Water Resources Research

Influence of permafrost distribution on groundwater flow in the context of climate-driven permafrost thaw: Example from Yukon Flats Basin, Alaska, United States



[1] Understanding the role of permafrost in controlling groundwater flow paths and fluxes is central in studies aimed at assessing potential climate change impacts on vegetation, species habitat, biogeochemical cycling, and biodiversity. Recent field studies in interior Alaska show evidence of hydrologic changes hypothesized to result from permafrost degradation. This study assesses the hydrologic control exerted by permafrost, elucidates modes of regional groundwater flow for various spatial permafrost patterns, and evaluates potential hydrologic consequences of permafrost degradation. The Yukon Flats Basin (YFB), a large (118,340 km2) subbasin within the Yukon River Basin, provides the basis for this investigation. Model simulations that represent an assumed permafrost thaw sequence reveal the following trends with decreasing permafrost coverage: (1) increased groundwater discharge to rivers, consistent with historical trends in base flow observations in the Yukon River Basin, (2) potential for increased overall groundwater flux, (3) increased spatial extent of groundwater discharge in lowlands, and (4) decreased proportion of suprapermafrost (shallow) groundwater contribution to total base flow. These trends directly affect the chemical composition and residence time of riverine exports, the state of groundwater-influenced lakes and wetlands, seasonal river-ice thickness, and stream temperatures. Presently, the YFB is coarsely mapped as spanning the continuous-discontinuous permafrost transition that model analysis shows to be a critical threshold; thus, the YFB may be on the verge of major hydrologic change should the current permafrost extent decrease. This possibility underscores the need for improved characterization of permafrost and other hydrogeologic information in the region via geophysical techniques, remote sensing, and ground-based observations.

1. Introduction

[2] Northern ecosystems containing permafrost (perennially frozen ground) are responding rapidly to climate warming that has persisted for several decades [Serreze et al., 2000; Hinzman et al., 2005]. Variations in wetland and lake areas [Riordan et al., 2006; Rover et al., 2012; Roach et al., 2011], in soil moisture, and in magnitude and seasonal patterns of stream discharge are hydrologic changes that directly influence vegetation [Jorgenson et al., 2001], biogeochemical cycling [Wickland et al., 2006; Frey and McClelland, 2009], and habitat sustainability [Prowse and Brown, 2010]. There is evidence of increasing base flow in streamflow records in the Canadian Northwest Territories [St. Jacques and Sauchyn, 2009], in northern Eurasian rivers [Smith et al., 2007], and in the Yukon River Basin (YRB) [Walvoord and Striegl, 2007]. The ice thinning on some YRB rivers observed by native Alaskans [Herman-Mercer et al., 2011] may be a reflection of increased base flow. Changes in aquatic carbon exports at the basin scale may also result from enhanced groundwater discharge to major rivers. Striegl et al. [2005]identify a historical decrease in discharge-normalized dissolved organic carbon (DOC) exported from YRB in summer and autumn. This trend is attributed to increases in water residence time and deepening flow paths [Walvoord and Striegl, 2007; Lyon and Destouni, 2010], increases in DOC stabilization in mineral soils [e.g., Kawahigashi et al., 2006], as well as the relatively high lability of newly exposed carbon in former permafrost.

[3] All of the above mentioned studies propose a causative relation between observed hydrologic change and permafrost thaw. Indeed, it has been found that permafrost is warming and thawing in some areas of Alaska [Osterkamp and Romanovsky, 1999; Jorgenson et al., 2001; Osterkamp, 2005; Lachenbruch and Marshall, 1986; Jorgenson et al., 2006]. Connections between permafrost degradation and local hydrologic changes have been established [e.g., Osterkamp et al., 2000; Yoshikawa and Hinzman, 2003; O'Donnell et al., 2012]. Yet, the linkage between permafrost degradation and large-scale changes in hydrologic fluxes is not fully understood, althoughMichel and van Everdingen [1994], Hinzman et al. [2005], and Rowland et al. [2010], have provided insightful discussions about such linkages. A challenge in substantiating inferred linkages in a quantitatively robust manner arises from the paucity of hydrogeologic and permafrost information in remote northern systems and from the complexities of regional-scale permafrost-climate-hydrology interactions.

[4] The observed increases in base flow in northern latitudes may be achieved via increases in flow through the suprapermafrost aquifer via active layer thickening (reduction in elevation of the top of permafrost) as demonstrated using coupled heat and fluid flow modeling on the century time scale by Bense et al. [2009] and on the decadal time scale by Ge et al. [2011]. Increases in base flow may also be achieved via enhanced deep upward groundwater flow within unfrozen zones below rivers. Increased groundwater recharge and vertical connectivity between recharge areas and river valleys achieved by spatial degradation of permafrost would support such enhancement.

[5] This study examines the influence of spatial patterns of permafrost and potential pattern changes due to a warming climate on groundwater flow. The impacts of permafrost distribution are considered for the Yukon Flats Basin (YFB) in east central Alaska that is the target of ongoing U.S. Geological Survey (USGS) research efforts to identify and understand historical changes in biogeochemical cycling, lake area, and biodiversity resulting from climate change [Riggins et al., 2011]. The standard groundwater flow model, MODFLOW-2000 [Harbaugh, 2002], is applied to the YFB, in which permafrost is represented as hydrogeologic layers with very low hydraulic conductivity. Groundwater flow systems that develop for different configurations of permafrost are elucidated. The permafrost distributions considered may be arranged in a permafrost thaw sequence to represent a possible temporal evolution due to climate warming. Moreover, the result of each distribution also provides a basis for evaluating the groundwater system in basins with similar permafrost coverage. The primary objectives of this scoping analysis are (1) to illustrate the control that permafrost exerts on regional groundwater flow systems and (2) to reveal potential trends and behavior in groundwater flow that might result from climate warming.

2. Study Area and Data

[6] The YFB is within the Yukon River Basin in north eastern Alaska (Figure 1) and is approximately 118,340 km2. The basin includes an alluvial lowland (90–353 m in elevation and occupying 24,080 km2) transected by the Yukon River (YR), a dissected to rolling marginal upland (30 – 150 m higher than the lowlands and occupying 15,460 km2), and bordering highlands (occupying 78,800 km2). The Cenozoic sedimentary basin comprising the lowlands consists of Pleistocene to recent alluvial deposits, forming alluvial fans, terraces, and floodplains of the Yukon Flats. Alluvial deposits are underlain by quiet-water silt and silty sand of late Tertiary to early Quaternary age [Williams, 1962]. The marginal upland is separated from the lowlands by a steep river-eroded escarpment and is mantled by ice-rich loess, i.e., yedoma, or extremely ice-rich syngenetic permafrost formed during the late Pleistocene [Kanevskiy et al., 2011] that is slow to thaw because of the latent heat of the high ice content. Highlands surrounding YFB are north, the southern Brooks Range; northwest, the Hodzana Highland; east, Porcupine Plateau; and south, part of the Yukon-Tanana Plateau, including the White Mountains and Crazy Mountains. These highlands are composed mainly of Paleozoic bedrock of various origins. For additional geological description, seeWilliams [1962].

Figure 1.

Location map of the Yukon Flats Basin (YFB) [after Williams, 1962] with surficial geology used to represent substrate above permafrost [after Karlstrom et al., 1964]. Gauging stations with winter discharge information available are noted. Numbers apply to stations described in Table 1. Inset shows YFB location.

[7] The most recent map of generalized permafrost distribution and characteristics in Alaska by Jorgenson et al. [2008] indicates that YFB spans the transition between continuous permafrost (defined as >90% coverage) to the north of the YR and discontinuous permafrost (50%–90% coverage) to the south. The YR and the Porcupine River (PR) are the water bodies in the YFB indicated on the Jorgenson et al. [2008] map and legend as having unfrozen conditions below. This follows the common, but generally untested, assumption (by field measurements) in Alaska that permafrost, in both continuous and discontinuous regimes, is absent below major water bodies such as rivers and large lakes [e.g., see Williams, 1970; Burn, 2002, 2005]. Where permafrost is present near Fort Yukon, AK, a deep borehole indicates a permafrost thickness of ∼90 m [Clark et al., 2009].

[8] Climate information, centrally located within the lowlands region of the YFB at Fort Yukon, Alaska, shows a mean annual air temperature of ∼−6°C and mean annual total precipitation of ∼250 mm yr−1 (National Climatic Data Center, http://www.ncdc.noaa.gov/oa/ncdc.html). YFB basin-wide precipitation, derived from PRISM data, ranges from 250 to 760 mm yr−1, with an area-weighted average of 490 mm yr−1 (PRISM Climate Group, Oregon State University, http://prism.oregonstate.edu, created 2/2000). The highlands receive greater summer rainfall than the lowlands, but winter precipitation is less spatially variable. The current average annual snowfall at Fort Yukon is 81 cm (SNOTEL 961, 2007–2010, Natural Resources Conservation Service, http://www.wcc.nrcs.usda.gov/snotel/Alaska/alaska.html).

[9] Winter streamflow data from gaging stations on the largest YFB rivers (Table 1 and Figure 1) allow estimation of base flow for 12 river reaches. (These values are compared with model-predicted net groundwater flux along these reaches later in this report.) Because of its remoteness, no other quantitative hydrologic data (such as measured hydraulic heads) exist for the area. Winter streamflow data serve as an adequate proxy for net groundwater flux to rivers in high-latitude environments, such as YFB, where surface runoff and near surface flow are negligible due to freezing conditions at the surface. Winter streamflow measurements are subject to low bias and greater uncertainty than flow measurements made in open water [Moore et al., 2002], but are the best available estimates of groundwater contribution to streamflow (referred to as base flow herein). The headwaters of tributaries are fully included in the area of analysis (i.e., the modeled area) with two exceptions. For the PR (station 1), average 1988–2009 winter flow from the gaging station at the international boundary (1.67 × 106 m3 d−1) was subtracted from measured flow at station 1 (1.97 × 106 m3 d−1 on 3/2002 by USGS; http://waterdata.usgs.gov/nwis) to provide base flow to the PR along the reach represented in this model analysis (Table 1). Winter streamflow measurements on the YR also had to be adjusted to represent inputs only within the modeled area. Average 1977–2009 winter flow on the YR at Eagle, Alaska (4.99 × 107 m3 d−1), 250 river km upstream of where the YR enters the area of analysis was subtracted from average winter flow over the same period at station 12 (YR at Stevens Village = 6.19 × 107 m3 d−1). The difference was multiplied by 61% (proportion of watershed area represented in the modeled area) to remove groundwater inputs to the river upstream of the modeled area. From this value, summed winter streamflow from all the tributary stations were also subtracted to achieve a value (Table 1, station 12) that reflects base flow to the YR within the reach represented in the modeled area plus base flow to tributary sections downstream of their stream gaging stations.

Table 1. Winter Streamflow for River Reaches in Model Domain
StationNameLatitude NAD 1927Longitude NAD 1927Agency and Period of RecordaAverage Winter (Jan–Mar) Streamflow (m3 d−1)
1Porcupine River66.9906−143.1378EC; 1988–2009 USGS; 20023.03 × 105
2Sheenjek River66.9102−144.3319FWS; 1994–19982.50 × 105
3Chandalar River67.0969−147.1844USGS; 1964–19734.92 × 104
4Hadweenzic River66.6633−146.9549FWS; 1994–19989.79 × 103
5Hodzana River66.6452−148.2597FWS; 1994–19981.42 × 105
6Upper Beaver Creek66.0457−146.5147FWS; 1995–19982.26 × 105
7Lower Beaver Creek66.2572−146.1436FWS; 1995–19988.56 × 105
8Birch Creek66.0997−144.7474FWS; 1994–19987.34 × 104
9Preacher Creek66.7750−145.5469FWS; 1994–19981.96 × 104
10Little Black River66.2786−143.1819FWS; 1994–19981.22 × 104
11Black River66.5377−142.0272FWS; 1994–19981.10 × 105
12Yukon River65.8756−149.7178USGS; 1977–20093.84 × 106

3. Hydrogeology of Permafrost

[10] The classical definition of permafrost is regions of ground at or below 0 °C for 2 or more years [van Everdingen, 1998]. When studying impacts of permafrost on groundwater flow, a more useful definition of permafrost for hydrogeologic studies is ground containing perennial ice or perennially frozen geologic substrate that impedes the flow of water. The hydrogeologic definition is used in this study. In the YFB region, where present, the permafrost table is generally encountered <1 to 4 m below the land surface, and frozen conditions are believed to continue to depths of tens to several tens of meters, perhaps exceeding 100 m in some areas [Jorgenson et al., 2008].

[11] The “suprapermafrost zone” is situated below ground above the top of permafrost. In some areas, the depth to permafrost may exceed the depth of seasonal frost, particularly in areas recently perturbed by wildfire, a common occurrence in interior Alaska [e.g., Yoshikawa et al., 2002]. In these areas, groundwater flow may occur perennially creating a relatively thin suprapermafrost aquifer. Elsewhere, the suprapermafrost zone is completely frozen for at least part of the year, impeding winter groundwater flow. Below the bottom of permafrost in the “subpermafrost zone,” should permeable layers exist, there are one or more subpermafrost aquifers that are perennially unfrozen.

[12] A talik is a perennial locally unfrozen region of the subsurface within an otherwise more extensive permafrost layer. As defined in this hydrogeologic study, a “closed talik” is an unfrozen zone that does not fully penetrate the permafrost layer. In contrast, an “open talik” is defined as a perennial locally unfrozen zone that fully penetrates an otherwise more extensive permafrost layer. Exchange between the suprapermafrost and subpermafrost aquifers and between surface waters and the subpermafrost aquifer can only occur within open taliks. The location of taliks depends on several factors including existence and depth of surface water bodies, existence of relatively warm groundwater discharge, and existence of relatively high solute content in otherwise cold groundwater discharge. Typically, closed taliks are believed to exist under shallow surface water bodies and open taliks are believed to exist below deeper bodies of water that do not freeze to the bottom in winter.

[13] The hydraulic conductivity (K) of frozen ground, where ice fills the entire pore space, is believed to be extremely low. Burt and Williams [1976] measured the K of saturated unconsolidated materials just below the freezing point using a simple permeameter and found a 4–5 order of magnitude reduction in K as temperature decreases from 0°C to −0.5°C. A steep reduction in K as soil temperature drops below the water freezing point has also been confirmed in other experimental work [i.e., Horiguchi and Miller, 1983; McCauley et al., 2002]. For regional groundwater flow, a continuous permafrost layer thus effectively isolates hydrologic flow systems that are above from those below the permafrost. In general, active flow paths are expected to remain shallow in regions of continuous permafrost. Likewise, active flow paths are anticipated to become much deeper in regions of progressively less permafrost [e.g., Bense et al., 2009]. However, changes in the routing of water as permafrost thaws are not expected to be obvious since head gradients, as well as spatial pattern of K, will also change in response to permafrost thaw. Determining system response and changes in flow patterns require implementation of a 3-D groundwater model.

4. Methods

[14] Hydrogeology of YFB is distilled to represent only the primary geologic controls on the system, in an analysis intended to elucidate understanding of the most important groundwater-permafrost processes. The impact on groundwater flow of a variety of permafrost distributions is evaluated via numerical simulation using MODFLOW-2000 [Harbaugh, 2002]. Groundwater flow paths are determined with postprocessing using MODPATH [Pollock, 1994]. Water budgets are calculated in a postprocessing step by ZONEBUDGET [Harbaugh, 1990]. Using ZONEBUDGET, total horizontal and vertical inflow/outflow are determined for each of the following hydrologic regions within YFB: (1) suprapermafrost aquifer (0 m to 1–4 m below surface in regions overlying permafrost), (2) talik shallow aquifer (0 m to 4 m below surface in permafrost-free areas, not directly overlain by river), (3) river zone (0 m to 4 m below surface of river), (4) talik deep aquifer (4 m to 90 m below surface in permafrost-free areas), (5) permafrost (1–4 m to 90 m below surface; i.e., regions containing permafrost), (6) subtalik aquifer (90 m below surface to aquifer bottom below taliks), and (7) subpermafrost aquifer (90 m below surface to aquifer bottom below permafrost).

[15] For the permafrost distribution considered plausible under present climate conditions, the simulation model's alluvium K is roughly calibrated by having model base flow results match the available data (consisting of several winter streamflow estimates in YFB rivers, Table 1). Calibration is accomplished automatically, using the inverse model (parameter estimation) functionality of MODFLOW-2000. Lack of typical hydrogeologic information, such as hydraulic head data, detailed hydrostratigraphy, and groundwater age data, precludes more detailed model calibration. Using the present-day roughly calibrated model as a reference or “base case,” permafrost distribution is varied in subsequent simulations, and the impacts of permafrost distribution on groundwater flow are evaluated to assess sensitivity, magnitude, and implications of differences. Primarily steady state groundwater flow is considered, but transient analysis is also carried out to check the magnitude of variation resulting from the simplifying steady state assumption. The reference base case and variation for thaw sequence cases are defined insections 4.1 and 4.2.

4.1. Base Case

[16] The reference base case model (PDISC1), the representation most compatible with current broad scale mapping [Jorgenson et al., 2008] and conceptualization of permafrost distribution at the continuous-discontinuous transition, is constructed to roughly provide a plausible representation of the current regional groundwater flow regime in YFB. The watershed boundary for YFB, encompassing an area of 118,340 km2, defines the model domain (Figure 1). Lateral grid spacing for all simulations presented here is 1500 m, and there are 26 vertical layers, ranging from 1 to 25 m thick. Modeled basin thickness is 400 m. The hydrogeology of the basin, based on descriptions by Williams [1962] and the State Surficial Geology Map of Alaska [Karlstrom et al., 1964] and data from a deep borehole at Fort Yukon [Clark et al., 2009], is here interpreted as belonging to six lithologic units with different hydraulic properties: upper alluvium (sand and gravel mix), lower alluvium (sand to fine silt mix), loess, bedrock, mountain alluvium and colluvium, and permafrost. Table 2 provides a list of parameter values in the base case model. A simplified surficial geology map is used to delineate suprapermafrost units (Figure 1), and a map of the sedimentary basin defines subpermafrost units [Kirschner, 1988; Troutman and Stanley, 2002]. The outline of the sedimentary basin approximately follows the contact between bedrock and unconsolidated material on the surficial geology map.

Table 2. Base Case Hydrogeologic Units and Parametersa
Hydrogeologic UnitKx (m s−1)Anisotropy Kx/KzMaximum Depth to Permafrost (m)Percent Area on Land Surface
  • a

    Kx is horizontal hydraulic conductivity; Kz is vertical hydraulic conductivity; n/a indicates not applicable. Porosity of all hydrogeologic units is set to 0.20. Anisotropy values are assumed.

  • b

    Consistent with measured K for frozen lensed silt and frozen unlensed fine sand [Burt and Williams, 1976].

Upper (coarse) alluvium1.8 × 10−410428
Lower (fine) alluvium1.8 × 10−510n/an/a
Loess1.2 × 10−81113
Bedrock1.2 × 10−71254
Mountain alluvium, colluvium9.3 × 10−6125
Permafrostb1.2 × 10−111n/an/a

[17] Permafrost configuration for PDISC1 follows the simple conceptual model that permafrost exists everywhere below the land surface to some depth, except below rivers that exhibit observed flow under river ice during the winter period, as noted on Figure 1. It is recognized that the simplified conceptual model of permafrost configuration is idealized in some areas. For example, Birch Creek upstream of the loess plateau is historically observed to completely freeze in the winter suggesting the absence of open taliks (H. Best, personal communication, 2011), but for consistency and the purposes of these sets of basic results, open taliks are assumed throughout the length of Birch Creek. The spatial coverage (percent of basin area) of permafrost for PDISC1 is 89%, and permafrost thickness ranges from 86 to 90 m thick depending on the lithologic unit in which it exists because of variable suprapermafrost aquifer thickness. Table 2provides depths to the top of permafrost for each near-surface unit. Permafrost K is set to ∼10−6 times the unfrozen alluvium vertical K value.

[18] A minor variation of the base case is constructed (PDISC1L) in which permafrost is also absent beneath small discrete areas in the lowlands (extent of upper alluvium surficial unit on Figure 1) representing the area of lakes with >100 m diameter, thereby decreasing basin permafrost coverage to 88%. The area affected by this adjustment is small because the overall spatial coverage of lakes is relatively minor and moreover because lakes in the marginal uplands and highlands, comprising another 1%, are not considered in this variation. Further, many of the lakes within the lowlands are already located within the river taliks, particularly the YR talik, and were thus already considered in the base case to be permafrost free. Additional modifications to the base case simulated to test model sensitivity include variable suprapermafrost layer thickness and K, permafrost thickness, and subpermafrost layer transmissivity (K* thickness). Results of these variations (not shown) are briefly discussed.

[19] The base case represents a steady state (SS) flow system. A water table is approximated by prescribing the water table as equal to the land surface elevation as derived from a 60 m resolution DEM [Gesch et al., 2002; Gesch, 2007]. This is a reasonable representation for consideration of regional-scale flow. Lateral and bottom model boundaries are no flow boundaries, representing conditions where groundwater divides may occur and where permeability becomes insignificant at geologic contacts at depth. For the boundaries crossing the Yukon and Porcupine catchments, no flow boundaries coincide with stream gauging stations to account for contributions outside of the model domain.

[20] Winter streamflow measurements at the gauging stations in Figure 1 are used for model comparison and for approximate base case model calibration of the upper alluvium K, the most sensitive estimated parameter to discharge measurements, as indicated by a series of both manual and automated sensitivity analyses. Because of the uncertainty of the winter streamflow data, the variation of measurement seasonal timing and periods of record among the stations, and given that this is a scoping modeling investigation with broad objectives, model calibration is considered sufficient when base flow, calculated by summing (area cumulative) net groundwater fluxes for river reaches, is approximated to the order of magnitude of the measurement. To reduce the number of parameters that need to be estimated, the lower (subpermafrost) alluvium K is set to be 10 times lower than upper alluvium K (based on geologic descriptions by Williams [1962] and Clark et al. [2009]). These values are linked through a multiplier function during calibration.

[21] The steady state simulation represents conditions of seasonal maximum thawed suprapermafrost layer thickness, and thus maximum flow in the suprapermafrost aquifer. This implies an inconsistency when using steady state simulation, because calculated base flow for summer maximum-flow conditions will be compared with winter measurements that perhaps represent the lowest yearly base flow because seasonal freezing of the suprapermafrost layer blocks all or part of groundwater recharge. To check the magnitude of the error made due to this discrepancy, transient simulation from maximum summer to minimum winter flow conditions is also carried out for the base case, as follows. Steady state results described above, representative of full suprapermafrost aquifer conditions, are used as initial conditions for transient simulation. To invoke a 5 month winter period, recharge is set to zero everywhere and prescribed heads are no longer applied at the ground surface. The exception to removal of prescribed heads is at river locations, since flow in these river reaches continues under river ice all winter, and both groundwater recharge and discharge can occur in these locations. A specific yield of 0.2 is assumed, and 1 day time steps are used for the transient period.

4.2. Thaw Sequence Cases

[22] A full sequence of cases from full permafrost (no taliks) to permafrost-free conditions allows the modeling analysis to determine the influence of permafrost on groundwater fluxes and flow paths for a full range of possible conditions. Boundary conditions and permafrost K for all cases are the same as for the base case. Only the internal configuration of permafrost in YFB differs among the cases.Table 3lists selected simulations within this sequence. The progression may either be considered as a collection of arbitrary coverages of different extent, or may be expressed (as described in the following) as a temporal thaw sequence that might occur during a long-term period of climate warming.

Table 3. List of Select Model Simulations and Corresponding Permafrost Distribution
ModelShort NamePermafrost Coverage (%)Description
Complete permafrostPALL100No open taliks
Continuous permafrostPCONT95Narrow open taliks only beneath two largest rivers, Yukon and Porcupine Rivers
Base case discontinuous IPDISC189Open taliks beneath all major rivers
Discontinuous I with lake taliksPDISC1L88Same as above plus open taliks beneath lowland lakes
Discontinuous IIPDISC267Open wide (+5 km on each side from base case) river taliks plus lake taliks
Discontinuous IIIPDISC355All alluvium unfrozen, wide river taliks plus lake taliks
Sporadic IPSPOR130All bedrock unfrozen, wide river taliks plus lake taliks
Sporadic IIPSPOR218Permafrost free except in yedoma and other patchy bedrock areas
Sporadic IIIPSPOR313Permafrost free except in yedoma
Permafrost freePNONE0Entirely unfrozen material

[23] The thaw sequence begins with full permafrost coverage and no open taliks (PALL, 100% permafrost). As coverage begins to decrease from the full permafrost coverage case, narrow open taliks first appear below just the Yukon and Porcupine Rivers (PCONT, 95% permafrost). Then narrow open taliks appear under the footprint of all major rivers (Figure 1) in the YFB (PDISC1, 89% permafrost). Lowland lake taliks appear beginning with the PDISC1L case (88% permafrost). Open river taliks are expanded to cover a maximum of 5 km on each side of the major rivers for the PDISC2 case with 67% permafrost. The transition from discontinuous to sporadic permafrost coverage cannot be inferred from location of surface water bodies as in the previously mentioned cases. Therefore, a variety of cases (not all presented here) with reduced permafrost coverage are considered to span this transition. Uncertainty in the results is captured by preferentially thawing areas of differing geologic properties and topographic position within the thaw sequence: PDISC3 (55% permafrost) represents the extreme in which all alluvium is thawed and PSPOR1 (30% permafrost) represents the extreme in which all bedrock is thawed. The yedoma is assumed last to completely thaw because of its high ice volume (PSPOR2 has only permafrost in yedoma and bedrock, 18% permafrost) (PSPOR3 has only yedoma permafrost, 13% permafrost). Finally, a permafrost-free case is also considered (PNONE, 0% permafrost).

5. Results and Discussion

5.1. Base Case: Steady State

[24] Results for the base case demonstrate the general regional groundwater flow system for the YFB at 89% permafrost coverage. Hydraulic head distribution and water budget analysis are used to highlight the influence of permafrost on groundwater flow. Base flow magnitude, flow paths, travel times, and source partitioning provide additional insight into subsurface water routing in a basin with extensive permafrost coverage. The base case results also serve as a frame of reference for other simulations in which spatial and vertical permafrost distributions are varied.

5.1.1. Estimated K Values

[25] For the base case permafrost configuration (PDISC1), an estimated K value of 1.8 × 10−4 m s−1 for the upper alluvium (and 1.8 × 10−5 m s−1 for lower alluvium, constrained to be one 10th of the upper alluvium value), allows the modeled base flow to best fit the observed winter streamflow data. Though these are the K values used in all simulations presented here, it should be noted that the estimated values depend somewhat on the assumed permafrost configuration. For example, when assuming greater permafrost coverage than PDISC1, the estimated K values must be higher to achieve the measured base flow quantity, and vice versa. Estimated upper alluvium K values obtained to best fit winter streamflow observations range from 1.9 × 10−3 m s−1 for the complete permafrost case (PALL) to 5.6 × 10−5 m s−1for the permafrost-free case (PNONE).

5.1.2. Modeled Base Flow

[26] Table 4provides model-predicted base flow for the base case (PDISC1 column) and observed values. Modeled PDISC1 base flow values are generally within 1 order of magnitude of the measurements. Disparities may be due to the existence of more complex K heterogeneity in the real system than modeled, and to imperfect winter streamflow estimates. Seen in a constructive light, the discrepancies at some stations offer some indication of the actual permafrost coverage in individual watersheds. Stations with overpredicted base flow may have a greater coverage of permafrost in the watershed than prescribed in PDISC1, and stations with underpredicted base flow may have greater areas of open taliks than represented in the model. This speculation is supported by qualitative observations. For example, open taliks are not thought to be present below Upper Birch Creek, yet are represented in PDISC1 to achieve a consistent conceptual model. This could explain the overprediction of modeled base flow for Birch Creek. A similar situation is also thought to exist in the Little Black River watershed, which has an overpredicted modeled value of base flow. In contrast, the extent of permafrost in the Beaver Creek watershed is thought to be less than PDISC1 on the basis of data from an airborne electromagnetic survey used to infer permafrost distribution [Ball et al., 2011], potentially explaining the underpredicted Beaver Creek base flow. Although no parallel empirical evidence of permafrost distribution is available for the Hodzana River watershed, the underprediction of modeled base flow would also suggest permafrost coverage to be less than assumed.

Table 4. Model-Predicted Base Flow for Select Permafrost Distributions
StationModel-Predicted Base Flow to Upstream River Segment (m3 d−1)
  • a

    PSPOR3 results differ only minimally from PSPOR2 and are thus not listed.

  • b

    Values from Table 1.

11.12 × 1051.15 × 1051.31 × 1051.35 × 1052.28 × 1051.49 × 1063.11 × 1063.19 × 1063.03 × 105
23.32 × 1042.97 × 1045.02 × 1043.20 × 1057.90 × 1053.88 × 1057.95 × 1057.94 × 1052.50 × 105
31.61 × 1047.50 × 1049.36 × 1041.28 × 1053.04 × 1053.05 × 1056.73 × 1056.73 × 1054.92 × 104
45.45 × 1035.40 × 1032.27 × 1041.92 × 1051.71 × 1051.58 × 1061.86 × 1055.14 × 1059.79 × 103
54.77 × 1044.89 × 1041.45 × 1046.12 × 1048.30 × 1048.30 × 1041.44 × 1061.44 × 1061.42 × 105
67.27 × 1047.45 × 1047.01 × 1044.96 × 1059.06 × 1051.00 × 1062.50 × 1062.52 × 1062.26 × 105
71.02 × 1041.03 × 1043.26 × 1045.51 × 1054.60 × 1059.04 × 1054.58 × 1054.86 × 1058.56 × 105
84.12 × 1044.20 × 1043.48 × 1051.14 × 1066.52 × 1051.41 × 1061.67 × 1061.71 × 1067.34 × 104
92.55 × 1042.61 × 1041.95 × 1043.96 × 1049.72 × 1041.02 × 1058.66 × 1058.79 × 1051.96 × 104
102.50 × 1042.55 × 1041.27 × 1059.03 × 1052.65 × 1052.04 × 1061.07 × 1061.19 × 1061.22 × 104
111.25 × 1051.28 × 1051.33 × 1051.39 × 1051.92 × 1051.97 × 1053.73 × 1063.77 × 1061.10 × 105
122.26 × 1051.22 × 1063.88 × 1066.92 × 1065.05 × 1068.84 × 1067.99 × 1068.37 × 1063.84 × 106

5.1.3. Modeled Heads

[27] Differences between the steady state head distribution above and below permafrost illustrate the influence of permafrost as a confining unit within the regional flow system (Figures 2a and 2b). Subpermafrost head variations are a greatly subdued version of the large lateral head variations in the suprapermafrost layer. Of particular interest are areas where the subpermafrost aquifer has higher head than the suprapermafrost aquifer above, indicating upward head gradients. In these locations, relatively warm groundwater from deep flow paths would discharge to surface water bodies should there be thawed vertical paths through the permafrost layer. Figure 2c illustrates areas of upward head gradients with positive head differences >0.1 m. Areas with simulated upward gradient occur both in valleys and in the lowlands. Upward gradients in upland valleys are generally high (often >0.1), while the upward gradient in the lowlands is lower, but still relatively high (0.001 to ∼0.3). It is remarkable that the locales of modeled subpermafrost overpressure in the lowlands coincide with regions of high lake density (Figure 2c, close-up). More than half of the >100 m diameter lakes occur in areas of model-predicted upward gradients exceeding 0.001. Should lakes in these regions possess open taliks, these may serve as conduits for upward flow from the subpermafrost aquifer.

Figure 2.

Steady state head distribution for (a) the suprapermafrost zone (for base case PDISC1) and (b) the subpermafrost zone. (c) Areas of upward gradients across the permafrost layer are indicated by the color scheme in the legend. The close-up for the lowlands (within green outline) shows the distribution of lakes >100 m in diameter.

5.1.4. Water Budget: Recharge

[28] Overall water budget results from PDISC1 indicate a groundwater-flow system that transports recharge equivalent to 55 mm yr−1 (Table 5), when expressed as a uniform distribution across the entire surface of the model domain. This amounts to 11% of the basin-wide PRISM-estimated precipitation. The budget results (schematically illustrated inFigure 3) demonstrate the important role of taliks in the circulation of groundwater in permafrost-dominated systems. For the base case, 76% of the total groundwater recharge through the ground surface occurs above open taliks, which comprise only 11% of the total area. The remaining 24% of recharge enters through the ground surface in areas overlying the permafrost layer. This recharge partitioning is explained by the contrast in transmissivity of the uppermost aquifer between areas that are permafrost-free and areas that are underlain by permafrost.

Figure 3.

Steady state zone water budget results (not drawn to scale) for the base case, PDISC1. Fluxes are reported in units of 105 m3 d−1. Values in parentheses represent the percent increase in flux for PDISC1L (with lake taliks) relative to PDISC1 flux.

Table 5. Summary of Steady State Results
ModelAverage Rechargea (mm yr−1)Total Base Flow (m3 d−1)Groundwater Discharged as Base Flow (%)Suprapermafrost Contribution to Base Flowb (%)Median Base Flow Travel Times (years)Median Base Flow Path Length (km)
  • a

    Value is calculated by the model, reflects an average across the entire model domain, and represents maximum recharge capacity for the prescribed hydrogeologic conditions.

  • b

    For cases in which permafrost is absent, this component represents the shallow (≤4 m) aquifer contribution.

PALL137.4 × 105189337125.72
PCONT282.3 × 10620471143.56
Base case/PDISC1555.4 × 10630193353.73
PDISC1L565.4 × 10630193203.68
PDISC21461.1 × 107238.54863.67
PDISC31669.2 × 10617104573.40
PSPOR12381.7 × 107225.63653.36
PSPOR24302.4 × 107183.84803.30
PSPOR34532.5 × 107173.84553.22
PNONE4582.6 × 107173.74613.26

5.1.5. Water Budget: Distribution of Groundwater Flow and Transit Times

[29] Flow through the suprapermafrost aquifer totals 43.7 × 105 m3 d−1 representing 25% of groundwater flow through the system. Although the suprapermafrost zone covers 89% of the basin in area, it comprises less than 1% of the basin by volume. Thus, the result that a relatively large proportion (25%) of groundwater flows through the suprapermafrost zone signifies relatively rapid flushing. Calculated mean transit time to flush one pore volume through the entire suprapermafrost aquifer (including the low K bedrock and loess) is ∼30 years. Of the total suprapermafrost flow, 76% is discharged to topographic lows (e.g., lakes, wetlands, and streams) located away from the major rivers and open taliks. Lateral flow through the suprapermafrost zone to shallow taliks associated with major rivers comprises another 21% of total flow in the suprapermafrost aquifer. The latter quantity discharges to rivers and comprises a component of river base flow. Only a minor fraction (3%) of flow in the suprapermafrost zone seeps downward through the permafrost layer, given the low permafrost K (Table 2).

[30] Most of the modeled recharge (95%) entering the subsurface above deep taliks flows deeper than 4 m below the surface, and 22% of this recharge continues to depths below the base of the surrounding permafrost layer (into the subtalik aquifer in Figure 3). Of the flow that enters from above into the subtalik aquifer, 36% moves laterally away from the subtalik zone, entering the region of the subpermafrost aquifer. This amount plus the relatively small amount of flow derived from transport through the thick permafrost eventually moves laterally back to the subtalik aquifer zone, discharging to the river above after moving upward through subriver deep taliks. The quantity of flow circulating in the subpermafrost aquifer is 12.3 × 105 m3 d−1, representing 7% of the total groundwater flow in the system for the base case. This quantity increases to 31.6 × 105 m3 d−1 when subpermafrost flow directly below taliks are considered, representing 18% of total groundwater flow being transmitted through the ∼80% of the basin volume encompassed by the subpermafrost and subtalik aquifers. Calculated mean transit times to flush one pore volume through the subpermafrost aquifer and the subtalik aquifer are 19,000 and 900 years, respectively.

[31] Base flow, summed for rivers represented in the model domain is 54.0 × 105 m3 d−1, or 30% of the total groundwater flow in the system. Base flow includes contributions from lateral suprapermafrost flow (19%), vertical upward flow from below the depth of permafrost (i.e., upward flow from the subtalik aquifer) (58%), and intratalik flow (23%). The ratio of subpermafrost to suprapermafrost contributions to base flow is more than 3 to 1, indicating a major contribution of water from deep groundwater flow. These results are composite for the entire YFB; individual subbasins will have differing proportions of suprapermafrost to subpermafrost contributions that are dependent on basin characteristics.

5.2. Incorporation of Discrete Lake Taliks and Other Base Case Variations

[32] When lakes in the lowlands, covering an area of the model domain consistent with the area of lakes with >100 m diameter, are simulated as having open taliks (PDISC1L), the calculated permafrost coverage is ∼1% lower than for the base case permafrost configuration (PDISC1). Percent increases in flow resulting from lake taliks (PDISC1L compared to PDISC1) are given in parentheses in Figure 3. Overall flow in the system increases only modestly, since lowland lakes generally coincide with areas of groundwater discharge. Fluxes to and from the river zone remain unchanged. Fluxes into and out of the shallow talik and deep talik zones increase by 3% to 18% because of the additional vertical conduits for flow. Lateral flow in the subpermafrost aquifer to and from the subtalik aquifer increases by 14 to 16%. In summary, lowland lake taliks comprising only ∼1% of the model domain, exert a proportionally large effect on groundwater circulation patterns, increasing some fluxes by nearly 20%. The greatest effects of lake taliks on lateral groundwater flow are localized and are thus, locally substantial.

[33] When transmissivity of the suprapermafrost zone is modified by varying its thickness (i.e., depth to permafrost), the effect on the YFB water budget is rather small, although changes in base flow to individual rivers may be large. For example, a rather extreme 50% reduction/enhancement in depth to permafrost yields minor a 12% reduction/enhancement in total groundwater flow and a 7% reduction/enhancement in total base flow in the YFB relative to the base case. However, the same 50% reduction/enhancement in depth to permafrost yields changes of 40 to 65% in base flow of individual rivers that receive a large proportion of suprapermafrost flow to total base flow (e.g., Porcupine R., Upper Beaver Cr., Preacher Cr.). These types of systems would be most likely to show changes in streamflow recession that could be attributed to active layer deepening [e.g., Lyon et al., 2009]. Similar results manifested in YFB base flow magnitude are generated for changes in suprapermafrost layer thickness as are induced by 50% reduction/enhancement in hydraulic conductivity, because both serve to change suprapermafrost aquifer transmissivity by the same factor.

[34] In contrast to the depth to permafrost top, results show very little sensitivity to the depth to permafrost bottom, so long as the ensuing change in subpermafrost aquifer transmissivity is low. A 50% thinner permafrost layer relative to the base case exhibits negligible changes in groundwater flow for the YFB; in this case, transmissivity of the subpermafrost layer is increased by <15%. However, order of magnitude changes in transmissivity of the subpermafrost aquifer can impact the YFB groundwater flow system. Increasing and decreasing the subpermafrost layer transmissivity by an order of magnitude from the base case (by increasing subpermafrost K in the model) notably influences the amount of groundwater circulating at depth and changes the proportion of base flow derived from suprapermafrost flow from 19% (PDISC1) to 10% and 22%, respectively.

5.3. Base Case: Transient

[35] Under transient conditions for the base case, during a 5 month winter period, simulated groundwater discharge to streams decreases with time in an amount that varies among the tributaries examined. The YR experiences a 25% reduction in simulated base flow over the 5 month period (1 November to 31 March). Decreases of similar proportion as the YR are noted for Chandalar and Black Rivers, 20% and 24%, respectively. Upper Beaver Creek and Birch Creek sustain the greatest modeled reductions in groundwater recharge, 62% and 47%, respectively. Because heads along the rivers are prescribed to be time independent in this simulation, while heads below the land surface everywhere else are allowed to change, reflective of a release of water from the unconfined aquifer, some of the river reaches examined actually become losing reaches (i.e., provide recharge to groundwater) during the simulated transient period. This change from net gaining to net losing is observed for the following river stretches: Sheenjek River, Hadweenzic River, Hodzana River, Middle Beaver Creek, and Preacher Creek. Spatially and temporally averaged over the YFB for the 5 month transient period, 13 mm yr−1 enters the groundwater system from these net losing river reaches and 21–35 mm yr−1 is released from storage as the water table declines. The range in calculated input from water table decline depends on whether or not a modeled reduction in K, equivalent to permafrost K, is invoked for the seasonal ice layer. The least amount of input water is yielded from the water table decline when a seasonal ice layer is represented with a permafrost K for the transient period. Future work may explore seasonal dynamics in greater detail. Ge et al. [2011]provide an excellent basis for this type of analysis. The transient results indicate that the steady state simplification does not adversely impact the primary modeled behavior and results because the estimated alluvium K that allows an order-of-magnitude fit to measured base flow has larger inherent approximations than those imposed by the steady state modeled base flow.

5.4. Thaw Sequence

[36] Selected results are presented from several permafrost configurations that may be considered to reflect key stages of a reducing permafrost sequence, including (1) continuous permafrost coverage (PCONT; 95% permafrost), (2) discontinuous I, cold (PDISC1; base case, 89% permafrost), (3) discontinuous II, warm (PDISC2; 67% permafrost), and (4) permafrost-free conditions (PNONE). Numerous other intermediate permafrost configurations were simulated and several of these additional outcomes are also presented in the results tables (Tables 4 and 5). The greatest hydrologic changes to the system occur within the continuous to discontinuous sections of the permafrost coverage spectrum, and results presented here focus on this range.

5.4.1. Impact of Permafrost Distribution on Groundwater Discharge Pattern

[37] The pattern of groundwater discharge is strongly controlled by the permafrost distribution (Figure 4). For the continuous permafrost case (PCONT), groundwater discharge is focused in the permafrost-free Yukon River valley and to upland river valleys with the greatest vertical relief (Figure 4a). The widening of taliks under the YR and opening of taliks beneath all major tributaries for PDISC1 has the effect of focusing discharge through river taliks in the lowlands (Figure 4b). The increase in vertical connectivity in the system, concurrent with progressive loss of permafrost (PDISC1 to PDISC2 to PNONE, shown in Figure 4) results in expanded regions of groundwater discharge in the lowlands (Figures 4b–4d). In all cases, a band of focused discharge occurs at the base of the southern loess plateau; discharge rates are also greatest in the same region for the permafrost-free simulation (PNONE). This persistence implies that the base of the loess plateau (and its lakes and streams) tends to be a major discharge area for regional groundwater flow from the highlands in the south irrespective of the state of permafrost.

Figure 4.

Model results for steady state groundwater discharge patterns with varying permafrost distributions.

5.4.2. Impact of Permafrost Distribution on Magnitude of Groundwater Flow

[38] Total groundwater flow increases with decreasing permafrost coverage. Average basin-wide recharge values (and thus, total groundwater flow for steady state) increases nearly an order of magnitude from the base case (PDISC1) to permafrost-free conditions (PNONE) (Table 5). This represents the absolute maximum increase in recharge that could be expected in the YFB for a future in which all permafrost has thawed. However, the value of recharge for PNONE approaches that of the current annual basin-wide precipitation indicating that truly predictive scenarios of future permafrost conditions must incorporate a more complex near-surface water balance representation in the model to accurately route and partition near-surface water. In other words, at some point in an actual permafrost thaw sequence, recharge is expected to be more limited by evapotranspiration and runoff than by overall basin transmissivity as is implicitly assumed in the current approach. Consideration of near-surface water dynamics is beyond the scope of this present work, but is an area of future research interest. Yet, for conditions from PCONT through PDISC2, the presumption that basin transmissivity is the primary limiter of recharge is supported by reasonable recharge/precipitation ratios of less than one third.

[39] The influence of permafrost coverage on the total groundwater flow (m3/d) and the ratio of base flow to total groundwater flow are illustrated in Figure 5. The increase in groundwater flow resulting from permafrost thaw depends on relative contrasts in K between the geologic materials that are thawing and their frozen counterparts. The greater the contrast between the frozen and unfrozen material, the greater the change manifested in flow rates. Therefore, the magnitude of the change in groundwater flow is dictated by the basin transmissivity in its frozen state versus its wholly unfrozen state. Further, the shape of the curve of groundwater enhancement with decreasing permafrost coverage is determined by the pattern of thaw with respect to K of the various geologic materials. The YFB thaw sequence between 100% and 67% permafrost coverage is based upon a logical progression of talik formation associated with the location of surface water bodies. But the thaw progression between 67% and 18% is less constrained; necessitating two curves within this interval to bound the possibilities, shown as the thickened gray line in Figure 5. The upper boundary represents conditions in which all the permeable alluvium progressively thaws prior to bedrock thaw. The lower boundary represents conditions in which all bedrock progressively thaws before the more permeable alluvium [Romanovsky et al., 2010]. There is rationale both for (1) permeable material thawing preferentially over less permeable material as a result of advection and (2) the south facing slopes of bedrock highlands showing preferential thaw. An actual progression between 67% and 18% permafrost coverage is thus expected to fall somewhere within the bounds given in the graph. Finally, the graph illustrates the relative unimportance of the existence of permafrost on groundwater flow as permafrost coverage nears the permafrost-free end of the spectrum. Small patches of permafrost have minimal influence on the regional flow regime. This effect is further augmented because the last material to thaw because of its thermal inertia is the ice-rich loess, which has the lowest unfrozen K of all geologic materials represented in the YFB aside from permafrost. A change from an extremely low K to a very low K alone has little consequence.

Figure 5.

Steady state model results for total groundwater flow and the ratio of base flow to total groundwater flow. The shaded zone bounds the range of the base flow ratio dependent on the spatial pattern of permafrost thaw between 67% and 18% permafrost coverage. Similar range in the total groundwater flow is shown by the thicker line.

5.4.3. Impact of Permafrost Distribution on Discharge Focus to Rivers

[40] Unlike the total groundwater flow trend, the trend in the ratio of base flow to total groundwater flow with decreasing permafrost is not monotonic (Figure 5). The transition from continuous to discontinuous permafrost coincides with a pivotal point in the base flow ratio trend from positive to negative. Near the 100% permafrost end of the spectrum, the groundwater system behaves primarily as a very thin suprapermafrost aquifer, 1–4 m in thickness. Base flow comprises just 0.18 of the total flow in the groundwater system. As permafrost coverage decreases, in the early thaw stages of the assumed sequence, discrete river taliks appear (i.e., PCONT, PDISC1). Discharge at these stages includes some subpermafrost aquifer flow that is focused to the rivers, thereby increasing the ratio of total groundwater flow that discharges to rivers to an all-sequence maximum of 0.30 for PDISC1. For cases with greater thawed zones (i.e., PDISC2), where river taliks are wider and lowland lake taliks are considered, discharge is more widely distributed (Figure 4), resulting in the trend of decreased base flow ratio shown in Figure 5, down to values of 0.17 near the permafrost-free end of the spectrum. The hatched zone inFigure 5 bounds the uncertainty in the base flow ratio given a range of thaw patterns described above. The analyses displayed in Figure 5 suggest that, for a region similar to YFB, the greatest changes in the distribution of discharge relevant to base flow will occur near the transition from continuous to discontinuous permafrost (practically defined at ∼90%; [van Everdingen, 1998]).

5.4.4. Impact of Permafrost Distribution on Base Flow Flow Paths

[41] Travel times and path lengths of groundwater comprising base flow, for various permafrost configurations, are provided in Table 5. Base flow flow paths are shown in Figures 6a–6dfor several stages of the thaw sequence. For the full permafrost stage (PALL), a large proportion of base flow is derived from the suprapermafrost zone, as confirmed by the water budget results described below. Base flow flow paths in this case through the suprapermafrost layer are short in length and most have relatively short travel times. However, the PALL model has the largest median base flow travel time of all cases considered, 3712 years, reflecting long, slow paths through the low permeable permafrost. Although still present, the influence of these long cross-permafrost paths is lessened in the continuous permafrost case (PCONT) since some deep flow paths through river taliks occur in addition to short, fast suprapermafrost flow paths (Figure 6a and Table 5). The median base flow travel time in this case is 114 years. The transition from continuous to discontinuous permafrost and the formation of an expanded network of river taliks promotes development of deeper and longer flow paths with greater travel times, as shown for PDISC1 (Figure 6b and Table 5). The base case simulation (PDISC1) yields a median base flow travel time of 335 years and 3.73 km median path length. Three-dimensional examination of these long flow paths indicates movement through the subpermafrost zone connecting up-gradient regions to down-gradient regions below river taliks. Flow path analysis also indicates that there is a reduction in flow path lengths and travel times from the base case to cases with less permafrost (Table 5 and Figures 6b–6d). With decreasing permafrost coverage beyond the base case, some of the long flow paths are truncated as vertical connectivity in the system increases. Maximum path lengths decrease twofold to threefold from PDISC1 (124 km) to permafrost coverages of 67% and less. Yet, transitioning away from shallow, fast suprapermafrost flow paths is reflected by a gradual decrease in the percentage of flow paths <100 years from 41% (PCONT) to 31% (PNONE). Histograms of base flow travel times with computed cumulative distribution functions are shown in Figure 7 for the four stages in the thaw sequence corresponding to base flow flow paths delineated in Figure 6. For PCONT, base flow travel times are clearly bimodal (Figure 7a); the younger travel times (centering on the 102 year timescale) correspond to short flow paths through the suprapermafrost layer, and the set of much older travel times (centering on the 106 year timescale) correspond to slow seepage through the 90 m thick permafrost layer and subpermafrost zone with maximum path lengths approaching 250 km. This bimodality continues for the discontinuous permafrost distributions (PDISC1, 2, Figures 7b and 7c), but the intermediate travel times (104–105years), absent at the complete permafrost stage (not shown) and minimally present at the continuous permafrost stage, are well represented because of the development of long intratalik and subpermafrost flow paths. When permafrost-free conditions are reached, base flow travel times are no longer bimodal, reflecting the lack of a confining layer capable of partitioning flow paths into shallow/young and deep/old components (Figure 7d).

Figure 6.

Model results for steady state base flow flow paths with varying permafrost distribution. Color shows elapsed travel time when groundwater flows to river.

Figure 7.

Histograms and cumulative distribution functions of modeled base flow travel times for varying permafrost distributions. Binned values are centered at 0.5 log 10 spacing and range ±0.25 log 10.

5.4.5. Impact of Permafrost Distribution on Base Flow Magnitude

[42] Simulated base flows in river reaches upstream of their gauging stations are reported for several of the permafrost distribution stages (Table 4). Base flow magnitude increases substantially with decreasing permafrost coverage, exceeding an order of magnitude increase from complete permafrost (PALL) to permafrost-free conditions (PNONE) (Figure 8). Pertinent substantial base flow change applicable to the near term (101–102 years), as a result of current warming trends, may be represented by the change between one of the stages depicted in Figure 8 and the following stage. Model results that show an overall trend in increased base flow with diminishing permafrost coverage are consistent with the hypothesis that permafrost degradation has led to the observed increases in groundwater input to rivers in northern basins, as noted by Smith et al. [2007], Walvoord and Striegl [2007], and St. Jacques and Sauchyn [2009].

Figure 8.

Model-calculated base flow for selected river reaches with a variety of permafrost distributions.

5.4.6. Impact of Permafrost Distribution on Base Flow Sources

[43] In addition to considering potential changes in base flow magnitude, it is important to address the sources and flow paths of base flow and how these may change with permafrost degradation. This is particularly important in evaluating potential changes in aquatic chemical exports as a consequence of permafrost thaw [Frey and McClelland, 2009]. Groundwater modeling results for YFB indicate a sharp decline in the proportional contribution of lateral flow from the suprapermafrost layer to base flow near the continuous to discontinuous permafrost transition (Figure 9; also reported in Table 5). The suprapermafrost contribution declines from 93% in PALL to 19% in PDISC1 and is ≤10% for permafrost coverages 67% and less. The magnitude of flow through the suprapermafrost layer remains relatively stable through the thaw sequence, but the flow from the subpermafrost aquifer increases by about 1.5 orders of magnitude along the full permafrost thaw sequence causing suprapermafrost flow to become increasingly less important from a surface water budget perspective. The major change in the proportion of suprapermafrost contribution to deep talik/sub permafrost contribution occurs at the cold end of the permafrost coverage spectrum (greater than 67% permafrost). Once again, model results indicate that conditions in the base case, assumed to be a reasonable proxy for current conditions in YFB, are positioned for major hydrologic transformation with continued degradation of permafrost. Changes in the magnitudes and ratios of organic to inorganic aquatic chemical exports would be expected to accompany the changes in the proportional contributions of suprapermafrost flow to base flow illustrated in Figure 9.

Figure 9.

Model-calculated percentage of suprapermafrost (or shallow aquifer in thawed cases) contribution to base flow. The range associated with thaw patterns beginning in late discontinuous stages is indicated by the thicker line.

6. Conclusions and Implications

[44] This study elucidates modes of groundwater flow for a variety of possible permafrost distributions in the Yukon Flats Basin (YFB), representative of other similar high-latitude groundwater systems. The base case pattern, derived from the conceptual model of permafrost existing everywhere except under large rivers, serves as a reference for the results of other configurations. Results for the spectrum of permafrost configurations reveal basic understanding of the influence of permafrost on groundwater flow and broad hydrologic consequences of potential permafrost degradation.

[45] The steady state base case illustrates that most groundwater recharge and discharge occurs through areas overlying open taliks, emphasizing the importance of understanding and characterizing talik extent and morphology. Modeled YFB base flow represents 30% of the total groundwater flow in the system. This base flow is composed of 19% suprapermafrost groundwater flow, 23% intratalik flow, and 58% subpermafrost flow.

[46] Permafrost functions as an effective confining layer in the model, causing relatively high head below the permafrost in low-lying regions of YFB. Large modeled regions of upward gradient across the permafrost layer coincide with locations of concentrated lake distributions in the low-lying Yukon Flats, suggesting a possible link between the existence and persistence of lowland lakes and upwelling of deep groundwater. The confining effect of permafrost also serves to partition flow paths into shallow/young and deep/old components. Residence times for pore water in the suprapermafrost aquifer (shallow) are on the order of 101 years, compared with residence times of 103–104 years for pore water in aquifers below permafrost. Bimodal distributions of base flow travel times exhibited by the model also highlight distinct differences in timescales associated with flow through the suprapermafrost and subpermafrost aquifers.

[47] Model results further suggest that open lake taliks have an important effect on enhancing groundwater flux through suprapermafrost and subpermafrost aquifers as well as through the deep talik system because of the addition of vertical conduits for flow. Assessment of resistivity and inferred permafrost patterns from an airborne electromagnetic survey in the Yukon Flats provides good support for the presence of open taliks associated with large lakes [Minsley et al., 2012]. Thus, geophysical indication of open lake taliks together with the results of the modeling work that demonstrates their potential influence on the groundwater circulation and groundwater–surface water interaction provide motivation for characterizing lake taliks (assessing open or closed and mapping 3-D extent) and incorporating open taliks below lakes in groundwater flow models of permafrost-impacted basins.

[48] The suite of model simulations represents a possible permafrost thaw sequence for YFB highlighting the strong control of permafrost configuration on groundwater flux and flow pattern, and locations of recharge and discharge. In general, overall groundwater flux increases with decreasing permafrost coverage. Discharge locations in the lowlands expand as more areas become permafrost free, suggesting that permafrost degradation may influence the state of lakes and wetlands whose water budgets are supported by groundwater discharge.

[49] The thaw sequence results also demonstrate the influence of permafrost distribution on two important factors: the magnitude of base flow and the proportion of various groundwater sources to base flow. The proportion of total groundwater flow in the system that becomes base flow is at a maximum for YFB for the conceptual model consistent with permafrost-free zones existing only beneath major surface water bodies. Modeling analysis demonstrates a substantial increase in base flow with decreasing permafrost coverage. This trend is consistent with the hypothesis that increases in base flow across the YRB and other northern river basins in the past several decades are a consequence of permafrost degradation. Given constant total stream discharge as has been observed in the YRB [Walvoord and Striegl, 2007], an increase in base flow implies a decrease in surface water runoff and thereby longer water residence times of composite river water due to the increased (older) base flow component of total streamflow. Adding to that effect, the reduction in the proportion of suprapermafrost flow to base flow with decreasing permafrost coverage, as revealed by model analysis, also serves to increase residence times of base flow and ultimately, composite river water. In addition to river water residence times, increased base flow and the shift toward deeper sources of base flow with decreased permafrost coverage have implications for the chemical composition of riverine exports [Frey and McClelland, 2009]. On the basis of what is generally understood regarding the chemistry of supra versus sub permafrost waters [i.e., Spencer et al., 2008; O'Donnell et al., 2010], a shift toward decreased dissolved organic exports and increased dissolved inorganic exports would be expected to accompany the model-predicted trends in base flow magnitude and source contribution. Hence, this work provides support for the hypothesis that the historical decrease in YRB discharge-normalized dissolved organic carbon export and increase in dissolved inorganic export [Striegl et al., 2005] might have been caused by changes in groundwater fluxes and flow paths resulting from permafrost thaw.

[50] As the proportion of base flow to total streamflow increases and the proportion of subpermafrost flow to base flow increases, a cascade of consequences can be anticipated, with a significance that depends on the magnitude of this shift. Possible implications of an increase in the proportion of groundwater flow within total streamflow include (1) decreased seasonal variability of stream discharge (subduing of the hydrograph thereby impacting sediment transport and river geomorphology), (2) changed seasonal stream temperatures (cooler summer temperatures and warmer winter temperatures, thereby impacting fish habitats), (3) decreasing river ice thickness and timing of ice breakup (thereby impacting winter transportation), and as mentioned above, (4) changed aquatic chemistry (thereby impacting flora and fauna) and basin aquatic exports (thereby impacting freshwater and carbon inputs to the Arctic Basin [McGuire et al., 2010]). Though vertical permafrost thaw via active layer deepening can also contribute to these consequences, wholesale spatial permafrost thaw intensifies these impacts because of the enhancement of subpermafrost groundwater–surface water interaction. Some of the above mentioned changes are already being observed in the Yukon River Basin.

[51] This groundwater investigation of the Yukon Flats Basin illustrates the important control of permafrost on regional-scale hydrology and identifies the continuous to discontinuous permafrost coverage transition as particularly vulnerable to hydrologic change. On the basis of the most recent map of large-scale permafrost distribution in Alaska [Jorgenson et al., 2008], the YFB spans this critical transition. Major alterations, including changes in the proportion of total groundwater flow that discharges to rivers and changes in base flow source waters, are predicted near this transition with only minor changes in permafrost distribution. Many studies regard regions of “warm” discontinuous permafrost as being well poised for change [Osterkamp and Romanovsky, 1999; Hinzman et al., 2005]. Though this may be true, with respect to rate and extent of permafrost thaw and ecosystem impact [Osterkamp et al., 2000; Jorgenson et al., 2001], results shown here suggest that the continuous-discontinuous transition may be a more important threshold for groundwater system transformation and its cascade of consequences. This work identifies a need for better characterization of permafrost and other hydrogeologic information in permafrost-impacted basins (e.g., airborne [Abraham, 2011; Ball et al., 2011; Minsley et al., 2012] and ground-based geophysical techniques in YFB [Nolan et al., 2011]) and integration of such information into site-specific hydrologic models. This study also indicates the need for better understanding of the impact of groundwater flow itself on freezing and thawing of seasonal ground ice and permafrost [e.g.,McKenzie et al., 2007; Bense et al., 2009; Ge et al., 2011; Rowland et al., 2011]. Integration of an improved understanding of the current permafrost distribution, its relation to surface water bodies, and time responses to changes in groundwater flow and thermal properties at the surface, including surface water redistribution such as river migration and lake appearance/disappearance, will serve as a critical step toward developing more reliable models that can predict hydrologic changes resulting from possible future changes in climate.


[52] We gratefully acknowledge support from the USGS National Research Program, the USGS Climate and Land Use Change and Water Mission Areas, and SERDP grant RC-2111. K. E. Johnson provided GIS support and analysis. We thank J. O'Donnell, J. Rowland, three anonymous reviewers, and the associate editor for their insightful comments.