Physical short-term changes after a tussock tundra fire, Seward Peninsula, Alaska



[1] The Kougarok area, situated on the central Seward Peninsula, Alaska, experienced a severe fire in August 2002. This may be the only tundra fire where high-quality prefire (1999–2002) and postfire (2003–2006) active layer and meteorology measurements have been collected in the same locations. After fire, near-surface soil showed increased moisture at the burned tussock site, remaining close to saturation throughout the thawed season 2003–2006. Despite wetter soil after the fire, freezing occurred earlier at the burned tussock site than at the control, indicating the importance of a reduced organic layer. Severe combustion of lichen and moss left 15–25 cm high tussocks, resulting in a doubling of the surface roughness coefficient. Average September temperature at the tussock site increased 2.3 ± 0.7°C throughout the 1 m soil profile, doubling the active layer depth, although this is due partly to favorable meteorological conditions. The shrubby control station experienced a mean annual temperature increase of 1.1 ± 0.3°C in the upper 0.5 m. A similar annual change was found at the burned tussock site. Cooler weather conditions in 2006 stagnated the soil-warming trend, which occurred after 2002. How the thermal and moisture regimes in tundra will be affected after fire is highly influenced by weather, fire severity, vegetation regrowth, prefire vegetation, and ground ice conditions.

1. Introduction

[2] The Arctic and sub-Arctic are particularly sensitive to climate change [Manabe et al., 1991; Houghton et al., 1996; Watson et al., 1998]. Increased fire frequency, severity, and area burned are among the anticipated effects of climate warming [Stocks et al., 1998; Rupp et al., 2000; Flannigan et al., 2001; McCoy and Burn, 2005]. The historical and potential future extent, effects, and feedbacks of fires in the boreal forest are better known [Kasischke and Turetsky, 2006; Lynch et al., 2002; Valeo et al., 2003] than in tundra regions [Chambers et al., 2005; Racine, 2004]. Dramatic changes may follow in the Arctic and sub-Arctic, where short-term thawing/freezing processes and longer-term permafrost dynamics shape the ecosystem through plant growth, evapotranspiration, infiltration, runoff, wetland dynamics [Hinzman et al., 2003], biogenic gas fluxes to the atmosphere [Waelbroeck and Monfray, 1997], and export of carbon and nutrients to rivers and seas [Vörösmarty et al., 2001].

[3] In Arctic regions, many physical and biological processes are confined to the active layer, the near-surface layer above permafrost that thaws each summer. Critical microclimatic variables interactively controlling the active layer include surface temperature, snow cover, the nature of the vegetation canopy, organic horizon thickness, soil moisture content, and ground ice conditions [Kane et al., 1992; Burn, 1997; Hinzman et al., 2003]. Previous studies have shown that short-term (<10 years) postfire effects on surface characteristics include (1) depletion of the insulating organic layer, increasing the near-surface soil thermal conductivity and heat flow into the ground [Yoshikawa et al., 2002]; (2) reduced albedo, increasing the energy absorption; and (3) increased insolation on the ground surface [Kasischke et al., 1995] and reduction in the surface-atmosphere coupling [Chambers et al., 2005] due to changes in the canopy structure. These fire effects are known to increase the active layer thickness [Heginbottom, 1976; Viereck, 1982; Mackay, 1995; Burn, 1997]. Continued degradation of permafrost over longer periods can result in drier near-surface soils [Swanson, 1996; Yoshikawa et al., 2002] due to increased storage capacity. Fire-induced changes can therefore have broad effects hydrological, biological, and morphological systems and land-atmosphere interactions in the Arctic.

[4] This study examines the short-term (4 years) postfire changes on the thermal and moisture regimes of the active layer. Unlike most previous studies, observations were made at fixed locations before and after the fire. Ground and surface temperatures, near-surface soil moisture, freezing and thawing degree-day sums, and seasonal n-factors were observed before and after the fire in 2002 at burned (K2) and control (K3) sites, to determine whether fire-induced effects, weather variations, or both explain postfire changes in the active layer. We also examine prefire and postfire radiation efficiency, albedo, surface roughness at K2, and broader scale (1 km2) thaw-depth observations. Information obtained about ground ice and snow conditions during the study period was limited, restricting the discussion of their contribution to observed changes.

2. Site Description

[5] The Kougarok area (also known as Quartz Creek), occupies an area of transition between the continuous and discontinuous permafrost zones [Brown and Péwé, 1973] in the central part of the Seward Peninsula, northwestern Alaska (Figure 1). Kougarok (65°25′N, 164°38′W), approximately 140 km north of Nome, has a continental climate and is locally underlain by thin (15–50 m) continuous permafrost (>90% areal cover). The rolling landscape supports low Arctic tundra dominated by sedge tussock communities, with shrubs in the valley bottoms. The Ruptic-Histic Aquiturbel soils are of eolian origin [Sturm et al., 2005], with thickness of a half-meter on the ridges to a few meters in the valleys. Frost boils, circular features resulting from freeze and thaw processes, exist in the area.

Figure 1.

(left) Location of Kougarok area (65°25′N, 164°38′W), central Seward Peninsula, northwestern Alaska. (center) Topographic map showing Mauze Gulch and Niagara Creek watershed, with the three meteorological stations, K1, K2, and K3, and the Circumpolar Active Layer Monitoring (CALM) grid.

[6] The Seward Peninsula has a relatively high fire frequency compared to other tundra regions [Racine et al., 1987] but longer return intervals than the boreal forest of Interior Alaska [Stocks et al., 2002]. The annual area burned in Seward Peninsula was episodic during 1954–2006 (Figure 2), consistent with previous studies of Alaska [Kasischke et al., 2002] but without the trends observed in the boreal regions of Alaska and Canada [Kasischke and Turetsky, 2006]. The largest fire year was 1977, both in number of fires larger than 2 ha (13) and area burned (471,578 ha) (

Figure 2.

Annual area burned at Seward Peninsula 1954 through 2006 based on fires larger than 2 km2 ( Although most of Seward Peninsula is represented by open tundra, some spruce forest exists in the east [Thayer-Snyder, 2002].

[7] The Kougarok area experienced three fires in the last 35 years: 1971, 1997, and 2002. Three meteorological stations (K1, K2, and K3) were installed in the area during 1999 under the Arctic Transitions in the Land-Atmosphere System (ATLAS) program [Sturm et al., 2005;] (Figure 3). Tables 1 and 2 contain background information about these sites. All three stations are located on land affected by the 1971 fire. The 1997 fire (311 km2, 22 July to 25 August) burned the location of the meteorological station K1 while the 2002 fire (87 km2, 4 August to 19 August) destroyed instrumentation at the K2 site. Instrumentation was replaced in October 2002. A principal component analysis of two Landsat images (band 3, 4, and 7) acquired in July 1997 and October 2002 shows the spatial distribution of the 1997 and 2002 burns (Figure 4). On the basis of the fire severity classification system of Viereck et al. [1979], the 2002 fire resulted in severe (soil organic material completely or nearly consumed down to mineral soil) to moderate burns (organic layer partially consumed) with a few unburned patches. Approximately 5–7 cm of the organic layer remained after the fire in August 2002, where the K2 thermistors and water content reflectometers were installed (Figure 5). The surface below the long- and short-wave radiation instrument experienced a partial burning, leaving unburned patches. Thermal and mechanical erosion initiated by the 2002 fire were observed 400 m to the east of K2 in the Niagara Creek streambed (Figure 5), resulting in a >3500 m3 gully by September 2006. The erosion exposed ice-rich permafrost and ice wedges, with active thermokarst processes extending several meters away from the bank. The ice wedges probably formed during the Holocene, although the ice-rich permafrost is of older age (V. E. Romanovsky, personal communication, 2006).

Figure 3.

(left) K2 station before fire, showing rain gauge, 10 m tower and radiation instruments in June 2001. (right) The 3 m high K3 station summer 2005 at Mauze Gulch watershed.

Figure 4.

A multispectral principal component analysis (PCA) based on Landsat (band 3, 4, and 7) of two overlaid images acquired in early July 1997 and October 2002. The 1997 and 2002 burns are shown in nonnatural colors. Also marked are the year of fire, the three meteorological stations (K1, K2, and K3), and the 1 km2 CALM grid. Darker lines across the area show ATV and truck trails.

Figure 5.

(left) K2 soil moisture sensors were re-installed in October 2002 at 5, 10, 15, 20, and 30 cm depth. The photo shows the fire severity through the reduced organic material between the tussocks. Right: Erosion of Niagara Creek with exposed ice-rich permafrost in September 2003. The gully has become wider and longer since then, reaching 275 m in length fall 2006. Substantial ground subsidence has occurred along the banks, resulting in more gentle sides. Note the recovery of the tussocks.

Table 1. Site Descriptions of Meteorological Stations K1, K2, and K3
Site DescriptionK1K2K3
Latitude65″ 26.42′65″ 25.70′65″ 27.58′
Longitude164″ 34.71′164″ 38.61′164″ 38.29′
Elevation290 m110 m260 m
SlopeLevel3 deg.6.5 deg.
LandformRidgeBroad, alluvial fanSlope
VegetationMoss, lichensEriphorum TussocksShrub
Parent materialLoessLoessLoess
Fires1971, 19971971, 20021971
WatershedNiagara Cr. (6.5 km2)No nameMauze Gulch (4.5 km2)
Table 2. Soil Classification Scheme of K2 and K3 Stations From Sturm et al. [2005] With Frozen Thermal Conductivities by Hinzman et al. [1991]
 HorizonDepth, cmClay<2 mm SiltSand>2 mmSaturation, % by wtBulk Density, g cm−3Soil ClassThermal Conductivity
k_thaw W/mKk_frozen W/mK
TussockOi/Oe0–4    518.30.08Organic0.6951
(K2)O/A4–14    336.80.24Organic0.6951
 Bg/Ajj14+21.657.620.8 35.71.52Fine1.5352.689
ShrubOi0–15    724.50.05Organic0.7451
(K3)O/A15–186434.41.6 157.80.58Organic0.6951
 BA25+21.639.638.8   Fine0.8432.919

[8] A Circumpolar Active Layer Monitoring (CALM) grid [Brown et al., 2000] was installed in 1999 adjacent to K3 in the Mauze Gulch watershed (Figures 1 and 4). The 1 km2 grid, with nodes at 100 m spacing, extends over the creek and includes both a south- and a north-facing slope. The 2002 fire burned 90 grid nodes, mainly represented by tussock communities. Most of the 31 control nodes (unburned in 2002) were in the moist valley bottom and vegetated by high shrubs (<2 m). A few unburned sites occupied areas of moss and lichen at the northwest corner of the grid, on a south-facing slope shoulder.

3. Methods

[9] Thaw-depth observations on the 1 km2 CALM site were made manually at the 121 nodes locations by inserting a metal probe to the depth of refusal during the end of September or early October. The CALM grid results were divided into two groups, burned and control.

[10] Meteorological and soil variables were collected from 1999 through 2006. Temperature, wind speed, precipitation, radiation, and moisture instruments are summarized in Table 3. Meteorological measurements were made every minute and averaged into hourly intervals at K1, K2, and K3. Soil temperature and moisture were measured every 5 min and recorded as a 3-hour mean at K2 and K3. Information was stored in a Campbell Scientific CR10X data logger. The depth of each ground sensor was defined as the distance from the surface at the time of observation. Therefore near-surface prefire and postfire data from K2 represent different soil material. Unfrozen soil moisture was estimated from volumetric water content (VWC) observations. Spring peak in VWC was assumed to represent saturated conditions (all micro and macro pore spaces filled with liquid water) and for winter conditions, organic and mineral soil was set to 8% and 10% saturation, respectively [Hinzman et al., 1991].

Table 3. Instrumentation Used at the Three Meteorological Stationsa
Meteorological VariableInstrument
  • a

    K3 station, 3 m tall, was equipped with soil water content reflectometers of model CS 615.

Air temperatureVaisala (HMP45C)
RainfallTexas Electronics TR-525M
Short-wave radiation/pyranometerCampbell Scientific Eppley PSP
Long-wave radiation/pyrgeometerCampbell Scientific Eppley PIR
Net radiationREBS Q-7.1 Net Radiometer (1.5 m)
Soil temperatureAlpha Thermistors 13A5001-C3
Soil moistureCampbell Scientific CS615 (pre-fire), Campbell Scientific CS616 (postfire)
Wind speedMet One 014A Anemometer (1 m) and R.M. Young Wind Monitors 05103 (3 and 10 m)

[11] Thermistors were calibrated each fall following the method described by Romanovsky and Osterkamp [1995] using the phase equilibrium temperature between 0°C and −0.1°C during freezing. Mean daily ground temperatures provided information about the beginning of thaw, freeze, freeze-up dates, and the length of thawing and freezing seasons. The beginnings of the thawing and freezing periods were defined by ground surface temperatures remaining consistently above and below 0°C, respectively. The freeze-up date for the 1 m profile represented the day when all temperatures began to decrease sharply after the “zero curtain” disappeared [Romanovsky and Osterkamp, 1995].

[12] The n-factor, the ratio of the accumulated seasonal ground surface thawing or freezing degree-day sums to the seasonal air freezing and thawing indexes, is a generalized representation of the various climatic and vegetative variables affecting the soil thermal regime. Originally developed for engineering purposes [Carlson, 1952], the n-factor has also been used to examine the surface energy balance in natural regimes [Klene et al., 2001; Karunaratne and Burn, 2004; Kade et al., 2006] and for active layer mapping [Shiklomanov and Nelson, 2002]. Because of continuous temperature records, the n-factor was approximated using measurements 5 cm below surface (K2) and directly under green vegetation (0 cm) at K3. The K2 n-factor calculations should therefore be seen as a modified parameter similar to that employed by Kade et al. [2006].

[13] Owing to the limited number of soil temperature measurements, permafrost table temperatures (Tps) were estimated following Romanovsky and Osterkamp [1995, equation (13)] where:

equation image

where Kt and Kf are the bulk thawed and frozen thermal conductivities, It and If are the ground surface thawing and freezing indices (degree-days), and P is the annual period (365 days). Information about active layer depth was interpolated from soil temperature measurements. The thermal offset, the difference between the mean annual permafrost and ground surface temperatures, was also calculated.

[14] End-of-winter snow observations were made in 1999 and 2000 between K2 and the Niagara Creek streambed. Snow depth was obtained by 50 probed measurements along a transect at 1 m intervals. Snow water equivalents were based on five samples using an Adirondack tube. Tussock and shrub tundra were observed separately.

[15] Net all-wave radiation (Q*) represents the amount of energy available at the surface, described during daytime as:

equation image

where K↓, K↑, L↓, and L↑ represent incident and reflected short-wave radiation, incoming long-wave radiation emitted by the atmosphere, and outgoing long-wave from the surface, respectively [Oke, 1987]. A slight negative feedback mechanism is present because a surface with low albedo absorbs energy well, increasing emission of long-wave radiation from the surface unless there is rapid heat dissipation [Oke, 1987]. An approach used by Chambers et al. [2005] was applied to reduce the influence of incoming short-wave radiation differences, energy budget components, and climatic conditions on the net all-wave radiation: (1) the net all-wave radiation was treated in terms of radiation efficiency (Re), which is net-radiation normalized to incoming short-wave radiation; (2) the period of interest was restricted to local noon ±2 hours; and (3) only data obtained during clear-sky conditions (>500 Wm−2) and not directly following rain were included. A wind speed cooling effect algorithm [Campbell Scientific Inc., 1996] was applied to the net-radiation observations, which were made early June to late September.

[16] The height and spacing of the microroughness and macroroughness elements of the surface, i.e., the surface roughness length (zo) was obtained under neutral conditions, defined as the Richardson number (Ri) lying between −0.01 and 0.01. The Ri relates the relative roles of free and forced convection and is positive for stable and negative during unstable atmospheric conditions [Oke, 1987]. Following Braun [1985], calculations of Ri and, roughness length (zo) were based on a wind speed and temperature profile between the surface and 10 m, calculated as:

equation image
equation image

where zo is in meters, g is the gravitational constant (m s−2), z is the height of the wind-speed and temperature measurement (m), Ta is air temperature at height z (°C), Tsurf is the effective surface temperature (°C), and uz1 and uz2 are wind-speed (m s−1) at heights 1 and 2. Emitted long-wave radiation represented effective surface temperature in surface roughness and air stability estimations, assuming an emissivity equal to one.

4. Result

4.1. Weather

[17] The area experienced a period of warmer weather beginning in 2002. The warmest mean annual air temperature was −3°C (2002 and 2004), while 2001 and 2006 were cooler (−5.5°C) (Table 4). Longer thaw seasons were observed at the control site (K3) in the 2002–2006 period than in 2000 and 2001. The warmest year (2004) experienced a thaw season nearly 100 days longer than in 2001. The cooler mean annual temperatures in 2001 and 2006 resulted primarily to lower summer temperatures. Hourly air temperatures in the 2000–2006 period ranged from −40°C to 30°C, with postfire mean daily temperature at K2 and K3 strongly correlated (R2 0.93, p < 0.001). Average rainfall from June through August during the study period was 94 mm, with the largest monthly proportion falling in August. The 2002 and 2006 summers experienced the minimum rainfall amounts (64 mm) and 2005 the maximum (134 mm). With regard to winter precipitation, early May snow water equivalent was 80 mm and depth 0.4 m at Niagara Creek 1999 and 2000 at the tussock tundra. Shrub tundra showed approximately doubled values. A 0.15–0.3 m thick depth hoar layer was usually found at the bottom of the snowpack.

Table 4. Summary Table of Observations Made at K1, K2, and K3 Stations With Calculated Thermal Offsets and Top of the Permafrost Annual Temperatures
  • a

    Air temp at K3 missing 9/13–10/7, replaced with K2 temperature.

  • b

    Beginning of freezing to end of year.

  • c

    No K3 air temperature in March, April, and May. Replaced with K2 temperature.

  • d

    First day of year to beginning of thaw.

  • e

    K3 estimated through linear interpolation regression between 0.3 and 0.55 m depth.

  • f

    K3 estimated through linear regression between 0.4 and 0.55 m depth.

Snow water equivalentmtuss./shrub0.08/0.210.08/0.24nananananana
Snow depthmtuss./shrub0.39/0.730.39/0.74nananananana
End of snowmelt (albedo <20%)Julian DayK2na138150na(<131)106125137
Beginning thawJulian DayK2/K3na156/154152/160na/141140/144123/127128/145139/146
Beginning freezingJulian DayK2/K3286294/280289/281na/291292/316305/343277/298298/310
Duration thawing season (−5 cm/0 cm)DaysK2/K3na138/126137/121na/150152/172182/216149/153159/164
Freeze-up (−1 m)Julian DayK2nananana615429
Duration freeze-up period (−1 m)DaysK2nananana881039787
Duration frozen period (−1 m)DaysK2nananana13410886130
n-factor thawednoneK2/K3na0.51/0.710.46/nana/0.610.7/0.710.74/na0.84/0.630.77/0.55a
n-factor frozen 2ndbnoneK2/K3na0.84/0.380.68/0.17na/0.300.82/0.110.68/0.18c0.49/0.090.66/0.13
n-factor frozen 1stdnoneK2/K30.470.31/0.080.38/0.19na/0.070.31/0.040.21/0.020.36/0.100.33/0.06
Thawed Air Degree-DaysDegreesK2/K3na919/917973/893na/11561121/9691508/na1218/11931108/995
Thawed Surface Degree-Days (−5 cm/0 cm)DegreesK2/K3na471/648447/nana/707788/6841122/7671019/748850/543
Freezing Surface Degree-Days 2ndb (−5 cm/0 cm)DegreesK2/K3na−1455/−638−1102/−272na/−473−1357/−163−1245/−195−942/−162−1457/−279
Freezing Surface Degree-Days 1std (−5 cm/0 cm)DegreesK2/K3na−156/−46−519/−234na/−39−253/−30−146/−7−395/−100−260/−42
Freezing Air Degree-Days 2ndbDegreesK2/K3na−1723/−1659−1618/−1638na/−1573−1653/−1502−1843/na−1921/−1777−2207/−2143
Freezing Air Degree-Days 1stdDegreesK2/K3na−495/−600−1354/−1256na/−558−822/−741−710/−319−1084/−1032−797/−765
Rainfall (June–Aug.)mmMean allna101102641207213463
Min air temp. (hourly)DegreesK1/K2/K3na−31/−40/−34na/−38/−32na/na/−38−33/−38/−34−33/−40/na−32/−37/−32−35/−40/−37
Max air temp. (hourly)DegreesK1/K2/K3na21/24/2419/22/22na/na/2724/25/2527/30/na26/27/2824/27/26
Annual mean air temp. (1 m)DegreesK1/K2/K3na−3.9/−3.6/−3.7na/−5.5/−5.6na/na/−2.7−3.6/−3.7/−3.5−3.1/−2.9/na−5/−4.9/−4.3−5.5/−5.2/−5.9a
Annual mean surface temp. (−5 cm/0 cm)DegreesK2/K3na−3.1/−0.1−3.2/nana/0.5−2.3/1.3−0.8/1.9−0.9/1.5−2.4/0.2
Annual mean ground temp. (−30 cm/−40 cm)DegreesK2/K3na−3.9/−1.3e−3.3/nana/−0.5−2.5/0.2−1.5/0.9−1.2/0.9−2.8/−0.5
September mean ground temp. (−50 cm)DegreesK2/K30.3/na0.2/1fna/nana/1.3f1.8/1.7f2.3/2.1f3/3.1f2.7/2.4f
Annual mean ground temp. (−1 m)DegreesK2na−4.4nana−2.8−2.6−1.5−2.9
Annual min ground temp. (−1 m)DegreesK2na−11.7nana−11.5−10.6−7.5−9.2
Annual max ground temp. (−1 m)DegreesK2−0.6−0.6nana0.
Min soil temp. (−30 cm/−40 cm)DegreesK2/K3na−15.9/na−12/nana/−4−16/−2.5−14.8/−2.8−10.21/−1.8−13.5/−3.1
Max soil temp. (−30 cm/−40 cm)DegreesK2/K3na1.7/na2.1/nana/4.57.4/68.6/9.37.8/6.94.2/4.8
Radiation Efficiency (June)noneK2na0.67 ± 0.020.69 ± 0.02nana0.68 ± 0.120.69 ± 0.03na
Radiation Efficiency (July)noneK2na0.66 ± 0.020.67 ± 0.02nana0.66 ± 0.130.68 ± 0.02na

4.2. Soil Temperature, Soil Moisture, and Active Layer Thickness

4.2.1. Soil Thermal Regime

[18] Both the burned tussock tundra (K2) and the shrub control site (K3) experienced warmer September and annual ground and surface temperatures after the 2002 burn (Figure 6). The colder weather in 2006 resulted in soil cooling and reduced thermal offsets at K2 and K3, compared to the previous year (Table 4). A comparison of 0.3 m (K2) and 0.4 m (K3) mean annual ground temperatures between 2001 and 2006 (2002 excluded) shows a total change of +1.1°C and +1.3°C, respectively. Mean September temperatures at 0.5 m depth in 2000, 2003–2006 show a 2.7°C (K2) and 1.5°C (K3) increase. The warming in September was generally less the deeper in the 1 m soil profile (+0.9°C at 0.95 m depth, K2), while the annual averages showed the opposite trend at both stations. Annual soil temperatures in the same years were always higher at K3 than at K2, with the upper 0.4 m at K3 above freezing except in 2000. Averaged September soil and surface temperatures were higher at K3 before the fire, while postfire temperatures were higher at K2.

Figure 6.

Mean annual and September ground temperatures at K2 (burn) and K3 (control) station during year 1999 through 2005 of prefire (gray) and postfire (black) observations. The depth is relative to the surface at the time of observation.

[19] Mean daily air, surface, and ground temperatures at K2 and K3 are shown in Figure 7. The effect of vegetation on ground temperatures is clearly visible at K3, especially during the winter season. K3 never experienced temperatures colder than −4°C at 0.4 m depth, compared to −16°C at K2 (0.3 m). K2 reached a summer maximum of 2.1°C at 0.3 m before the fire, while the postfire maximum was 8°C in 2004. Similar changes between the summers can be found at K3 station at 0.4 m depth. The cumulative thawing degree-days at the surface can be seen as a measure of the total amount of energy received at the ground surface during the thawed season (Table 4). During all 4 years after the fire, K2 has higher cumulative thawed surface degree-days than K3, despite a shorter thaw season as a result of an earlier freezing.

Figure 7.

Mean daily air, surface, and ground temperatures at K2 and K3 stations 1999–2006.

[20] The thaw season n-factor shows a higher prefire value at K3 than at K2 (Table 4). After the burn, the K2 n-factor was equal to or higher than K3. When averaging all years, the K2 winter/spring n-factor (first day of the year until beginning of thaw) shows a higher value of (0.70 ± 0.12) than fall/winter (0.32 ± 0.06). K3 experienced similar partitioning but with lower ratios and relatively higher interannual variation 0.18 ± 0.11 (winter/spring) and 0.08 ± 0.06 (fall/winter). No significant difference was evident between prefire and postfire years.

[21] The postfire freeze-up date at K2 occurred in January, except after the warm summer of 2004, which resulted in freeze-up on 11 February 2005. The duration of the freeze-up period ranged between 87 to 130 days, a length similar to the frozen period of 86 to 134 days. Calculated mean annual temperature at the top of the permafrost was −3.7°C (K2) before the fire and reached −2.1°C in 2004 and 2005, with a cooling in 2006 (−3.4°C). K3 showed similar trends, but under thawing conditions at 1 m depth (Table 4). Thermal offsets varied around −1°C at K2. The negative offset was usually larger at K3.

4.2.2. Soil Moisture

[22] The fire resulted in relatively wetter near-surface soils. Before the fire and directly following the spring peak saturation, observations of near-surface soil moisture at K2 showed a drastic decrease to ∼40% saturation of unfrozen water content (Figure 8), which remained relatively stable during the rest of summer. After the fire, the near-surface soil was nearly saturated throughout the thawed season (70–100% saturation). This clear prefire and postfire difference did not occur at the control site (K3). A rapid increase during the beginning of thaw was observed postfire a month earlier at both stations. After the 2002 fire the decrease in unfrozen moisture content due to soil freezing occurred 19–30 days (K2) and approximately a month (K3) later than prior the burn.

Figure 8.

Mean daily unfrozen soil moisture presented as percent saturation at K2 (−10 cm) and the control K3 (−5 cm) 2000–2006. Gray and black lines represent observations made before and after 2002.

4.2.3. Active Layer Thickness

[23] Active layer depths at control and burned grid nodes, exhibit a similar trend from 2000 to 2002 (Figure 9) with the control 0.11 to 0.16 m deeper than the burned group. The control showed an active layer depth of 0.55, 0.67, and 0.64 m, while the group that later would be affected by fire 0.44, 0.53, and 0.48 m. Fall 2003, 1 year after the burn, resulted in a 0.11 m shallower thaw in the control group compared to 2002, while the burned area active layer depth increased with 0.06 m. The years following the tundra fire were characterized by consecutive deepening of the active layer over the entire CALM grid in 2004–2005 (control 0.68, 0.75 m, burned 0.60, 0.75 m) with a slight reduction after the cooler summer of 2006 (control 0.72 m, burned 0.70 m). The prefire difference (0.11–0.16 m) between mean of the two groups was reduced drastically after fire to ≤0.02 m, except in 2004 with 0.08 m deeper thaw in the control. Standard deviations indicate larger variability of thaw depths within the control group (≤0.21 m) than in the burned (≤0.15 m). In the shrubby valley, a few control thaw depth observations exceeded the 1.23 m long probe in 2004–2006. In the NW corner of the CALM grid, some control sites had rocks present, complicating thaw depth measurements.

Figure 9.

Observed mean CALM grid active layer depths acquired in late September or early October 2000–2006 of control (gray) and 2002 burned sites (black).

4.3. Surface Observations

4.3.1. Albedo

[24] Short-wave noontime absorbed insolation at K2 increased slightly after the fire, owing to reduced albedo. Mean albedo from the end of snowmelt to solstice was 16% (2000), 15% (2001) 11% (2004), 13% (2005), and 14% (2006) with standard deviation of ±2. From solstice to end of August, the observations were 17% and 16% before fire and 13%, 15%, and 17% in 2004–2006, respectively. A 5-week earlier snowmelt in 2004 compared to 2001 at K2 (Table 4) was evident through the drastic reduction in surface albedo. Despite the disappearance of snow cover on Julian day 106 in 2004, the thaw did not start until day 123.

4.3.2. Radiation Efficiency and the Richardson Number

[25] At K2, there was no prefire to postfire difference in radiation efficiency, Re, which varied between 0.66 and 0.69 (Table 4). More pronounced air instability was found after fire through lower Richardson numbers. The average for June was −0.45 (2000–2001) and −0.75 (2004–2005) and July changed from −0.39 to −0.67. Surface roughness increased after fire, from a June-July average of 0.02 in 2000–2001 to 0.04 m 2004–2005, enhancing turbulent mixing.

5. Discussion

5.1. Soil Thermal and Moisture Regime

[26] Four major factors influencing spatial and temporal changes in the thermal and hydrological regimes are weather, vegetation, fire effects, and ground ice conditions. The effect of fire on the sub-Arctic physical and biological systems is influenced by the presence of permafrost. Ice-rich permafrost promotes thermal stability at depth owing to latent-heat effects, while thawing of the ice-rich transition zone (the uppermost permafrost that alternates in status between seasonally frozen ground and permafrost over subdecadal to centennial timescales) results in release of water to the active layer [Shur et al., 2005]. Unfortunately, information about ground ice conditions was not obtained in this study. The potential response of permafrost to fire, and the dependence of surface conditions on ground ice, will therefore be discussed through observed changes in the active layer.

[27] The increased thaw depths and warmer soils at both burned and control sites after the fire show the dependence on weather. Although the changes in mean annual soil temperature 2001–2006 were very similar in the upper half meter of the soil, the change in September was nearly twice larger at K2 than at K3. Higher vegetation resulting in more efficient snow trapping and winter insulation [Sturm et al., 2001] at K3, offset the warming fire effects during summer at K2, on the annual scale.

[28] The postfire soil warming at K2 can be explained by several factors: (1) Approximately 50% of the 14 cm organic layer was destroyed by fire, causing drastic changes in near-surface thermal conductivity. Destruction of the insulating organic layer increased the heat conducted into the ground during summer, resulting in deeper thaw and increased n-factors. (2) Earlier snowmelt and onset of thaw after the fire was confirmed by albedo observations (K2) and ground surface temperature (K2 and K3). In a period of high incoming short-wave radiation, such a shift in snowmelt can have a significant effect on the energy available for soil heating and sensible heat flux. To what degree the earlier snowmelt was caused by warmer weather or reduced snow cover cannot be evaluated due to lack of postfire snow observations. (3) Summer albedo decreased after the fire (2004–2006) but not as drastically as observed in previous studies [Rouse and Mills, 1977], reducing the possible magnitude in soil temperature change. The limited burn and efficient regrowth of the well-developed tussocks resulted in little change in albedo, with the canopy shading moderately to severely burned ground between tussocks. No information on albedo during the first summer following fire existed, a time period that most likely experienced the lowest value. It is possible that a low albedo was a strong factor in 2003. This due to the largest postfire difference between K2 and K3 in 2003 September soil temperatures, and the increase in active layer thickness from the previous year in the burned CALM grid group compared to a decrease in the control group.

[29] A negative correlation should be expected between snow depth and n-factors, with lower ratios in the latter half of the winter as snow depth increases. However, lower n-factors were usually observed during the first half of the winter (beginning of freezing to end of year) than the second half, indicating a strong and efficient dissipation of heat from the ground toward the surface. Unfortunately, there were no postfire observations of snow depth, density, and timing to evaluate the influence of postfire snow, which plays a significant role through its insulating effects [Sturm et al., 2001; Stieglitz et al., 2003].

[30] The thaw season was longer at K3 than at K2 after the fire (the reverse was true before the burn), through a consistently earlier freezing of K2 near-surface soil. K2 experienced an earlier postfire thaw than K3, which also occurred in 2000. The observations were somewhat different compared to those of Burn [1998], who found a longer thaw period after the fire, which was explained by a more effective heat transfer (depletion of surficial organic soil), increased energy absorption (earlier thaw and reduced albedo), and increased heat capacity (wetter soils). Reduced organic layer and increased soil moisture may be competing mechanisms during freezing of the near-surface soil. At K2, reduced thermal conductivity was the dominant factor over the increased latent heat storage, resulting in relatively earlier freezing at K2 than K3. To what degree K2 was affected by a larger ice volume in the near-surface soils during thaw cannot be determined through these observations. Increased soil moisture favors a higher ice content, which can slow the descent of the thawing front [Woo, 1986].

[31] Possible contributors to the more moist postfire near-surface include modified soil properties at 10 cm depth, thawing of the upper part of the transition zone, and reduced evapotranspiration. (1) Soil moisture was presented as relative saturation of unfrozen water, which does not describe the absolute change in total volume of water. Before the fire, a porous organic layer represented the soil at 10 cm depth. After the fire, the observations were made in an organic-rich mineral soil, with a relatively larger surface area per volume and different pore shape, resulting in higher water-retaining efficiency. (2) The fire resulted in deepening of the active layer and partial thaw of the ice-rich transition zone. Owing to the enrichment of ice in the transition zone relative to the active layer, a disturbance (fire) may be accompanied by increased moisture throughout the thawed soil profile. (3) The fire combusted vegetation between tussocks, resulting in reduced evapotranspiration capacity. Mosses and lichens will probably require many decades to regenerate [Racine, 2004], while the tussocks recovered well within the 4-year post-fire observation period.

5.2. Boundary Layer and Radiation Efficiency

[32] The postfire soil warming at K2 caused increased convective heat loss from the surface to the atmosphere during local noon throughout the summer, as shown by the Richardson number (Ri). A 40% and 43% reduction after fire in Ri occurred in June and July, respectively, resulting in stronger buoyancy forces. This change in Ri indicates an increased partitioning of net radiation into sensible heat, seen at both burned spruce stands and tundra surfaces [Chambers et al., 2005]. At local scales, this may have resulted in a warming of unburned areas from adjacent burned areas. With stronger buoyancy forces, a large burn area can produce local to regional circulation patterns [Vidale et al., 1997], which include the potential to affect the overall atmospheric heat and water budgets even above the atmospheric boundary layer [Eugster et al., 2000].

[33] Postfire June-July surface roughness length (z0) doubled. Prefire estimates of z0 (0.02 m) are in agreement with studies made in northern Alaska tussock tundra of 0.02 m [Hinzman et al., 1996]. Field observations support a post-fire increase in microscale surface topography, since the fire-affected landscape was composed of 15–25 cm high tussocks separated by predominantly bare soil, owing to an efficient burn between the tussocks. Chambers et al. [2005] found a decrease in fire-affected, sedge-dominated tundra.

[34] Radiation efficiency (Re) was relatively stable throughout the observation period (66–69%), differing from previous studies. Directly after fire, boreal forest has shown a decrease [Yoshikawa et al., 2002; Chambers et al., 2005; Liu et al., 2005], while Chambers et al. [2005] found a Re increase at a tundra regime. The slight decrease in albedo at K2 might have been balanced by increased emitted long-wave radiation, which may be partly explained by a higher emissivity of the surface due to wetter conditions and relatively bare ground. Further, the inconsistency in Re and z0 compared to the study of Chambers et al. [2005] may be due to the different observation times after fire and the vegetation type and recovery. Observed increases in ground temperatures, buoyancy forces, and wetter near-surface soils are all indicators of changed partitioning in the net radiation among sensible, latent, and ground heat fluxes.

6. Conclusions

[35] Short-term fire effects at a sub-Arctic tussock tundra area included increased soil temperatures and near-surface soil moisture, altered energy partitioning through increased ground heat and sensible heat flux, changes in surface characteristics such as increased roughness length and slightly reduced albedo. The soil warming after the fire was the combined result of weather and the thermal evolution of the system induced by the fire. Reduction in the organic layer was a dominating mechanism during near-surface freezing in fall, despite the increased latent heat of wetter soil. The study shows the importance of postfire effects in influencing the physical status of the tundra ecosystem. Continuing long-term observations of the area would better characterize both fire impacts and recovery, as well as give insight to the future of tundra systems under a warming climate when fire frequency may increase.


[36] Financial support for this research was provided through the National Science Foundation, grants OPP-03332964 and OPP-0328686. Any opinions, findings, conclusions, or recommendations expressed are those of the authors and do not necessarily reflect the views of NSF. Mention of specific product names does not constitute endorsement by NSF. Ken Irving, Stefan Kooman, Crane Johnson, Matt Stone, Kristin Susens, Andy Monaghan, Jeremy Kasper, and Robert Bolton provided field assistance. The authors are grateful to Claude Duguay for advice in remote sensing and the Alaska Fire Service for providing information. Two reviewers and Associate Editor made many helpful suggestions that improved the paper.