Hydrogeology Journal

, Volume 21, Issue 1, pp 93–106

Rapid runoff via shallow throughflow and deeper preferential flow in a boreal catchment underlain by frozen silt (Alaska, USA)

Authors

    • US Geological SurveyAlaska Science Center
  • S. A. Ewing
    • Department of Land Resources and Environmental SciencesMontana State University
  • R. Striegl
    • US Geological SurveyBranch of Regional Research
  • D. M. McKnight
    • Department of Civil, Environmental, and Architectural EngineeringUniversity of Colorado
Paper

DOI: 10.1007/s10040-012-0934-3

Cite this article as:
Koch, J.C., Ewing, S.A., Striegl, R. et al. Hydrogeol J (2013) 21: 93. doi:10.1007/s10040-012-0934-3

Abstract

In high-latitude catchments where permafrost is present, runoff dynamics are complicated by seasonal active-layer thaw, which may cause a change in the dominant flowpaths as water increasingly contacts mineral soils of low hydraulic conductivity. A 2-year study, conducted in an upland catchment in Alaska (USA) underlain by frozen, well-sorted eolian silt, examined changes in infiltration and runoff with thaw. It was hypothesized that rapid runoff would be maintained by flow through shallow soils during the early summer and deeper preferential flow later in the summer. Seasonal changes in soil moisture, infiltration, and runoff magnitude, location, and chemistry suggest that transport is rapid, even when soils are thawed to their maximum extent. Between June and September, a shift occurred in the location of runoff, consistent with subsurface preferential flow in steep and wet areas. Uranium isotopes suggest that late summer runoff erodes permafrost, indicating that substantial rapid flow may occur along the frozen boundary. Together, throughflow and deep preferential flow may limit upland boreal catchment water and solute storage, and subsequently biogeochemical cycling on seasonal to annual timescales. Deep preferential flow may be important for stream incision, network drainage development, and the release of ancient carbon to ecosystems.

Keywords

Rainfall/runoffPermafrostPipeflowSolute transportAlaska (USA)

Ecoulement rapide par conduit superficiel et flux profond préférentiel dans un bassin versant boréal sous-jacent à un silt gelé (Alaska, USA)

Résumé

Dans les bassins versants de haute latitude où le permafrost est présent, les dynamiques de lessivage et d’écoulement sont compliquées par le dégel saisonnier des couches productives, qui peut causer une modification des chenaux principaux si l’eau entre progressivement en contact avec des sols minéraux à conductivité hydraulique basse. Une étude de deux ans, menée sur un bassin versant de hautes terres en Alaska (USA) sous-jacent à silt éolien bien calibré gelé, a examiné les changements d’infiltration et d’écoulement avec la dégel. On a pris comme hypothèse qu’un écoulement rapide serait soutenu par un flux à travers des sols peu épais au cours du début de l’été et par un flux préférentiel profond plus tard durant l’été. Les variations saisonnières d’eau du sol, infiltration et intensité du ruissellement, localisation et chimie, suggèrent que le transport est rapide, même au maximum d’extension du dégel. Entre juin et septembre, l’emplacement de l’écoulement change, en rapport avec l’écoulement préférentiel de subsurface, dans les zones en pentes et humides. Des isotopes de l’uranium suggèrent que l’écoulement d’été tardif érode le permafrost, indiquant qu’un écoulement rapide substantiel peut savoir lieu le long de la limite gelée. Simultanément, l’écoulement superficiel et l’écoulement préférentiel profond peuvent limiter le bassin versant du plateau boréal et l’emmagasinement de soluté, et subséquemment le cycle biochimique aux échelles saisonnière à annuelle. L’écoulement profond préférentiel peut être important pour la coupure du flot, le développement du réseau de drainage et la restitution de carbone ancien à l’écosystème.

Escurrimiento rápido vía flujo horizontal somero y flujo preferencial más profundo en una cuenca boreal subyacente a sedimentos congelados (Alaska, EEUU)

Resumen

En las cuencas de altas latitudes donde el permafrost está presente, la dinámica de lixiviación y escurrimiento se complican por el deshielo estacional de la capa activa, que puede causar un cambio en las trayectorias dominantes del flujo de agua cada vez más en contacto con suelos minerales de baja conductividad hidráulica. Un estudio de dos años de duración, realizado en una cuenca alta de Alaska (EEUU) sustentada por limos eólicos congelados, bien ordenados, examinó los cambios en la infiltración y el escurrimiento con el deshielo. La hipótesis fue que el escurrimiento rápido podría ser mantenido por el flujo a través de suelos someros a principios del verano y por el flujo preferencial más profundo después del verano. Los cambios estacionales en la humedad del suelo, infiltración, la magnitud del escurrimiento, la ubicación y la química sugieren que el transporte es rápido, incluso cuando los suelos están descongelados en su máxima extensión. Entre junio y septiembre se produjo un cambio en la ubicación del escurrimiento, consistente con el flujo subsuperficial preferencial de zonas escarpadas y húmedas. Los isótopos de uranio sugieren que a finales del verano el escurrimiento erosiona el permafrost, lo que indica que un sustancial flujo rápido puede ocurrir a lo largo del límite congelado. Conjuntamente, el flujo horizontal somero y el flujo profundo preferencial pueden limitar el agua de la cuenca boreal alta y el almacenamiento de solutos, y posteriormente el ciclo biogeoquímico en escalas de tiempo estacionales a anuales. El flujo profundo preferencial puede ser importante para la incisión corriente, para el desarrollo de la red de drenaje, y para la liberación de carbono antiguo a los ecosistemas

美国阿拉斯加州冻土上北向流域内通过浅部径流和深部优先流实现的快速径流

摘要

高纬度流域内的冻土永久存在,季节性活动层消融导致的淋滤及径流的动力学复杂,会引起由于水与低水力传导系数的矿质土壤接触增加而发生的主导流径的变化。美国阿拉斯加州山地流域下伏分选好的风成冻土,在该区对融水的渗透及径流变化监测两年。假定快速流可通过夏季初的浅层土壤以及夏季晚期的深层优先流维持。土壤水分、渗透量、径流量、位置以及化学结果的季节性变化表明即使土壤解冻到最大程度,径流也是迅速的。六月到九月,径流的位置存在转变,与陡湿区地下优先流相一致。铀同位素显示晚夏的径流消融了永久冻土,表明大量快速流在冻土边界发生。同时,浅部径流及深部优先流限制了山地北面流域中水及溶质的存储以及随后的季节及多年时间尺度上的生物地球化学循环。深部优先流对于河流切割、排水网络发育、以及古代碳向生态系统中的排放是有意义的。

Escoamento rápido via escoamento subsuperficial e fluxo preferencial profundo numa bacia boreal em siltes congelados (Alasca, EUA)

Resumo

Nas bacias localizadas a altas latitudes onde ocorre permafrost, a dinâmica de lixiviação e de escoamento é dificultada pelo degelo sazonal da camada ativa, o qual pode causar uma mudança nos sentidos de fluxo dominantes enquanto a água aumenta o contacto com solos minerais de baixa condutividade hidráulica. Um estudo de dois anos realizado na cabeceira de uma bacia hidrográfica no Alasca (EUA) coberta por material congelado, siltes eólicos bem calibrados, observou variações na infiltração e no escoamento durante o degelo. Admitiu-se a hipótese que o escoamento rápido seria mantido pelo fluxo através dos solos superficiais durante o princípio do verão e que o fluxo preferencial mais profundo ocorreria no final do verão. As variações sazonais na grandeza da humidade do solo, da infiltração e do escoamento, na localização e no quimismo sugerem que o transporte é rápido, mesmo quando os solos sofrem descongelação na sua máxima extensão. Entre junho e setembro ocorreu um deslocamento da posição do escoamento, consistente com fluxo subsuperficial preferencial em áreas declivosas e húmidas. Os isótopos de urânio sugerem que o escoamento do fim do verão erode o permafrost, indicando que este fluxo rápido substancial pode ocorrer ao longo da fronteira de congelação. O escoamento subsuperficial e o fluxo preferencial profundo podem em conjunto limitar o armazenamento de água e de solutos nas bacias superiores em zonas boreais e, subsequentemente, o ciclo biogeoquímico às escalas sazonal a anual. O fluxo preferencial profundo pode ser importante para a incisão das linhas de água, o desenvolvimento da rede de drenagem e a libertação do carbono antigo para os ecossistemas.

Introduction

Northern hemisphere boreal soils contain as much as one third of the world’s carbon (Dixon et al. 1994; Schuur et al. 2008). Much of this carbon is stored in frozen soils and is thus unavailable to ecosystems, while the remainder exists in the vegetation and shallow organic soils. Many studies have considered carbon storage and transport in boreal ecosystems, invoking deeper thaw as the cause of trends in stream chemistry (Petrone et al. 2007; Prokushkin et al. 2005; Striegl et al. 2005; Walvoord and Striegl 2007). While these studies seem to implicate a deepening active layer or thawing permafrost as the cause of biogeochemical trends, few have specifically examined the hydrologic flowpaths that move water and solutes from catchments into streams. Physically based investigations are critical in order to predict how the disproportionately large effect of climate warming on high-latitude catchments will affect water and heat flow, and subsequently catchment geomorphology and carbon cycling.

High latitude and high elevation catchments commonly contain frozen soils with a shallow, unfrozen “active” layer that expands downward as the summer progresses. Shallow thaw depths in the early summer promote flooding and flashy hydrographs (Roulet and Woo 1986). Subsurface storage capacity and deeper flowpaths may develop during the summer due to increased thawing of soils (Prokushkin et al. 2005) and near-stream environments (Brosten et al. 2006; Zarnetske et al. 2007). This seasonal deepening of the active layer may promote storage and subsequent mineralization of carbon that is leached into the deeper layers from the upper organic horizons (Ågren et al. 2007; Carey 2003; Petrone et al. 2006), thereby leading to a positive feedback of greenhouse gas production and subsequently greater inorganic carbon release to surface waters (Striegl et al. 2005). While deep soils may support carbon storage and mineralization, much of the transport and biogeochemistry in upland boreal systems occurs in shallow organic soil horizons because of substantially higher hydraulic conductivity in the overlying peaty material (Hinzman et al. 1991). The hydraulic conductivity of the organic soil horizons decreases with depth, leading to higher runoff potential in the shallowest horizons (Quinton and Marsh 1999). An abrupt change from organic to mineral soil promotes lateral flow at this boundary, and may preclude significant infiltration into the deeper mineral soil (Carey and Woo 2001; Hinzman et al. 1991; Quinton and Marsh 1999).

Preferential flow may provide an important mechanism of water and solute transport through deeper soils where hydraulic conductivity is otherwise limited. Soil pipes are an example of a preferential flow mechanism known to be important in many high-latitude catchments. Soil pipes are large conical voids aligned parallel to the slope in the subsurface that allow flow to bypass the soil matrix (Beven and Germann 1982; Jones 2010). Soil pipes exist in both temperate and polar environments, and may originate in macropores caused by animal burrows and/or root decay following tree death or fire. Pipes are formed when macropores erode due to positive pore pressures that entrain or erode mineral grains (Carey and Woo 2000). In cold regions, pipes occur predominantly on steeper slopes (Quinton and Marsh 1998), when soils are saturated (Carey and Woo 2002; Dingman 1971), upon thaw of ice-rich sediments, and often at the organic/mineral soil boundary (Carey and Woo 2002). Soil pipes may explain rapid hyporheic transport beneath Antarctic streams (Cozzetto 2009; Koch et al. 2011).

An integrated view of catchment runoff processes can be gleaned using multiple physical and chemical methods. Seasonal changes in the ratio of precipitation to discharge and the time scale over which flood hydrographs recess may provide evidence of soil thaw and development of soil pipes upon catchment drainage. Previously, soil thaw has been invoked to explain decreased runoff ratios (Wang et al. 2009) and longer hydrograph recessions (Jones and Rinehart 2010; Lyon et al. 2009). Changes in the magnitude and location of inflows can be observed using stream tracer dilution (Kilpatrick and Cobb 1985), and may provide evidence of shifts in the dominant runoff processes. Tracers can also be used to calculate inflow chemistry, which provides information on the interactions between water and catchment soils.

Under favorable circumstances, uranium (U) concentrations and isotope ratios can be a useful tool for constraining the dominant runoff mechanism through mineral soils, because U activity ratios (UARs, 234U/238U) in water may be used to indicate water/sediment contact time over timescales greater than 1,000 years (Kigoshi 1971). Variation in aqueous UARs results from alpha decay of 238U, which may increase 234U/238U activity ratios (UARs) in the surrounding porewater by direct ejection of the daughter product (DePaolo et al. 2006) or increased leaching of the daughter at mineral surfaces due to recoil damage (Andersen et al. 2009; Fleischer 1980). Uranium isotope analysis is particularly well suited for studies in the silt-loess of interior Alaska, because activity ratios are dependent on mineral surface area and hence sediment size and shape, which are predictable and relatively consistent in the silt loess deposits of Alaska (Ewing et al. 2010; Shur et al. 2010). Recent work has used UARs in large arctic rivers to detect hydrologic variation in sub- and supra- permafrost water sources (Kraemer and Brabets 2012). Permafrost U concentrations vary greatly, likely related to dissolution variability that depends on permafrost water content and organic matter concentrations. UARs generally increase with depth in permafrost (Ewing et al. 2010; S. Ewing, Montana State University, personal communication, 2012).

It was hypothesized that in catchments underlain by frozen silt-loess, the dominant mechanism of runoff shifts from shallow throughflow to deeper preferential flow as the active layer deepens seasonally. Soil-moisture data and infiltration modeling results are used to consider mineral-soil-matrix flow and transport. Runoff ratio and stream-hydrograph analysis are used to quantify the seasonal trends in runoff. By quantifying stream inflows and chemistry, temporal and spatial patterns in inflow magnitude and U signatures are identified. The present study improves our understanding of the catchment processes that contribute to the U signature observed at the integrated, river scale. These results provide improved understanding of silt-upland-catchment hydrology relevant to predicting carbon transport and reactivity and landscape evolution as upland catchments in the Arctic and sub-Arctic evolve in response to climate warming.

Site description

The Richardson Catchment is located in the Hess Creek watershed, which is approximately 150 km north of Fairbanks, Alaska, USA (Fig. 1). This catchment is drained by Richardson Tributary, which appears as an unnamed stream on the US Geological Survey (USGS) 1: 63,360-scale quadrangle. This stream is approximately 10 km long and flows into Richardson Creek, then Hess Creek, and eventually to the Yukon River. This catchment burned in 2003 and previously, which possibly lead to the 2nd-order stream incising up to 10 m into the valley fill. These investigations focused on a 1.2-km reach of Richardson Tributary that flows along the base of a steep, north-facing hill, which is generally representative of open black spruce Picea mariana (Mill.) ecosystems in the discontinuous permafrost zone of Alaska (Kane et al. 2005).
https://static-content.springer.com/image/art%3A10.1007%2Fs10040-012-0934-3/MediaObjects/10040_2012_934_Fig1_HTML.gif
Fig. 1

a Map of Alaska with the study area indicated by a triangle. b The study region, with a star indicating the precipitation gage near the junction of the road (line trending north–south) and Hess Creek, (meandering river trending east–west). The bold black polygon represents the drainage area above the lower gage, and the rectangle indicates the study hillslope. Contour intervals are 100 ft. c Map of the study hillslope delineated in part b with 5-m contours

Vegetation at this site is described by O’Donnell et al. 2011 according to the Alaska Vegetation Classification System (Viereck et al. 1992). In the recently burned black spruce stands, vegetation is dominated by standing dead Picea mariana, and living Vaccinium vitis-idaea, V. uliginosum, Ledum groenlandicum, and Equisetum spp. Burned organic soil surfaces are colonized by Ceratodon purpureus. Mature black spruce stands still exist in some unburned areas immediately adjacent to the stream channel. Feather mosses (Pleurozium schreberi and Hylocomium splendens), sphagnum (Sphagnum fuscum), and reindeer lichens (Cladonia stellaris and C. arbuscula) dominate ground cover.

Throughout the catchment, surface organic soil horizons are approximately 10 cm thick (varying up to 25 cm in swales), below which the soils are homogeneous silt-loess. These soils are seldom saturated, which is common in boreal soils due to high evapotranspiration potential (Bolton et al. 2004), related to low relative humidity and high productivity in the organic soils and mosses. Organics grade from live vegetation at the surface to decaying plant material and thicker peat at the organic-mineral boundary. Soils are commonly frozen to the surface in May, and thaw to approximately 70 cm by early September. Below the active layer is ice-rich wedge ice (yedoma) that is well documented in this area and extends to depths of 26 m or more due to formation with syngenetic permafrost (French and Shur 2010; Kanevskiy et al. 2012). Soil pits were dug in multiple locations on the hilltop, toe slope, and riparian area of the north-facing catchment and in the abandoned floodplain of the south-facing catchment during the summers of 2007 through 2009. During wet periods, north-facing catchment soil pits often filled with water flowing at the organic/mineral soil boundary, or from discrete mineral-pit-wall failures and piping of water into the hole. Soil pipe outlets were observed at the base of the steepest hillslope (just upstream of tributary T4 in Fig. 1), and were characterized by artesian conditions, and upwelling of silty water from discrete holes.

Several tributaries flow into the stream from the north-facing slope. Two inflows (T4 and T6) persisted throughout the summer, and have eroded the organic soils and flow on top of the mineral soil. Three smaller tributaries (T1, T2, and T5) have greater contact with organic soils and were observed flowing only after a large storm period beginning June 27, 2008. Dissolved organic matter concentrations are high in pore and stream water relative to many streams, a common occurrence in boreal systems related to the substantial organic matter cover. Stream and soil waters are generally oxic, consistent with shallow flow above the permafrost. There is evidence of small landslides and slumps along steeper slopes at the toe of the hillside near T3 and T4. Soil pits in these areas display a cohesionless thixotropic silt, that easily collapses upon disturbance. Tracer addition and synoptic sampling results indicate that surface and subsurface inflows are well mixed in the middle of the summer, but that at the end of the season as much as 22 % of the inflowing water may contain a significantly elevated solute load, indicating substantial inflow subject to mineral soil contact (unpublished data).

Several features indicate that inflows to the stream are dominated by drainage of the steeper north-facing hillslopes, and that the floodplain to the north does not contribute a significant amount of water. Dry soils, vegetation changes, and thermokarst features on the northern bank suggest that the floodplain was ‘abandoned’, related to decreased drainage and hydrologic connectivity with the stream following the 2003 fire. One small tributary, T3, flows regularly from the north side. This small tributary flows through a large thermokarst feature, which is characterized by reworked silts. Early in the season, the stream remains on the surface, while later in the summer this tributary is a disappearing stream, flowing intermittently in the surface channel and along subsurface pathways. Given that the watershed is dominated by the north-facing hillslopes and because the south-facing floodplain is very dry, it is assumed that the contribution of the south-facing catchment inflows to Richardson Tributary streamflow is minimal. T3 does, however display elevated solute and U concentrations that may be useful to interpreting U signatures from elsewhere in the catchment.

Methods

Field measurements

Precipitation was measured at the Hess Creek Gage, operated by the USGS and located several kilometers from the catchment. Stream stage was monitored at two locations in Richardson Tributary and in three persistent inflows (T3, T4, and T6) during the summers of 2008 and 2009 using pressure transducers placed in stilling wells in or above flumes. Pressure was logged every 15 min, corrected for barometric pressure, and field calibrated by independently reading flume staff plates during site visits. Discharge was calculated using flume rating curves and pressure logs. Discharge often exceeded 67.4 L s−1, which is the maximum flume capacity. During these high flows, discharge was modeled using Manning’s equation and verified with wading rod and pygmy measurements

Soil moisture and temperature were measured in three locations in the watershed in 2008 (S1–S3 in Fig. 1) and four locations in 2009 (S4 added). In 2008, loggers were installed at the end of June, once the active layers had deepened. The three hillslope locations (S1–S3) were chosen to follow a surface depression from the hill into T4 to track water accumulation in the soils of the depression. Soil moisture was measured with ECHO EC-5 probes (Decagon Devices, Inc.) and temperature was measured with a 12-bit temperature Smart Sensor (Onset Computer Corporation). Comparison of sensor and gravimetric moisture readings suggests that sensor errors were on the order of 0.03 m3 m−3 in the mineral soils. Measurements were collected from paired, shallow and deep locations (Table 1) to investigate infiltration characteristics. Data were logged with an Onset HOBO Micro Station. Soil moisture was converted to capillary pressure using class average van Genuchten 1980 parameters based on United States Department of Agriculture particle-size classifications (Carsel and Parrish 1988), and converted to hydraulic conductivity assuming relationships reported by Mualem 1976. Soil temperature records were used primarily to validate apparent perturbations of the soil moisture, which occurred due to precipitation or thawing.
Table 1

Depth (cm) of soil-moisture sensors at four locations delineated in Fig. 1

 

Soil-moisture sensors

 

S1

S2

S3

S4

2008

    

 Shallow

20

5

20

 Deep

70

28

55

2009

    

 Shallow

20

40

20

10

 Deep

70

79

55

25

Hydrologic calculations

The Green-Ampt model was used to simulate infiltration into the mineral soil on the catchment hillslopes during storms in 2008 and 2009 in order to demonstrate that precipitation is unlikely to reach the deep soils, and so the thin, often burned organic layer was disregarded. Because the organic layer is likely to absorb and transport some water, the results therefore overestimate infiltration into the silt. This model uses an approximation of Darcy’s Law and a continuity equation to estimate infiltration into a soil. Cumulative infiltration (F) and infiltration rates (f) were calculated as:
$$ F=Kt+\psi \Delta \theta \ln \left( {1+\frac{F(t) }{{\psi \Delta \theta }}} \right),\;\mathrm{and} $$
(1)
$$ f(t)=K\left( {\frac{{\psi \Delta \theta }}{F(t) }+1} \right), $$
(2)
where K is hydraulic conductivity [m s−1], t is time [s], ψ is the wetting front pressure head [m], ∆θ is the difference between measured and saturated-water content. Saturated-water content was estimated as 0.45 [m3 m−3] based on soil-moisture sensor observations. Simulations were performed at half-hour timesteps. The appropriate hydraulic conductivity for Green-Ampt modeling is uncertain (Chow et al. 1988; Dingman 1994; Risse et al. 1994), and for our simulations was approximated as silt-saturated hydraulic conductivity (1.65 cm h−1), consistent with boreal soil simulations by Dingman 1994. This hydraulic conductivity value is within the range of measurements for silt soils reported in Freeze and Cherry 1979 and is consistent with permeameter measurements on this hillslope (unpublished data). Moisture content was not logged in early June 2008 and so values were based on inter-annual trends and early 2009 moisture-content data. Because this analysis does not account for water lost to the organic horizons above the silt, the model may over-predict the amount and rate of water infiltrating the mineral soil. Saturation of desiccated boreal organic soils can absorb as much as 1.5 cm of the incoming precipitation (Kane et al. 1989), which would decrease the cumulative infiltration, depending on the antecedent moisture conditions. Therefore, our simulations represent the greatest potential amount of infiltration into a silt-loess.
Trends in runoff were quantified using stream discharge and precipitation data. Runoff lags were defined as the time between maximum precipitation rate and peak discharge at the downstream flume and were calculated for 2008 and 2009. Runoff ratios (RR) were calculated for each major storm, following:
$$ RR=\frac{{\int {{Q_{{\mathrm{stor}{{\mathrm{m}} \, \, {dt }}}}}} }}{{A{P_{\mathrm{tot}}}}}, $$
(3)
where ∫Qstorm is the total volume of water that entered the stream between the upstream and downstream flumes during the flood recession [m3], Ptot is the total precipitation [m] measured at the Hess Creek Gaging Station, and A is the drainage area [m2] between the two stream gages. Runoff ratios were only calculated when Ptot was greater than 0.1 cm.
Discharge recession analysis was conducted by identifying storm responses in the 2008 and 2009 hydrographs, and recession coefficients were calculated assuming first-order decrease of flood waves:
$$ {Q_{\mathrm{b}}}={Q_{\mathrm{f}}}{e^{{-\lambda t}}}, $$
(4)
where Qb and Qf are the discharge as the stream returns to baseflow and at the peak of the floodwave [L s−1], respectively, e is the exponential function, λ is the recession coefficient [s−1], and t is the time between the flood peak and the return to baseflow [s]. Hydrograph recessions were usually truncated by a subsequent storm event, and one event on June 14, 2009 was excluded from the analysis because it was a significant outlier and may have been affected by antecedent conditions from a proximal storm.

Tracer dilution

Continuous sodium bromide (NaBr) tracer additions were added to Richardson Tributary on June 30 and September 1, 2008. Addition rates were maintained using a CR10 datalogger and two flow-rate sensing FMI pumps. Bromide samples were collected at multiple locations along the 1.2-km-long reach in June and September 2008. Samples were filtered with a Gelman capsule filter, stored in 60-ml Nalgene bottles and chilled. Bromide concentrations were analyzed on a Dionex ion chromatograph at the USGS laboratory in Boulder, Colorado. Discharge at each sample location was calculated as:
$$ {Q_{\mathrm{j}}}=\frac{{{C_{\mathrm{i}}}{Q_{\mathrm{i}}}}}{{{C_{\mathrm{j}}}}}, $$
(5)
where Q represents the flow rate, C represents the bromide concentration [μM], ‘i’ is the injection location, and ‘j’ is the location of interest. Qi and Ci are known given the pumping rate and injectate concentration, and Cj was measured in collected samples. Partial pump failure in June and low flows in September resulted in many instances where collected samples were not well mixed, leading to outlying or impossible (negative) discharge calculations. Discharge outliers that deviated significantly from the trend, corroborated by the flume records, of increasing stream discharge with distance were excluded from the analysis. Diffuse inflows represent a combination of subsurface and small surface inflows and were quantified as the difference in discharge for each stream segment minus any gaged surface tributaries (T3, T4, and T6). Diffuse inflows were compared to catchment parameters including ‘drainage area’ and ‘proximity to a tributary channel’ in order to compare the relative importance of the wetter parafluvial zones to catchment runoff. Drainage areas for a given inflow were determined in ArcMap, where the drainage extent was defined as the upslope area that drained into the stream segment. Distance from a tributary is defined as the stream length between the reach midpoint and the nearest of the five tributaries draining the north-facing slope.
Uranium samples were collected from surface waters in 2008 and 2009. In 2008, nine samples were collected from the stream and three samples from tributary outlets at T3, T4, and T6 during the late summer tracer experiment. In 2009, additional samples were collected on two dates from T3 (August and September). At T4, an upstream location and neighboring soil pit (at S3) were sampled during the tracer experiment in September 2008 and in September 2009. All samples for U isotope analysis were stored in 1-L acid-rinsed bottles, acidified with ultra-pure nitric acid, spiked with 236U, dried down and dissolved completely using HF and HNO3, purified using standard ion exchange chemistry, and analyzed on a Triton solid-source mass spectrometer at the USGS laboratory in Lakewood, Colorado. Uncertainty in U concentrations and activity ratios are 0.002 ppb and 0.004 (dimensionless), respectively. Diffuse U inflow concentrations and UARs (Cdiff) were calculated based on bromide tracer dilution gaging, tributary gage discharge, and stream U concentrations, following:
$$ {C_{\mathrm{diff}}}=\frac{{{Q_{\mathrm{j}}}{C_{\mathrm{j}}}-{Q_{\mathrm{i}}}{C_{\mathrm{i}}}-{Q_{\mathrm{trib}}}{C_{\mathrm{trib}}}}}{{{Q_{\mathrm{j}}}-{Q_{\mathrm{i}}}}}, $$
(6)
where ‘i’ and ‘j’ represent the sampling locations at the top and bottom of a stream segment, respectively, and the subscript trib denotes a gaged tributary.

Results

Precipitation and discharge

Total precipitation values of 9.7 and 11.7 cm were measured at the Hess Creek gage for the summers (June 1 through September 1) of 2008 and 2009, respectively. The summer of 2008 was characterized by several large convective storms in late June/early July, and low precipitation after mid-August. Storms in 2009 were generally larger than in 2008 and more evenly distributed, except for a dry spell in late July. Daily mean stream discharge at the Richardson flumes varied substantially over the course of the summers and responded rapidly to precipitation (Fig. 2). Discharge decreased from high early season flows into the middle of the summer. July discharge remained fairly constant in 2008 and decreased steadily in 2009. This difference was concurrent with the lack of precipitation in 2009. In general, stream discharge increased between the upstream and downstream flumes, indicating that the hillslope was continuously providing water to the stream. Discharge in the T4 and T6 tributaries followed similar trends to the Richardson Tributary.
https://static-content.springer.com/image/art%3A10.1007%2Fs10040-012-0934-3/MediaObjects/10040_2012_934_Fig2_HTML.gif
Fig. 2

Precipitation from the Hess Creek Gage, located approximately 1.5 km from the center of the Richardson Catchment, and discharge records from the downstream Richardson Tributary gaging station during the summers of 2008 and 2009

Soil moisture and hydraulic conductivity

Mineral soils were generally unsaturated throughout the 2008 and 2009 summers, suggesting that matrix flow is not a significant runoff transmission pathway (Fig. 3). During both years, shallow soil moisture increased and approached saturation (approximately 0.40–0.45 m3 m−3) following early-summer storms, decreased from near saturation in June to 0.25–0.30 m3 m−3 in early August, and leveled or increased slightly towards the end of August. Deeper sensors were frozen for much of the summer. Thawing of the deep soil corresponds to a rapid increase in liquid water and soil moisture in the mid to late summer. Deeper sensors displayed significant noise and only minor responses to storms. Sensors at depths less than 30 cm deep generally displayed much greater variability than deeper sensors (Fig. 3b). Recorded soil moisture corresponded to hydraulic conductivities of 0.008 (highly unsaturated) to 1.65 cm hr−1 (assumed saturation) in shallow soils and a smaller range, from 0.04 to 0.4 cm hr−1, in deeper soils.
https://static-content.springer.com/image/art%3A10.1007%2Fs10040-012-0934-3/MediaObjects/10040_2012_934_Fig3_HTML.gif
Fig. 3

Soil moisture varies seasonally and due to precipitation, with a greater response in shallow soils. a Precipitation and soil moisture at S-1 in 2008 and S-3 in 2009. For a silt-loess soil, 0.40–0.45 represents saturation, which is only approached by the shallow soils following large, early-summer storms. The large increase in moisture content at 55 cm towards the end of the 2009 season results from thawing of initially frozen soils. b Mean summer moisture content in unfrozen soils. For all locations, moisture content is highly variable at shallow depths, while deeper soils are more constant, and often well below saturated conditions. Bars represent two standard deviations from the mean of each sensor for the thawed time period. The dashed line indicates the approximate boundary between the organic and mineral soils

Infiltration simulations

Mean Green-Ampt cumulative infiltration depths for the simulated storms (Table 2) were 7.1 and 6.9 cm in 2008 and 2009, respectively. Infiltration depths were inversely correlated to initial moisture content (R2 = 0.91, p < 0.00005, n = 19, Fig. 4a), consistent with other observations from fine-grained Alaskan soils (Hinzman et al. 1991; Kane and Stein 1983). Cumulative infiltration depths tended to be deeper later in the summer when soils were drier, and never exceeded 12 cm into the silt (22 cm into the soil column). Because this analysis does not consider the organic horizons above the silt, these simulations represent the greatest potential amount of infiltration into a silt-loess and indicate that precipitation is not likely to move through the silt matrix to greater depths. The modeled infiltration depth is consistent with the observed depth of moisture content variability in Fig. 3b.
Table 2

Green Ampt infiltration modeling results for major storms

Year

Size rank

Date

Initial moisture

Duration

Precipitation

Cumulative infiltration

Total

Max rate

(m3 m−3)

(hr)

(cm)

(cm hr−1)

(cm)

2008

8

6/4

0.35

1.5

0.20

0.20

4.3

11

6/9

0.35

4.5

0.05

0.25

4.5

10

6/11

0.35

1.0

0.08

0.10

4.2

9

6/22

0.35

1.5

0.13

0.20

4.3

4

6/27

0.35

1.5

0.70

1.3

4.9

2

6/29

0.42

12

1.9

0.61

3.5

3

7/1

0.42

7

1.0

0.25

2.6

1

7/8

0.32

1.5

2.2

2.6

7.7

7

7/21

0.26

2

0.13

0.10

9.5

6

7/29

0.24

3.5

0.41

0.15

12

5

8/10

0.27

13

0.69

0.30

9.4

2009

3

6/6

0.38

5.0

1.2

0.46

4.1

10

6/13

0.38

1.5

0.33

0.36

3.2

2

6/21

0.38

4.0

1.2

0.97

4.2

6

6/28

0.38

1.5

0.76

1.3

3.7

8

7/9

0.36

2.5

0.41

0.20

3.7

4

8/13

0.3

18

1.1

0.41

7.6

1

8/16

0.3

10

1.2

0.51

7.6

11

8/23

0.31

4.5

0.33

0.25

7.6

9

8/24

0.31

6.5

0.41

0.20

6.4

12

8/25

0.31

5.0

0.30

0.10

6.3

5

8/29

0.31

12

0.84

0.15

6.9

7

8/31

0.32

8.0

0.48

0.15

6.1

https://static-content.springer.com/image/art%3A10.1007%2Fs10040-012-0934-3/MediaObjects/10040_2012_934_Fig4_HTML.gif
Fig. 4

Moisture content is a dominant control on whether precipitation is stored or exported in this silt-loess catchment. a Soil moisture versus infiltration. Low moisture content leads to greater infiltration of precipitation in Green-Ampt simulations (p < 0.00005). b Soil moisture versus runoff ratio. Similarly, runoff is greater when moisture content is high (p < 0.005), although variability from the trend indicates the importance of other factors

Runoff analysis

Runoff timing and magnitude varied greatly seasonally and between years. Runoff lags increased over the 2008 summer, and varied greatly in 2009. The overall mean for 2008 and 2009 was 18.8 (±8.9) hr. Runoff ratios varied from 0.03 to 1.59, indicating a very large range in the stream response to storms. Runoff ratios greater than one often occurred during large convective storms. Runoff ratios displayed significant positive correlation with soil moisture (R2 = 0.46, p < 0.005, n = 20; Fig. 4b), which is consistent with the infiltration modeling results. Generally, higher runoff ratios occurred in the early and late summer, and the lowest values were in mid-July to mid-August. There was a distinct decrease in the magnitude of runoff ratios during the dry mid-summer period, but no obvious change in runoff lags across the season (Fig. 5).
https://static-content.springer.com/image/art%3A10.1007%2Fs10040-012-0934-3/MediaObjects/10040_2012_934_Fig5_HTML.gif
Fig. 5

Runoff trends in early, mid, and late summer. a Runoff lags, which increase over the summer in 2008 and not in 2009. b Runoff ratios, which are smaller during the dry, mid-summer time period

Discharge recession coefficients

Discharge recession coefficients were calculated for eleven of the larger storms that displayed acceptable (R2 > 0.80) exponential regressions. Regression coefficients varied from 0.0184 to 0.0896 hr−1, indicating a five-fold range in the rate at which water can run off the catchment and enter the stream. Recession coefficients were inversely correlated to the date (Fig. 6; R2 = 0.84, p < 0.0005, n = 10).
https://static-content.springer.com/image/art%3A10.1007%2Fs10040-012-0934-3/MediaObjects/10040_2012_934_Fig6_HTML.gif
Fig. 6

Discharge recession over the summer. Recession coefficients decrease significantly (p < 0.0005) over the period of observation, indicating that runoff exits the catchment more slowly in the late summer. One outlier was omitted from this analysis, and may have been affected by two proximal storms

Tracer dilution

Tracer dilution gaging indicated a significant difference in the location and magnitude of inflows to Richardson Tributary between the wet early-season period and the dry late-season period of 2008 (Fig. 7). Inflows within the study reach accounted for 18 % in June and 79 % in September of the total discharge measured at the downstream end of the reach. Pump failure in June and incomplete mixing of the tracer in September resulted in inability to calculate inflows between each and every sampling location. Eleven and nine measurements were used for discharge calculation in June and September, respectively. Samples that deviated significantly from the dominant increasing discharge trend or that conflicted with flume records or observations were excluded. During both periods, inflows occurred mainly in reach 5 and beyond. Diffuse inflows were larger than tributary inflows, accounting for 61 and 79 % of the total stream discharge increase in June and September, respectively. There was also a seasonal shift in the spatial distribution of diffuse inflows (Fig. 7). In June, diffuse inflows correlated to catchment area, although the trend is only significant at the 90 % level (R2 = 0.63, p = 0.058; Fig. 8a). In September, inflow magnitude displayed a significant power-law relationship to the proximity to tributary channels (R2 = 0.72, p < 0.001; Fig. 8b).
https://static-content.springer.com/image/art%3A10.1007%2Fs10040-012-0934-3/MediaObjects/10040_2012_934_Fig7_HTML.gif
Fig. 7

Stream discharge and inflows along the study reach. Discharge increases along the study reach as a result of catchment runoff. Dashed lines indicate the locations of the surface-water tributaries (T1 to T6), whose input is insignificant relative to diffuse inflows. Numbers (15) and line segments indicate the reaches specified in Fig. 1. Stream discharge is much lower later in the season and there is a shift in the location and magnitude of diffuse inflows between a June and b September 2008

https://static-content.springer.com/image/art%3A10.1007%2Fs10040-012-0934-3/MediaObjects/10040_2012_934_Fig8_HTML.gif
Fig. 8

Regressions of June and September 2008 diffuse inflows with topographic variables, indicating a seasonal shift in how water reaches the stream. Lines and R2 values indicate significant trends. a June diffuse inflows are controlled by catchment area and more evenly distributed (p < 0.1), while September inflows (b) are greatest proximal to surface-water tributaries (power law transform, p < 0.001)

Uranium concentrations and activity ratios

Variability in U concentrations and UARs provides a means of constraining the flowpaths of diffuse and tributary inflows. Stream U concentrations and activity ratios varied only slightly from 0 to 932 m in the study reach (range of 0.033 ppb and 0.757, respectively) before dropping sharply at 1,015 m. Generally, tributary inflows had higher U and lower UARs than the mainstem. Therefore, mass balances (Eq. 6) indicated that diffuse inflows must have had U concentrations that are similar or lower than the mainstem, and UARs that are lower in reaches 1 and 3 and higher in reaches 2 and 5 (Table 3 and Fig. 9). Uranium ARs did not change significantly from the top to bottom of reach 4, implying that the UAR of diffuse inflows was similar to that of the stream. Based on the calculated UARs for reaches 2 and 5 (1.376 and 1.434, respectively) and permafrost core data (S. Ewing, Montana State University, personal communication, 2011), diffuse inflows in those locations may be accessing water that attained its UAR signature from as deep as 5–15 m below the soil/permafrost boundary.
Table 3

Discharge, U concentrations and activity ratios (UAR) from the late-season synoptic sampling of the stream and tributaries. Numbers following stream sample types indicate samples taken at reach endpoints. Diffuse inflow values are calculated based on the stream endpoint and tributary discharge and chemistry. Dashes indicate locations where tracer dilution provided unrealistic discharge values, likely because of incomplete mixing in the stream

Sample type

Site/distance

Stream discharge

Tributary inflow

Uranium

UAR

Reach

Diffuse inflows

Discharge

Uranium

UAR

(m)

(L s−1)

(L s−1)

(ppb)

(234U 238U−1)

(L s−1)

(ppb)

(234U 238U−1)

Stream: 1

0

1.96

 

1.2236

1.332

1

0.97

1.241

1.298

Stream

20

 

1.2191

1.324

    

Stream

73

 

1.2393

1.321

    

Stream: 1–2

173

2.93

 

1.2295

1.321

2

0.70

1.214

1.376

Tributary

190

2.96

0.03

1.2690

1.289

    

Stream: 2–3

372

3.66

 

1.2269

1.331

3

1.37

1.156

1.250

Stream

493

 

1.2256

1.312

    

Stream: 3–4

726

5.03

 

1.2076

1.310

4

0.19

0.736

1.471

Stream: 4–5

829

5.22

 

1.1904

1.313

5

1.39

0.957

1.434

Tributary

855

5.27

0.05

1.9044

1.301

    

Stream: 5

932

6.66

 

1.1472

1.334

    

Tributary

1023

0.14

1.3498

1.223

    
https://static-content.springer.com/image/art%3A10.1007%2Fs10040-012-0934-3/MediaObjects/10040_2012_934_Fig9_HTML.gif
Fig. 9

a Uranium concentrations and b activity ratios in the Richardson (streamflow), tributaries, and calculated diffuse inflows. An activity ratio was not calculated for reach 4 because analytical uncertainty was greater than the difference in reach endmembers

Figure 10 indicates the U and UAR variability between different types of samples and suggests how low U and UAR precipitation may evolve as water mixes with mineral soils and frozen ground while travelling downslope towards the mainstem stream. In the T4 drainage, porewaters at S3 displayed the lowest measured U concentrations and UARs, with values suggesting dilute precipitation that has interacted only slightly with mineral soils. Tributary water proximal to S3 displayed slightly higher U and UARs, and streamwater at the outlet of the tributary displayed the highest measured U concentrations and UARs similar to the mainstem. T3 displayed a large range in U and UAR, potentially related to variable contact with deeper mineral soils depending on seasonal thaw and its tendency to ‘disappear’ into the subsurface later in the summer.
https://static-content.springer.com/image/art%3A10.1007%2Fs10040-012-0934-3/MediaObjects/10040_2012_934_Fig10_HTML.gif
Fig. 10

Uranium concentrations and 234U/238U activity ratios in the stream, tributaries, diffuse inflows, and porewaters, indicating contact between water and mineral soils. Bars represent two standard errors. Stream bars represent nine samples collected along the study reach during the August, 2008 tracer. T3 bars indicate variability between three samplings in August, 2008 and 2009 and September, 2009. Filled markers indicate U chemistry at the outlet of the tributaries. The dashed line represents the increasing U and UAR in surface water flowing from the upper hillslopes of T4 to the outlet. The upstream T4 site is several meters from site S3, where porewater samples were collected. Both upstream T4 sites were sampled in September of 2008 and 2009. Diffuse inflows were calculated based on tracer dilution during the August 2008 experiment and are labeled corresponding to the reaches delineated in Fig. 1. Reach 4 is not included because analytical uncertainty was greater than the difference in reach endmembers. ab present similar data, but b is plotted in inverse U space so that mixing trends are linear. Bold lines labeled with letters represent potential mixing lines between a dilute precipitation endmember with shallow mineral soils (low UAR, high U, line A) and permafrost (high UAR, high U, line B). Different proportions of these lines can mix (line C) to produce the observed stream water

Discussion

Our results support the hypothesis that flowpaths through boreal catchments underlain by frozen silt evolve seasonally, with increasing potential for rapid subsurface flow later in the summer. Despite substantial active layer thawing, water continued to reach the stream quickly (Figs. 5 and 6), implicating preferential flow and not porous media flow through the silt as the primary transport mechanism.

Seasonal variability in runoff

The seasonal shift in the location and magnitude of diffuse inflows indicates a shift in the primary runoff mechanism, from rapid flow through shallow soils in the early season to deeper, preferential flow later in the season. Similar inflow variations in a temperate catchment in Montana have been attributed to a switch from topographic controls during wet periods to geologic controls in drier times (Jencso et al. 2009; Payn et al. 2009). Similarly, inflow magnitude in the Richardson Catchment correlates to drainage area during the wet June sampling and proximity to tributaries in the dry September sampling (Fig. 8). Focused late-season upwelling of deep, sub-permafrost groundwater could explain this trend, but is considered unlikely, given evidence that permafrost acts as an aquitard between supra- and sub-permafrost flow. This lack of deep, sub-permafrost groundwater inflow is supported by (1) the inability of the stream to maintain baseflow without constant precipitation inputs, (2) aerobic conditions and relatively dilute solute chemistry in all catchment waters, and (3) the presence of shallow subsurface ice in every soil pit and frost probe measurement. Supra-permafrost preferential flow is therefore the likely explanation for this late-season flow.

Potential mechanisms of preferential flow

Tracer dilution gaging provides evidence that subsurface preferential flow in the Richardson Catchment occurs in conditions similar to those that promote the shallow pipe flow described by Carey and Woo 2000. Theoretically, preferential flow is likely to occur when positive pore pressures are greater than lithostatic pressures, allowing erosion of silt grains and rapid flow of water through voids (Carey and Woo 2000). This may occur in the high inflow regions of this catchment for several reasons: (1) convergent topography may maintain positive pore pressures, allowing extensive preferential flow networks to develop as suggested by the correlation between diffuse inflow magnitude and proximity to the surface tributaries (Fig. 8); (2) the streams themselves may be important, providing erodible material via ice-rich saturated margins, as well as positive pressures and/or heat from hyporheic exchange to further promote incision and erosion of these margins; (3) steep slopes near T1, T2, and T4 may develop preferential flow networks by promoting high pore pressures and high velocities, similar to observations in previous studies (Quinton and Marsh 1998) and as suggested by the sharp increase in diffuse inflows near the steep, 30° slopes at 800-m stream distance (in Fig. 7b).

While pipes have been previously observed at the organic/mineral boundary, preferential flow in this catchment was found deeper in the mineral soils. Shallow pipes likely form at the organic/mineral soil boundary because of the relative impermeability of the underlying clayey mineral soils (Carey and Woo 2000). It is possible that a similar situation exists at depth, and that preferential flowpaths may be forming at the boundary between the thawed mineral soils and the impermeable ice. Fire is known to increase seasonal active layer depths (O’Donnell et al. 2011), and the recent fire in this catchment may have allowed subsurface water to begin accessing and eroding the transition zone, which is a layer of structurally weak, ice-rich soils at the boundary between the seasonal active layer and underlying permafrost (Bockheim and Hinkel 2005; Shur et al. 2005). Fire may also have created the original macropores (i.e. burned roots) allowing water to move vertically through the mineral soils that would otherwise exhibit low hydraulic conductivities and infiltration potential.

Chemical evolution of runoff and a link to erosion

Trends and variability in U concentrations and activity ratios between the different water sources in Richardson Catchment provide evidence that diffuse runoff and tributary water is in contact with thawing subsurface permafrost before it reaches the stream. Given the lack of connectivity with sub-permafrost groundwaters, the most likely explanation of high and increasing U concentrations and activity ratios is mixing between a low, precipitation/runoff endmember (similar to S3 porewater) and waters with elevated UAR reflecting progressive thaw or cryoturbation along pathways defined by thawing ice-wedge networks. Ice wedges were directly observed in nearby cores (Kanevskiy et al. 2012; O’Donnell et al. 2011) and ongoing ice-wedge thaw is implied by extensive hummocky topography on surrounding hillslopes. While cores indicate contact between water and silt well below the permafrost boundary (5–15 m), processes such as hillslope slumping, cryoturbation, or stream incision may have decreased the overburden, allowing active layer water to contact these soils. Mixing between runoff and permafrost with high UARs could occur due to stream incision and diffuse runoff traveling through preferential pathways at the base of the active layer and along frozen boundaries. Elevated UARs at these boundaries could be maintained by cryoturbation, which inmixes high UAR silts from permafrost zones not previously exposed to thaw. Thus, preferential flowpaths may be modified annually, rather than runoff flowing along the same contact surfaces each year. Constant high UAR runoff may also indicate continual thermo-mechanical erosion by flowing water derived from precipitation. This erosion is likely the explanation for increasing UARs along the tributary flowpaths. Such incision is evident on the landscape and likely related to geomorphologic reorganization as the hillslopes slump and tributaries downcut towards the mainstem stream channel, which may itself be actively incising.

The data suggest variable U concentrations in the high UAR permafrost. High UARs are consistent with increasing UAR (and age) with depth in permafrost cores up to 20 m deep from this area (Ewing et al. 2010; S. Ewing, Montana State University, personal communication, 2011). Uranium variability may be related to surface and subsurface runoff contacting permafrost with depth-variable properties (i.e. water content, organic matter concentrations, etc.). Tributaries likely interact with the top of the permafrost low on the slope and very near the stream, coincident with the greatest downcutting, whereas diffuse inflows are likely contacting the top of the permafrost on the steeper sections of the catchment, which tend to be on the toe slopes and further from the stream. Different mixing lines between precipitation and high UAR diffuse inflows (reaches 2 and 5 in Fig. 10b) may indicate differences in the amount and duration of water/permafrost contact, which may be related to differences in topography in these reaches. For example, the higher UAR and lower U concentrations in reach 5 likely indicate rapid flow related to the steep slopes adjacent to the stream. Some inflows appear to have little contact with high U water, as evidenced by the decrease in U concentrations and UARs associated with large diffuse inflows near the bottom of the stream reach (near T6).

Evidence of preferential flow through silts

Discharge recession coefficients provide further evidence for preferential flow in the Richardson Catchment. Discharge recession coefficients are inversely correlated to date (Fig. 6), indicating slower drainage of the catchment later in the season, similar to what has been witnessed in other catchments with increasing thaw depth (Jones and Rinehart 2010; Lyon et al. 2009; McNamara et al. 1997; Wang et al. 2009). However, because calculated runoff responses are more rapid than can be reasonably expected from matrix flow through a silt loess, preferential flow is a likely mechanism explaining trends in runoff. A trend of seasonally increasing specific conductance in the surface waters of this catchment provides further evidence that late-summer runoff is in contact with the mineral soils.

Evidence for shallow storage and minimal matrix flow through the silt-loess soils

Soil-moisture trends (Fig. 3) and Green-Ampt modeling (Table 2) indicate that matrix flow through silt soils is not a significant pathway of hillslope runoff due to low hydraulic conductivity inhibiting both infiltration and subsurface matrix flow. Soil moisture varies minimally at depth and does not respond to storms (Fig. 3), indicating that a substantial portion of the storm water does not infiltrate to greater than approximately 25 cm. Minimal infiltration precludes significant flow in deeper soils due to the constantly low hydraulic conductivities (0.04–0.4 cm hr−1). Shallow mineral soils approach saturation in the early summer and following storms, but even at saturation, these soils exhibit low hydraulic conductivities (approximately 1.65 cm hr−1). These data indicate that regardless of thaw depth in the silt soil, the hydraulic conductivity is too low to explain observed short runoff lag times (Fig. 5a) and recession curves (Fig. 6). In fact, given that shallower soils are often drier than the deep soils, calculations using silt soil properties from Carsel and Parrish 1988 suggest that evapotranspiration creates an upward hydraulic head gradient that indicates movement of water towards the surface rather than down the hillslope.

Infiltration into the mineral soil matrix may significantly decrease and retard catchment runoff. Green-Ampt modeling indicates that the greatest infiltration occurs in dry soils (Fig. 4a), and runoff coefficients confirm that less water runs off dry soils (Fig. 4b). In the early and late summer, wet soils preclude significant subsurface storage resulting in high runoff ratios, while in the dry mid-season, soil storage potential increases (Fig. 5b).

Implications

These results present a view of boreal catchment runoff that focuses on the seasonal shift from shallow, topographically controlled runoff to deeper flow through preferential flowpaths that form in moist areas and on steep slopes. Given that late-season preferential flow is accessing and thawing the top of the permafrost, and that permafrost contains high concentration of ancient carbon, runoff may be delivering ancient carbon to ecosystems. Catchments underlain by frozen silt may be particularly important to climate change in the coming decades, given that silts are highly erodible, display significant thaw settlement because of their ice rich nature (Kanevskiy et al. 2012), contain vast stores of frozen carbon, and in interior Alaska are undergoing geomorphic reorganization related to an accelerating fire cycle and permafrost thaw.

Conclusion

The data from this study provide compelling evidence that drainage of silt-dominated permafrost-bound catchments of interior Alaska shifts from near-surface topographically controlled throughflow to deeper pipeflow later in the season. Moisture content and infiltration modeling indicate that silts are typically unsaturated, and in the absence of preferential flow, are incapable of rapidly conducting water. Seasonal trends in runoff coefficients and flood recessions indicate greater potential for storage, and decreased runoff velocities later in the summer. The change in location of runoff indicates that late summer flow is associated with steeper and wetter near-stream slopes, and is likely related to soil pipes that have been observed in these locations. Uranium isotopes indicate that preferential flowpaths are renewed seasonally, achieving rapid transport of water and solutes and thawing the top of the permafrost. The data and observations highlight steep slopes, the near-stream environment, and recently exposed silts along thawing margins as likely locations of preferential flow and erosion. The near-stream environment has been recognized as an important location of biogeochemical activity, and this work provides evidence that in high latitude regions, such areas may be of significance for catchment geomorphic evolution via erosion of permafrost as well.

Acknowledgements

This work was supported by a National Science Foundation grant, OCE-BE 0628348 to Q. Zhuang. We thank P. Schuster and C. Hart for laboratory analysis of the bromide tracer, and J. Paces for assisting with the U isotope analysis. B. Rajagopalan and K. Bencala provided invaluable discussions on the data and results. Field support was provided by K. Wickland, R. Runkel, K. Kelsey, M. Bourret, D. Halm, M. Dornblaser, G. Aiken, and R. Spencer. The manuscript was significantly improved through reviews by B. Ebel and two anonymous reviewers. Any opinions, findings, conclusions, or recommendations expressed in this material are those of the authors and do not necessarily represent the view of the National Science Foundation or US Geological Survey. Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the US Government.

Copyright information

© Springer-Verlag Berlin Heidelberg (outside the USA) 2012