Synoptic scale controls on the δ18O in precipitation across Beringia

Oxygen isotope records of precipitation (δ18Oprecip) from Beringia are thought to reflect synoptic‐scale circulation changes associated with the Aleutian Low. To delineate the spatial pattern of δ18Oprecip associated with the two dominant modes of Aleutian Low circulation, we combine modern δ18Oprecip and deuterium excess data with climate reanalysis and back trajectory modeling. Aleutian Low strength and position are revealed to systematically affect regional moisture source and δ18Oprecip; whereby a strengthened Aleutian Low causes lower (higher) δ18Oprecip in western (eastern) Beringia. We compare a new 100 year long δ18O record from the Aleutian Islands with the North Pacific Index, the primary indicator of Aleutian Low strength, and find a significant positive relationship (r = 0.43, P < 0.02, n = 28) that tracks late twentieth century change. This study demonstrates synoptic‐scale circulation controls on our isotope record and provides a coherent framework for interpreting existing and emerging paleoisotope data from the region.


Introduction
Oxygen isotope records of precipitation (δ 18 O precip ) from Beringia have been shown to reflect changes in atmospheric circulation associated with the Aleutian Low-a large-scale feature of mean low sea level pressure (SLP) [e.g., Anderson et al., 2005Anderson et al., , 2007Anderson et al., , 2011Fisher et al., 2004;Schiff et al., 2009;Clegg and Hu, 2010;Chipman et al., 2012;Jones et al., 2014]. As the leading feature of North Pacific climate, this system influences regional temperature and precipitation patterns, including the heat and moisture flux between the extratropical Pacific and Arctic [Zhu et al., 2007]. Collectively, these records are important for assessing the spatiotemporal heterogeneity of climate evolution prior to the instrumental record.
However, most isotope records are located on the eastern edge of this system, and there are discrepancies with interpreting the climate-isotope signal. For example, lower δ 18 O in the Mount Logan ice core [Fisher et al., 2004] and Jellybean Lake record [Anderson et al., 2005] are interpreted as an Aleutian Low intensification, while an isotope-enabled circulation model suggests the opposite [Field et al., 2010]. These inconsistencies call for an observation-based analysis of modern atmospheric processes affecting δ 18 O precip . This study advances these goals in two ways: 1. We use the available δ 18 O precip data from long-term monitoring stations across Beringia to delineate the spatial pattern of δ 18 O precip associated with two contrasting states of Aleutian Low circulation. The resulting pattern provides a framework for interpreting modern and paleoisotope records in context of regional atmospheric circulation. 2. We present the first 100 year long δ 18 O record from Adak in the Aleutian Islands, at the heart of the Aleutian Low. Fluctuations in δ 18 O are compared with instrumental records and examined in context of the mapped δ 18 O precip anomalies, deuterium excess values, and storm-track trajectories. This analysis provides a calibrated approach to interpreting downcore changes in δ 18 O and contributes to the growing network of records from this region that is needed to investigate long-term change in atmospheric circulation.
~9°E of its mean position ( Figure 1a) and is associated with the eastward expansion of the Asian trough and a stronger North American ridge [Rodionov et al., 2007]. Extratropical cyclones forming east of Japan propagate along a strengthened midlatitude westerly storm track and pump warm moist air and abundant precipitation into the Gulf of Alaska [Rodionov et al., 2007]. As the Aleutian Low shifts east, the northerly winds that circulate within its northwest quadrant draw polar air to the central and western Aleutian Islands (Figure 1b). During weak Aleutian Low winters the pressure minimum shifts west and often separates into two centers-one east of Kamchatka and one in the Gulf of Alaska ( Figure 1c) [Rodionov et al., 2007]. The North American ridge shifts to the central North Pacific, effectively blocking the typical west-to-east propagation of cyclones, and storm systems are steered northward over the Aleutian Islands and into the Bering Sea ( Figure 1d) [Rodionov et al., 2007].
These circulation patterns vary on interannual to decadal time scales and induce characteristic responses in precipitation, sea surface temperature (SST), and surface air temperature (SAT) [Mock et al., 1998]. Climate anomalies are well expressed in coupled modes of the North Pacific Index (NPI) [Trenberth and Hurrell, 1994] and the Pacific Decadal Oscillation (PDO) [Mantua et al., 1997] and are linked to remote El Niño-Southern Oscillation forcing via rossby wave propagation from the tropical Pacific [e.g., Bjerknes, 1966;Trenberth et al., 1998]. Specifically, a strong Aleutian Low (ÀNPI/+PDO) induces positive SSTs, SATs, and precipitation anomalies in the Gulf of Alaska and negative anomalies in the central North Pacific and the reverse during a weak Aleutian Low (+NPI/ÀPDO) [Mantua et al., 1997;Trenberth and Hurrell, 1994] ( Figure S1 in the supporting information). We hypothesize and test how these circulation patterns affect δ 18 O precip values across Beringia through systematic changes in moisture source and storm-track trajectory.  (Table S1 in the supporting information). Monthly deuterium excess (d-excess) values were calculated as δD-8.δ 18 O [Craig, 1961], and raw  [Trenberth and Hurrell, 1994]. A negative (positive) NPI is a strong (weak) Aleutian Low. Heart Lake is marked by a star; the black diamonds indicate Mount Logan (ML) and Jellybean Lake (JL). The black arrows highlight the main storm tracks. Data obtained from NCEP/NCAR V1 reanalysis [Kalnay et al., 1996].

Geophysical Research Letters
where P i and δ i denote the amount (mm) of each monthly precipitation sample and its measured oxygen isotope composition, respectively, and n represents the number of months. The North Pacific Index (NPI) is an instrumental index of Aleutian Low strength calculated as the area-weighted mean SLP over 30°N-65°N and 160°E-140°W [Trenberth and Hurrell, 1994]. NPI anomalies were calculated as departures from the 1928-1989 winter mean and used to determine whether each winter of GNIP data was characterized by a strong (ÀNPI) or weak (+NPI) Aleutian Low. Winter month δ 18 O precip values at each GNIP site were averaged for both Aleutian Low modes and are presented as anomalies from the long-term winter mean δ 18 O precip for each site.
To spatially interpolate δ 18 O precip the data were kriged [Keckler, 1995] using (a) the long-term winter average δ 18 O precip at each GNIP site and the average δ 18 O precip anomalies for (b) strong and (c) weak Aleutian Low winters at each GNIP site (see Text S1 in the supporting information). Cross validation was used to quantify the quality of prediction and the spatial distribution of errors within interpolated δ 18 O precip values [e.g., Davis, 1987] (Text S1 and Figure S2 in the supporting information). We note that interpolated values do not consider smaller-scale geographic parameters influencing δ 18 O precip (e.g., orography and temperature) and cannot be considered meaningful across regions with sparse data (e.g., central Alaska). Confidence intervals are greatly reduced in the vicinity of GNIP stations (±0.10‰, 1σ).
Trajectories of precipitating air masses were computed using the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model [Draxler and Hess, 1997]. Gridded hourly meteorological data [Kalnay et al., 1996] were used to calculate daily back trajectories to Adak for all months of available GNIP δ 18 O precip data (1962-1967 and 1972-1973) on the assumption that each trajectory contributed precipitation to the monthly composite δ 18 O precip sample. Back trajectory errors are an estimated 15-30% of the trajectory length and relate to how well the numerical fields estimate the true flow field in space and time. For full details, see Text S1 in the supporting information.

Sedimentary Analyses
Heart Lake is a small, freshwater throughflow system on Adak Island (51.87°N, 176.63°W). Lake volume is 8 × 10 5 m 3 , and average outflow is 5 × 10 4 m 3 /d (~15 day retention). Bottom lake water samples were collected for isotope analysis in June 2009 and 2010, with two surface gravity cores (09-AS-1A and 09-AS-1B) from the deepest part of the lake (7.6 m). The upper 12 cm of 09-AS-1B was sampled continuously at 0.5 cm resolution for radiometric dating ( 210 Pb, 226 Ra, 137 Cs, and 241 Am), and both cores were sampled at 0.5 cm resolution for biogenic silica (BSi) and the oxygen isotope composition of diatom silica (δ 18 O diatom ) (see Text S2 and Table S3 in the supporting information).
Adak mean annual δ 18 O precip is À9.00‰, while mean winter (December-February) and summer (June-August) values are À10.41‰ and À8.40‰, respectively. There is no correlation between monthly δ 18 O precip and SAT (r = 0.15, n = 72) or precipitation amount (r = 0.03, n = 72). Seasonal δ 18 O precip differences (±2.01‰) broadly reflect progressive loss of 18 O in water vapor during the poleward transport and condensation of marine air Geophysical Research Letters 10.1002/2015GL063983 masses from the midlatitudes to Adak during winter, leading to a climatological decrease of δ 18 O in vapor (and thereby precipitation) with temperature [e.g., Dansgaard, 1964].
The spatial distribution of mean δ 18 O precip during strong and weak Aleutian Low winters are presented as departures from the long-term winter means (Figures 2b and 2c). During a strong Aleutian Low, winter δ 18 O precip anomalies exhibit a dipole pattern over western Alaska (negative) and the Yukon (positive) (Figure 2b). Higher δ 18 O precip in the Yukon can be attributed to enhanced cyclonic circulation and meridional flow delivering warm moist Pacific air enriched in 18 O (Figure 1b), consistent with modeled δ 18 O precip output by Field et al. [2010]. The spatial δ 18 O precip pattern during a weak Aleutian Low demonstrates the opposite: higher δ 18 O precip in western Alaska and lower δ 18 O precip in the Yukon and Kamchatka (Figure 2c). Differences between strong and weak Aleutian Low δ 18 O precip values are significant (P < 0.05) and greatly exceed interpolation error ( Figure S2 in the supporting information).
Barrow δ 18 O precip anomalies highlight the sensitivity of Arctic Alaska to North Pacific circulation patterns, which likely relates to latitudinal migration of the polar front. Persistent winter sea ice implies a distal moisture source; hence, a 52% increase in total precipitation during a weak Aleutian Low (data available Values in Figures 2b and 2c are expressed as anomalies from the long-term winter mean for each GNIP site (circle). The spirals mark mean Aleutian Low positions; the dotted curves plot the mean moisture trajectories during strong (blue) and weak (red) Aleutian Low circulation during the corresponding months and were modeled using HYSPLIT [Draxler and Hess, 1997]. The dots represent hourly intervals along each trajectory. Sites referred to in text are marked by diamonds: Heart Lake, Mount Logan (ML), and Jellybean Lake (JL). Data and all trajectories are listed in Table S1 and Figure S2 in the supporting information.

Processes Controlling δ 18 O precip on Adak Island
Mean winter δ 18 O precip is significantly (P < 0.05) lower on Adak during a strong Aleutian Low (À1.14‰, n = 14) compared to a weak Aleutian Low (+0.44‰, n = 7) (Figures 2b  and 2c). Furthermore, Adak monthly back trajectories indicate systematic differences in air mass source and trajectory. During a strong Aleutian Low, back trajectories at 48 h originate in the north and span an area from 48-85°N and 154-230°E (Figure 2b), and during a weak Aleutian Low, back trajectories at 48 h originate in the west/southwest and span an area from 29-65°N and 140-200°E (Figure 2c).
We propose that changes in the strength and position of the Aleutian Low shift the moisture source location for precipitation falling at Adak and across Beringia. When the SLP minimum is near Adak (strong Aleutian Low), precipitating air masses from the north bring vapor and precipitation relatively depleted in 18 O (Figures 1b and 2b), as well as lower-than-average winter temperatures (À0.5°C) and increased snowfall (>556 mm). Unlike liquid precipitation, snow (and other solid precipitation) will not equilibrate with ambient moisture by continuous exchange during the descent from cloud base to the ground [Dansgaard, 1964], and hence, any moisture source "effects" will be retained.
Alternatively, increased temperatures and rainfall (>148 mm) during a weak Aleutian Low reflect more prevalent southwesterly storm tracks (Figure 1d) [e.g., Rodionov et al., 2007]. These storms carry warm 18 O-enriched vapor to Adak from as south as 29°N (48 h prior) (Figure 2c and Figure S3 in the supporting information). Precipitation has relatively high δ 18 O values, and these are likely amplified by postcondensational exchange with the ambient moisture as it falls to the ground [e.g., Dansgaard, 1964;Pfahl et al., 2012].

Deuterium Excess and Moisture Source
Deuterium excess (d-excess [Craig, 1961]) is determined by evaporative conditions of the source region (e.g., relative humidity and SST) [see Pfahl and Sodemann, 2014], as well as kinetic processes during vapor transport and condensation (e.g., ice-crystal formation and reevaporation) [Dansgaard, 1964;Jouzel and Merlivat, 1984;Gat, 1996]. Adak winter month d-excess and δ 18 O precip negatively correlate (r = À0.56, P < 0.01, n = 48) (Figure 3), and HYSPLIT output shows that d-excess values of Arctic-derived precipitation (+9.6‰) are significantly (P < 0.01) higher than Pacific derived (+4.6‰). These differences reflect the slower diffusivity of H 2 18 O relative to HDO during evaporation above the ocean surface [Dansgaard, 1964]. Specifically, high d-excess values arise when there is insufficient time for vapor to equilibrate between the saturated ocean surface layer and the subsaturated atmosphere [e.g., Pfahl and Sodemann, 2014]. Evaporated moisture entrained in the Arctic (as opposed to Pacific) is subject to such conditions, whereby large-humidity gradients between the ocean surface and the dry atmosphere above-particularly at the sea ice margin [e.g., Kurita, 2011]-will lead to strong nonequilibrium (kinetic) fractionation and an evaporate characterized by relatively high d-excess and low δ 18 O [e.g., Gat et al., 2003;Uemura et al., 2008;Pfahl and Sodemann, 2014]. Furthermore, Adak receives increased snowfall during months with Arctic-derived moisture, and this will also be characterized by high d-excess due to nonequilibrium condensation during ice particle growth [Jouzel and Merlivat, 1984].

. Heart Lake Aleutian Low Proxy Record
We use our δ 18 O diatom record from Adak to determine whether changes in the Aleutian Low are captured in natural archives of δ 18 O precip . Diatom silica is thought to precipitate in isotopic equilibrium with the surrounding water, and therefore, δ 18 O diatom resembles the δ 18 O of the lake water (δ 18 O LW ) in which the diatom grows at a specific temperature [Leng and Barker, 2006]. Heart Lake is a throughflow system with short retention (~15 days), and its primary component is precipitation it receives directly, and indirectly through surface water runoff and groundwater input. Accordingly, Heart Lake δ 18 O LW resembles measured δ 18 O precip at Adak ( Figure S6 in the supporting information). Previous diatom and δ 18 O diatom analyses in Heart Lake rule out potential disequilibrium effects [Bailey et al., 2014]; hence, δ 18 O diatom should reflect δ 18 O precip changes associated with the Aleutian Low.
The North Pacific Index (NPI) is a direct instrumental index of Aleutian Low strength [Trenberth and Hurrell, 1994]. Each 0.25 cm thick sample analyzed for δ 18 O diatom represents approximately 1-5 years, so our δ 18 O diatom time series is compared with the 5 year running mean of winter NPI values (monthly NPI values averaged over NovemberÀFebruary) from 1900 to 2009. We use winter NPI values because in late spring, when the main diatom bloom occurs [Bailey et al., 2014], we estimate that winter snowpack melt exceeds spring precipitation input. This estimate is based on local long-term  mean March snow depths (250 mm) which, assuming uniform snow distribution across the catchment (8 km 2 ) and a mean density of 250 kg/m 3 for settled snow [Cuffey and Paterson, 2010], equate to~5 × 10 5 m 3 water equivalent entering Heart Lake during the spring melt. This volume exceeds spring precipitation inputs (~4 × 10 5 m 3 ); hence, Heart Lake spring δ 18 O LW values will trend toward winter δ 18 O precip values [e.g., Kortelainen and Karhu, 2004;Jonsson et al., 2009]. This is evidenced by Heart Lake δ 18 O LW samples taken in June (mean = À9.5‰), which are more comparable to mean January (À9.3‰) than mean June (À8.1‰) δ 18 O precip values.
Raw δ 18 O calcite data from Jellybean Lake [Anderson et al., 2005] (available from http://www.ncdc.noaa.gov) negatively correlate with the NPI over the 100 year instrumental period (r = À0.48, P < 0.02) (Figure 5d), although the inverse correlation (r = 0.48) is stated in Anderson et al. [2005]. Our revised calculation is supported by the mapped GNIP δ 18 O precip anomalies ( Figure 2) and by Field et al. [2010], whose isotopeenabled general circulation model simulates higher δ 18 O precip around the Gulf of Alaska during a strengthened Aleutian Low.

Conclusions
We demonstrate a west-east dipole pattern in δ 18 O precip associated with synoptic-scale atmospheric circulation over the North Pacific. Analyses of δ 18 O precip reveal that the dominant effect of a strengthened Aleutian Low is higher (lower) δ 18 O precip in the east (west) and this can be attributed to water vapor origin and moisture trajectory . This new δ 18 O record from sedimentary diatoms in Heart Lake demonstrates their ability to capture changes in synoptic-scale atmospheric circulation. Furthermore, our interpolation scheme demonstrates the paucity of δ 18 O precip data across Beringia and highlights target areas for future sampling. Figure 5. Comparison of (a) twentieth century North Pacific Index (NPI) [Trenberth and Hurrell, 1994] with (b) Heart Lake δ 18 O diatom (this study), (c) Jellybean Lake δ 18 O calcite [Anderson et al., 2005], and (d) Mount Logan ice δ 18 O [Moore et al., 2002]. Values are expressed as anomalies from the 100 year mean for each record (black lines).