The dominance and growth of shallow groundwater resources in continuous permafrost environments

Significance

High-latitude environments are called polar deserts because of their lack of liquid surface water and limited potential for groundwater flow and storage. Though limited, groundwater is a crucial resource, providing pockets of perennial liquid water and contributing to river baseflow. Climate change is thawing the Arctic and altering the timing and availability of water, with implications for ecosystems, human communities, and industry. This work provides a revised conceptual model of water sources and flow paths in continuous permafrost regions, with a focus on shallow groundwater. This model suggests the potential for increased shallow groundwater flow as the Arctic continues to change due to warming and thaw.

Abstract

Water is a limited resource in Arctic watersheds with continuous permafrost because freezing conditions in winter and the impermeability of permafrost limit storage and connectivity between surface water and deep groundwater. However, groundwater can still be an important source of surface water in such settings, feeding springs and large aufeis fields that are abundant in cold regions and generating runoff when precipitation is rare. Whether groundwater is sourced from suprapermafrost taliks or deeper regional aquifers will impact water availability as the Arctic continues to warm and thaw. Previous research is ambiguous about the role of deep groundwater, leading to uncertainty regarding Arctic water availability and changing water resources. We analyzed chemistry and residence times of spring, stream, and river waters in the continuous permafrost zone of Alaska, spanning the mountains to the coastal plain. Water chemistry and age tracers show that surface waters are predominately sourced from recent precipitation and have short (<50 y) subsurface residence times. Remote sensing indicates trends in the areal extent of aufeis over the last 37 y, and correlations between aufeis extent and previous year summer temperature. Together, these data indicate that surface waters in continuous permafrost regions may be impacted by short flow paths and shallow suprapermafrost aquifers that are highly sensitive to climatic and hydrologic change over annual timescales. Despite the lack of connection to regional aquifers, continued warming and permafrost thaw may promote deepening of the shallow subsurface aquifers and creation of shallow taliks, providing some resilience to Arctic freshwater ecosystems.

Water resources are limited in high-latitude environments, where continuous permafrost typically exists close to the land surface, restricting subsurface water storage and groundwater flow. In these regions, snowmelt in the spring (May–June) accounts for the majority of the total annual runoff (1). By mid-summer (July–August), surface conditions are drier, with sporadic rainfall that temporarily bolsters stream and river baseflow. Groundwater availability is limited in continuous permafrost regions because permafrost acts as an impermeable barrier to deep water infiltration (2). As a result, runoff can only move through shallow suprapermafrost flow paths that tend to occur preferentially at the organic-mineral soil boundary or the frozen boundary (1–3). Hydrologic connectivity across broader regions is seldom demonstrated, because of studies focused on individual research catchments, and limited ability for three-dimensional modeling of frozen systems (2). Groundwater has been shown to be a minor source of freshwater, supplying ponds and lakes (4, 5), rivers (6), and the coastal ocean (7). However, the presence of large aufeis in many continuous permafrost regions (8) including Alaska (9–11), Canada (12), and Russia (13, 14) provides evidence that groundwater flow is nonnegligible. Aufeis develop over the winter months (Nov–April) as groundwater continuously reaches the surface and freezes (11, 14). Aufeis then melts gradually over the summer, thereby changing the timing of the groundwater contribution to rivers (11). The presence of aufeis and small perennial springs that are often their source (15, 16) has given rise to the hypothesis that groundwater is passing through deep, regional, subpermafrost aquifers (8, 11, 16). An alternate hypothesis is that springs and aufeis are sourced from shallow, suprapermafrost aquifers.

Determining whether groundwater emanates from deep, regional groundwater or from shallow, suprapermafrost aquifers aids in quantifying water resources, considering the resilience of ecosystems, and supporting management decisions of high-latitude regions underlain by continuous permafrost. Quantifying water resources in continuous permafrost zones is particularly relevant now given the impact of global climate change, which is disproportionately warming Arctic regions (17), and has the potential to drastically alter water availability (18, 19). For example, in the Arctic National Wildlife Refuge (ANWR) within the continuous permafrost zone of northern Alaska, sporadic springs that discharge into the rivers provide a critical habitat for overwintering fish (20, 21). This same region has the potential to be leased for oil and gas extraction, which would require substantial amounts of water to be pumped from the ground, rivers, or sparse lakes (20). If the groundwater that contributes to river flow and aufeis is sourced from deep, regional flow, this may indicate connection to a large, regional subpermafrost aquifer (22, 23) that could supply water for various demands of the limited water resources in continuous permafrost landscapes. Conversely, if groundwater, springs, and aufeis are sourced from more localized, shallow, and/or suprapermafrost waters, these resources may be limited and more susceptible to meteorological extremes, interannual variability, and a warming climate.

Although suprapermafrost groundwater flow is limited in the continuous permafrost zone, climate change may be increasing near-surface permafrost thaw (24), creating taliks that increase shallow subsurface groundwater availability and transmissivity. Taliks are increasingly common in the discontinuous permafrost zone south of the Brooks Range (25). Deeper and more prevalent taliks in the continuous permafrost zone could potentially explain observed hydrologic changes in Arctic Alaska over the last several decades, including greater winter and shoulder season river discharge (24, 26, 27), variability in interannual runoff (6, 28), increasing solute loads in springs and streams (10), and changes in aufeis extent (9, 13).

The goal of this work is to determine the sources and residence times of waters contributing to rivers in environments with continuous permafrost and limited groundwater flow. Springs and large aufeis fields are common in many continuous permafrost regions and provide a means to study water resources and consider the impacts of climate change on the rapidly changing cryosphere. We focused our study on ANWR, where springs and aufeis are prevalent and water resources demands include unique ecosystems and potential oil and gas exploration. In this region, several small rivers flow northward to the Beaufort Sea (Fig. 1) and often contain pockets of water in the winter (20, 21). We hypothesized that surface waters are derived from local precipitation and a shallow, suprapermafrost aquifer. We tested this hypothesis using 1) hydrochemistry and age tracers of springs, streams, and rivers sampled in the summer and in April prior to snowmelt, and 2) interannual trends in aufeis size and relationships to hydroclimatic variables. Using these data, we developed a conceptual model of continuous permafrost hydrology (Fig. 1), which indicates relatively short flow paths and transmission to aufeis fields through shallow taliks and highlights rapid responses to climate dynamics and thaw. These findings can help to evaluate the sensitivity of water resources in continuous permafrost regions to climate change and increased anthropogenic activity.

Fig. 1.

(A) Map of the eastern North Slope of Alaska, indicating aufeis fields and sampling locations, where “Sx” indicates spring samples presented in Table 1, and 'x)' indicates the location of images shown in figure parts. (B) and (C) conceptual model of surface and subsurface sources of runoff to rivers and aufeis in summer and winter, respectively, based on the results of this work, (D) aerial view of the Katakturuk aufeis field in August, 2021 (Photo courtesy of H Best, U.S. Geological Survey), and (E) the Sadlerochit Spring where it discharges from a hillside (Photo courtesy of J. Koch, U.S. Geological Survey).

1. Results and Discussion

1.1. Modern Water Dominates in Springs, Streams, and Rivers.

Water samples from springs, streams, and rivers (29) contained high concentrations of tritium (³H) and chlorofluorocarbons (CFCs) (Fig. 2 and SI Appendix, Supplemental Results 2.1), indicating the dominance of young, Anthropocene-aged recharge (30) in the region’s surface waters. Surface water tritium concentrations are consistent with modeled precipitation concentrations for this region (10 to 14 TU between 2008 and 2018) (31), indicating that rivers and streams are fed predominantly by modern runoff. Spring water displayed a broader distribution of tritium concentrations, indicating mixing of water with short and long subsurface residence times (Table 1). Four of the seven springs had mean tritium concentrations averaging 9.6 ± 1.0 (SD) TU, consistent with short subsurface residence times (ie. young water). Tracer data from the four young springs were best fit by an exponential mixing model of subsurface flow (Fig. 2 B–D), which indicates a mean subsurface residence time of less than 30 ybp. The remaining three springs contained lower tritium concentrations. Sadlerochit Spring (n = 3) and Red Hill Spring (n = 1) had mean tritium concentrations of 4.3 ± 0.5 (SD) TU and elevated terrigenic helium (He_terr) (SI Appendix, Fig. S1), indicating longer subsurface residence times. These springs were best fit with a binary mixing model of a shallow piston flow aged 26 to 33 ybp and deep exponential mixing flow aged 15,000 ybp. Sadlerochit Spring samples contained < 10 to ~80% old water and the Red Hill Spring sample contained <10% old water.

Fig. 2.

(A) Tritium concentrations in rivers, streams, and springs (29). The thin black lines in (A) and (B) bound the expected recent (2008 to 2018) tritium concentration in precipitation (31). (B–D) He_terr, tritium, and CFC concentrations in springs (points), separated into young (squares) and old (diamond) samples based on the bimodal tritium distribution, air correlation ratios, and helium sources (31; SI Appendix, Fig. S1). He_terr points represent measured values and bars are uncertainty estimated from DGMeta (32) modeling results. CFC and tritium points are means and bars indicate the full range of measured concentrations in 2 or 3 samples. The orange line indicates subsurface residence times based on an exponential mixing model (years before 2020). The blues line indicate binary mixing of a shallow piston flow and deep exponential mixing given a young age of 26 (dark blue line) or 33 y (light blue line) and an old age of 15,000 y. Numbers along the blue, binary mixing lines indicate the proportion of young water. The dashed line identifies modern water in the binary mixing space. “Sx” indicates spring samples presented in Table 1.

Table 1.

Spring chemistry and maximum subsurface residence times (29) compared to previously reported ages (16)

Spring	Date	Chloride (mg L−1)	Tritium (TU)	CFC-11 (pptv)	Heterr (cm3 g−1)	Age range (ybp)	Previously-reported age
S1. CHS-3	8/4/21	0.18	10.0	173	9E-10	<15	–
S2. Shublik	7/4/22	0.14	6.5	–	ND	–	7,230
S3. Red Hill	4/22/21	81.0	4.6	200	5.0E-8	179 to 1,227	18,650
S4. Katakturuk	4/22/21	0.20	8.3	164	3.5E-10	<15	–
S5. Sadlerochit	7/29/20	2.9	4.0	–	–	–	11,440
S5. Sadlerochit	4/22/21	3.1	3.5	141	6.7E-8	9,311 to 11,706	11,440
S5. Sadlerochit	8/3/21	3.4	4.2	61	8.4E-7	329 to 4,670	11,440
S5. Sadlerochit	8/16/21	3.2	–	-	3.0E-8	–	11,440
S6. Hulahula	4/21/21	0.35	12.5	191	1.6E-10	<15	2,360
S7. Okerokovik	4/21/21	1.06	11.1	148	4.4E-9	25 to 110	–

Whereas two of the samples from Sadlerochit spring were collected during baseflow conditions, the third was collected near the peak of a large storm event (SI Appendix, Fig. S2) and produced our oldest water age (~11,000 ybp). This may indicate that the storm produced a subsurface hydrologic pulse that allowed water to exchange between adjacent permeable and impermeable layers (33). Such layers are commonly formed by boundaries of soil and bedrock (34), and in permafrost regions at the organic-mineral (3), and frozen boundary (2). These exchanges typically lead to older ages than would be expected given diffusion from uranium and thorium decay alone, thus our estimates based on H_terr represent an upper limit of subsurface residence time (33, 35). Diffusion from mantle sources is another potential influence on apparent age (36), which we do not believe is important in our system given helium isotope and excess air ratios (SI Appendix, Fig. S1). Thus, despite the challenges of sampling groundwater (2) and interpreting tracer chemistry (33, 35, 37), our age findings remain consistent across tracers (SI Appendix, Fig. S3 and Supplemental Text 2.2 for more discussion of variability and uncertainty in the age dates).

1.2. Surface Waters Dominated by Local Meteoric Water.

Biplots of δ²H versus δ¹⁸O and chloride (Fig. 3) (29) indicate that stream and river water is primarily composed of meteoric water, with limited influence from springs, which have elevated chloride related to longer subsurface residence times (e.g., Sadlerochit and Red Hill springs). All surface water samples fell along the global meteoric water line. Isotopic composition of summer river water was well constrained by mean summer rain and mean annual precipitation measured 200 km to the southwest at a similar elevation (750 m asl) and geographic setting (38–40). Mountain streams were slightly isotopically lighter than mean annual precipitation, suggesting that they are predominantly fed by snowmelt. Foothill streams were comparable to summer rain. Isotopically heavier values for small coastal streams may reflect their lower elevation and proximity to the coast. Winter river water was isotopically similar to springs and mountain streams.

Fig. 3.

(A) Water stable isotopes measured in springs, streams, and rivers (29), compared to mean and summer precipitation from ref. 40 and the Global Meteoric Waterline (GMWL). Large colored symbols represent means, with SDs indicated by whiskers. Winter river samples were collected in April from under ice. (B) Log chloride versus δ²H from the same sites, but with springs represented by individual points due to the broad range and nonnormal distribution. Precipitation chloride data are from ref. 39. “Sx” indicates spring samples presented in Table 1.

Biplots of chloride and δ²H from surface waters (Fig. 3B) further support short, shallow flow paths and the limited presence of old water. Chloride is an effective hydrologic tracer in many settings due to its high solubility in water and biologically conservative nature (41). The strong inverse relationship of tritium versus chloride and magnesium (SI Appendix, Fig. S4) are consistent with greater solute loads associated with longer subsurface flow paths, as has been observed in other Alaskan surface waters (19, 42). Streams and river concentrations including winter river water were consistently low (<1.0 mg chloride L⁻¹) and easily explained by mixing of summer and annual precipitation. Dilute concentrations in mountain streams and winter rivers suggest that they are sourced from snowmelt. Chloride concentrations in the springs ranged several orders of magnitude, with the highest concentrations in the two springs with old ages. Slightly elevated mean chloride concentrations in winter rivers (0.47 ± 0.17 (SD) mg L⁻¹) is consistent with inputs from springs with short subsurface residence times and other suprapermafrost talik flow that has 1) leached solutes from soils before reaching the rivers, 2) mixed with some small fraction of old, solute-rich spring water (33), and/or 3) been freeze-concentrated upon reaching the river. Given that all three of these processes may be occurring, only a very small fraction of old spring water is needed to explain the observed chloride concentration of the winter rivers. For summer rivers, only one river sample collected within 15 km of the coast had a chloride concentration greater than 1 mg L⁻¹, likely indicating the influence of marine aerosols.

1.3. Variability in Aufeis Size and Temporal Trends.

Aufeis fields are unique features of many permafrost environments and provide novel insight into the regional hydrology that supports our conceptual model. Aufeis are expected to thaw faster in a warmer climate, which has been demonstrated in northern Alaska (9, 10). Our 37-y time-series analysis of four aufeis fields within the study region (43) revealed that two of these fields have decreased in size over time, while the other two have increased in summertime areal extent from 2006 to 2022 (Fig. 4). All four of these aufeis fields show evidence of cyclical interannual variability, suggesting some control by current meteorological conditions, which has also been noted in Russian (13) and Canadian (44) fields. We propose three hypotheses for why aufeis fields may be decreasing in size: 1) upwelling water is more likely to enter unfrozen stream reaches and remain unfrozen, 2), accumulated aufeis thaws faster due to warmer spring and summer air temperatures, 3) winter groundwater flow is decreasing. Our data show strong (P < 0.05) inverse correlations between aufeis size and the concurrent year’s air temperatures for three of four fields (Table 2 and SI Appendix, Table S1), which is consistent with smaller aufeis size or faster melt in warm years. However, aufeis fields could also be bolstered by warm summer temperatures in the previous year, if increasing ground thaw and enhanced suprapermafrost groundwater flow deliver more water to areas of aufeis formation in the winter. Widespread increases in subsurface taliks have been observed in the discontinuous permafrost zone (25) and subriver aquifer development has been found on the Alaska North Slope (45), both of which could deliver water to the aufeis fields over the winter months. We found that aufeis size correlated positively with the previous year’s temperatures, with significant (P < 0.05) correlations for two of four aufeis fields (Table 2 and SI Appendix, Table S2), supporting the idea that warm years lead to more aufeis in the subsequent year. We used multiple linear regression to determine the predictors of aufeis size. We found that previous year and concurrent year temperature significantly predicted aufeis size, often with a positive response to previous year temperature and negative response to concurrent year temperature (Table 3 and SI Appendix, Table S3). Precipitation is likely another important predictor of aufeis size. Climate indices were used in lieu of precipitation, given the paucity of accurate precipitation data in remote, cold, and windy locations. Two aufeis fields displayed correlations to climate indices, with weak (P < 0.10) positive correlations to the ENSO Precipitation Index and negative correlations to the Pacific-North American pattern (Table 2). Together, our observations of cyclical variability in aufeis field size, disparate trends across fields, and statistically significant response to concurrent year and previous year air temperatures make a strong case that aufeis is sourced from local water moving through shallow flow paths. The cyclical aufeis sizes are consistent with observations of high interannual variability in discharge of a nearby river related to runoff and shallow subsurface flow (6), and inconsistent with a connection to deep regional groundwater flow (16), which would likely buffer interannual variability. Although aufeis may melt faster in a warming climate, its areal extent and volume may vary on short timescales depending on climate conditions and may even increase as warming continues to drive active layer deepening and suprapermafrost talik flow.

Fig. 4.

Aufeis field areal extent over 37 y from 1986 to 2022 (43). Values represent means of calculated daily values. Trend lines and test statistics (tao) are indicated when Mann–Kendall tests were significant at P < 0.05.

Table 2.

Significant Spearman correlations between aufeis size (43), air temperature (46), and climate indices (47)

		Air temperature		This year’s climate indices
Aufeis field	Year	This year	Last year	ESPI	AO	PNA	PDO
Canning	++	−−	++
Hulahula		+	++	+		−−
Okerokovik		−−
Sadlerochit		−−		+		-

⁺⁺positive correlation at P < 0.05.

⁺positive correlation at P < 0.10.

⁻⁻negative correlation at P < 0.05.

⁻negative correlation at P < 0.10.

Table 3.

The best multiple linear regression models incorporating aufeis size (43) and a positive relationship to last year’s air temperature and negative relationship to this year’s air temperature (46)

Aufeis	Model	r2	P	Intercept	Last year’s Tair slope	This year’s Tair slope	b1 * b2
Canning early summer	Avg area ~ last June Tair * this summer Tair	0.64	0.02	68.2	1.8	−1.7	−0.6
Hulahula early summer	Avg area ~ last annual Tair * this June Tair	0.79	0.02	6.8	0.29	−0.14	0.00
Sadlerochit late summer	Avg area ~ last summer Tair * this summer Tair	0.55	0.01	−10.2	1.11	−1.33	0.22
Okerokovik late summer	Avg area ~ last Jun Tair * this summer Tair	0.78	0.01	2.64	0.20	−0.07	−0.06

1.4. An Updated Conceptual Model—Local Water Sources and Rapid Runoff through Shallow Flow Paths.

Previous research in continuous permafrost zones has focused on specific catchments, highlighting surface and shallow subsurface flow in regions with ice-rich permafrost (1, 2). However, hydrologic connectivity through suprapermafrost aquifers or through-going taliks has not been demonstrated on broad scales (2), yet connectivity to deep regional groundwater flow has been argued to be an important process across broad regions (8, 16, 48). The ambiguity regarding connections between shallow and deep flow and equivocal consideration in the literature limits our understanding of Arctic hydrology and ecosystems (2, 49, 50) and our ability to predict how ecosystems will be impacted by change and disturbance (21, 51). Results from this study, including water residence times, hydrochemistry, and aufeis data highlight the utility of a revised conceptual model of hydrologic flow in continuous permafrost regions, emphasizing local meteoric water sources and short, suprapermafrost flow paths between recharge and discharge areas (Fig. 1 B and C).

Our data indicate substantial water storage in suprapermafrost aquifers on short timescales along a gradient from the mountains to the coastal plain. Additionally, we provide evidence of talik development, which until now has only been recognized as a ubiquitous process in the warmer, discontinuous permafrost zone to the south (25). We demonstrate that shallow subsurface aquifers are hydrologically connected on broad enough scales to regulate water storage in aufeis (11). Surface water samples from springs, streams, and rivers collected in this study generally had short subsurface residence times, low solute loads, and were well constrained by precipitation from a nearby monitoring station, suggesting that river water is a mixture of local snowmelt and rain following fast flow paths to the river. In previous research, ¹⁴C dating and the presence of aufeis have been inferred as evidence of springs emanating from long, deep, flow paths (8, 16). Our findings challenge these previous conclusions, instead indicating that 1) springs comprise water with predominantly short subsurface residence times and 2) aufeis responds rapidly to interannual climatic variability. Furthermore, some springs emanate far from the mountain front with no evidence of geologic faults (52), challenging the notion that unseen through-going taliks allow deep groundwater to transit the thick permafrost of this region. Age tracers and hydrochemistry provide evidence of longer subsurface residence times in three springs, and the potential for rapid shifts in spring water source on short timescales. These findings are best explained by interactions of runoff with small amounts of old water residing in shallow aquitards (3, 33). Regardless of the ultimate source of this long-residence time water, spring discharge is not sufficient to substantially influence surface water chemistry (Fig. 3) or to explain the observed aufeis volume (SI Appendix, Table S4). Observations of cyclical trends in aufeis size and differences between the fields are consistent with the conclusion that aufeis water is sourced from shallow soil horizons. Differences in trends between aufeis size and year (Fig. 4) may potentially be explained by individual basin characteristics such as geology, size, slope, and morphology.

2. Conclusion

Continuous permafrost regions are changing rapidly in response to climate change. Warming air temperatures and decreased sea ice extent are altering precipitation regimes (28, 53) and promoting vegetation growth (54), with implications for runoff dynamics, surface water discharge (6, 19, 27, 53, 55), and potentially groundwater flow. In the continuous permafrost zone of northern Alaska, we find that surface water and groundwater feeding springs and aufeis fields are dominated by young water derived from local precipitation. This water is relatively dilute with an isotopic composition similar to local precipitation. Trends in aufeis field size indicate substantial interannual and between-site variability, which is most easily explained by meteorological conditions impacting shallow subsurface aquifers. Together these findings lead to our updated conceptual model (Fig. 1 B and C), which illustrates how rapid runoff and shallow suprapermafrost flow control surface water availability in ANWR and other continuous permafrost environments of similar geographic settings, often typified by small perennial springs and large aufeis fields. The implications of this shallow, fast flow are that perturbations from ongoing climatological shifts with climate change will impact surface waters on short timescales, and contamination from anthropogenic activities may impact surface waters within decades or even faster given that contaminants generally move faster than mean ages determined from lumped parameter models (30). In the short term, active layer thaw and suprapermafrost talik development may increase groundwater contributions to the rivers. On longer timescales, surface water availability will be determined by the interactions of warming and thaw of the tundra, which may increase landscape hydrologic connectivity and surface water drainage (18), infiltration, vegetation growth, and evapotranspiration (19).

3. Materials and Methods

3.1. Site Description.

This study took place in the mountains, foothills, and coastal plain of the ANWR (Fig. 1). This region is bounded by the Brooks Mountain Range to the south and the Beaufort Sea to the north. Whereas the western portion of Alaska’s North Slope is flat and lake-rich, conditions are much drier in the east due to the steeper topographic relief and coarser sediments associated with the short distance between the mountain front and the coast. The geology of the northeastern Brooks Range is characterized by a Cenozoic fold-thrust belt, where pre-Mississippian metasedimentary rocks and Carboniferous carbonate and shale formations have been thrust up and over Triassic-Cretaceous shales (56). This intense structural deformation has led to a high degree of fracturing, particularly in carbonate units, which may serve as important aquifers in higher-elevation areas of the landscape (57). The region is underlain with continuous permafrost (58) that may be up to hundreds of meters thick (16, 58). Mean annual air temperatures in this region average −10 °C and the region receives 200 to 350 mm of precipitation a year depending on elevation, with 40 to 50% falling as snow (59). The active layer is typically less than 1 m thick in the foothills and coastal plain, where the permafrost is relatively ice-rich. In the Brooks Range, coarser substrate and fractured bedrock may lead to ice-poor, permeable permafrost and deeper active layers (58). The rivers of this region flow from May to October and are ice-covered with very low discharge the rest of the year. These rivers generally run from south to north, with aufeis fields forming in the channels of some of the rivers (43) (Fig. 1A). Snowmelt runoff is a substantial annual hydrologic flux in Arctic watersheds (1). Liquid water exists in pockets of the rivers in the winter, but is of limited volume, sometimes occurring at junctions with tributaries (20). Springs have been used as evidence of deeper flow, potentially following active faults to form taliks through the permafrost (16). Despite the harsh conditions and limited water availability, this region provides critical habitat for polar bears, wolves, caribou, geese, and hosts resident and anadromous fish in many of the rivers (21, 51).

3.2. Sample Collection and Analysis.

We collected water samples to assess water sources and to determine subsurface residence times using hydrochemistry and age tracers. Samples were collected between April 2019 and July 2022 from major rivers, mountain, foothill and coastal plain streams, and springs. In April, most surface waters are frozen at this latitude, meaning that our samples represent groundwater (10, 15). April samples were collected near known springs (Fig. 1A), or by drilling into the frozen Canning River. We define streams as smaller, usually first-order tributaries that feed into the rivers. Small streams are often well-connected to their catchments, and thus provide the most direct information about runoff water sources (19). Several gases had elevated atmospheric concentrations in the mid-twentieth century due to nuclear testing (tritium) and new industrial technologies (CFCs), providing timeseries data against which water sample gas concentrations can be compared to determine the time at which recharge entered the subsurface (60). These gases are effective at determining ages for water that infiltrated since about the 1950s. To date older samples, we relied on noble gases and helium. Helium concentrations and isotopic ratios can provide age estimates for much older waters, as their concentration in water samples is related to 1) atmospheric concentrations at the time of infiltration, which is dependent on recharge elevation and temperature (37), 2) the decay of tritium in the water, and 3) the decay of uranium and thorium within rocks in the ground or degassing from the mantle (35, 61). Springs usually could not be sampled directly at their source, either because of ice and snow cover or other logistical challenges. As a result, the age tracers may have been impacted by gaseous exchange between the spring water and modern atmosphere, or by inflows of snowmelt or groundwater between the spring discharge and sampling locations. Contamination by modern air would not lead to a consistent bias in apparent ages, thus we have confidence that our age dates are accurate, given consistent results across hydrochemical and age tracers collected multiple times from several springs. Noble gases were used to determine recharge temperatures and excess air concentrations using DGMeta (32), using a range of recharge elevations relevant to the topography of the Brooks Range (29). The mean recharge elevation was then applied to CFC and helium concentration calculations. Terrigenic helium, He_terr, was determined by subtracting atmospheric helium and excess helium from the sample concentration (32). Helium isotopes and terrigenic versus total concentrations were used to attribute He to bedrock or mantle sources (32). Additional details on sample collection and analysis can be found in SI Appendix, Supplemental Methods.

Age tracers including tritium, He_terr, and CFCs were then imported into TracerLPM, which allows tracer data to be fit by lumped parameter models of subsurface transport and mixing (62). Tritium and CFC concentrations of the youngest samples were fit to a diffusion model based on atmospheric input histories from the closest available dataset to identify whether concentrations in northern Alaska fell near expected atmospheric concentrations. For CFCs, the input history is based on data for the northern hemisphere, primarily based on the record from Ottawa, Canada, compiled by the U.S. Geological Survey Chlorofluorocarbon Laboratory (62). For tritium, the input history is based on the gridded data for the contiguous United States (62). We chose the grid closest to our study site (Latitude: 47° to 49° N and Longitude 120° to 125° W). Scaling factors for CFCs and tritium were then adjusted to minimize differences between the young samples and the modeled atmospheric concentrations (SI Appendix, Fig. S5). Once scaling factors were determined, several common lumped parameter models of subsurface residence time distributions (63) were fit to the data. For the exponential mixing model that fit young samples, unsaturated zone residence time was the only calibrated parameter. For the binary mixing model that fit old samples, calibrated parameters included unsaturated zone residence time, soil porosity, uranium, and thorium concentrations, and the age of the young and old water sources. This approach neglects potential helium release from aquifer solids (35), meaning that our subsurface residence times based on He_terr represent the oldest possible subsurface residence times.

Biplots of geochemical parameters were used to determine the relative proportions of various water sources contributing to the rivers. This was done first by calculating mean and SDs of stable isotopes of water and major ions and using solute biplots to identify solutes that bound the water samples. Existing data on summer and annual precipitation chemistry from the region (38–40) were used as possible endmembers.

3.3. Aufeis Area and Melt.

We quantified aufeis areal extent at four persistent aufeis fields of varying size distributed across the study area, which together account for the majority of the observed, persistent aufeis in this region (43). To estimate aufeis area, we utilized Landsat scenes from May to September of 1986 to 2022 with less than 10% cloud cover. We further restricted our analysis to the floodplains of four rivers containing persistent aufeis fields. We classified aufeis features using the Normalized Difference Snow Index and assessed accuracy by comparing with independent estimates based on panchromatic, high-resolution commercial satellite platforms with submeter spatial resolution for dates where concurrent observations were available (10). To quantify mean aufeis area for the summer months of each year despite limited scene availability, we fit sigmoidal decay curves to the available imagery. Aufeis area was regressed versus year using Mann Kendal tests and Thiel Sen slopes and correlated to climate indices (47), the present and previous year’s air temperatures (46), and other aufeis field areas using Spearman correlations. Multiple linear regressions were run to test for the combined importance of the current and previous year’s air temperatures on aufeis field size. Additional details on aufeis detection, decay modeling, and statistical methods are available in SI Appendix, Supplemental Methods.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All data in this study are publicly available. Water quality and age tracer data are available in ref. 29 and aufeis data in ref. 43.