1. Introduction
The Gulf of Alaska region, which includes Alaska, Yukon and parts of British Columbia, contains over 40 mm of sea level equivalent in its many alpine glaciers and icefields (Farinotti and others, Reference Farinotti2019). The region also contains one of the planet’s most dramatic barriers to inland transport of coastal moisture, with peaks exceeding 5000 m a.s.l. located less than 100 km from the coast (Newman and others, Reference Newman, Clark, Wood and Arnold2020). Orographic precipitation resulting from this topography feeds the icefields of coastal mountain ranges and the glaciers that drain them (Marcus and Ragle, Reference Marcus and Ragle1970). High-alpine glaciers are also found in the region’s interior, most notably in the Alaska Range (central Alaska, Fig. 1).
Figure 1. Study sites in the Gulf of Alaska region. Instrumental accumulation records are marked with yellow stars; ice-core locations discussed in this study are marked with blue diamonds. Peaks 
$ \gt 4,000$ m a.s.l. are marked with gray triangles. The Alaska (AK), Yukon Territory (YT) and British Columbia (BC) borders are shown in gold in panel (a). Panel (c) shows study sites in the Alaska Range; panel (d) shows study sites in the St. Elias Range. Stars in the Alaska Range and St. Elias Range in panel (b) correspond to 14 Camp and Divide, respectively.
The proximity of coastal and interior glaciated alpine sectors in the Gulf of Alaska region has presented opportunities to collect a unique suite of ice-core climate records located close to one another, but spanning distinct climate regimes (Fisher and others, Reference Fisher2004; Yalcin and others, Reference Yalcin, Wake, Kreutz, Germani and Whitlow2007; Bieniek and others, Reference Bieniek2012; Zdanowicz and others, Reference Zdanowicz2014; Tsushima and others, Reference Tsushima2015; Osterberg and others, Reference Osterberg2017; Winski and others, Reference Winski2017; Fang and others, Reference Fang2023). Such cores provide hundreds of years of climate data at sub-annual resolution, with additional deeper, lower-resolution ice reaching as much as 30 000 years in age (Fisher and others, Reference Fisher2004; Zdanowicz and others, Reference Zdanowicz2014; Fang and others, Reference Fang2023). However, the heterogeneity of the terrain challenges interpretation of such records; sites in relatively close proximity may be subject to different atmospheric processes, moisture sources and depositional mechanisms. This may result in dissimilar climate records that can, when used together, provide a more nuanced picture of regional climate than the record from any one site alone.
For example, a doubling of snow accumulation recorded in an ice core from the Begguya summit plateau (We use the indigenous name ‘Begguya’, rather than the more recent ‘Mt. Hunter’, Fig. 1c) between approximately 1840 and 2016 has been interpreted to reflect a combination of regional atmospheric warming and concurrent strengthening of the Aleutian Low (Winski and others, Reference Winski2017), which could occur due to greater duration or strength of individual storms, higher storm frequency, greater coherence between individual storm trajectories to a well-defined track, or some combination. However, an unresolved question is why the Begguya accumulation record shows such a distinct increase, while a concurrent regional ice-core record from Eclipse Icefield (Kelsey and others, Reference Kelsey, Wake, Yalcin and Kreutz2012) does not (Figs. 1d and 2)?
Figure 2. Ice-core accumulation records from Eclipse Icefield (blue; Kochtitzky and others, Reference Kochtitzky2020) and Begguya (orange; Winski and others, Reference Winski2017). Panel (b) shows a close-up of the dotted inset in panel (a) for the time period over which the two records overlap. Bolded lines show an 11-year smoothing.
Here, we combine in situ snow accumulation data and back trajectory modeling to examine the air-parcel trajectories associated with snowfall in the St. Elias (e.g., Eclipse Icefield) and Alaska (e.g., Begguya) Ranges, and their implications for historical accumulation records contained in the two ranges’ glacier ice. Our work offers a new justification for interpreting multiple regional ice-core records in the context of one another, rather than as standalone climate histories.
2. Methods
2.1. Instrumental records of snow accumulation
Accumulation data were obtained from snow sounders at two different sites: the Kaskawulsh/Hubbard Divide (St. Elias Range) and Denali 14 Camp (Alaska Range). We use the term ‘snow sounder’ to refer to instruments used to measure changes in the snow surface position, from which accumulation and ablation can be calculated. To quantify snowfall conditions at each site, we define non-storm accumulation days (referred to as ‘snow’ days) to be days (defined midnight to midnight, local time) on which a surface change greater than 2 cm and less than 20 cm was detected, and ‘storm’ days when at least 20 cm of snow accumulated. We use a value of 20 cm within 24 hours as it is the threshold used by both Environment Canada and the United States National Weather Service for issuing public snowfall warnings (National Oceanic and Atmospheric Administration, 2009; Environment and Climate Change Canada, 2024). We use a threshold of 2 cm as our minimum detection limit for snowfall to account for instrumental error (noted below) and wind redistribution.
We calculate total snowfall by summing measured surface height increases, ignoring surface drops (e.g., due to surface melt, snow densification or wind redistribution). Although wind distribution may contribute to our measured surface height increases, its effect is likely small relative to snowfall, given the locations of our snow sounder sites away from any defined lee or stoss side of a ridge. Additionally, although we observe the effects of settling and compaction following snowfall events, we assume that these processes primarily affect snow surface height after a precipitation event is over, such that seasonal differences in these processes can be neglected when focusing, as we do, on surface changes during the event. We recognize, however, that the effects of both surface settling and wind redistribution remain uncontrolled sources of uncertainty in snowfall estimates for each snowfall event.
To evaluate the representativeness of our temporally limited snow sounder data, we compare monthly accumulation from our snow sounder records to monthly ERA5 total precipitation over each site from 1940 to 2025. ERA5 has a footprint of 31 km, so it provides a spatially integrated estimate of precipitation in each of our two mountain ranges, rather than point measurements like our snow sounder data. We compare the two datasets, not to assess the accuracy of specific snow accumulation values, but to evaluate whether our snow sounder data are representative of the seasonal distribution of snowfall in each mountain range. Although ERA5 is not able to capture specific extreme precipitation events, it can generally capture spatial and temporal patterns of observed extratropical precipitation (Lavers and others, Reference Lavers, Simmons, Vamborg and Rodwell2022).
2.1.1. The Kaskawulsh/Hubbard divide (St. Elias Range)
The Kaskawulsh/Hubbard Divide (
$60^{\circ}42' \mathrm{N}$, 
$139^{\circ}47' \mathrm{W}$; 2603 m a.s.l.) is situated in the accumulation zone near the top of Kaskawulsh and Hubbard Glaciers (Fig. 1). The Kaskawulsh/Hubbard snowfall record comprises twice-daily readings from a Campbell Scientific SR50 snow depth sounder covering the period from June 2003 to July 2012 obtained by . Instrumental accuracy is 
$\pm1\,\mathrm{cm}$ or 0.4% of the distance to the snow surface, whichever is greater (McConnell, Reference McConnell2019). Six (2004, 2005, 2006, 2008, 2009, 2011) of the 10 hydrological years in the record have at least partial (0–90%) coverage in all months (Fig. 3). We define hydrological years to run from 1 September to 31 August and designate them based on the calendar year in which they end (e.g., hydrological year 2003 runs from September 2002 through August 2003). We begin our hydrological years in September rather than October to better capture the autumn storm season (approximately September through December) that is characteristic of the region, as discussed below (Fig. 3).
Figure 3. Instrumental records of snow accumulation from the Kaskawulsh/Hubbard Divide (a) and Denali 14 Camp (b). Accumulation is shown in meters relative to the first snow accumulation data point from each hydrological year represented. Panels (c) and (d) show the temporal coverage of the Divide and 14 Camp records, respectively.
2.1.2. Denali 14 camp (Alaska Range)
We obtained hourly accumulation data from a snow sounder at Denali’s 14 200 ft climbers’ camp (‘14 Camp’; 
$63^{\circ}04' \mathrm{N}$, 
$151^{\circ}04' \mathrm{W}$; 4328 m a.s.l.; Fig. 1), which is located on the Kahiltna Glacier and sits 
$\sim$1700 m higher in elevation than the Kaskawulsh/Hubbard Divide. The 14 Camp snow accumulation record comprises hourly readings from a Campbell Scientific SR50a snow depth sounder covering the period from June 2018 to September 2023 (Fig. 3d). Instrumental accuracy is 
$\pm 1\,\mathrm{cm}$ or 0.4% of the distance to the snow surface, whichever is greater. Of the seven hydrological years included in the 14 Camp record (2018–24), four of them (2019–22) have at least partial coverage during all months.
2.2. HYSPLIT atmospheric back trajectories
In order to determine whether and/or how air transport patterns differ between sites in the Alaska Range vs the St. Elias Range, we used NOAA’s HYSPLIT trajectory model to run back trajectories for the atmospheric transport of moisture to three sites of interest in the St. Elias and Alaska Ranges (Fig. 1): the Kaskawulsh/Hubbard Divide, Eclipse Icefield (‘Eclipse’) and the Begguya summit plateau (‘Begguya’). Eclipse and Begguya were chosen as they are sites in the region where ice cores have been recovered (Wake and others, Reference Wake, Yalcin and Gundestrup2002; Yalcin and others, Reference Yalcin, Wake, Kreutz, Germani and Whitlow2007; Kelsey and others, Reference Kelsey, Wake, Yalcin and Kreutz2012; Winski and others, Reference Winski2017; Reference Winski2018. Because the NCEP/NCAR meteorological data input to the HYSPLIT model has a resolution on a 
$2.5^{\circ} \times 2.5^{\circ}$ lat/lon grid and 17 pressure levels from 1000 to 10 mb, we interpret trajectories for Begguya to be representative of trajectories to 14 Camp; the two sites are 15 km apart with 300 m difference in elevation (Fig. 1c). We also consider Eclipse and the Kaskawulsh/Hubbard Divide climatically analogous given our input data, as Eclipse is located 30 km from the Divide with a 400 m difference in elevation (Fig. 1d).
Interpretation of both Begguya and Eclipse ice-core records has to date relied on the assumption that snowfall at each of these sites is associated with transport of moisture from the Gulf of Alaska to the ice-core site. However, accumulation and chemistry records differ between the two ice-core sites, implying that this simple model of inland moisture transport is not sufficient to describe snowfall and resultant chemical signals across the Gulf of Alaska region. Here, we use HYSPLIT to provide additional detail about air-parcel trajectories associated with snowfall in order to inform differences in the interpretation of ice-core records, focusing on accumulation records, from different sites in the Gulf of Alaska region.
Air-parcel back trajectories (four per day) were computed by the HYSPLIT atmospheric trajectory model between 1 January 1979 and 31 December 2019 using NCEP/NCAR v1 meteorological inputs. Each back trajectory position was computed at hourly time steps for a length of 10 days, beyond which we consider increasing model uncertainties to render outputs insufficiently meaningful.
Model outputs (latitude, longitude and air-parcel elevation values) were examined by month and by season for three snow conditions based on our snow sounder records: days without accumulation (‘no-snow’, 
$ \lt 2\,\mathrm{cm}$), days with non-storm accumulation (‘snow’, 2–20 cm) and storm days (‘storm’, 
$\geq$20 cm). In order to better isolate patterns associated with accumulation at the Kaskawulsh/Hubbard Divide, we identified no-snow, snow and storm days in the site’s snow sounder record, both over the course of the entire year and also only within the months of September through December (SOND), when most accumulation occurs (Fig. 4). At 14 Camp, in contrast, we did not isolate September through December data, as snow accumulation is more evenly distributed throughout the year, as discussed below.
Figure 4. Monthly snow sounder accumulation from the St. Elias (Kaskawulsh/Hubbard Divide, (a)) and Alaska (14 Camp, (b)) Ranges compared with monthly ERA5 precipitation value from 1940 to 2025. Snow sounder data are shown with filled-in scatterpoints. ERA5 data are shown with boxplots with diamond-shaped outliers. We use a typical new-snow density value of 
$200\,\mathrm{kg\ m}^{-3}$ when converting snow sounder accumulation to m w.e. (Cuffey and Paterson, Reference Cuffey and Paterson2010).
Finally, to contextualize our sites of interest in the broader Gulf of Alaska climate regime we identify years since 1979 with a particularly strong or weak Aleutian Low, and compare the HYSPLIT back trajectories for these years. The center position of the Aleutian Low is the average position of extratropical cyclones moving west to east across the North Pacific, and its location and intensity are primary indicators of wintertime North Pacific climate (Trenberth, Reference Trenberth1990; Overland and others, Reference Overland, Adams and Bond1999; Rodionov and others, Reference Rodionov, Bond and Overland2007). We define years since 1979 with the bottom 20% of November–March North Pacific Index (the area-weighted sea level pressure over the region 
$30^{\circ} \mathrm{N}-65^{\circ} \mathrm{N}$, 
$160^{\circ} \mathrm{E}-140^{\circ} \mathrm{W}$; ) values as strong Aleutian Low years, and years with the top 20% of North Pacific Index values as weak Aleutian Low years.
3. Results
We report results for each of our snow sounder sites individually, then discuss them in relation to one another. No-snow, snow and storm days at each snow sounder site are summarized in Table 1. When reporting and discussing results, we refer to the St. Elias Range (or ‘St. Elias’) and Alaska Range rather than individual snow sounder or ice-core sites since, as stated above, we view the snow sounder and ice-core sites in each range as climatically analogous. We use this language for clarity, recognizing that our results are not representative of each mountain range as a whole. Therefore, in the context of this paper, the term ‘Alaska Range’ should specifically be taken to mean 14 Camp and Begguya, and the term ‘St. Elias Range’ to mean the Kaskawulsh/Hubbard Divide and Eclipse.
Table 1. No-snow, snow and storm days at snow sounder sites: Kaskawulsh/Hubbard (K/H) Divide and 14 Camp. No-snow days are defined as days with a surface increase of less than 2 cm. Snow days are defined as days with a surface increase between 2 and 20 cm. Storm days are defined as days with a surface increase of at least 20 cm in 24 hours.
3.1. Instrumental records of snow accumulation
Precipitation in the St. Elias is predominantly controlled by coastal storm systems moving inland from the Gulf of Alaska. Forty-four percent of annual accumulation in the St. Elias occurs in September through December, 40% of this during discrete storm events, despite storms occurring on only 2.6% of total September–December days on record (Fig. 3a).
Accumulation in the Alaska Range is more evenly distributed throughout the year than in the St. Elias; August receives, on average, the most snowfall, but there is no distinct fall/winter influx of accumulation (Figs. 3 and 4).
We do not have continuous in situ surface density measurements for converting monthly accumulation values from our snow sounders to meters w.e. for comparison with ERA5 precipitation (1940–2025). We therefore convert to m w.e. using three different values: 
$200\,\mathrm{kg\ m}^{-3}$ (typical new-snow density; Cuffey and Paterson, Reference Cuffey and Paterson2010), 
$470\,\mathrm{kg\ m}^{-3}$ (surface density measured at Eclipse; McConnell, Reference McConnell2019) and 
$400\,\mathrm{kg\ m}^{-3}$ (surface density measured at the Kaskawulsh/Hubbard Divide; Ochwat and others, Reference Ochwat, Marshall, Moorman, Criscitiello and Copland2021). When converted using 
$200\,\mathrm{kg\ m}^{-3}$, monthly precipitation values from our snow sounder records generally fall within the range of ERA5 precipitation from 1940 to 2025. However, snow sounder monthly precipitation tends to be higher than ERA5 precipitation when converted using 470 and 
$400\,\mathrm{kg\ m}^{-3}$. This may be a result of surface compaction between snowfall and measurement of surface density, an underestimation of precipitation in ERA5, or some combination. Regardless, seasonal patterns of snow accumulation are consistent between snow sounder data and ERA5 precipitation values at both sites (Fig. 4).
3.2. HYSPLIT atmospheric back trajectories
HYSPLIT results demonstrate that the St. Elias tends to receive air parcels uplifted from lower elevations as they travel from the southwest, an effect that is amplified locally by the high relief coastline surrounding the Gulf of Alaska (Fig. 5). During the spring and summer, air parcels tend to arrive in the St. Elias from the Gulf of Alaska and northern Pacific (Fig. S1a and b). In the fall, the area from which air parcels originate elongates across the Pacific toward Asia, then shifts south in the winter (Fig. S1c and d). We report the following patterns for no-snow vs snow vs storm days specifically for the months of September through December, when the St. Elias receives most of its snowfall. During no-snow days in the St. Elias, arriving air parcels do not follow any well-defined track (Fig. 6a). In contrast, during snow, and especially during storm, days in the St. Elias, arriving air parcels follow a distinct track inland from the Gulf of Alaska, with a more southerly source region during storm days, implicating the Gulf of Alaska and North Pacific as dominant moisture sources for St. Elias fall and winter storm events (Fig. 6c and e). Snow and storm days in most individual years conform to this pattern, with 2004 being the most notable exception (Figs S2 and S3). Individual storm tracks are shown in Fig. S4.
Figure 5. Fraction of HYSPLIT air trajectories passing through each 
$1^{\circ} \times 1^{\circ}$ (a–b) or 
$1^{\circ} \times 1\,\mathrm{m}$ (c–f) parcel en route to the St. Elias Range and Alaska Ranges. Panels (a), (c) and (e) show latitude 
$\times$ longitude, latitude 
$\times$ altitude and longitude 
$\times$ altitude parcels for air parcels en route to the Kaskawulsh/Hubbard Divide. Panels (b), (d) and (f) show latitude 
$\times$ longitude, latitude 
$\times$ altitude and longitude 
$\times$ altitude parcels for air parcels en route to Begguya. The locations of the Kaskawulsh/Hubbard Divide and Begguya are marked by yellow stars in their respective panels. Note the log scale on the colorbar.
Figure 6. September through December air-parcel transport to the St. Elias and Alaska Ranges based on snowfall conditions at the Kaskawulsh/Hubbard Divide (St. Elias Range). For both the St. Elias and Alaska Ranges, HYSPLIT air-parcel trajectory anomalies over each 
$1^{\circ} \times 1^{\circ}$ parcel are shown leading to no-snow (a–b), snow (c–d) and storm (e–f) days recorded at the Kaskawulsh/Hubbard Divide (marked by a yellow star). The anomaly for each parcel is calculated relative to the fraction of air mass trajectories passing over that parcel under all snow conditions. The dominant patterns of airflow (as visually identified) that precede each snowfall condition at the Kaskawulsh/Hubbard Divide are shown by white arrows. Note the log scale on the colorbar.
The Alaska Range also tends to receive air parcels uplifted from lower elevations as they travel from the southwest, indicating that most precipitation at the site is orographically induced (Fig. 5). Westerly trajectories toward the Alaska Range are more common in the summer than during other seasons (Fig. S1). The area from which air parcels arriving in the Alaska Range originate elongates across the Pacific toward Asia during the fall and winter relative to the summer (Fig. S1).
The strength of the Aleutian Low is an indicator of North Pacific wintertime climate, and HYSPLIT trajectories show that strong Aleutian Low conditions are associated with greater funneling of air parcels northwest from the Gulf of Alaska, while under a weaker Aleutian Low, the overall back-trajectory pattern is far more scattered, with comparatively more air parcels originating from a westerly direction in the Bering Sea region (Fig. 8).
4. Discussion
4.1. Regional snowfall comparisons: St. Elias vs Alaska Range
The results presented above highlight many similarities, but also some notable differences, between snowfall regimes and moisture sources in the Alaska and St. Elias Ranges. In both ranges, each large snowfall event is followed by a drop in snow surface height over the subsequent days, reflecting settling and compaction of the snowpack (Fig. 3a and b). Both the Alaska and St. Elias Ranges tend to receive air parcels uplifted from lower elevations as they travel from the southwest, indicating orographic precipitation. For both ranges, the area from which arriving air parcels originate elongates across the Pacific toward Asia during the fall and winter relative to the summer.
Accumulation zone snowfall records from the St. Elias and Alaska Ranges show a clear distinction between maritime and continental climate regimes, respectively. The St. Elias Range’s proximity to the Gulf of Alaska makes it more sensitive to the fall and winter weather systems that develop offshore and move inland; we find that 44% of annual snow accumulation in the St. Elias falls in the months of September through December, often accumulating in discrete storm events. In contrast, the Alaska Range is located in the Alaskan interior and shows an annual snowfall distribution less dominated by fall and winter storms originating from the Gulf of Alaska, with only 35% of annual accumulation occurring during September–December.
Overall mean annual accumulation (given in meters water equivalent calculated with a snow density value of 
$200\,\mathrm{kg\ m}^{-3}$; Cuffey and Paterson, Reference Cuffey and Paterson2010) is not significantly different between the St. Elias (
$0.59 \pm 0.17$ m w.e.) and Alaska Ranges (
$0.45 \pm 0.13$ m w.e.), though without temporal overlap in the snow sounder records from the two sites no direct comparison can be made. Modeled regional precipitation in Alaska and the Yukon (Newman and others, Reference Newman, Clark, Wood and Arnold2020) shows near-maximum levels in both the St. Elias and Alaska Ranges, although the zone of maximum precipitation is much more restricted in the Alaska Range compared to the St. Elias.
Combined HYSPLIT results from the St. Elias and Alaska Ranges suggest that while both locations are impacted similarly by large storm systems moving inland from the Gulf of Alaska, non-storm accumulation in the two ranges differs and is largely governed by orography (Figs. 1b, 6 and 7). During snow days in the St. Elias, trajectories for the range come predominantly from the south, moving inland from the Gulf of Alaska, similar to during storm days but less defined (Fig. 6c). Concurrent trajectories for the Alaska Range, in contrast, are more dispersed (Fig. 6d), suggesting that while weather systems moving in from the Gulf of Alaska dominate both storm and non-storm accumulation in the St. Elias, many smaller coastal systems weaken before reaching the interior.
At Cook Inlet, however, there is an approximately 100 km gap in the modeled coastal high-precipitation band, north of which sits the Alaska Range, where another, smaller zone of high precipitation is located (Newman and others, Reference Newman, Clark, Wood and Arnold2020). Our results support the idea that when non-storm coastal systems do reach the Alaska Range, they do so where Cook Inlet cuts through the coastal mountains (Fig. 7). Because there is no comparable feature cutting through the coastal peaks between the Gulf of Alaska and the interior St. Elias icefields, small weather systems may be blocked from reaching the Kaskawulsh/Hubbard Divide even if they are able to access the Alaska Range. The lack of southerly air-parcel trajectories in the St. Elias associated with non-storm snowfall in the Alaska Range is consistent with this idea.
Figure 7. Air-parcel transport to the St. Elias and Alaska Ranges based on snowfall conditions at 14 Camp (Alaska Range). For both the St. Elias and Alaska Ranges, HYSPLIT air-parcel trajectory anomalies over each 
$1^{\circ} \times 1^{\circ}$ parcel are shown leading to no-snow (a–b), snow (c–d) and storm (e–f) days recorded at 14 Camp (marked by a yellow star). The anomaly for each parcel is calculated relative to the fraction of air mass trajectories passing over that parcel under all snow conditions. The dominant patterns of airflow (as visually identified) that precede each snowfall condition at 14 Camp are shown by white arrows. Note the log scale on the colorbar.
4.2. The Aleutian Low and Gulf of Alaska precipitation regime
HYSPLIT trajectories for storm days recorded at our two snow sounder sites show similar trajectories in both the St. Elias and Alaska Ranges (Figs. 6e and f and 7e and f). The importance of large Gulf of Alaska storm systems to regional climate can also be seen in the HYSPLIT trajectories during periods of a strong vs weak Aleutian Low (Fig. 8). Each year, the Aleutian Low undergoes a seasonal westward migration; in early fall, storm deepening occurs primarily in the Gulf of Alaska, while in late fall it shifts south of Kamchatka (Pickart and others, Reference Pickart, Moore, Macdonald, Renfrew, Walsh and Kessler2009). This migration is reflected in the seasonal westward shift in HYSPLIT trajectories to both the St. Elias and Alaska Ranges (Fig. S1).
Figure 8. Air-parcel trajectories during years with a strong and weak Aleutian Low defined by anomalies from the 1925–89 mean (Trenberth and Hurrell, Reference Trenberth and Hurrell1994b). Anomaly plots of the fraction of trajectories passing over each 
$1^{\circ} \times 1^{\circ}$ parcel are shown for the St. Elias Range (Kaskawulsh/Hubbard Divide; a, c) and to the Alaska Range (Begguya; b, d). The anomaly for each parcel is calculated relative to the fraction of air mass trajectories passing over that parcel under all snow conditions. Each back-trajectory starting cell is marked by a yellow star. Note the log scale on the colorbar, which is different than the log scale on all the snow condition anomaly plots.
A stronger Aleutian Low can occur as a result of increased duration or strength of individual storms, increased storm frequency, greater coherence between individual storm trajectories to a well-defined track, or some combination of the three, complicating the relationship between the Aleutian Low and regional snow accumulation. Generally, strong Aleutian Low conditions are associated with more storms of high intensity and a well-defined storm track compared to weak Aleutian Low conditions (Rodionov and others, Reference Rodionov, Bond and Overland2007; Zhu and others, Reference Zhu, Sun, Liu, Liu and Martin2007). HYSPLIT trajectories to both the St. Elias and Alaska Ranges are indeed much more coherent under a strong Aleutian Low, when air parcels are funneled in from the south and southeast, off the Gulf of Alaska (Fig. 8). Conversely, under a weak Aleutian Low, air-parcel trajectories are more dispersed. Our results suggest that a ‘typical’ strong Aleutian Low, characterized by more storms of high intensity, leads to snowfall in the Gulf of Alaska region, more defined by the size and frequency of storm events and less moderated by regional topography. Under this type of strong Aleutian Low, the position of the dominant storm track may be less important, as large storm systems affect both coastal and interior sectors. Conversely, a ‘typical’ weak Aleutian Low, characterized by a higher proportion of non-storm accumulation, leads to greater topographic moderation of snowfall in the Gulf of Alaska region. Under this type of weak Aleutian Low, storm track position and interaction with topography play a comparatively more important role in controlling the regional distribution of snow accumulation.
The future of the Gulf of Alaska precipitation regime will be largely determined by the future of stormy weather in the region. A northward shift and intensification of the Aleutian Low, and specifically an intensification of extreme Aleutian Low events, is projected under IPCC scenario RCP8.5 (Gan and others, Reference Gan2017; Giamalaki and others, Reference Giamalaki, Beaulieu, Henson, Martin, Kassem and Faranda2021). A northward shift of the Aleutian Low center aligns slightly more closely with storm trajectories toward the St. Elias than the Alaska Range (Figs. 6 and 7). A more intense Aleutian Low likely means more storms of higher intensity funneled inland from the Gulf of Alaska, which would increase snowfall throughout the region, even on areas sheltered by orographic barriers during less intense snow events.
4.3. Implications for ice-core records
The difference we see in St. Elias and Alaska Range accumulation patterns suggests that interpreting ice-core accumulation records from the two locations in the context of one another can prove a powerful tool for untangling past patterns of air-parcel movement. Our HYSPLIT results show the most distinct air-parcel trajectories on storm days, which are associated with southerly storm tracks moving inland from the Gulf of Alaska toward both the St. Elias and Alaska Ranges. Although neither of the sites show a single defined air-parcel trajectory for non-storm snow days, they do demonstrate patterns of air transport distinct from each other: snow days in the St. Elias are associated with southerly trajectories in both the St. Elias and Alaska Ranges, indicating southerly transport of storm systems strong enough to carry moisture over the first band of coastal high peaks, reaching the rest of the St. Elias and the Alaska Range. In contrast, snow days in the Alaska Range are associated with southerly trajectories in the Alaska Range but not the St. Elias, indicating blockage of air parcels by the first band of high coastal peaks except where Cook Inlet cuts through.
Using two or more accumulation records in combination therefore provides insight into air parcel trajectories beyond the major storm events that dominate the accumulation signal during stormy years. For example, combining our HYSPLIT results with the ice-core accumulation records from both Begguya (Alaska Range) and Eclipse (St. Elias Range) can narrow the list of likely mechanisms for the increase in Begguya accumulation reported by . Non-storm snowfall in the St. Elias is more reliant on system strength, whereas non-storm snowfall in the Alaska Range is more reliant on system trajectory, in particular on where storm systems encounter the coastal mountains. Eclipse is thus likely affected by a higher proportion of midsize systems not strong enough to impact the entire region (i.e., not strong enough to make it to the Alaska Range), but strong enough to cross the high coastal peaks without dropping all their moisture. In contrast, small and midsize systems can both reach Begguya so long as their trajectories cut through the coastal mountains at Cook Inlet. Our HYSPLIT results suggest the increase in Begguya accumulation, and concurrent decrease in Eclipse accumulation, from 1400 to 2000 C.E. may be explained by a shift in dominant air-parcel trajectory, or a coalescing of air-parcel trajectories to a well-defined track that cuts through Cook Inlet, rather than an increase in average storm strength or frequency.
4.4. Data limitations and value
The biggest limitation of our results arises from the restricted temporal coverage and lack of overlap between our two snow sounder datasets. This precludes direct comparison of snowfall values themselves, as well as of atmospheric conditions associated with snowfall in any given year. However, we consider each record individually a unique and valuable characterization of local snow accumulation, as only in situ measurements such as these can provide daily, event-resolved accumulation data. The Kaskawulsh/Hubbard Divide record, in particular, provides an unparalleled quantification, with seven years of near-complete coverage of the local snowfall regime dominated by fall and winter accumulation, a phenomenon that is well known anecdotally, but to date has received limited documentation. Similarly, 14 Camp has been well observed anecdotally by climbers to receive significant summertime accumulation, but this, to date, has remained largely unquantified. In fact, local event-resolved accumulation data are largely unavailable in alpine regions worldwide because of the challenges associated with deploying and maintaining instrumentation. Our results, therefore, not only provide a valuable quantitative basis for previous anecdotal understanding of local conditions in alpine terrain, but are unique in doing so.
Furthermore, monthly accumulation values from our snow sounder records are generally consistent with ERA5 precipitation from 1940 to 2025, and seasonal patterns of snow accumulation are consistent between snow sounder data and ERA5 precipitation values at both sites (Fig. 4).
Finally, the majority of individual years covered by the Kaskawulsh/Hubbard Divide accumulation record show consistent atmospheric back trajectories associated with snow and storm days. Based on the consistent trajectories associated with snow days at the Kaskawulsh/Hubbard Divide combined with the reasonable seasonal distribution and monthly accumulation totals at both sites, we consider the patterns observed over our in situ records representative of general snowfall conditions at each site.
5. Conclusions
Orographic effects are a defining feature of atmospheric transport in the Gulf of Alaska region and account for a clear distinction between maritime and continental climate regimes. For example, snow accumulation in the St. Elias Range is dominated by September–December storm events, while accumulation in the Alaska Range is more evenly distributed throughout the year. Although orographic effects are ever present, their influence is greatest during non-storm accumulation events, during which southerly air parcels tend to either bring moisture to the Kaskawulsh/Hubbard Divide (St. Elias Range) but dry out before reaching 14 Camp (Alaska Range), or bring moisture to 14 Camp via Cook Inlet but be blocked from reaching the Kaskawulsh/Hubbard Divide by the nearby high coastal St. Elias peaks. In contrast, storm-day accumulation at sites in both the St. Elias and Alaska Ranges is associated with southerly airflow inland from the Gulf of Alaska.
Results suggest that the future of the Gulf of Alaska precipitation regime will be largely determined by the evolution of stormy weather in the region. Projected intensification of the Aleutian Low (Giamalaki and others, Reference Giamalaki, Beaulieu, Henson, Martin, Kassem and Faranda2021) likely means more storms of higher intensity funneled inland from the Gulf of Alaska, generally increasing precipitation in the region. However, local high-alpine increases in snow accumulation associated with more storms of higher intensity may either be amplified or offset by changes in non-storm accumulation associated with changes in the dominant pathways of air-parcel transport.
Finally, results support the interpretation of ice cores from the St. Elias Range in terms of regional (maritime) climate, and provide insight into differences between St. Elias and Alaska Range ice-core accumulation records. In particular, our HYSPLIT results suggest that the divergent behavior between Begguya and Eclipse ice-core accumulation records from 
$\sim$1400 to 2000 C.E. may be explained by a shift in dominant air-parcel trajectory, or a coalescing of air-parcel trajectories to a well-defined track that cuts through Cook Inlet, rather than an increase in average storm strength or frequency.
Our study highlights the value of in situ accumulation records from alpine regions and their contribution of daily, event-resolved snowfall data from regions where such observations are exceedingly rare.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/jog.2026.10142.
Acknowledgements
We dedicate this paper to Bea Alt (formerlyTaylorBarge), a pioneering researcher in alpine and polar climatology, who first carried out field investigations in the St. Elias Mountains in the mid-1960s during the Icefield Ranges Research Program. We thank the Tanana people, as well as the Kluane, Champagne and Aishihik, and White River First Nations, on whose traditional territories our snow sounders were deployed. We thank Denali National Park, Kluane National Park and Reserve (ParksCanada), Icefield Discovery, Kluane Lake Research Station, Talkeetna Air Taxi, Jean Bjornson and NSF logistics providers for collaboration and support. We thank the Denali National Park and Preserve South District Mountaineering Rangers, with whom ML has partnered for many years to collect the Kahiltna Glacier weather data.
Funding statement
Funding for this work was provided by the U.S. National Science Foundation, including awards 2002470,OPP-0136005, 0240878, 0713974, and AGS1203838, 1502783, 1806422, and 2002483. We are also grateful for funding from the Natural Sciences and Engineering Research Council of Canada, ArcticNet Network of Centres of Excellence Canada, Geological Survey of Canada, Polar Continental Shelf Program and University of Ottawa. Snow sounders at 14Camp and Base Camp were supported in large part by the Glacier Monitoring Program of the NPS Inventory and Monitoring Program.
Open access
