] Parameterizations for the stable boundary layer caseof the integrated universal functions are taken from Gra-chev et al. [2007]

�M ¼ �3aM

bMx� 1ð Þ

þ aM BM

2bM

2lnxþ BM

1þ BM

� �� ln

x2 � xBM þ B2M

1� BM þ B2M

� �

þ2ffiffiffi3p

arctan2z� BMffiffiffi

3p

BM

� �� arctan

2� BMffiffiffi3p

BM

� �� �26664

37775ð3Þ

where x ¼ 1þ �ð Þ1=3, �¼ z/L, BM ¼ 1�bM

bM

1=3> 0,

aM¼ 5, and bM¼ 5/6.5 and

�H ¼ �E ¼ �bE

2ln 1þ cE� þ �2� �

þ � aE

BEþ bEcE

2BE

� �

� ln2� þ cE � BE

2� þ cE þ BE

� �� ln

cE � BE

cE þ BE

� �� � ð4Þ

where BE ¼ffiffiffiffiffiffiffiffiffiffiffiffiffic2

E � 4p

¼ffiffiffi5p

, aE¼ 5, bE¼ 5, and cE¼ 3.These are only valid if RiB<RiBcrit � 0.2, where RiB is thebulk Richardson number, otherwise the moisture flux esti-mate is discarded and the mean is calculated from a smallersample.

[19] In the summer, the sea ice is covered with meltponds and leads that enhance the turbulent momentumtransfer at the surface [Andreas et al., 2010a]. The edges ofthe ice floes create vertical surfaces that the winds comeinto contact with Andreas et al. [2010a] found CD to be afunction of the ice concentration (ci).

ln z0ð Þ ¼ ln zð Þ � kC1=2D ð5Þ

BOISVERT ET AL.: MOISTURE FLUXES IN THE ARCTIC: 2003–2011

3

where CD � 10�3 ¼ 1:5þ 2:233ci� 2:333ci2 in equa-tion (5).

[20] In the winter, the sea ice pack is compact and cov-ered with snow and thus requires a different parameteriza-tion for z0 than in the summer [Andreas et al., 2010b].

z0 ¼ 0:135v

u�þ 2:30� 10�4tan h3 13u�ð Þ ð6Þ

[21] The roughness length accounts for gustiness and pla-teaus with increasing u� [Andreas et al., 2010b]. z0 was calcu-lated from the SHEBA data over the Arctic ice in the winter,thus being more accurate than previous parameterizations.

[22] Andreas et al. [2005] requires different polynomialcoefficients for zT and zq, which are calculated on the basisof the roughness Reynolds number. This is because the mo-lecular diffusivity of heat in air is less than the moleculardiffusivity of water vapor in air [Andreas et al., 2005].

[23] The effective wind speed, Sr, includes a parameterfor gustiness. In stable conditions

Sr ¼ Vz þ 0:5sech Vzð Þ ð7Þ

where Vz is the wind speed at the reference height. Forunstable conditions, Sr includes a term for enhanced turbu-lent exchange:

Sr ¼ V 2z þ �2

gw2�

1=2ð8Þ

where w� ¼ u� � 600kL

� �1=3, is the convective velocity scale

and �g¼ 1.25 [Fairall et al., 1996]. By using the effectivewind speed, we prevent the transfer coefficients fromapproaching infinity when the wind speed approaches zero[Andreas et al., 2010b].

[24] The moisture flux was calculated over the study area(Figure 1) using this updated algorithm for each day from 1January 2003 to 30 September 2011.

4. Results

[25] In Figure 2a, we present the moisture flux per monthintegrated over the entire Arctic for 2003–2011 (the resultsfrom equation (1) are multiplied by the number of secondsin a month, by the pixel area of 625 km2, and by the num-ber of pixels. The area around the North Pole, where nodata are available, is excluded). Figure 2a also shows sepa-rately the integrated moisture flux from the open ocean (0–15% ice concentration) and from the sea ice zone (15–100% ice concentration). The total moisture flux (blackline) and the moisture flux from the ocean (red line) followan annual cycle, but the moisture flux over the ice (blueline) does not. The amount of moisture exchanged betweenthe surface and atmosphere is lowest in July because this isthe time of the year when the air temperature and specifichumidity are often close in magnitude, the average temper-ature, and specific humidity differences being 3.9 K and0.013 kg/kg, respectively. When this occurs, there is littleexchange of moisture. The amount of moisture exchangedin July is on average only 17.2% of what is exchanged inOctober (October is the tallest peak each year in Figure 2aand July is the lowest).

[26] The total moisture fluxes are highest in October,with the majority of the moisture coming from the areas ofopen water (see red line in Figure 2a). The reason for this isthat in October there is still a large area of open water, spe-cifically in the Chukchi/Beaufort and Laptev/E. SiberianSeas (regions 9 and 10 in Figure 1), which is warmer thanthe surface of the solid ice pack. In October, the air temper-atures begin to rapidly decrease, whereas the SSTs drop offmuch more slowly due to the large heat capacity of theocean, creating temperature and specific humidity differen-ces between the sea surface and air of around 7 K and0.080 kg/kg and substantial moisture fluxes. But this trendis starting to shift toward warmer air temperatures extend-ing into October, possibly due to larger areas of open waterin the summer and warmer SSTs creating warmer air tem-peratures. In 2010, the air temperatures were larger thanthe surface values in an area extending from the E. SiberianSea to the north coast of Greenland and the magnitude ofthe temperature differences are decreased (Figure 3,bottom).

[27] These moisture fluxes remain elevated in the wintermonths because of the vertical differences in specific hu-midity that arise due to areas of leads and polynyas andlarger areas of open water. They drop off rapidly in thespring and are the lowest in the summer months when theair is warmer than the surface and the air specific humidityis only slightly lower than that of the surface.

[28] The amount of moisture put into the lower atmos-phere in the Entire Arctic each year is decreasing 1.9%/yrbetween 2003 and 2010, but it is not statistically significant(Figure 4). There are changes seen on a regional level, forinstance, in the Chukchi/Beaufort, Laptev/E. Siberian Seas,Canadian Archipelago, and the Central Arctic there hasbeen 35.97, 20.6, 19.15, and 30.31 kg/m2 of excess wateradded to the lower atmosphere between 2003 and 2010,

Figure 1. Arctic study area. Entire Arctic refers toregions 4–10. (1) Sea of Okhotsk, (2) Bering Sea, (3) Hud-son Bay, (4) Baffin Bay, (5) East Greenland Sea, (6) Kara/Barents Seas, (7) Central Arctic, (8) Canadian Archipelago,(9) Laptev/East Siberian Seas, (10) Chukchi/Beaufort Seas.

BOISVERT ET AL.: MOISTURE FLUXES IN THE ARCTIC: 2003–2011

4

respectively. Conversely, regions like the Kara/BarentsSeas, East Greenland Sea, and Baffin Bay are not releasingmoisture as much as in the previous years, losing 20.6, 2.8,and 0.5 kg/m2, respectively, between 2003 and 2010. Fig-ure 4 shows the anomalies of yearly moisture flux, inte-grated over pixels 625 km2 large, for 2003–2010. Figure 4shows that in years 2003, 2004, 2009, and 2010 the mois-ture flux anomalies were negative over most of the Arctic,whereas in years 2005, 2006, 2007, and 2008 they werepositive. Since 2007, the regions of Baffin Bay, E. Green-land, and Kara/Barents Seas have changed from a positiveto a negative anomaly. Reasons behind these changes arediscussed in the following section.

4.1. Monthly Trends

[29] Looking at how the aerially integrated monthlymoisture flux changes on a yearly basis, one sees a slightlydifferent story. A linear regression was applied to themonthly total moisture flux and average ice concentration(15–100% ice) for each month from 2003 to 2011. For Oc-tober–December, the data records are only for 2003–2010.A statistical t test was then performed on each linear regres-sion to determine whether or not the trend seen in the datawas due to actual changes in the variables as opposed tojust random chance. Table 1 shows the moisture flux trendsfor each month. The total moisture flux, summed over theocean and sea ice, has decreased at the 95% confidence

level in January (by 8.1%/yr), February (5.5%/yr), and De-cember (7.1%/yr) (see Figure 2b for January). During thesemonths the moisture flux is driven by areas where there isopen water, specifically the Kara/Barents Seas, E. Green-land Sea, and Baffin Bay regions. In these regions, the 2 m-air temperature has increased by 6.0 K and the specific hu-midity difference between the surface and the air hasdecreased by 0.055 kg/kg between 2003 and 2011, whereasthe SSTs have increased less, only 1.3 K (black line inFigure 5).

[30] Figure 5 shows many interesting anomalies betweenthe 2 m air and surface temperatures and specific humid-ities, which warrant more discussion. On average, insummer the air is much warmer than the sea surface, butthe surface specific humidity is still larger than the air spe-cific humidity in the Kara/Barents Seas, E. Greenland Sea,and Baffin Bay regions (black line), whereas in the fall andwinter months the open sea surface is much warmer thanthe air and the specific humidity at the surface is muchlarger than at 2 m. However, there has been a major anom-aly during this study period in the winter of 2010–2011,when the temperature differences in these regions werevery small compared to the other years. In the other regionsof the Arctic (red line), the surface and air temperatures areclose to each other in summer, because both are close tothe freezing point. In winter, when regions 7–10 (Figure 1and red line in Figure 5) are covered by sea ice, the surface

Figure 2. (a) The total moisture flux for the entire Arctic for January 2003 until September 2011. Theblue line is the moisture flux from the solid sea ice pack (15–100% ice concentration), the red line is themoisture flux from the ocean (10–15% ice concentration), and the black line is the moisture from boththe sea ice pack and the ocean. Gray vertical lines separate the different years. (b) January (black line)and August (red line) moisture flux from 2003 to 2011. Their trend lines have corresponding colors.

BOISVERT ET AL.: MOISTURE FLUXES IN THE ARCTIC: 2003–2011

5

is often much colder than the air above, due to the negativeradiation balance and generation of a surface-based temper-ature inversion [e.g., Serreze et al., 1992]. Recently therehave, however, been large interannual variations. Thesemight be related to changes in the heat advection, cloudcover, and ice concentration [e.g., Stroeve et al., 2012] butthese are out of the scope of our paper.

[31] The large increase in air temperature has beenaccompanied by a decrease in the specific humidity differ-ence between the surface and air in January (Figure 6, ex.2010 and 2011), which drastically decreases the moisture

flux. The temperature and specific humidity differences inthe winter over the ice pack have also decreased over thetime period, and in 2010 and 2011 they were on average�2.0 K and 0.002 kg/kg. Since the monthly mean air spe-cific humidity is practically the same as the ice surface spe-cific humidity, there is very little exchange of moistureoccurring. The positive moisture flux anomalies increasedfrom 2003 to 2007 and these positive anomalies changed tomore negative anomalies from 2007 to 2011 (Figure 6).The same was seen in February and December, but onlyJanuary data are shown here (Figure 6).

Figure 3. (top) The 2 m specific humidity and surface specific humidity (kg/kg) differences for themonth of October for each year 2003–2010 for the entire Arctic. (Bottom) The 2 m air temperature andsurface temperature (K) differences for the month October for each year 2003–2020 for the entire Arctic.The gray is either the land or no data in both images.

BOISVERT ET AL.: MOISTURE FLUXES IN THE ARCTIC: 2003–2011

6

[32] A decrease in the moisture flux, like that seen in2010 over the North Atlantic regions (regions 4–6 in Figure1), can be explained by either one or a combination of thesethree factors: (1) anomalously low SSTs, (2) anomalouslyhigh air specific humidity, and (3) anomalously low winds.

[33] SST anomalies from AMSR-E in the East GreenlandSea (region 5 in Figure 1) are shown in Figure 7. From thisfigure, it is clear that SSTs are much colder in 2010 and2011 than in previous years. The largest negative anomaliesoccur in the winter months and are around �1�C. This fig-ure strongly supports factor 1 and can explain why mois-ture fluxes in 2010 and 2011 are much lower than inprevious years because of low SSTs in the North Atlanticregions of the Arctic.

[34] Figure 8 shows 1000 hPa specific humidityanomalies for January 2010, October 2010, and January–

December 2010 from the 1981 to 2010 climatology as wellas the evaporation rate anomalies. All of these images wereproduced from ERA-Interim reanalysis data using theWeb-based Reanalysis Intercomparison Tools (WRIT)maps (https://reanalyses.org/atmosphere/web-based-reanal-ysis-intercomparison-tools-writ). There are positive 1000hPa specific humidity anomalies in both January and Octo-ber 2010 (Figures 8c and 8d) around the southern tip ofGreenland and into Baffin Bay. Positive anomalies are alsoseen in the Kara/Barents Seas. There are also 5�C 2 m tem-perature anomalies seen in these same regions (not shown).Looking at the 1000 hPa specific humidity anomalies forthe entire year in 2010, again ERA-Interim produces largenegative anomalies in East Greenland Sea and Baffin Bay(Figure 8f), which again supports smaller moisture fluxes.Figure 8a shows the evaporation rate anomaly produced byERA-Interim for January 2010. Here we see negative evap-oration rates in the East Greenland Sea, which correspondswith our data. This is also seen for the entire year evapora-tion rates (Figure 8c).

[35] We also looked into the observed differencesbetween the sea surface and 2 m air specific humidity in2003–2011. Monthly mean 2 m specific humidities werecalculated from air temperature and relative humidityobservations at weather stations in Longyear airport, Sval-bard (http://eklima.met.no) and Keflavik airport, Iceland(http://en.vedur.is), and the sea surface saturation specifichumidities were calculated from AMSR-E SST data fromthe sea pixels nearest to the weather stations. At the Sval-bard station (Figure 9a), it is evident that the specific hu-midity differences in 2010–2011 are smaller than inprevious years creating a decrease in the moisture flux dur-ing these times. Figure 8b shows the same for Iceland,although this decrease is not as evident. It is important tokeep in mind that the atmospheric observations were takenfrom coastal weather stations, as opposed to our AIRS-

Table 1. Trends in Monthly Moisture Flux From the Entire ArcticDuring 2003–2011a

MonthTrend (kg/yr)

(All)Trend (kg/yr)

(Ice)Trend (kg/yr)

(Ocean)

Jan. 24.04 3 1013 �2.46 � 1012 23.79 3 1013

Feb. 22.40 3 1013 �1.53 � 1012 22.25 3 1013

Mar. 8.31 � 1012 2.88 � 1012 5.44 � 1012

Apr. 7.08 � 1012 2.37 � 1012 4.71 � 1012

May �2.57 � 1012 1.32 � 1012 �3.91 � 1012

Jun. 4.82 � 1011 2.23 � 1012 �1.74 � 1012

Jul. �1.59 � 1012 1.12 � 1012 22.71 3 1012

Aug. 7.40 � 1012 7.47 � 1011 6.65 � 1012

Sep. 4.43 � 1011 �7.32 � 1011 1.18 � 1012

Oct. �1.68 � 1013 �3.04 � 1012 �1.37 � 1013

Nov. 6.58 � 1012 3.11 � 1012 3.97 � 1012

Dec. 23.94 3 1013 �4.11 � 1012 23.53 3 1013

aNumbers highlighted in bold are statistically significant in the 95%confidence level.

Figure 4. The differences between the yearly moisture flux and the average moisture flux from 2003 to2010. Positive anomalies are shown in pinks, negative in blues. The gray is either the land or no data.

BOISVERT ET AL.: MOISTURE FLUXES IN THE ARCTIC: 2003–2011

7

based estimates being made over the ocean where specifichumidities could be different.

[36] Finally, we looked at wind speeds over the EastGreenland Sea using ERA-Interim wind speed data and theKeflavik station data to see if there have been any signifi-cant changes in their magnitudes. The average monthlywind speeds in both the East Greenland Sea and Icelandfollow annual cycles with the strongest winds in the wintermonths, but their magnitudes show no change in 2010 or2011 from the earlier years (not shown).

[37] By looking at these three factors that could affectthe magnitude of the moisture flux in more detail, we wereable to determine that in 2010 the SSTs were much coolerthan normal, the 1000 hPa specific humidity was higherthan normal, and there were no changes in the wind speed.The lower SSTs in this year as well as higher specifichumidities created smaller specific humidity differencesbetween the surface and the air than seen in previous years.These two changes worked together to decrease the magni-tude of the moisture flux in 2010, supporting our findingsbased on AIRS data.

[38] The moisture flux is increasing in March–Septem-ber, although none of these increases are statisticallysignificant. On average, the total moisture flux duringthese months is small, accounting for 32% of the totalmoisture flux in October. In these spring and summer

months, the ice concentration is also decreasing at the95% confidence level. In spring (April, May, and June),the ice pack is losing 0.4% ice concentration per year,meaning that the ice pack is beginning to break up ear-lier exposing some of the ocean to the atmosphere.When this occurs, the SSTs are very close to the freez-ing point of seawater and in the spring this is muchwarmer than the overlying air. This is enough to causethe moisture flux to increase because the moisture fluxover the solid ice pack is so small. However, theincreasing trends in the March–September moisture fluxare not significant, demonstrating that interannual vari-ability in the moisture flux does not only rely onchanges in the sea ice pack; changes in humidity, windspeed, and air temperatures are other essential factors.

[39] In the summer months (July, August, and Septem-ber), the area of the open ocean has been increasing andthis has allowed for warmer surface temperatures in Sep-tember because the ocean has a lower albedo than the iceand absorbs more heat. Ocean area has increased 4.8% andthe ice cover has become 6.9% less compact between 2003and 2011 (Figure 10, left). Surface temperatures over theChukchi/Beaufort, Laptev/E. Siberian Seas, Canadian Ar-chipelago, and Central Arctic have increased 1.53 Kbetween 2003 and 2011. Although the 2 m air specific hu-midity has increased a small amount, the larger increases in

Figure 5. (top) Average temperature differences for the surface and 2 m, monthly from January 2003to September 2011. (bottom) Average specific humidity differences for the surface and 2 m, monthlyfrom January 2003 to September 2011. The black line is the average difference for regions 4, 5, and 6 inFigure 1. The red line is the average difference for regions 7, 8, 9, and 10 in Figure 1. The gray lines dif-ferentiate between years.

BOISVERT ET AL.: MOISTURE FLUXES IN THE ARCTIC: 2003–2011

8

area of the open ocean and surface temperatures are themain cause of the changes in the moisture fluxes in thesummer (Figure 10, right). In August, the moisture flux hasbeen increasing (Figure 2b), but its trend is not significant.

[40] The moisture flux in October from 2003 to 2010 ona whole is decreasing 1.8%/year (Figure 11, bottom). Themain reason behind this is that the air specific humidity isincreasing (together with air temperature), which createssmaller specific humidity differences (see Figure 3). Theair temperature normally drops off rapidly in this month,but in recent years this has not happened so fast. Air tem-peratures remaining warmer in October could be due to ex-cessive sensible heat fluxes from the ocean surface, relatedto the reduced and less compact ice cover (see Figure 11,top), that warm up the lower atmosphere, and also due to

the increased cloud cover, which traps the outgoing long-wave radiation [Overland and Wang, 2010].

4.2. Moisture Flux and Ice Compactness

[41] One important factor to determine is whether or notthe compactness of the sea ice pack affects the moistureflux. The sea ice pack, defined as the zone where the iceconcentration is at least 15%, is becoming less compactduring each month and each year over the 2003–2011 pe-riod (Table 2) and each trend is statistically significant atthe 95% confidence level. For each month, we calculatedthe percentage of the total moisture flux that is supplied bythe ice pack (ice concentrations between 15 and 100%) inorder to determine the effect that a less compact ice packhas on the amount of moisture evaporated (or sublimated)

Figure 6. (left) Specific humidity (kg/kg) differences between the surface and 2 m for January for2003–2011. (right) Moisture flux anomalies (yearly minus the average moisture flux) from 2003 to 2011.Pinks are positive anomalies and blues are negative anomalies. Gray is either land or no data.

Figure 7. Average monthly SST anomalies for the East Greenland Sea from January 2003 to Septem-ber 2011. Black line is the SST anomaly from AMSR-E SST product. Gray line delineate between years.

BOISVERT ET AL.: MOISTURE FLUXES IN THE ARCTIC: 2003–2011

9

(Table 2). From 2003 to 2010, the moisture flux from theice pack was increasing because the compactness of the icewas decreasing. Over the 2003–2010 time period, the icehas become 4.8% less compact and the percentage of thetotal moisture flux contributed by the sea ice pack hasincreased by 3.0%. Also, during this time period, except in2010, in areas where the ice is less compact than average,the moisture flux has increased and vice versa (see Figure11, for an example). This is generally the case, but thereare also exceptions. When there is no ice, or the ice pack isless compact, then the surface is subsequently warmer thanthe air.

[42] In the winter months (November–February), althoughthe solid ice pack only contributes on average 12.2% of thetotal moisture flux, this percentage has increased 2.6 and0.6% in November and December between 2003 and 2010,and 5.0 and 2.6% in January and February between 2003

and 2011, respectively. During these cold months when theice pack is less compact, there are more areas of open waterthat are exposed. The SSTs are near the freezing point, butthe air temperatures are much colder, creating large tempera-ture differences and excessive evaporation. This is essen-tially what happens when a lead or polynya opens. Theamount of these increases would be even larger if 2 m-airtemperature and specific humidity during these months werenot increasing (see Figure 6, left).

[43] In June and July, the contribution of the moistureflux from the ice pack to the total moisture flux is alsoincreasing. This is because the ice pack is becoming muchless compact, creating a warmer surface and decreasing thespecific humidity differences between the surface and theair. The 2 m air temperatures during June and July are onaverage warmer than the surface, especially the openocean, which is near the freezing point of sea water,

Figure 8. (a) ERA-Interim January 2010 evaporation rate (mm/day) anomaly for the northern hemi-sphere, (b) ERA-Interim October 2010 evaporation rate (mm/day) anomaly for the northern hemisphere,(c) ERA-Interim January–October 2010 evaporation rate (mm/day) for the northern hemisphere, (d)ERA-Interim January 2010 1000 mb specific humidity (g/g) anomaly for the northern hemisphere, (e)ERA-Interim October 2010 1000 mb specific humidity (g/g) anomaly for the northern hemisphere, (f)ERA-Interim January–December 2010 1000 mb specific humidity (g/g) anomaly for the northern hemi-sphere. All of these images were produced using WRIT at https://reanalyses.org/atmosphere/web-based-reanalysis-intercomparison-tools-writ.

BOISVERT ET AL.: MOISTURE FLUXES IN THE ARCTIC: 2003–2011

10

whereas the snow/ice surface temperatures are close to0�C. The amount that the ice pack is contributing to thetotal moisture flux is decreasing in August–October andthis is because air temperatures are remaining warmer,

decreasing the magnitude of the temperature and specifichumidity differences (see Figure 3). Even though the per-centage of moisture coming from the ice pack is increasingduring June and July (Figure 2a, blue line), it is important

Figure 10. (left) July, August, and September ice concentration anomaly with respect to the average of2003–2011. (right) July, August, and September moisture flux anomaly with respect to the average of2003–2011. Positive anomalies are in pinks and negative anomalies are in blues. Gray is either land orno data.

Figure 9. Monthly mean specific humidity differences between the sea surface and air at 2 m height at(a) Longyear airport, Svalbard and (b) Keflavik airport, Iceland. Gray lines delineate between years.

BOISVERT ET AL.: MOISTURE FLUXES IN THE ARCTIC: 2003–2011

11

to keep in mind that the exchange of moisture during thesemonths is only 20% of what it is in the fall and wintermonths.

5. Validation and Error Estimates

[44] In order for our estimate of the 2 m specific humid-ity to be accurate, the 1000 hPa relative humidity fromAIRS must be accurately observed and the Monin-Obukhov similarity theory must be valid at the 1000 hPalevel. This validation was done using 2 m specific humidityobservations from the Tara drifting station [Vihma et al.,

2008], which drifted in the central Arctic (region 7 in Fig-ure 1). The 1000 hPa relative humidity from AIRS wastaken for each 625 km2 pixel that corresponded to the dailylatitude and longitude of the Tara drifting station from 1April to 20 September 2007. Since the 1000 hPa geopoten-tial heights during this period ranged from �101 to 243 m,the relative humidity and air temperature were used to cal-culate the specific humidity, which was then converted tothe standard 2 m height using the Monin-Obukhov theory.

[45] Time series of the observed and calculated 2 m spe-cific humidity are shown in Figure 12; the RMS error forthe AIRS 2 m specific humidity was 4.57 � 10�4 kg kg�1,

Figure 11. (top) October ice concentration anomaly with respect to the average of 2003–2010. (bot-tom) October moisture flux anomaly with respect to the average of 2003–2010. Positive anomalies are inpinks and negative anomalies are in blues. Gray is either land or no data.

BOISVERT ET AL.: MOISTURE FLUXES IN THE ARCTIC: 2003–2011

12

equating to a 20% error from the Tara 2 m specific humid-ity. This is very encouraging for two reasons: (a) the pointmeasurements at Tara are compared against the 625 km2

pixel values of AIRS and still the errors are small and (b)the average daily specific humidity from Tara was used tocompare to AIRS. Since Aqua only passes over the exactlocation a few times a day, a more detailed validation couldbe done comparing the observations at exactly matchingtimes, but this is out of the scope of this study. Our result issimilar to the 20% error that Tobin et al. [2006] foundusing version 4 water vapor profiles from AIRS when com-paring them to observations from Barrow, Alaska.

[46] We also compared the saturation specific humidity atthe surface using the surface temperature from AIRS and theTara drifting station during the same time period, and foundan error of 13%. For the moisture flux, however, the mostessential AIRS error is that of the difference between thesurface and 2 m specific humidity. The Tara data showedthat AIRS has a 16% error in this difference. This numberwas used in the error estimation of the moisture flux.

[47] Other factors causing errors in the moisture flux arethe uncertainties in the wind speed, air density, and thewater vapor transfer coefficient. An error estimate of 0.8 ms�1 was chosen for the ERA Interim wind speed followingJakobson et al. [2012], also based on Tara data. The accu-racy of the water vapor transfer coefficient is probably nobetter than 620% [Cronin et al., 2006]. The error of the air

density was estimated from the errors in the surface tem-perature from AIRS (2.7 K) and the SSTs (0.58 K) follow-ing Wentz and Meissner [2000]. Average values of thecalculated sensitivities, estimated uncertainties, and thefinal uncertainties for the entire Arctic from 2003 to 2011are shown in Table 3. We assumed that the variables areuncorrelated, and this has allowed for us to make an errorestimate by using

�2E ¼

X�2

x dE=dxð Þ2 ð9Þ

[48] Using this method and the error estimates the errorof the moisture flux in the Arctic was found to be 2.90 �10�3 g m�2 s�1, which corresponds to about a 20.3% erroroverall. This average error of 20.3% can also be expressedas 8.33 � 1013 kg of moisture when compared to the totalintegrated moisture flux from Figure 2a. This is a smallerror when compared to the range of variability in the mois-ture flux in the Arctic during the 2003–2011 time period,and we assume that conclusions on the interannual and spa-tial variations in the moisture flux are solid.

6. Conclusions

[49] Using geophysical data sets from multiple sensorson a single satellite like NASA’s EOS Aqua have allowedestimates of daily surface moisture flux to be produced

Figure 12. The 2 m specific humidity (kg kg�1) from AIRS (black line) and Tara drifting station (redline) from 1 April to 20 September 2007.

Table 2. Monthly and Yearly Trends and Means in Sea Ice Concentration and the Percentage of Moisture Supplied From the Sea IcePacka

Trend of Moisture FromSea Ice Pack (%/yr)

Average % of MoistureFrom Sea Ice Pack

Trend of IceConcentration (%/yr)

Average IceConcentration (%)

Jan. 0.55 12.81 20.59 90.66Feb. 0.29 12.15 20.44 92.35Mar. 0.35 13.88 20.33 92.88Apr. 0.24 19.82 20.27 92.59May 1.03 32.63 20.30 91.08Jun. 1.18 61.35 20.61 85.04Jul. 1.40 56.76 20.80 74.89

Aug. �0.21 15.79 20.80 71.41Sep. �0.18 7.28 20.70 76.68Oct. �0.25 11.13 20.69 84.45Nov. 0.32 11.51 20.72 90.60Dec. 0.07 12.43 20.55 91.47

Yearly 0.38 34.77 20.60 84.05

aNumbers highlighted in bold are statistically significant in the 95% confidence level.

BOISVERT ET AL.: MOISTURE FLUXES IN THE ARCTIC: 2003–2011

13

from 2003 to 2011. Evaluating this 9 year data set hasgiven us a clearer picture of how moisture is beingexchanged between the ocean and atmosphere in the Arctic.Moisture fluxes are also available from atmospheric reanal-yses. Although the fluxes have not been extensively vali-dated in the Arctic, the air moisture includes large errors inreanalyses, often larger than those we found for AIRS[Cullather et al., 2000; Jakobson and Vihma, 2010; L€upkeset al., 2010; Jakobson et al., 2012], and validation resultsfrom the Antarctic sea ice zone show large errors in thereanalyses moisture flux itself [Tastula et al., 2013]. We donot claim that our results are free of errors, but ourapproach of calculating the moisture fluxes has a strongbenefit for a study on interannual variations. In our method,the air moisture is always based on the same sensorsonboard the NASA’s EOS Aqua satellite: AMSR-E andAIRS. Hence, the errors are expected to be systematicrather than varying from year to year. Instead, the amountof observations assimilated into reanalyses varies from yearto year, depending on the availability of various in situ andremote sensing data.

[50] The flux of moisture from the surface to the loweratmosphere in the Arctic follows an annual cycle, with thelargest fluxes in the fall and winter months and the smallestones in the spring and summer months. Instead of anincrease in the moisture flux due to a declining sea icepack, there has actually been a 15% decrease between 2003and 2010 as an average over the area shown in Figure 1.This decrease is mostly due to an increase in the 2 m airspecific humidity over the Arctic, in particular over theBaffin Bay, E. Greenland Sea, and Kara/Barents Seas. Thedecrease in sea ice concentration tends to increase themoisture flux, i.e., to oppose the effect of increasing airspecific humidity (except during the melting season of seaice when the snow/ice surface is often warmer than theopen ocean). This effect was not significant over areasdefined as the open ocean, with sea ice concentration lessthan 15%. Hence, the zone where the sea ice concentrationhas decreased, but is still more than 15%, is contributingincreasingly more to the total moisture flux from the Arctic.The locations and magnitudes of the positive moistureanomalies are changing based on changes seen in the seaice pack and air humidity, but the total moisture flux in theentire Arctic (defined as in Figure 1) is becoming less.

[51] On a regional scale the Chukchi/Beaufort Seas, Lap-tev/E. Siberian Seas, Canadian Archipelago, and CentralArctic are seeing a slight increase, between 2.1 and 4.8%/yr, in the amount of moisture flux each year. In theseregions, the changes in the moisture flux are due mostly to

the changes in the ice compactness, which allows for thesurface temperatures to increase substantially in the falland winter months when the amount of moisture exchangedis the largest. On a regional scale, the Kara/Barents Seas,E. Greenland Sea, and Baffin Bay are seeing a decrease,between 0.53 and 9.2%/yr in the amount of moisture eachyear. The regions have areas of open water year round andtheir exchanges of moisture are due mostly to smaller dif-ferences in surface and 2 m specific humidities.

[52] Comparison of our results to previous studies is sen-sitive to the exact region and time period addressed. Screenand Simmonds [2010b] looked at the change in the latentheat flux for January–October over the Arctic from 1989 to2009 using ERA Interim reanalysis data and in situ obser-vations. They found increasing decadal trends in the Chuk-chi/Beaufort and Laptev/E. Siberian Seas regions, similarto our increases, as well as decreasing trends in the E.Greenland Sea region. In this study, they did not includethe Barents Sea or the months of November and December,which in our study are locations and times of significantfluxes of moisture. Because of this, they concludedincreases over the Arctic, whereas we have seen decreases.On the basis of ERA-40 reanalysis, Jakobson and Vihma[2010] concluded that evaporation in the circumpolar Arc-tic north of 70�N has not had any significant trend in theperiod 1979–2001, although the sea ice extent decreased alot already during this period.

[53] Recent studies by Palm et al. [2010], Eastman andWarren [2010], and Kay and Gettelman [2009] have lookedinto the changing sea ice extent and thickness and how thishas affected the amount of clouds over the Arctic. Theyfound that there were increases in clouds in all months withthe largest increases in the fall. This is when large differen-ces in temperatures between the ocean surface and atmos-phere enhanced the turbulent fluxes, which help to producelow-level clouds [Klein et al., 2009]. Further studies areneeded to better understand the relationships between seaice concentration, surface fluxes, and large-scale advectionof heat and moisture, as well as cloud cover in the Arctic.

[54] Acknowledgments. The authors thank the ACE academic sup-port team for their support on this project. The work of Linette Boisvertwas supported by ESSIC Cooperative Agreement NNX12AD03A, Task624. The work of Timo Vihma was supported by the Academy of Finlandvia the CACSI project (contract 259537). The authors would also like tothank the suggestions and input from two anonymous reviewers. We thankTimo Palo, Erko Jakboson, and Jaak Jaagus from the University of Tartu,Estonia, for providing us with the Tara observations.

ReferencesAndreas, E. L. (2002), Parameterizing scale transfer over snow and ice: A

review, J. Hydrometeorol., 3, 417–432.Andreas, E. L., R. E. Jordan, and A. P. Makhtas (2005), Parameterizing tur-

bulent exchange over sea ice: The Ice Station Weddell results, BoundaryLayer Meteorol., 114, 439–460.

Andreas, E. L., T. W. Horst, A. A. Grachev, P. O. G. Persson, C. W. Fairall,P. S. Guest, and R. E. Jordan (2010a), Parameterizing turbulent exchangeover summer sea ice and the marginal ice zone, Q. J. R. Meteorol. Soc.,136, 927–943, doi:10.1002/qj.618.

Andreas, E. L., P. O. G. Persson, R. E. Jordan, T. W. Horst, P. S. Guest, A.A. Grachev, and C. W. Fairall (2010b), Parameterizing turbulentexchange over sea ice in winter, J. Hydrometeorol., 11, 87–104.

Aumann, H. H., et al. (2003), AIRS/AMSU/HSB on the Aqua mission:Design, science objectives, data products, and processing systems, IEEETrans. Geosci. Remote Sens., 41, 253–264.

Table 3. Sensitivity of the Ocean-Atmosphere Moisture Flux inthe Arctic to Different Input Uncertaintiesa

Variable (x) dE/dx �x �x dE/dx

� (kg m�3) 3.22 � 10�3 0.83 2.67 � 10�3

CE 2.70 3.20 � 10�4 8.64 � 10�4

U (m s�1) 6.48 � 10�4 0.80 5.18 � 10�4

qs – qa (g kg�1) 1.43 � 10�2 4.80 � 10�2 6.86 � 10�4

rE (g m22 s21) 2.90 � 1023

hEi (g m22 s21) 1.43 � 1022

aBold indicates the standard deviation of the moisture flux error andhEi represents the average moisture flux error.

BOISVERT ET AL.: MOISTURE FLUXES IN THE ARCTIC: 2003–2011

14

Beljaars, A. C. M., and A. A. M. Holtslag (1991), Flux parameterizationover land surfaces for the atmospheric models, J. Appl. Meteorol., 30,327–341.

Boisvert, L. N., T. Markus, C. L. Parkinson, and T. Vihma (2012), Moisturefluxes derived from EOS Aqua satellite data for the North Water polynyaover 2003–2009, J. Geophys. Res., 117, D06119, doi:10.1029/2001JD016949.

Businger, J. A., J. C. Wyngaard, Y. Izumi, and E. F. Bradley (1971), Flux-profile relationships, J. Atmos. Sci., 28, 181–189.

Cavalieri, D. J., and C. L. Parkinson, (2012), Arctic sea ice variability andtrends, 1979–2010, The Cryosphere, 6, 881–889.

Cheng, Y., and W. Brutsaert (2005), Flux-profile relationships for windspeed and temperature in the stable atmospheric boundary layer, Bound-ary Layer Meteorol., 114, 519–538.

Cronin, M. F., C. W. Fairall, and M. J. McPhaden (2006), An assessment ofbuoy derived and numerical weather prediction surface heat fluxes in thetropical Pacific, J. Geophys. Res., 111, C06028, doi:10.1029/2005JC003324.

Cullather, R. I., D. H. Bromwich, and M. C. Serreze (2000), The atmospherichydrologic cycle over the Arctic basin from reanalyses. Part I: Compari-son with observations and previous studies, J. Clim., 13, 923–937.

Dee, D. P., et al. (2011), The ERA-Interim reanalysis: configuration andperformance of the data assimilation system, Q. J. R. Meteorol. Soc.,137, 553–597, doi:10.1002/qj.828.

Dyer, A. J., and B. B. Hicks (1970), Flux-gradient relationships in the con-stant flux layer, Q. J. R. Meteorol. Soc., 96, 715–721.

Dyer, A. J., and E. F. Bradley (1982), An alternative analysis of flux-gradient relationships at the 1976 ITCE, Boundary Layer Meteorol., 22,3–19.

Eastman, R., and S. G. Warren (2010), Interannual variations of Arcticcloud types in relation to sea ice, J. Clim., 23, 4216–4232.

Fairall, C. W., E. F. Bradley, D. P. Rogers, J. B. Edson, and G. S. Young(1996), Bulk parameterization of air-sea fluxes for Tropical Ocean-Global Atmosphere Coupled-Ocean Atmosphere Response Experiment,J. Geophys. Res., 101, 3747–3764.

Forrer, J., and M. W. Rotach (1997), On the turbulence structure in the sta-ble boundary layer over the Greenland ice sheet, Boundary Layer Mete-orol., 85, 111–136.

Garratt, J. R. (1992), The Atmospheric Boundary Layer, 316 pp., Cam-bridge Univ. Press, Cambridge, U. K.

Grachev, A. A., E. L. Andreas, C. W. Fairall, P. S. Guest, and P. O. G. Pers-son (2007), SHEBA flux-profile relationships in the stable atmosphericboundary layer, Boundary Layer Meteorol., 124, 315–333.

Högström, U. (1988), Non-dimensional wind and temperature profiles inthe atmospheric surface layer: A re-evaluation, Boundary Layer Mete-orol., 42, 55–78.

Holtslag, A. A. M., and H. A. R. de Bruin (1988), Applied modeling of thenighttime surface energy balance over land, J. Appl. Meteorol., 37, 689–704.

Jakobson, E., and T. Vihma (2010), Atmospheric moisture budget over theArctic on the basis of the ERA-40 reanalysis, Int. J. Climatol., 30, 2175–2194, doi: 10.1002/joc.2039.

Jakobson, E., T. Vihma, T. Palo, L. Jakobson, H. Keernik, and J. Jaagus(2012), Validation of atmospheric reanalyses over the central ArcticOcean, Geophys. Res. Lett., 39, L10802, doi:10.1029/2012GL051591.

Kay, J. E., and A. Gettelman (2009), Cloud influence on and response toseasonal Arctic sea ice loss, J. Geophys. Res., 114, D18204,doi:10.1029/2009JD011773.

Klein, S. A., et al. (2009), Intercomparison of model simulations of mixed-phase clouds observed during the ARM Mixed-Phase Cloud Experiment.Part I : Single-layer cloud, Q. J. R. Meteorol. Soc., 641, 979–1002,doi:10.1002/qj.416.

Klipp, C. L., and L. Mahrt (2004), Flux-gradient relationship, self-correlation and intermittency in the stable boundary layer, Q. J. R. Mete-orol. Soc., 130, 2087–2103.

Launiainen, J., and T. Vihma (1990), Derivation of turbulent surfacefluxes—An iterative flux-profile method allowing arbitrary observingheights, Environ. Software, 5, 113–124.

Launiainen, J., and T. Vihma (1994), On the surface heat fluxes in the Wed-dell Sea, in The Polar Oceans and Their Role in Shaping the GlobalEnvironment, Geophys. Monogr. Ser., vol. 85, Nansen Centennial Vol-ume, edited by O. M. Johannessen, R. Muench, and J. E. Overland, pp.399–419, AGU, Washington, D. C.

Laxon, S. W., et al. (2013), CryoSat-2 estimates of Arctic sea ice thicknessand volume, Geophys. Res. Lett., 40, 732–737, doi:10.1002/grl.50193.

L€upkes, C., T. Vihma, E. Jakobson, G. Konig-Langlo, and A. Tetzlaff(2010), Meteorological observations from ship cruises during summer tothe central Arctic: A comparison with reanalysis data, Geophys. Res.Lett., 37, L09810, doi:10.1029/2010GL042724.

L€upkes, C., V. M. Gryanik, J. Hartmann, and E. L. Andreas (2012), Aparameterization, based on sea ice morphology, of the neutral atmos-pheric drag coefficients for weather prediction and climate models,J. Geophys. Res., 117, D13112, doi:10.1029/2012JD017630.

Markus, T., and D. J. Cavalieri (2000), An enhancement of the NASA Teamsea ice algorithm, IEEE Trans. Geosci. Remote Sens., 38, 1387–1398.

Monin, A. S., and A. M. Obukhov (1954), Dimensionless characteristics ofturbulence in the surface layer, Trudy Geofiz. Inst. Akad. Nauk., 24, 163–187.

Obukhov, A. M. (1971), Turbulence in an atmosphere with a non-uniformtemperature, Engl. Transl., Boundary Layer Meteorol., 2, 7–29.

Olsen, E. T. (Ed.) (2011), AIRS/AMSU/HSB Version 5 Level 2 ProductLevels, Layers and Trapezoids, Rep., Jet Propul. Lab., Pasadena, Calif.

Overland, J., and M. Wang (2010), Large scale atmospheric circulationchanges are associated with the recent loss of Arctic sea ice, Tellus, Ser.A, 62, 1–9.

Palm, S. P., S. T. Strey, J. Sphinhirne, and T. Markus (2010), Influence ofArctic sea ice extent on polar cloud fraction and vertical structure andimplications for regional climate, J. Geophys. Res., 115, D21209,doi:10.1029/2010JD013900.

Parkinson, C. L. (2003), Aqua: An Earth-observing satellite mission toexamine water and other climate variables, IEEE Trans. Geosci. RemoteSens., 41(2), 173–183.

Paulson, C. A. (1970), The mathematical representation of wind speed andtemperature profiles in the unstable atmospheric surface layer, J. Appl.Meteorol., 9, 857–861.

Polyakov, I. V., J. E. Walsh and R. Kwok (2012), Recent changes of Arcticmultiyear sea ice coverage and likely causes, Bull. Am. Meteorol. Soc.,93, doi:10.1075/BAMS-D-11-00070.1.

Rinke, A., C. Melsheimer, K. Dethloff, and G. Heygster (2009), Arctic totalwater vapor: Comparison of regional climate simulations with observa-tions, and simulated decadal trends, J. Hydrometeorol., 10, 113–129,doi:10.1175/2008JHM970.1.

Screen, J. A., and I. Simmonds (2010a), The central role of diminishing seaice in recent Arctic temperature amplification, Nature, 464, 1334–1337,doi:10.1038/nature09051.

Screen, J. A., and I. Simmonds (2010b), Increasing fall-winter energy lossfrom the Arctic Ocean and its role in Arctic temperature amplification,Geophys. Res. Lett., 37, L16707, doi:10.1029/2010GL044136.

Serreze, M. C., J. D. Kahl, and R. C. Schnell (1992), Low-level temperatureinversions of the Eurasian Arctic and comparisons with Soviet driftingstation data, J. Clim., 5, 615–629.

Serreze, M. C., A. P. Barrett, and J. Stroeve (2012), Recent changes in tro-pospheric water vapor over the Arctic as assessed from radiosondes andatmospheric reanalyses, J. Geophys. Res., 117, D10104, doi:10.1029/2011JD017421.

Stroeve, J. C., M. C. Serreze, M. M. Holland, J. E. Kay, J. Maslanik, and A.P. Barrett (2012), The Arctic’s rapidly shrinking sea ice cover: Aresearch synthesis, Clim. Change, 110, 1005–1027, doi:10.1007/s10584-011-0101-1.

Tastula, E.-M., and T. Vihma (2011), WRF model experiments on the Ant-arctic atmosphere in winter, Mon. Weather Rev., 139, 1279–1291,doi:10.1175/2010MWR3478.1.

Tastula, E.-M., T. Vihma, E. L. Andreas, and B. Galperin (2013). Valida-tion of the diurnal cycles in atmospheric reanalyses over Antarctic seaice, J. Geophys. Res., 118, 4194–4204, doi:10.1002/jgrd.50336.

Tisler, P., T. Vihma, G. M€uller, and B. Br€ummer (2008), Modelling ofwarm-air advection over Arctic sea ice, Tellus, 60(4), 775–788.

Tjernström, M., et al. (2005), Modeling the Arctic boundary layer: Anevaluation of six ARCMIP regional-scale models using data from theSHEBA project, Boundary Layer Meteorol., 117, 337–381.

Vihma, T., J. Jaagus, E. Jakobson, and T. Palo (2008), Meteorological con-ditions in the Arctic Ocean in spring and summer 2007 as recorded onthe drifting ice station Tara, Geophys. Res. Lett., 35, L18706,doi:10.1029/2008GL034681.

Walsh, J. E., and W. L. Chapman (1998), Arctic cloud-radiation-temperature associations in observational data and atmospheric reanaly-ses, J. Clim., 11, 3030–3045.

Wentz, F., and T. Meissner (2000), AMSR Ocean Algorithm, AlgorithmTheoretical Basis Document, Version 2, Remote Sens. Syst., Santa Rosa,Calif.

BOISVERT ET AL.: MOISTURE FLUXES IN THE ARCTIC: 2003–2011

15