Significance
There has been a growing consensus that a decrease in sea ice would cause an increase in Arctic precipitation because of the potential for increased local evaporation. We quantify the effect of sea ice on the percentage of moisture sourced from the Arctic, using measurements of the isotopic composition of precipitation at six sites across the Arctic. These moisture proportion changes are important in that they indicate systematic adjustment and/or reorganization of the global hydrological cycle with climate change and provide validation for climate models. We explore how much these changes may increase Arctic precipitation and its impact on the energy balance.
Keywords: water cycle, precipitation, sea ice, climate change, deuterium excess
Abstract
Global climate is influenced by the Arctic hydrologic cycle, which is, in part, regulated by sea ice through its control on evaporation and precipitation. However, the quantitative link between precipitation and sea ice extent is poorly constrained. Here we present observational evidence for the response of precipitation to sea ice reduction and assess the sensitivity of the response. Changes in the proportion of moisture sourced from the Arctic with sea ice change in the Canadian Arctic and Greenland Sea regions over the past two decades are inferred from annually averaged deuterium excess (d-excess) measurements from six sites. Other influences on the Arctic hydrologic cycle, such as the strength of meridional transport, are assessed using the North Atlantic Oscillation index. We find that the independent, direct effect of sea ice on the increase of the percentage of Arctic sourced moisture (or Arctic moisture proportion, AMP) is 18.2 ± 4.6% and 10.8 ± 3.6%/100,000 km2 sea ice lost for each region, respectively, corresponding to increases of 10.9 ± 2.8% and 2.7 ± 1.1%/1 °C of warming in the vapor source regions. The moisture source changes likely result in increases of precipitation and changes in energy balance, creating significant uncertainty for climate predictions.
There is increasing interest in the response of the Arctic hydrologic cycle to changing climate because of its potential to influence, or feedback to, future climate change. Modeling studies have identified enhanced transport of subtropical moisture to the Arctic as well as increased Arctic evaporation as potential mechanisms of augmentation of the water cycle (1–3). The enhanced hydrologic cycle may feedback to climate change either positively or negatively; both the sign and the magnitude are yet to be determined.
Observational evidence for hydrological acceleration during the past few decades is limited. Direct measurement of precipitation is difficult in the Arctic because of its cold, windy environments (4). Despite these difficulties, increasing precipitation has been reported for some Arctic locations (5, 6), and it has been hypothesized that changes in sea ice extent may have significantly influenced precipitation both in the past (7) and today (8–10). We report a study of changes in the isotopic composition of precipitation to understand the larger-scale changes of the hydrologic cycle, focusing on moisture source changes. The objective of this work is to assess observationally the effect of sea ice and the moisture transport regime on Arctic precipitation from 1990 to 2012, using the isotopic composition of precipitation from six Arctic stations. In particular, we quantify how the fraction of the total Arctic precipitation that is sourced in the Arctic responds to the sea ice extent. We then use these empirically established sensitivities of precipitation isotope ratios to sea ice change to project potential future precipitation changes and to evaluate impacts of these changes on the energy balance.
Our approach is based on the premise that Arctic precipitation is composed mostly of water from two marine evaporation regions or “moisture sources”—one subtropical and one local—and that the relative contributions of the two sources to the precipitation can be determined from the stable isotopic ratios of the precipitation. We partition the two proportions, using the precipitation deuterium excess (d-excess, defined as d = δD − 8δ18O, where δD and δ18O are the parts per thousand deviation of deuterium/hydrogen and 18O/16O atomic ratios, respectively, from those of the standard mean ocean water), which is an indicator of moisture source conditions, principally the sea surface temperature (SST) and relative humidity (RH) (11–13). Moisture from subtropical regions has high d-excess values, indicative of relatively high SST and low RH at the source, whereas locally evaporated Arctic moisture has low d-excess values (14), indicating low SST and high RH. We hypothesize that precipitation d-excess is positively associated with sea ice area as a consequence of increasing local evaporation and thus increasing proportion of Arctic-sourced moisture with reduction of sea ice.
We use the North Atlantic Oscillation (NAO) index as a proxy for general climate conditions to quantify effects that are independent of the sea ice influence on precipitation. Most importantly, the NAO is associated with the strength of meridional transport (15), which in turn affects precipitation d-excess by changing the proportion of subtropical moisture in the total precipitation. For example, if winds from the south strengthen, the proportion of moisture transported from the subtropics would increase, thus increasing the d-excess. In addition to meridional transport, the NAO also influences other variables, such as location of the subtropical moisture source region, temperature and humidity along the storm track, etc., all of which may affect the d-excess of precipitation. When holding the NAO constant (statistically), we also effectively remove the influences of these variables, achieving limited contamination to the signal of the direct precipitation–sea ice relationship.
The six sites included in this work were from two regions, the Canadian Arctic (Alert, Eureka, and Cambridge Bay, Canada) and Greenland Sea (Reykjavìk, Iceland, Ny-Ålesend, Norway, and Danmarkshavn, Greenland) (Fig. 1). We consider all sites within a region to share similar local moisture sources. The Canadian Arctic sites receive most of their local moisture from Baffin Bay (16) and the Greenland Sea sites receive it from the Greenland Sea (17).
Fig. 1.
Location of sites for monthly precipitation isotope ratio measurements. Shown are Canadian Arctic sites Alert, Eureka, and Cambridge Bay, Canada and Greenland Sea sites Reykjavìk, Iceland; Ny-Ålesend, Norway; and Danmarkshavn, Greenland. Local moisture sources Baffin Bay (BB) and Greenland Sea (GS) are labeled as well.
Precipitation Isotope Ratio Observations
For each region, a multiple regression was conducted of the annual d-excess of each region (site mean removed) against the local sea ice extent and the NAO index (Fig. 2). In the Canadian Arctic, the overall r2 is 0.39 (P < 0.0001) with a sea ice partial coefficient of 1.38 ± 0.35‰/105 km2 (P = 0.0003) and an NAO partial coefficient of −4.3 ± 0.9‰/unit NAO (P < 0.0001). In the Greenland Sea region, the overall r2 is 0.33 (P = 0.0006) with a sea ice partial coefficient of 0.91 ± 0.30‰/105 km2 (P = 0.0048) and an NAO partial coefficient of 2.3 ± 0.7‰/unit NAO (P = 0.0011). The uncertainties of the above values are the SEs of the partial regression coefficients, which will be propagated through subsequent calculations. Sources of errors are discussed in more detail in a later section. The sensitivities reported here reflect both long-term temporal trends and multiyear variations in all variables. Although the partial coefficient of sea ice is similar in both regions, the sign of the NAO partial coefficient reverses with region; these contrasting results are critical to the interpretation of physical mechanisms. The partial coefficient for the cross effect of sea ice and the NAO is not significant, and the regression residual does not have a significant serial correlation.
Fig. 2.
(A and B) Leverage plots (43) for multiple regressions of d-excess against local sea ice extent and the NAO index for (A) Canadian Arctic and (B) Greenland Sea regions. Solid circles are observations for individual years and sites with site mean removed. The regressions yield overall r2 and P values of 0.39, <0.0001 and 0.33, 0.0006 for A and B, respectively.
These results show that sea ice extent significantly influences the moisture source of Arctic precipitation. As sea ice extent decreases, d-excess decreases, indicating a relative increase in Arctic sourced moisture contribution. This effect is further demonstrated by the different d-excess sensitivities of Reykjavìk and Ny-Ålesend, which share similar storm tracks (18) but with the latter located at a significantly higher latitude. Of the two, Ny-Ålesend (NYA) should experience a stronger sensitivity of Arctic moisture increase to sea ice reduction than Reykjavìk (REY) as the sea ice margin retreats. Indeed, a multiple regression of d-excess difference (NYA − REY) against sea ice extent and NAO has a positive partial coefficient of sea ice of 1.47 ± 0.48‰/105 km2 (P = 0.014) (Fig. 3).
Fig. 3.
Leverage plots (43) for multiple regression of Ny-Ålesend (NYA) – Reykjavìk (REY) d-excess difference against Greenland Sea ice extent and the NAO index. The results are r2 = 0.64 and P = 0.010.
The results in Fig. 2 also show the importance of the NAO phase to the moisture source contributions. However, the relationship between the NAO and precipitation d-excess reverses signs between the two regions. In the Greenland Sea region, a positive NAO is associated with relatively high d-excess values, corresponding to reduced contributions of Arctic-sourced moisture. The opposite relationship occurs in the Canadian Arctic.
The differences between the Greenland Sea and Canadian Arctic responses to the NAO likely result from the opposite NAO influences on meridional moisture transport in the two regions. Fig. 4 shows the differences of the meridional wind anomalies at the 500-hPa level for the five strongest positive and negative NAO years during our study period. When the NAO phase is positive, there are stronger than average southerly winds in the Greenland Sea region, but stronger northerly winds in the Canadian Archipelago. The reverse wind anomalies are seen during the negative NAO phase. This pattern is seen at all pressure levels. With a stronger local flow from the south (north), advection will increase the proportion of subtropical (Arctic) moisture with relatively high (low) d-excess values. The relationship between meridional transport and the NAO phase is consistent with the observed negative (positive) association between d-excess and the NAO at Canadian Arctic (Greenland Sea) sites (Fig. 2). This relationship has been shown to cause differences in precipitation amount between the eastern and western coastal regions of Greenland (19) and has also been demonstrated to affect the isotopic ratios and d-excess in Greenland ice cores (20). The same argument applies to the negative partial correlation between the NAO and the NYA − REY d-excess difference (−2.1 ± 1.1‰/unit NAO, P = 0.083, Fig. 3), because the northern site is more strongly affected by enhanced meridional transport during the negative NAO phase. Similar results were obtained using the strength of the meridional wind component, rather than the NAO index, as an independent variable in the regression. However, we explain later the advantages of using the NAO.
Fig. 4.
(A and B) Contours of the mean meridional wind speed anomaly at 500 hPa for (A) the five strongest positive (1990, 1992, 1994, 1999, and 2011) and (B) negative (1998, 2005, 2008, 2009, and 2010) NAO years during the study period. Positive (negative) anomaly values represent anomalously high southerly (northerly) wind speeds. The plots show the opposite wind directions in the Greenland Sea and Canadian Arctic regions and the wind reversal with NAO phase. Image provided by the NOAA/ESRL Physical Sciences Division, Boulder, CO, at www.esrl.noaa.gov/psd/; data from ref. 44.
Arctic Moisture Proportion
The observed effects of changes in sea ice extent on d-excess can be used to determine the sensitivity of the proportion of Arctic-sourced moisture to sea ice change. We define the “Arctic moisture proportion” (AMP) as the fraction of the total Arctic precipitation that originated as Arctic evaporation. We further assume that the balance, 1 − AMP, is attributed to subtropical evaporation. To obtain AMP requires an estimate of the d-excess end members for the subtropical- and Arctic-sourced moisture. Following the recommendation of Jouzel et al. (21), we used the isotopically enabled general circulation model (GCM) ECHAM 4 (22) to determine the d-excess value of the vapor originating from the respective sources (Methods). Assuming that the precipitation d-excess is a mass-weighted mean of the moisture evaporated from two end member source regions and that these source signatures are preserved to the precipitation site, the Arctic moisture proportion in the total precipitation can be calculated (using Eq. 1 as described in Methods). The AMP increases by 18.2 ± 4.6% and 10.8 ± 3.6% in the Canadian Arctic and Greenland Sea regions, respectively, per 105 km2 sea ice loss.
To convert these sensitivities to percentage change in AMP per degree centigrade of source area warming, we divide them by the regression coefficients of source temperature against sea ice extent (discussed in Methods) for the Canadian Arctic and Greenland Sea, respectively, obtaining AMP changes of 10.9 ± 2.8% and 2.7 ± 1.1%/°C at the vapor sources.
The effect of the NAO, alone, on the AMP can be similarly assessed using the NAO partial regression coefficient. In the Canadian Arctic and Greenland Sea regions, the AMP changes by 56 ± 12% and −28 ± 8%, respectively, per unit NAO increase. If both sea ice extent and NAO were changing, the total sensitivity of the AMP to both variables would differ from the earlier estimates. Although still under debate, there is some evidence that the future warmer climate will have larger-amplitude Rossby waves (23, 24) and thus potentially a more persistent negative NAO phase (25). If the NAO were to decrease by 0.25 units, for example, a 105-km2 sea ice reduction would result in increases of AMP by 4.2% and 17.8% in Canadian Arctic and Greenland Sea regions, respectively.
The concept of AMP is important in that it, in one way, indicates systematic adjustment and/or reorganization of the global hydrological cycle with climate change. We emphasize that responses of AMP to sea ice extent and the NAO reported here are based on robust observations of precipitation isotope ratios and a simple mass balance model. This result may be used to validate complex models on the global climate and hydrological cycle. In particular, it would allow for the testing of model output of changes in Arctic evaporation in response to sea ice reduction and meridional transport under different climate scenarios. The AMP also has implications for changes in Arctic precipitation amounts and consequent changes in radiation balance. Quantifying these outcomes requires knowledge of variables that are subject to known and unknown uncertainties. We nevertheless provide the following sensitivity experiments to demonstrate how changes of AMP may act as feedback mechanisms to climate change.
Implications
The first sensitivity experiment is to convert a change in AMP to the corresponding change in the total annual precipitation amount. Obviously, this conversion requires that we know how the subtropical-sourced precipitation amount changes with sea ice extent and with climate change in general. With absence of that information, we compute the response of the total precipitation amount to sea ice change with a 0%, +5%, or −5% change in the subtropical-sourced precipitation amount. Under a scenario where sea ice decreases by 105 km2, and the subtropical-sourced precipitation amount stays constant (0% change), the precipitation across our sites would increase by 21.1 ± 9.1% (among-site mean ± 1 SD; example calculation described in Methods). If moisture transported from the subtropics were to increase by 5%, the total precipitation increase would rise to 28.2 ± 9.5%. The opposite transport scenario of a 5% decrease yields a precipitation increase of 16.0 ± 8.6%.
The above sensitivities of percentage of precipitation increase to sea ice reduction can be converted to the sensitivities to temperature warming at the Arctic moisture source region, using the method described earlier. We obtained 8.1 ± 1.7%, 10.6 ± 2.5%, and 5.6 ± 1.4% /°C with subtropical-sourced moisture changes of 0%, 5%, and −5%, respectively. Bintanja and Selten (3) reported a projected sensitivity of 4.5%/°C for Arctic-wide precipitation responses to warming. Our sensitivity estimates, although not under conditions equivalent to those by Bintanja and Selten (3), are, within error, similar to or slightly greater than their value.
The predicted change in the precipitation amount over the past two decades does not show up consistently in the precipitation records. During the study period, the annual average sea ice extent has decreased by ∼1.5 × 105 km2 in Greenland Sea and 3 × 105 km2 in Baffin Bay. This should correspond to ∼30–60% increase in precipitation if holding other effects constant. However, the precipitation records do not show a significant increasing trend at all six sites. Some sites, such as Reykjavìk, Cambridge Bay, and Eureka, do have an increasing tendency, and similar observations have been reported for other Arctic locations (5, 6). For example, Wong (26) reported a significant 13% per decade precipitation increase based on the weather record at Thule Air Base, Greenland. However, considerable interannual variability and limited record length prevent most apparent trends from being statistically significant. We point out that measuring Arctic precipitation is extremely challenging because of high winds (4). Our method may provide an alternative for assessing regional changes in precipitation amount.
Our second sensitivity experiment is to address how the projected precipitation changes impact the energy balance of the Arctic and the global climate system. Precipitation has an effect on land surface albedo (27). If the additional precipitation falls as snow, it could potentially increase glacial mass and the number of days of high land surface albedo, thus having a cooling effect. Under the scenario of 0% change of the subtropical-sourced moisture, this negative feedback to global warming could have an annualized effect of −1.1 ± 0.2 W/m2 per degree centigrade of warming (Methods), compared with +3.7 W/m2 per carbon dioxide doubling as discussed by the Intergovernmental Panel on Climate Change. However, if the increased precipitation fell as rain, it would cause earlier spring melt and/or later onset of autumn snow coverage and a longer low albedo period. This positive feedback would enhance atmospheric warming (27), likely by a magnitude similar to that of the potential negative feedback. In either case, the resulting radiative forcing likely has an order of magnitude similar to that of the forcing from doubling CO2, thus demonstrating that the sea ice feedback to radiation balance through the Arctic hydrological cycle is potentially a major component of climate change. A more thorough study of the seasonal distribution of precipitation increase and the radiative effects of this forcing across different regions of the Arctic is needed to better constrain the sign and magnitude of this effect.
There are other ways by which Arctic evaporation and precipitation may impact the energy balance. Some are well known, such as the impacts from the changing latent heat flux and the radiative effects of increased water vapor; others are less well known, such as the impact on cloud formation and properties (28). Although detailed discussions on these impacts are beyond the scope of this paper, they further demonstrate the uncertain nature of feedbacks of Arctic sea ice to climate change.
Robustness and Uncertainties
The multiple regressions shown in Fig. 2 explain less than 40% of the total variance in d-excess variations. The remaining 60% unexplained variance contributes to our computational uncertainty and may be derived from several sources. Physically, sources of uncertainty include factors affecting sea ice extent, d-excess, and climate variables that are not fully included in the NAO.
In the analysis, we limit the Arctic moisture source regions to the Greenland Sea and Baffin Bay. It is likely that other Arctic basins do contribute to the precipitation to some degree. These Arctic sources would likely have a similar d-excess end member and thus would not alter the moisture proportions, but they have different sea ice extent variability and trends that could alter the regression analyses. Combining the sea ice extents of other regions in the regression analysis yields significant results, but the strongest relationships occur with the analysis presented. That does not exclude the possibility that alternate moisture sources and their sea ice extents contribute to the precipitation of our sites in some years.
There are a number of factors that affect d-excess but were not explicitly included in our analysis. Different local source conditions, variable evaporation conditions (29, 30), entrainment of moisture along the storm track, location of the site relative to the storm or storm track/jet stream (31), snow formation (32), and evaporation/sublimation or isotopic exchange between vapor and the condensed phase below the precipitating cloud base (33) all have the potential to alter the d-excess in ways that might degrade the simple two-component mixing relationship. The relative importance of these effects remains to be quantified. A single strong storm event could significantly dominate the measured monthly precipitation d-excess value, such as during an atmospheric river event (34). Kurita (35) showed that the Arctic water vapor during the sea ice growth season had relatively high d-excess values, rather than the hypothesized low value, potentially adding uncertainty to the simple mixing relationship. These high d-excess scenarios are typically not present within the isotopically enabled GCMs but are an important consideration, especially when trying to observe the d-excess–moisture source relationship (or other mixing relationships) on seasonal or even shorter timescales. Furthermore, the relative humidity in the Arctic has been increasing with time (36). If this change is not correlated with the NAO, it may contribute to the unexplained variation in the d-excess.
To understand the pure sea ice–d-excess relationship, we considered the NAO index to be a better explanatory variable than only meridional winds to be used in the regression analysis, even though the latter also yielded significant results. This is because the NAO represents the general atmospheric state that includes a number of other meteorological variables, such as condensation temperature, ridging and blocking patterns, and source moisture location. When holding the NAO constant in the multiple-regression analysis, we also effectively minimize the influences of these variables.
There are a number of factors that affect atmospheric transport regimes that are poorly represented by the NAO index. Fig. 4 shows that Cambridge Bay and Ny-Ålesend reside near the margins of the major NAO wind changes; thus these sites may have different sensitivities than others in their respective regions (r2 values of regressions increase significantly if these sites are removed). Other circulation patterns, components of which are not quantified by the NAO, can also affect transport regimes and moisture sources of the Arctic, such as El Nino, the Pacific Decadal Oscillation, the Arctic Oscillation, or the Pacific–North American pattern.
We conclude that the validity of our results is supported by the strong statistical significance in the hypothesized relationships, that the results are unlikely to be biased by climate factors not explicitly included in the analysis, and that they are broadly consistent with an independent modeling assessment (3). The method can be applied to and further tested in other Arctic regions as well as on different spatial and temporal scales, such as on a storm-by-storm basis or with ice core data.
Methods
Precipitation was collected monthly from 1990 to 2012 at the Canadian Arctic sites and from 1990 to 2009 at the Greenland Sea sites or a subset of those years. Each precipitation sample was measured for its hydrogen and oxygen isotopic compositions (expressed as δD and δ18O, respectively), and the d-excess is calculated as d = δD – 8δ18O. Data from the Greenland Sea region were obtained from the Global Network of Isotopes in Precipitation database (37). The d-excess was annually averaged (January to December) to remove the annual cycle to assess the interannual variations. Years with missing data were removed from the analysis. When data are pooled for the regional analysis, site means were removed from the annual averages.
To compute the d-excess end members at the three respective sources, the isotopically enabled GCM ECHAM 4 (22) data were used. The moisture source regions were the oceanic areas from 60° to 80°N and 50°–80°W for Baffin Bay and from 62° to 82°N and 24°W–10°E for the Greenland Sea. The subtropical moisture source was assumed to be at the subtropical high (STH) (38). For each source, the specific humidity value of each isotopologue was extracted from the ECHAM 4 output and converted to δD and δ18O values, from which d-excess values were obtained. Annual averages were obtained from monthly data for 1990–1999 that overlap with our study period. The Baffin Bay and Greenland Sea sources had d-excess values of 3.2 ± 0.2‰ and 2.5 ± 0.1‰, respectively, whereas the North Atlantic STH d-excess was 10.7 ± 0.1‰. Small variations in the location of these Arctic sources do not change the d-excess by more than 1‰.
Assuming that the precipitation d-excess (d) is a mass-weighted mean of the two end members, the AMP in the total precipitation is given by
[1] |
Eq. 1 and the partial regression coefficients of sea ice extent were used to determine the sensitivities of AMP to sea ice change.
Daily sea ice extents, based on SMMR (Nimbus-7 Scanning Multichannel Microwave Radiometer) and SSM/I-SSMIS [Defense Meteorological Satellite Program (DMSP) Special Sensor Microwave/Imagers and DMSP-F17 Special Sensor Microwave Imager/Sounder] (39), were obtained from the National Snow and Ice Data Center for each basin, and the data were then averaged annually to compare with the isotope ratio data. NAO index data were also acquired daily from the National Oceanic and Atmospheric Administration (NOAA) Climate Prediction Center over the study period and averaged on annual timescales.
The relationship between sea ice and temperature was obtained to transform the AMP–sea ice partial coefficient to the AMP–temperature sensitivity. The temperature for each respective basin was taken as the ERA-Interim [European Center for Medium-Range Weather Forecasting (ECMWF) Reanalysis Interim] 2-m air temperature (40) within the source regions for Baffin Bay and the Greenland Sea defined in the text. The regressions between temperature and sea ice extent yielded sensitivities of –5.99 × 104 ± 0.39 km2/°C (r2 = 0.78, P < 0.0001) and –2.45 × 104 ± 0.62 km2/°C (r2 = 0.21, P = 0.0002) for Baffin Bay and the Greenland Sea, respectively. The AMP sensitivities to sea ice were then multiplied by the above respective regression coefficients to yield AMP sensitivities to temperature.
To convert a change of AMP to that of precipitation amount, we use Ny-Ålesend as an example. During our analysis period, Ny-Ålesend received 482 mm/y precipitation. If we use the annual d-excess before 1995, before the large sea ice decline during our study period, the corresponding AMP was 2.6%. Holding precipitation amount from the subtropics constant, a sea ice decrease of 105 km2, which corresponds to an increase of the AMP of 10.8%, would yield a precipitation increase of 53 mm or 12%. If the amount of subtropical transported moisture increased by 5%, the Ny-Ålesend precipitation would increase by 77 mm or 18%. If there were a 5% decrease of subtropical transported moisture, the precipitation would have increased by 29 mm or 7%. These calculations were done for each site with the mean and SD reported. To convert these sensitivities to a per degree centigrade change, we take the precipitation–sea ice sensitivity multiplied by the regional sea ice–temperature sensitivity obtained earlier.
To estimate the effect of changing precipitation on the energy balance, we used the results of Stone et al. (41) on the North Slope of Alaska and the difference in the surface albedo for snow cover and tundra (27). Stone et al. (41) reported 14 d of melt date delay per centimeter water equivalent of snow. Using this value with the current precipitation amounts at the study sites and the zero transport change scenario projected precipitation increases per degree centigrade of warming, we obtained 1.0–10.8 d of delay in spring melt. At all sites, we assumed that the incoming shortwave radiation reaching the surface is 200 W/m2, a conservative value for the Arctic during the spring melt season (42). The surface albedo change (41) from open tundra (0.17) to a snow-covered surface (0.75) results in a change in the annualized net radiative forcing of −1.1 ± 0.2 W/m2 (mean and SE among the six sites). These results are based on only a single study conducted in Alaska; applying it to our location introduces some error. However, the result provides an order of magnitude estimate that may serve as a reference point against which other scenarios can be considered.
Acknowledgments
Discussions with Leslie Sonder and Annie Putman were helpful for the development of this work. We thank Environment Canada and their staff at the Alert and Eureka weather stations and the staff of Community Aerodrome Radio Service (CARS) at the Cambridge Bay weather station for collecting precipitation samples. Robert Drimmie at Isotope Tracer Technologies conducted analytical work for the Canadian samples. This paper has been improved by the comments of three anonymous reviewers. The work was supported by the National Science Foundation (NSF) under Award 1022032 for the Isotopic Investigation of Sea Ice and Precipitation in the Arctic Climate System (iisPACS) project and partially under Award 0801490 for Dartmouth’s Integrative Graduate Education and Research Traineeship (IGERT) for Polar Environmental Change.
Footnotes
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
References
- 1.Kug JS, Choi DH, Jin FF, Kwon WT, Ren HL. 2010. Role of synoptic eddy feedback on polar climate responses to the anthropogenic forcing. Geophys Res Lett 37(14):L14704.
- 2.Bengtsson L, Hodges KI, Koumoutsaris S, Zahn M, Keenlyside N. The changing atmospheric water cycle in Polar Regions in a warmer climate. Tellus Ser A Dyn Meterol Oceanogr. 2011;63(5):907–920. [Google Scholar]
- 3.Bintanja R, Selten FM. Future increases in Arctic precipitation linked to local evaporation and sea-ice retreat. Nature. 2014;509(7501):479–482. doi: 10.1038/nature13259. [DOI] [PubMed] [Google Scholar]
- 4.Liston GE, Sturm M. The role of winter sublimation in the Arctic moisture budget. Nord Hydrol. 2004;35(4–5):325–334. [Google Scholar]
- 5.Forland EJ, Hanssen-Bauer I. Increased precipitation in the Norwegian Arctic: True or false? Clim Change. 2000;46(4):485–509. [Google Scholar]
- 6.Kattsov VM, Walsh JE. Twentieth-century trends of arctic precipitation from observational data and a climate model simulation. J Clim. 2000;13(8):1362–1370. [Google Scholar]
- 7.Gildor H, Tziperman E. Sea ice as the glacial cycles’ climate switch: Role of seasonal and orbital forcing. Paleoceanography. 2000;15(6):605–615. [Google Scholar]
- 8.Liu J, Curry JA, Wang H, Song M, Horton RM. Impact of declining Arctic sea ice on winter snowfall. Proc Natl Acad Sci USA. 2012;109(11):4074–4079. doi: 10.1073/pnas.1114910109. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Noone D, Simmonds I. 2004. Sea ice control of water isotope transport to Antarctica and implications for ice core interpretation. J Geophys Res 109(D7):D07105.
- 10.Weatherly JW. Sensitivity of Antarctic precipitation to sea ice concentrations in a general circulation model. J Clim. 2004;17(16):3214–3223. [Google Scholar]
- 11.Merlivat L, Jouzel J. Global climatic interpretation of the deuterium-oxygen-18 relationship for precipitation. J Geophys Res. 1979;84:5029–5033. [Google Scholar]
- 12.Petit JR, White JWC, Young NW, Jouzel J, Korotkevich YS. Deuterium excess in recent Antarctic snow. J Geophys Res. 1991;96:5113–5122. [Google Scholar]
- 13.Vimeux F, Masson V, Jouzel J, Stievenard M, Petit JR. Glacial-interglacial changes in ocean surface conditions in the southern hemisphere. Nature. 1999;398(6726):410–413. [Google Scholar]
- 14.Klein ES, et al. Arctic cyclone water vapor isotopes support past sea ice retreat recorded in Greenland ice. Sci Rep. 2015;5(10295):10295. doi: 10.1038/srep10295. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Hurrell JW. Decadal trends in the North Atlantic Oscillation: Regional temperatures and precipitation. Science. 1995;269(5224):676–679. doi: 10.1126/science.269.5224.676. [DOI] [PubMed] [Google Scholar]
- 16.Koerner R, Russell RD. Delta-O-18 variations in snow on the Devon Island Ice Cap, Northwest-Territories, Canada. Can J Earth Sci. 1979;16(7):1419–1427. [Google Scholar]
- 17.Sodemann H, Schwierz C, Wernli H. 2008. Interannual variability of Greenland winter precipitation sources: Lagrangian moisture diagnostic and North Atlantic Oscillation influence. J Geophys Res 113(D3):D03107.
- 18.Hoskins BJ, Hodges KI. New perspectives on the Northern Hemisphere winter storm tracks. J Atmos Sci. 2002;59(6):1041–1061. [Google Scholar]
- 19.Bromwich DH, Chen QS, Li YF, Cullather RI. Precipitation over Greenland and its relation to the North Atlantic Oscillation. J Geophys Res. 1999;104(D18):22103–22115. [Google Scholar]
- 20.Barlow LK, White JWC, Barry RG, Rogers JC, Grootes PM. The North-Atlantic Oscillation signature in deuterium and deuterium excess signals in the Greenland Ice-Sheet Project-2 ice core, 1840-1970. Geophys Res Lett. 1993;20(24):2901–2904. [Google Scholar]
- 21.Jouzel J, et al. Climatic interpretation of the recently extended Vostok ice records. Clim Dyn. 1996;12(8):513–521. [Google Scholar]
- 22.Hoffmann G, Werner M, Heimann M. Water isotope module of the ECHAM atmospheric general circulation model: A study on timescales from days to several years. J Geophys Res. 1998;103(D14):16871–16896. [Google Scholar]
- 23.Overland JE, Wang MY. Large-scale atmospheric circulation changes are associated with the recent loss of Arctic sea ice. Tellus Ser A Dyn Meterol Oceanogr. 2010;62(1):1–9. [Google Scholar]
- 24.Francis JA, Vavrus SJ. 2012. Evidence linking Arctic amplification to extreme weather in mid-latitudes. Geophys Res Lett 39(6):L06801.
- 25.Seierstad IA, Bader J. Impact of a projected future Arctic Sea Ice reduction on extratropical storminess and the NAO. Clim Dyn. 2009;33(7–8):937–943. [Google Scholar]
- 26.Wong GJ. 2015. Environmental and climatic controls on glaciochemistry of near-surface snow in Greenland. PhD thesis (Dartmouth College, Hanover, NH)
- 27.Chapin FS, 3rd, et al. Role of land-surface changes in Arctic summer warming. Science. 2005;310(5748):657–660. doi: 10.1126/science.1117368. [DOI] [PubMed] [Google Scholar]
- 28.Curry JA, Rossow WB, Randall D, Schramm JL. Overview of Arctic cloud and radiation characteristics. J Clim. 1996;9(8):1731–1764. [Google Scholar]
- 29.Steen-Larsen HC, et al. Climatic controls on water vapor deuterium excess in the marine boundary layer of the North Atlantic based on 500 days of in situ, continuous measurements. Atmos Chem Phys. 2014;14(15):7741–7756. [Google Scholar]
- 30.Benetti M, et al. Deuterium excess in marine water vapor: Dependency on relative humidity and surface wind speed during evaporation. J Geophys Res. 2014;119(2):584–593. [Google Scholar]
- 31.Lawrence JR, Gedzelman SD, White JWC, Smiley D, Lazov P. Storm trajectories in eastern-United-States D/H isotopic composition of precipitation. Nature. 1982;296(5858):638–640. [Google Scholar]
- 32.Jouzel J, Merlivat L. Deuterium and O-18 in precipitation - modeling of the isotopic effects during snow formation. J Geophys Res. 1984;89:1749–1757. [Google Scholar]
- 33.Lee JE, Fung I. “Amount effect” of water isotopes and quantitative analysis of post-condensation processes. Hydrol Processes. 2008;22(1):1–8. [Google Scholar]
- 34.Bonne JL, et al. The summer 2012 Greenland heat wave: In situ and remote sensing observations of water vapor isotopic composition during an atmospheric river event. J Geophys Res. 2015;120(7):2970–2989. [Google Scholar]
- 35.Kurita N. 2011. Origin of Arctic water vapor during the ice-growth season. Geophys Res Lett 38(2):L02709.
- 36.Boisvert LN, Markus T, Vihma T. Moisture flux changes and trends for the entire Arctic in 2003-2011 derived from EOS Aqua data. J Geophys Res. 2013;118(10):5829–5843. [Google Scholar]
- 37. International Atomic Energy Agency and World Meteorological Organization (IAEA/WMO) (2015) Global Network of Isotopes in Precipitation. The GNIP Database. Available at: www.iaea.org/water. Accessed November 2, 2012.
- 38.Feng XH, Faiia AM, Posmentier ES. 2009. Seasonality of isotopes in precipitation: A global perspective. J Geophys Res 114(D8):D08116.
- 39.Stroeve J. Sea Ice Trends and Climatologies from SMMR and SSM/I-SSMIS, Version 1. NASA National Snow and Ice Data Center Distributed Active Archive Center; Boulder, CO: 2003. [Google Scholar]
- 40.Dee DP, et al. The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Q J R Meteorol Soc. 2011;137(656):553–597. [Google Scholar]
- 41.Stone RS, Dutton EG, Harris JM, Longenecker D. Earlier spring snowmelt in northern Alaska as an indicator of climate change. J Geophys Res. 2002;107(D10):ACL 10-1–10-13. [Google Scholar]
- 42.Perovich DK. 2005. On the aggregate-scale partitioning of solar radiation in Arctic sea ice during the Surface Heat Budget of the Arctic Ocean (SHEBA) field experiment. J Geophys Res 110(C3):C03002.
- 43.Sall J. Leverage plots for general linear hypotheses. Am Stat. 1990;44(4):308–315. [Google Scholar]
- 44.Kalnay E, et al. The NCEP/NCAR 40-year reanalysis project. B Am Meteorol Soc. 1996;77(3):437–471. [Google Scholar]