Fire effects on surface-atmosphere energy exchange in Alaskan black spruce ecosystems: Implications for feedbacks to regional climate



[1] Although fire is crucial to the functioning and diversity of boreal forests, the second largest biome on Earth, there are few detailed studies of the effects of disturbance on surface-atmosphere interactions in these regions. We conducted tower-based micrometeorological measurements in summer over six recently burned black spruce stands that varied in age between 0 and 14 years. Results are presented for nonprecipitating conditions. There was an initial reduction of minimum albedo following fire from ∼0.09 to 0.06 followed by a rapid increase to 0.135 as the cover of successional vegetation increased. For clear-sky conditions near local solar noon, the combined effects of increased surface temperature due to a reduced surface-atmospheric coupling and albedo reduced midday net radiation by 9.3% (∼70 W m−2) for the first decade of succession. The average daily net radiation declined by 5.5%. Near noon, the relative partitioning of net radiation into ground heat flux doubled compared to nearby unburned stands, although on a daily basis the increase was only significant for the first few years following fire. Reduced net radiation, enhanced ground heat flux, and reduced Bowen ratio values following the first decade of succession could lead to a midday reduction of approximately 80 W m−2 in sensible heat flux compared to an unburned stand. Across a fire-scar boundary, the contrast in sensible heating and roughness is sufficiently large to induce mesoscale circulations and possibly trigger convective development. Because fire scars frequently exceed 104 ha, locally cooler conditions would prevail with suppressed planetary boundary layer development. Given the abundance of fire disturbance throughout interior Alaska, these effects may contribute to the regional climate patterns.

1. Introduction

[2] The boreal forest is the second most extensive biome on Earth, so changes in its interactions with the climate system have potentially important global implications. Previous modeling studies have focused on the climatic impact of changes in the extent of boreal forest in response to high-latitude warming [Bonan et al., 1992; Thomas and Rowntree, 1992; Foley et al., 1994] rather than changes in patterns of exchange within boreal forest. Climate warming has been implicated in recent and projected future increases in fire probability [Stocks et al., 1989; Flannigan and Van Wagner, 1991; Flannigan et al., 1998; Kasischke et al., 1999]. For example, regional warming in western North America in the past two decades has been associated with a doubling of area burned [Kasischke and Stocks, 2000]. However, the climatic implications of the resulting land-surface changes have not been explored.

[3] Fire alters many ecosystem properties that have potentially large effects on energy exchange. The initial impact of fire in northeastern Canada was to create a black surface, which absorbed more radiation, due to its low albedo [Rouse 1976; Johnson, 1992]. The loss of vegetation reduced evapotranspiration. Both of these changes tend to increase sensible heat flux. However, the loss of trees also reduces surface roughness and the coupling of turbulent fluxes between the land surface and the atmosphere [McGuffie et al., 1998], causing an increase in surface temperature, ground heat flux, and the depth of soil thaw [Viereck, 1973]. The warmer surface will, in turn, enhance long-wave radiation from the surface [Pielke, 1984; Eltahir, 1996]. In contrast to these initial effects of fire, mid-successional deciduous forests in western Canada had higher albedo, lower sensible heat flux and greater evapotranspiration than did late-successional conifer forests [Baldocchi et al., 2000]. Although the effects of fire disturbance on surface-atmosphere interactions have been the focus of some recent attention [Amiro et al., 1999; Litvak et al., 2000], we know of no study that has examined in detail the impacts of fire on energy exchange through a range of early successional stages.

[4] Lowland black spruce (Picea mariana) forests dominate 40% of interior Alaska [Van Cleve et al., 1986]. Following fire, herbaceous vegetation covers the charred surface within a few years, followed by a shrub stage, deciduous forests, and eventually after 40–80 years a return to black spruce [Viereck et al., 1983; Van Cleve et al., 1991]. At the landscape scale, the major effect of fire is to increase the relative abundance of these early and mid-successional vegetation types, relative to the late-successional black spruce [Chapin et al., 2000]. The purpose of this paper is to describe the effects of fire on water and energy exchange in these initial stages of postfire succession in the boreal forest of interior Alaska.

2. Measurement Site Descriptions

[5] To maximize the number of sites compared in this 3-year study, we used a “roving tower” approach [Eugster et al., 1997] rather than seasonal or annual tower deployments at one or a few sites. We characterized each site for the range of synoptic meteorological conditions that occurred during deployments of 1–3 weeks. Discounting rain events and equipment malfunctions, this resulted in 5–18 measurement days per site. Where possible we made simultaneous measurements over paired burned and unburned stands to account for the effects of seasonality, synoptic climate or soil moisture conditions.

[6] The sites were generally flat expanses of lowland black spruce on alluvial deposits or glacial outwash in the Tanana River drainage of interior Alaska (Figure 1). Burned sites were selected in large homogeneous expanses of burns of known ages [Kasischke et al., 1995]. Unburned control sites were selected near each burn in black spruce forests that appeared similar in tree density and height to the burned forests. We selected the largest most homogenous sites in interior Alaska that were road-accessible. In most sites, Kasischke and colleagues [Kasischke et al., 1995; O'Neill et al., 2002] had previously documented the fire history, patterns of soil environment and respiration, and remote sensing signatures, so the sites could be placed in a larger-scale regional context. The surroundings and characteristics of each site are summarized below, with more detailed descriptions provided at

Figure 1.

Regional map of Alaska indicating the locations of the fire scars studied during this experiment.

[7] One caveat of short-term tower deployments is that results may not be representative of long-term averages. Climate anomalies that occur before and during the measurement period may affect fluxes. An example of interannual variation in stand characteristics of the same unburned black spruce stand measured in consecutive years under different climatic conditions is presented in Figure 2. For the periods of measurement, conditions in 1999 were drier than in 1998, resulting in large changes in the partitioning of net radiation into turbulent fluxes. Unless otherwise stated, observations over unburned stands in this study represent an average over five sites, some of which were repeated over the 3-year period of this study. Consequently, the results span a wide range of climatic conditions. However, since each burned stand reported in this study was visited only once, we cannot estimate representativeness errors for these observation sites. For this reason, details of the climatic conditions during the observation periods are presented below with each site description.

Figure 2.

A comparison of radiation efficiency (Rn/Si) and the partitioning of net radiation into sensible, latent and ground heat fluxes for an unburned black spruce stand during climatically wet (1998) and dry (1999) conditions. These results are based on daily totals over 6 and 9 days for the wet and dry cases, respectively.

2.1. Age Class: 0-Year Post Burn

[8] The Donnelly Flats fire occurred just south of Delta Junction in June 1999 and burned approximately 7,600 ha. Since strong southerly winds are characteristic of this region we located the measurement site on the northwest side of the fire scar (63°54′N, 145°44′W). Except for a small unburned patch 1 km north-west of the tower, there were several km of fetch in all directions. The surface organic mat was largely consumed by the fire such that the mineral soil horizon was typically 3–5 cm beneath the surface. It was black except for patches of light brown wind-blown silt that covered 15–20% of the surface. Characteristics of the remaining spruce canopy are summarized in Table 1. Measurements were made simultaneously with those at the Granite Creek 12-year burn (see below).

Table 1. Canopy Characteristics of Burned Black Sprucea
 Stem Density (>1 m), ha−1Height (>1 m), m
  • a

    Data are means ± SE.

Donnelly Flats4500 ± 11253.9 ± 0.3
Hajdukovich Creek4520 ± 6473.0 ± 0.2
Tok Junction9780 ± 10203.6 ± 0.3
Granite Creek3200 ± 13293.9 ± 0.6
Rosie Creek4070 ± 7852.9 ± 0.4

[9] Measurements began approximately 2-weeks after the fire and spanned the period 3 July to 1 August 1999. Vegetation cover increased from 0% to 5% cover during the measurement period, as vegetation began to resprout. Weather was mostly clear during these observations. Climate statistics for 2 hours either side of local solar noon (1400 Alaska daylight time (ADT)) are presented in Table 2. Based on wind speed, measurement height and surface roughness, we calculated the 90% effective flux footprint and the region of maximum flux influence on flux estimates [Gash, 1986] (Table 2). The Gash footprint dimensions are based on neutral atmospheric conditions and overestimate distances by about 35% for conditions of moderate instability [Horst and Weil, 1992]. Since conditions were always unstable near local solar noon the area of maximum influence on the flux observations would lie well within the 100 m vegetation survey radius, and the 90% effective flux footprint was considerably less than the available fetch in all directions.

Table 2. Near-Noon (1100–1600) Climate and Fetch Data for Measurements at the Donnelly Flats (DF) 0-Year, Hajdukovich Creek (HC) 4-Year, Tok Junction (TJ) 7-Year, Rosie Creek (RC) 14-Year, and Granite Creek (GC) 12-Year Burn Sitesa
Site/TimeDF Burn (1999)HC Burn (1997)HC Burn (1998)TJ Burn (1997)RC Burn (1997)GC Burn (1999)
  • a

    Data are means ± SE. Si, incoming shortwave radiation (W m−2); Ts, near-surface (0.3 m) temperature (°C); VPDs, vapor pressure deficit (hPa), Uson, wind speed at sonic height (m s−1); X90, the 90% effective flux footprint (m); Xmax, the region of maximum influence on flux measurements (m); and z/L the stability parameter.

Si448 ± 14337 ± 18324 ± 16485 ± 18625 ± 13520 ± 20
Ts19.3 ± 0.419.1 ± 0.515.2 ± 0.419.4 ± 0.325.7 ± 0.324.6 ± 0.5
VPDs12.2 ± 0.69.3 ± 0.66.9 ± 0.510.0 ± 0.520.2 ± 0.719.7 ± 1.0
Uson2.2 ± 0.1 (6 m)1.6 ± 0.1 (7 m)1.9 ± 0.1 (9 m)1.8 ± 0.1 (7 m)1.8 ± 0.1 (7 m)2.7 ± 0.1 (9 m)
X90840 ± 1051760 ± 1331720 ± 145713 ± 561090 ± 781280 ± 36
Xmax44 ± 693 ± 790 ± 838 ± 358 ± 468 ± 2

2.2. Age Class: 3–4 Years Post Burn

[10] The 1994 Hajdukovich Creek fire burned an area of about 8900 ha 25 km southeast of Delta Junction. The 1997 measurement site was located in the southeastern quadrant of the fire scar (63°48′06″N; 145°06′28″W) so the shortest fetch was approximately 1.5 km to the southeastern boundary. To the north and west, fetch over relatively uniformly burned black spruce was in excess of 5 km. The average 90% effective flux footprint was less than the available fetch in most directions. New and resprouting vegetation covered roughly 16–20% of the surface and was typically less than 0.5 m tall. Herbaceous species (primarily Calamagrostis canadensis) constituted more than half of the regrowth, with the remainder dominated by the shrubs Vaccinium uliginosum, Ledum groenlandicum, and Salix spp. The charred organic horizon averaged 18.5 cm in depth. During the two measurement periods for this site (4–17 August 1997 and 3–13 August 1998), conditions were typically overcast and occasionally quite wet (Table 2).

[11] For the 1997 measurements at the Hajdukovich Creek burn we made simultaneous measurements at an unburned stand with similar preburn canopy characteristics that was located 6 km northeast of the fire scar near the Gerstle River. The unburned stand (Table 3) measured approximately 1.6 km (N-S) by 3.2 km (E-W). The tower was located near the southern boundary of the unburned stand to maximize the fetch in the direction of the prevailing winds (east, north, and west). Due to its limited size, the average 90% effective flux footprint of this site was less than the available fetch for wind directions from the north through east (320–130°T). The understory had approximately 60% cover, of which 5% was herbaceous species, the remainder dominated by Vaccinium vitis-idaea, V. uliginosum, and Ledum groenlandicum. The stem density of trees (>1 m) other than spruce at this site was 1040 ha−1 (primarily Salix spp.). The ground cover was dominated by feathermoss (Hylocomium splendens), Sphagnum spp., and foliose and fruticose lichens. The average thickness of the organic mat was 33 cm. The local midday climate statistics for this site are summarized in (Table 4).

Table 3. Unburned Black Spruce Canopy Characteristics
 Stem Density (All), ha−1Stem Density (>1 m), ha−1Average Height (>1 m), mAverage Height (10,000 ha−1),a m
  • a

    Height for the tallest 10,000 trees ha−1.

Rosie Creek66,900 (se ± 7,200)20,200 (se ± 2,000)2.02.5
Hajdukovich Creek35,800 (se ± 3,000)18,400 (se ± 1,600)2.73.4
Tok Junction43,500 (se ± 5,500)23,600 (se ± 2,600)3.04.3
CPCRW North24,000 (se ± 2,900)13,800 (se ± 3,600)2.53.0
CPCRW South23,500 (se ± 4,300)16,300 (se ± 880)2.94.0
Table 4. Near-Noon (1100–1600) Means ± Standard Errors of Climate and Fetch Data for Measurements at the Hajdukovich Creek, Tok Junction, Rosie Creek, and CPCRW Preburn (North) and Preburn Control (South) Sitesa
Site/TimeHC Control (1997)TJ Control (1997)RC Control (1997)CPCRW North 1998CPCRW North 1999CPCRW South 1999
  • a

    Symbols and abbreviations as in Table 2.

Si, W m−2327 ± 22457 ± 22622 ± 16365 ± 17530 ± 19532 ± 20
Ts, °C17.5 ± 0.618.7 ± 0.427.4 ± 0.317.3 ± 0.421.2 ± 0.421.2 ± 0.4
VPDs, hPa8.4 ± 0.78.4 ± 0.726.9 ± 0.48.9 ± 0.617.5 ± 0.816.4 ± 0.9
Uson, m s−11.7 ± 0.1 (9 m)2.6 ± 0.1 (12 m)1.9 ± 0.1 (11 m)2.0 ± 0.1 (10 m)2.1 ± 0.1 (9 m)2.3 ± 0.1 (10 m)
X90, m1230 ± 441930 ± 581650 ± 571320 ± 591020 ± 271180 ± 28
Xmax, m65 ± 2102 ± 387 ± 469 ± 354 ± 262 ± 2

2.3. Age Class: 7-Years Post Burn

[12] The Tok Junction fire occurred in 1990 and covered an area of about 41,800 ha. The measurement site was located in the northwest quadrant of the burn scar (63°21′N, 142°54′W) 2–3 km from the nearest burn boundaries. There was a 69% cover of regrowth up to 1.5 m tall interspersed with open patches of charred moss. Slightly less than half of the regrowth was herbaceous (Carex spp. and Calamagrostis canadensis); the remainder was dominated by Salix spp, Vaccinium vitis-idaea, and Ledum groenlandicum. The average depth of the charred plus regrowing organic mat was 10.3 cm. Weather conditions during the 16 July to 2 August 1997 measurement period were fine (Table 2) except for four overcast days with heavy rain in the middle of the observation period.

[13] The only unburned stand of black spruce forest near the Tok burn scar with similar canopy characteristics was 12 km to the east near the eastern boundary of the fire scar (63°18′N, 142°41′W). This patch of unburned spruce was small (2.2 km east-west and 0.8 km north-south), with the tower located in the center of the measurement site. The average 90% effective flux footprint (Table 4) was of a similar magnitude to the size of the patch, with the fetch maximized for wind directions from either the east or west. The canopy contained only 440 ha−1 trees >1 m that were not spruce, and these were primarily willow. There was a 51% cover of understory vegetation, 10% of which were herbaceous species; the remainder was dominated by Vaccinium vitis-idaea, Salix spp, and Ledum groenlandicum. The dominant ground cover was feathermoss and lichen. The average depth to mineral soil was 27.1 cm.

2.4. Age Class: 12-Years Post Burn

[14] The Granite Creek fire occurred southeast of Delta Junction in 1987 and covered an area of approximately 20,000 ha. The measurement site was located in the northeast quadrant of the fire scar (63°56′N, 145°27′W). The poorest fetch was roughly 750 m to the south-east, and the best fetch was at least 5 km for the sector 190–315°T. The average 90% effective flux footprint was less than the available fetch for wind directions excluding the southeast quadrant (Table 2). The site was characterized by a heterogeneous deciduous overstory (dominated by Populus tremuloides and Salix spp.) with a maximum height of 3–3.5 m, a cover of almost 30% of understory vegetation (predominantly Rosa spp, Ledum groenlandicum and Epilobium angustifolium), separated by open patches of charred moss. Measurements coincided with those at the 0-year Donnelly Flats burn. Weather conditions during the 2–14 July 1999 observation period were mostly fine. Fair-weather cumulus clouds were common in the afternoons, with occasional light showers, but there were no heavy or extended periods of rainfall.

2.5. Age Class: 14-Years Post Burn

[15] The Rosie Creek fire occurred in 1983 in the Bonanza Creek Experimental Forest (BCEF; 64°41′N, 148°20′W), 20 km southwest of Fairbanks, and covered an area of 3480 ha. The measurement site was located in the center of a strip of burned land between the southwest margin of the Tanana Uplands and an unburned strip of spruce, beyond which to the southeast, was the floodplain of the Tanana River. The flat portion of the burned black spruce stand was 500–600 m wide, and about 4 km long, oriented roughly north-east. The average 90% effective flux footprint was greater than the width of the burn at this site. The best fetch conditions were realized for wind directions of southwest or northeast. A substantial canopy of juvenile black spruce 0.5–2 m tall (stem density of 36700 ± 4800 ha−1) had developed. The regrowth covered about 80% of the surface, of which about 25% was herbaceous (Calamagrostis canadensis), and the remainder was dominated by Vaccinium uliginosum, Ledum groenlandicum, and Betula glandulosa. Some of the deciduous species were >2 m tall. The average depth of the charred plus regrown organic horizon was 20.9 cm. Observations at this site spanned the period 20 Jun to 5 July 1997, during which time the weather was typically fine and dry (Table 2).

[16] The unburned black spruce stand was slightly shorter and sparser than in the burned area, and the understory species reflected a moister environment (tussocks present). There was a 47% cover of understory species, of which about a third was herbaceous (Eriophorum vaginatum and Calamagrostis canadensis) and the rest dominated by Ledum groenlandicum, Vaccinium vitis-idaea, and V. uliginosum. This strip of unburned black spruce was 300–500 m wide, and over 4 km long, oriented parallel to the burned strip. The average 90% effective flux footprint was greater than the width of the unburned stand. The best fetch conditions were realized for wind directions of southwest or northeast. The stem density of trees >1 m other than spruce in this canopy was 220 ha−1. The dominant ground cover species were feathermoss, lichens, and Sphagnum. The average depth to mineral soil was 27.5 cm. This unburned site was located approximately 500 m SSE of the burn site, so weather conditions were similar (Table 4).

[17] Two additional unburned stands were observed in 1998–1999 on Helmer's Ridge in preparation for a prescribed burn within the Caribou Poker Creek Experimental Watershed (CPCRW; 65°10N, 147°30W), 45 km northeast of Fairbanks. The top of the ridge is broad and relatively flat and extends roughly 1.5 km from east to west. The sites were positioned 40–50 m to the north and south of the ridge centerline. Optimum wind directions for measurements at these sites are either from the western or eastern quadrants taking into account both fetch and topographical constraints. Differences in climate statistics between these two stands was negligible, values for the 1999 observations are presented in Table 4 as an example. In general, the 1998 measurement period at Helmer's Ridge (9–28 July) was cooler and wetter than the 1999 period (6–21 June; Table 3). For both cases, the area of maximum influence on fluxes lay within the 100-m vegetation survey radius. During conditions when the wind direction was from the east or west, the available fetch was suitable.

3. Equipment and Methods

[18] We conducted tower-based microclimatic and eddy-correlation flux measurements [Baldocchi et al., 1988] over each of the above stands during the summers of 1997–1999. Towers were located to maximize the fetch of flat homogeneous stands of target vegetation in the direction of the regional prevailing wind. We used two 15 m aluminum towers (Climatronics Corp., Bohemia, NY, U.S.A.) instrumented almost identically [Eugster et al., 1997; McFadden et al., 1998]. On each tower we measured the parameters required to calculate turbulent fluxes of heat and water vapor using a 3-D ultrasonic anemometer (Solent 1199HSH, Gill Instruments Ltd., England) in conjunction with a closed-path, infrared-absorption H2O/CO2 gas analyzer (IRGA: LI-6262, LI-COR Inc., Lincoln, NB, U.S.A.). Data were recorded on a laptop PC at a rate of 10 Hz. We used a diaphragm pump (UN815-KTDC, KNF Neuberger, Trenton, NJ, U.S.A.) to aspirate the IRGA via 3.3 mm internal diameter “bev-a-line” intake tube (Thermoplastic Processes Inc., Sterling, NJ, U.S.A.). We mounted the sonic anemometer and intake tube on a 3 m boom typically 4–8 m above the canopy (depending on the site). We oriented the sonic anemometer within a 180° sector that represented the largest region of most homogeneous fetch of the targeted ecosystem type in the presumed direction of prevailing wind. We calibrated the gas analyzer for water vapor every 48 hours using ultra-high-purity nitrogen as a zero gas, and a precision dew-point generator (LI-610, LI-COR Inc.). Power to each system was supplied by a gasoline-powered 650W Honda generator, situated roughly 80 m from the tower in the direction opposite to the presumed direction of the prevailing wind. The generator charged a bank of four 12-V deep-cycle batteries that served as a power buffer if the generator failed or required maintenance.

[19] We also recorded microclimatic data as 10-min averages of 20-s readings using a data logger (Cr10x, Campbell Scientific Inc., Logan, UT, U.S.A.). These observations included incoming and reflected short-wave radiation (Precision Spectral Pyranometers, Eppley Laboratories, Providence, RI, U.S.A.), long-wave radiation (Pyrgeometers, Eppley Inc.) respectively, net radiation (Fritschen type Q-7.1, Radiation and Energy Balance Systems [REBS], Seattle, WA, U.S.A.), temperature and relative humidity (Vaisala-type HMP45C, Campbell Scientific Inc.), pressure (PTB101, Campbell Scientific Inc.), wind speed (R.M.Young 03101-5, Campbell Scientific Inc.), soil temperature (STP-1, REBS), soil moisture (Hydra Probe, Vitel Inc.) and ground heat flux (HFT-3.1, REBS). Albedo was calculated as the ratio of outgoing to incoming short-wave radiation. Since we did not have suitable calibrations for the Vitel probes in organic matter, only relative soil moisture conditions at each site are discussed.

[20] In the unburned stands we made radiation measurements both above and below the canopy. The subcanopy measurements were made above the understory vegetation at 0.5 m. Although instrument heights varied slightly among installations, we always mounted the above-canopy radiometers as close as practical to the top of the 15 m tower, regardless of canopy height, to minimize shading errors and maximize the footprint of the downward facing radiometers. We installed the temperature/humidity probes at three levels per site, near-surface (0.2–0.3 m), 1.5–2 m (mid canopy for the unburned sites) and at the sonic height as an independent check of average temperature and humidity reported by the eddy-correlation instruments. The cup-anemometers were colocated with the net radiation measurements to correct for wind speed-induced dome cooling [REBS, 1994]. We derived wind direction from the sonic anemometer, as well as specific humidity and air density from the temperature, humidity, and pressure measurements. We calculated “total” ground heat flux via the combination method [Oke, 1987] using volume fractions of mineral, organic and water content of the surface soil to estimate surface heat capacities for the surface energy storage calculations.

[21] We characterized the vegetation in the tower footprint along five 100 m transects radiating from the tower at 45° intervals in the 180° sector of best fetch. Every 5 m along each transect we characterized the dominant overstory, understory and ground cover species in 1 m2 plots. Measurements included live/dead tree density >1 m, <1 m and seedlings, tree height, basal diameter and diameter at 1.4 m (DBH), leaf area index (LAI), % cover of understory and groundcover vegetation by species, % bare ground, % dead wood, active layer depth and depth to mineral soil (the latter in every second plot). Based on the vegetation surveys, we installed the soil temperature, moisture and ground heat flux instrumentation at sites representative of the four most common ground-cover types within the predicted tower footprint.

[22] The placement of the sonic anemometer for the reported measurements was based on a visual estimate of the canopy height (local average tree “skyline”). Even in the mature (>100 year) Alaskan black spruce stands there is an abundance of relatively small trees, which can result in large differences between statistical and visual estimates of canopy height (Figure 3b). Trees of height <1 m account for approximately 50% of the stem density (Table 3), but even the average height of trees >1 m was often considerably less than the visually estimated tree height (2.5 m Rosie Creek, 3.5 m Gerstle River, 4.5 m Tok Junction, 3.5 m and 4.5 m for the unburned black spruce stands Caribou Poker Creek sites). Although the smaller trees within the canopy play an important role in radiation/water interception and the exchange of H2O/CO2, they have less influence on the aerodynamic roughness of the surface. In order to derive a physically meaningful estimate of the average height of the primary roughness elements of the canopy, we assumed that, on average, not more than one mature spruce tree per square meter would contribute to the overstory surface roughness elements. We sorted the tree data in order of descending height and calculated the average canopy height using the tallest trees up to a stem density of 10,000 ha−1 (1 tree m−2) (Figure 3a). These values closely matched the visual estimates (Table 3).

Figure 3.

(a) Height-sorted stem density of black spruce in three unburned stands with different stand architecture, and (b) tree height histogram for the Tok Junction unburned black spruce stand, showing the statistical average height and the height estimated visually from the tower, when installing the sonic anemometer (see methods).

4. Measurement and Processing Considerations

[23] We omitted data collected during rainfall events, and checked the original time series data for spiking/noise. Events exceeding 3σ were checked for consistency with neighboring data and corrected via linear interpolation when necessary. Prior to calculating the turbulent fluxes we performed a coordinate rotation of the wind vector such that the u component was inline with the sonic axis [Baldocchi et al., 1988]. Cumulative cospectra (“ogive” plots, Figure 4b) indicated that turbulence on temporal scales of up to 5 min constituted the majority of the flux. However, we only linearly detrended the 30-min data to remove contributions from scales of turbulence too large to be sampled within that time period. Tests indicated that removing the contributions from scales of turbulence between 5 and 15 min reduced turbulent flux estimates by 4–6% and would have affected budget closure estimates. We calculated flux estimates as 30-min averages over the diurnal cycle and used 3-point averaging to reduce the sampling frequency of the climate data to match the flux data. In postprocessing, we omitted data obtained from wind directions originating from behind the sonic anemometer (±45°), and from regions of poor fetch at sites where this was a concern. To calculate daily averages for sites, we replaced short gaps in the turbulent flux data due to site maintenance via linear regression estimates derived from the available data.

Figure 4.

Representative example of (a) a w′t′ cospectrum and (b) an ogive plot generated from 30 min of data (11:30-1200 ADT) collected at 7.2 m over an unburned black spruce stand near the Gerstle River, interior Alaska. The roughly monotonic increase of the w′t′ ogive ceases at a frequency of 0.003, corresponding to 5.5 min (shown with a vertical line).

[24] When using a closed-path analyzer for eddy-covariance flux measurements, turbulent flow of the sample air through the IRGA must be maintained in order to minimize attenuation of turbulent constituent fluctuations [Leuning and Judd, 1996]. For our system, this would require a flow rate of ≥5 L min−1. However, to fully utilize the fast response time of the IRGA, the 11.9 cm3 optical bench must be flushed at roughly 10 Hz, requiring a flow rate of ≥7 L min−1. As an added precaution, we installed a 1.3L PVC buffer volume between the diaphragm pump and the IRGA to reduce pressure-pulsing and constructed 1.5 m of the intake from insulated copper tubing to minimize temperature-induced density fluctuations in the sample air stream [Webb et al., 1980; Leuning and Judd, 1996]. To account for slight variations in flow rate due to varying pump performance and ambient wind, we calculated and corrected for the lag time between the sonic anemometer and IRGA signals for each 30-min block of data. Under conditions when the lag time was poorly defined (typically near neutral stability), the daily modal lag time was adopted.

[25] Many factors contribute independently to underestimated flux estimates including the distance between instruments, instrument response time and signal attenuation [Leuning and Moncrieff, 1990; Rißmann and Tetzlaff, 1994; Leuning and Judd, 1996; Horst, 1997]. Historically the preferred method to correct covariance estimates was to address each issue independently and subsequently apply the collection of correction factors [Philip, 1963a, 1963b; Moore, 1986]. Contemporary methods are converging on an approach that applies a single correction of a suitable form to account for all problems collectively [Eugster and Senn, 1995].

[26] We utilized the latter approach to avoid the potential risk of compounding several sources of error from the independent estimates. This technique requires the calculation of an induction factor to account for the discrepancy between an observed and idealized cospectrum [Kaimal and Finnigan, 1994, p. 55] as a result of all the system inadequacies. As an example, Figure 5a compares a normalized w′t′ cospectrum, from a fast-response open-path sonic anemometer, to a w′q′ cospectrum, derived in part from a closed-path IRGA with a slower frequency response. The w′q′ cospectrum diverges from the w′t′ cospectrum beyond frequencies of 0.5–0.6 Hz (<1.6–2 s).

Figure 5.

(a) High frequency damping of a w′q′ cospectrum from a closed path IRGA (a composite of three 30-min cospectra with the plot resolution reduced to 20 points per logarithmic decade), (b) an example of the diurnal variation in flux correction factor required to account for underestimated w′q′ and w′c′ covariances. The correction factor curves were derived from 11 days of observations at 9.2 m over a 3.5-m stand of black spruce. Data are means ± standard error.

[27] The induction factor is unique to a given instrumental setup, and once determined can be applied to the whole measurement period. The final factor used to correct the covariance estimates was then determined for each 30-min block as a function of the induction factor, measurement height, wind speed and atmospheric stability (Eugster and Senn [1995], equations 32 and 33 for unstable and stable conditions, respectively). For daytime measurements over unburned canopies, 30-min average latent heat fluxes were commonly underestimated by ∼5% (Figure 5b). This estimate agrees well with the findings of Leuning and Judd [1996], who suggest that 95% of a covariance estimate is contributed by frequencies of turbulence up to f = 2U/(zd), where U is the mean wind speed, z and d are the measurement and displacement heights, respectively. The daytime measurement errors increased slightly with decreasing surface roughness and measurement height.

[28] Wet canopies confound radiation balances by drawing heat from the surrounding air mass to evaporate water [Blyth et al., 1994] leading to uncharacteristically high values of net radiation (Figure 6). Although soil moisture probes had been installed at each site at depths of ∼8 cm, components of the radiation budget appeared to be more sensitive to surface or canopy moisture status. Consequently for near-noon stand net radiation estimates we separated our measurements into periods of “wet” or “dry” canopy states based on near-surface (0.2–0.3 m) measurements of vapor pressure deficit. For values of vapor pressure deficit that exceeded 3hPa, we assumed that liquid water was no longer freely available upon the canopy elements, although the vegetation was not expected to be experiencing water stress. Separation of the data in this fashion (Figures 6b and 6c) reduced the scatter in the relationship of net radiation to global radiation (Figure 6a).

Figure 6.

Rn versus Si over a black spruce stand for all data and for suspected “wet” and “dry” canopy moisture conditions (see methods). Plots have been derived from 16 days of daylight (Si > 0) observations with incomplete or erroneous readings removed. Each point represents a 30-min average of 20-s readings.

[29] Albedo estimates have been confined to nonprecipitating and predominantly clear-sky conditions such that the indicated variability for a given surface type is primarily a function of changing surface moisture status from day to day. Over vegetated surfaces the overcast albedo is typically lower by 0.02 [Eugster et al., 2000] due to the increased penetration of diffuse light into the canopy. Unless otherwise stated, all other results presented in this paper are representative of the complete range of weather conditions (excluding rainfall events) experienced near local solar noon (1400 hours ± 2 hours) during the measurement period. All times quoted in this manuscript refer to Alaskan daylight time (ADT).

[30] Average midday budget closure errors (ΔQ = RnH − LE − G) were approximately 10%, which are typical of eddy covariance measurements where site and stability conditions meet the assumptions of eddy covariance [Eugster et al., 1997]. Based on daily totals, the average budget closure error was slightly larger (∼14%), with the poorest closure being achieved at the youngest burn scars (Figure 7), despite abundant fetch. This may be a result of underestimated ground heat fluxes due to representativeness errors of the selected soil microsites. The unburned black spruce stands have the most complex canopy, so we will briefly discuss the energy budgets of these sites for the measurement conditions near local solar noon (Table 5). On average across the control sites 50.8%, 30.8% and 7.9% of net radiation were partitioned into sensible, latent and ground heat fluxes near noon, accounting for a total of 89.5% of net radiation. Based on 10-min temperature measurements at two heights within the canopy airspace (∼0.3 m, ∼2 m) and biomass (0.5 m, 2 m: trunks and branches of four trees) peak energy storage by the canopy roughly accounted for an additional 4% of net radiation for the unburned black spruce stands. Based on CO2 assimilation rates by the unburned stands (data not shown) photosynthesis would have accounted for ≤1% of net radiation [Oke, 1987].

Figure 7.

Budget closure estimates, (RnG)/(H + LE) as a percentage, based on daily totals for 0-, 4-, and 12-year-old burn scars compared to an average of the unburned stands. Results are based on 6–9 complete days of observations, and error bars denote the standard error.

Table 5. Daytime (Sin >50 W m−2) Regressions of Net Radiation and Surface Fluxes for Control Sites for a Range of Weather Conditions, as Well as Near-Noon Partitioning of Net Radiation Into Ground Heat Fluxes
Unburned Black Spruce SiteH versus RnLE versus RnG/Rn (Noon ± 2 hours)
Tok 19970.502−43.310.7460.3709.730.7070.0920.011
Delta Junction 19970.583−43.990.8960.31019.380.7620.0920.006
CPCRW preburn site0.520−24.090.7570.32321.770.6210.0610.003
CPCRW control before burn0.731−47.100.9270.16423.840.6550.0640.002
CPCRW control after burn0.524−23.230.9340.2297.550.6760.0840.006

[31] We performed all postprocessing including data quality control, calibrations, signal time lag and spectral corrections using the FORTRAN time series analysis package “RAMF” version 8.0 [Chambers et al., 1996].

5. Results

[32] Minimum (midday) albedo values were similar among all the unburned black spruce stands that we observed, averaging 0.088 (Figure 8a). Immediately after fire the average stand albedo was lower (0.06) and more variable. This average albedo of the site was intermediate between the albedo of charred organic material (as low as 0.04) and mineral soil, which in a plowed agricultural field varied from 0.08 to 0.12 from moist to dry, respectively. Three and five years after fire average stand albedo had increased to 0.081 and 0.098, respectively, values similar to those in unburned stands. Beyond 5 years after fire, however, albedo was higher in the burned than in the unburned stands, with values as high as 0.135. Although the deciduous species at the 12-year old burn scar were often twice the height of those within the 7-year old burn site, the deciduous canopy was considerably more dense and homogeneous at the 7-year old burn. This contrast in canopy characteristics is most likely responsible for the greater radiation absorption (lower albedo) values in the 12-year old burn (Figure 8a). The establishment of early successional vegetation also increased the diurnal variation in albedo compared to the unburned or recently burned site (Figure 8b).

Figure 8.

A comparison of (a) midday (minimum) albedo between unburned and early successional postfire black spruce stands, and (b) the diurnal cycle of albedo in unburned and two burned stands (2 weeks and 12 years old). Columns are averages between the hours of 1100 and 1500 ADT. Error bars denote one standard deviation to more clearly indicate the day-to-day variability in stand albedo associated with factors such as surface moisture levels.

[33] The unburned black spruce stands had a high radiation efficiency (the ratio of net radiation to incoming short-wave radiation, Re = Rn/Si) with a midday value of 75.5 ± 0.8% (Figure 9a). Radiation efficiency of unburned stands was typically highest in stands with greatest canopy height and surface roughness (the CPCRW stands), indicating the importance of canopy complexity in radiation absorption. At the Tok Junction control site the canopy in the immediate vicinity of the tower was more heterogeneous than that at the CPCRW south site, which may have contributed to the observed difference in stand radiation efficiency. Midday radiation efficiency decreased significantly after fire (Figure 9b). Although somewhat erratic, there appears to be a minimum value reached at an age of 7 years before values increase again in the 12- to 14-year burns to an average value about 7% less than in the unburned stands. The 14-year burn scar already had spruce saplings overtopping successional shrubs, and appeared to be skipping the deciduous overstory phase. This may have been responsible for the differences in radiation efficiency observed between the 12- and 14-year burn sites. Overall, for the postfire stands that we studied, fire reduced midday radiation efficiency by 9.3%, from 75.5 to 66.2% for this early successional period.

Figure 9.

Radiation efficiency for clear-sky (dry canopy) conditions of (a) 5 unburned black spruce stands plotted in order of decreasing canopy height, and (b) 6 burned black spruce stands between 0 and 14 years of age. Values represent averages between the hours of 1100 and 1500 ADT. Error bars denote the standard error.

[34] Based on daily totals, values of radiation efficiency for each of the sites were somewhat lower (Figure 10). Of the burned sites, measurements at the 0, 4 and 12-year old burn scars have been selected because they had the greatest number of consecutive days of mostly continuous measurements. Compared to Figure 9b, the daily totals show a more consistent decline in net radiation with successional stand age. The large decrease in daily averaged radiation efficiency of the 12-year old burn is likely a result of the change in the diurnal albedo for that site (Figure 8b). The average reduction in the radiation efficiency of these three burn sites compared to the unburned sites is 5.5% (from 67.5 to 62%).

Figure 10.

Radiation efficiency for 0-, 4-, and 12-year-old burn scars and an average of the control sites based on daily totals. Values have been determined from 6 to 9 days of continuous measurements. Error bars denote the standard error.

[35] The Donnelly Flats fire removed much of the aboveground biomass and dramatically reduced surface roughness (Figure 11). Although a significant amount of the organic mat was also consumed, the remaining material was still a poor thermal conductor compared to the underlying mineral soil. When the surface is dry, a charred organic layer can sustain large near-surface below ground temperature gradients, even in regions underlain by permafrost. In addition, more than 90% of incoming solar radiation reached the ground surface, compared to 35% in the CPCRW north unburned stand. As a result of the increased insolation and reduced surface roughness, there was a dramatic increase in surface temperature, an increase in longwave radiation from the surface, and consequently a reduction in net radiation. As the cover of successional vegetation increased (12–14 years), promoting evapotranspiration and shading the surface, there was a reduction of surface temperatures and recovery of midday net radiation values (Figure 9b).

Figure 11.

A comparison of average daytime (local solar noon ±6 hours) U/u* (where U is the wind speed, and u* the friction velocity) between the Gerstle River unburned stand and the Donnelly Flats, for an unburned stand and three ages of burned stands. Error bars indicate standard error.

[36] The change in surface roughness also had a marked affect on the aboveground near-surface (<10 m) temperature profile. In the short-statured (2.5–3 m) moderate-density (13800 ha−1) control stand (CPCRW North), the average 0.3 m temperature (21.2 ± 0.4°C) was not significantly different from the “surface” temperature (21.1 ± 0.3°C; ground + canopy elements) calculated from the downward-facing long-wave radiometer data, using an emissivity value of 0.979. In contrast, spot measurements at the 0-year burn scar using a hand-held infrared thermometer were between 35° and 45°C for a period when the 0.3 m temperature was 27.5 ± 0.5°C. There is a similar contrast in temperature gradient between the burned and unburned stands between 0.3 and 1.5 m (Figure 12). Further from the surface where turbulent transport is more effective (between 1.5 m and sonic height), there is a much stronger temperature gradient over the rough forest surface than the freshly burned surface. Consequently, the increase in midday radiation efficiency that we observed from 7 to 12 years may be attributable to the development of a taller deciduous tree canopy, which partially shades the surface and increases the efficiency of surface-atmosphere coupling.

Figure 12.

Comparison of temperature profile above the surface for a moderate density black spruce canopy and a recently burned surface. Error bars denote standard error.

[37] Based on daily energy totals, the partitioning of net radiation into ground heat flux increased as a result of fire (Figure 13). The change is most evident in the early years before the development of a significant deciduous canopy. As previously indicated, it is possible that the reported values of G/Rn for the 0 and 4-year old burn sites have been underestimated based on the poor budget closure for these sites (Figure 7). Looking at shorter timescales, fire also increased the diurnal amplitude of ground heat flux (Figure 14). The near-noon “total” ground heat flux (G7cm + storage) in the early successional burned sites (16.3% of net radiation) was more than twice that in the unburned sites (7.9% of net radiation). Burning also increased the rate of nocturnal cooling due to the loss of biomass. At 12-years postfire, when a substantial canopy had developed, there was a pronounced decrease in both near-noon heat flux and nocturnal cooling rates.

Figure 13.

Comparison of G/Rn between unburned and recently burned stands derived from daily totals. Values have been derived from 6 to 9 days of observations, error bars denote standard error.

Figure 14.

Comparison of daily composite diurnal ground heat flux at 7–8 cm (not including near surface storage) immediately after fire, 12 years postfire and an unburned stand (>80 years postfire).

[38] Fire and postfire succession had a marked affect on the diurnal patterns of sensible and latent heat flux. Weather conditions during the periods represented by Figures 15a–15c were typically fine and dry. It should be noted that the data presented in Figures 15a and 18a are from a single unburned site (not an average over all the available unburned sites as in most other figures) chosen for meteorological conditions that most closely match those experienced for the corresponding 0 and 12-year old burn data.

Figure 15.

Comparison of the diurnal course of sensible and latent heat flux between 0- and 12-year-old burn scars and an unburned stand. Error bars denote standard error.

[39] The low evapotranspiration rate in the unburned site (Figure 15) reflected the low stomatal conductance that is typical of black spruce [Baldocchi et al., 2000]. Evapotranspiration declined in the 0-year burn, due to the absence of transpiring vegetation. By 12 years postfire, evapotranspiration had increased to values higher than in the unburned stand due to development of rapidly transpiring deciduous vegetation with a deeper root system. The average diurnal pattern of sensible and latent heat fluxes was similar among all sites, with maximum values near local noon (1400 ADT). In the unburned site during the day (Si > 50 W m−2), for example, net radiation correlated closely with both sensible heat flux (R2 = 0.93) and latent heat flux (R2 = 0.70). Sensible heat flux declined earlier in the day in the 12-year burn than at other sites due to the frequent development of cumulus clouds in the afternoon. In summary, variation in evapotranspiration was a major cause of changes in energy exchange during postfire succession.

[40] The relative partitioning of net radiation into sensible and latent heat also differed substantially among sites (Figures 16 and 17). Sensible heat flux accounted for more than half (H/Rn = 54.7 ± 2.6%) of the net radiation averaged over all unburned black spruce stands, whereas latent heat flux accounted for 32% of net radiation. This resulted in a high Bowen ratio (β = H/LE; β ≈ 2.7). The immediate effect of fire was to increase the relative partitioning of net radiation into sensible heat flux (Figure 16) at the expense of latent heat flux (Figure 17), causing an increase in the Bowen ratio (β ≈ 3.5). Within 4 years after fire, however, the relative partitioning of net radiation into sensible heat flux declined below levels in unburned stands, and the relative partitioning to latent heat flux increased to values that were somewhat higher than those in unburned stands, leading to a low Bowen ratio (β ≈ 1.3).

Figure 16.

Midday average relative partitioning of net radiation into sensible heat flux. The value for the unburned stand represents an average over all 5 unburned stands. Error bars denote standard error.

Figure 17.

Midday average relative partitioning of net radiation into latent heat flux. The value for the unburned stand represents an average over all 5 unburned stands. Error bars denote standard error.

[41] We analyzed midday values of LE/Rn as a function of soil moisture at the 0, 4 and 12-year burn sites as well as one of the unburned stands (Figure 18). Since a suitable calibration of the soil moisture probes in organic soils was not available we have arbitrarily defined three moisture classes (wet, moist and dry) according to the upper, middle and lower 33% of soil moisture readings over the observation period. The soil moisture measurements were made at a depth of 8 cm. At the unburned site this depth was within the fibric organic layer, and at the 4-year burn probes were located near the humic organic and mineral soil horizons. We observed a clear positive correlation between observed volumetric soil moisture and latent heat flux for the unburned site and 4-year burn. In an unburned canopy, the moss mat can contribute substantially to the stand evaporation, and the rooting depth of black spruce is fairly shallow. The rooting depth of the early successional vegetation in the 4-year burn was also fairly shallow, and as indicated by the site descriptions, there was still a significant organic layer at this site from which water could freely evaporate. At the 0- and 12-year burn sites the soil moisture probes were located beneath the remaining surface organic mat, several cm into the mineral soil. Consequently, these measurements would be less closely linked to “surface” moisture status. At the 0-year burn we observed no clear trend in latent heat flux with 8 cm soil moisture. At the 12-year burn with a fairly well developed deciduous canopy, there is a tendency for higher values of LE/Rn when the 8 cm soil moisture levels are relatively low. This result would be consistent with the deciduous canopy having ready access to a deeper moisture source despite relatively dry atmospheric conditions. Indeed, at this site (Granite Creek), the vapor pressure deficit at times when the 8 cm soil moisture was “dry” was twice as high as when the soil was considered “moist.”

Figure 18.

Midday values of LE/Rn as a function of relative soil moisture for 0-, 4-, and 12-year burn sites and one of the unburned stands. Error bars denote standard error.

6. Discussion

[42] Vegetation can strongly influence local and regional climate through modification of the radiation balance and surface energy budget [André et al., 1989; Bonan et al., 1992; Blyth et al., 1994; Pielke and Vidale, 1995; Bonan, 1997]. Consequently, extensive changes in terrestrial ecosystems have the potential to alter climate at global scales [Eltahir, 1996; McGuffie et al., 1998; Betts, 1999; Chase et al., 2000]. Contemporary dynamic vegetation models simulate changes in vegetation distribution, based on the relationship of vegetation to climate, and generally ignore disturbance [Kittel et al., 2000]. These models assume implicitly that climatically induced differences among ecosystems in climate feedbacks are more important than those induced by disturbance. Here we show that, for boreal forest, where fire is an important natural disturbance, disturbance causes changes in surface energy exchange that are large enough to be climatically important.

[43] Boreal forests are the second most extensive biome on Earth, and approximately 3.5 × 106 ha of North American boreal forest burns annually [Kasischke and Stocks, 2000]. A rate that is comparable to that of global tropical rain forest destruction [Eltahir, 1996]. The short growing seasons typical of these regions contribute to an extended successional phase of several decades such that at any given time a substantial proportion of the landscape would be in early to mid postfire succession.

[44] The average albedo that we measured in unburned Alaskan black spruce stands (0.088) was higher than that of black spruce stands in central Canada (0.083 [Betts and Ball, 1997]), probably reflecting the greater canopy height (>10 m) in the Canadian forests [Jarvis et al., 1997]. In addition, a greater proportion of the highly reflective ground cover may be visible through the sparse canopies typical of interior Alaska. The immediate effect of fire was to substantially reduce albedo, after which albedo increased within 4 years to values exceeding those in unburned stands. The stand-average albedo immediately after fire was strongly affected by the heterogeneous pattern of burning, which, in turn, was a function of local topography, vegetation, and soil moisture conditions.

[45] Mid-successional deciduous forests typically have a higher albedo than conifer forests [Betts and Ball, 1997], just as we observed. Compared to the 40- to 80-year duration of postfire succession [Viereck, 1973; Dyrness et al., 1986], the 5–10 years that it takes for albedo to exceed that of an unburned stand is relatively short. Consequently, the net effect of fire is to increase regional minimum albedo to at least 0.135 and accentuate the diurnal fluctuation of albedo. Any change in successional trajectory that prolongs the deciduous successional phase would further increase the regional albedo. A shift from a shrub-dominated deciduous phase to one dominated by taller, more long-lived deciduous trees such as aspen, for example, would prolong the period of high albedo (≤0.156 [Betts and Ball, 1997]) prior to black spruce regaining dominance.

[46] The surface-to-sky temperature difference also influences the radiation balance through its effect on longwave radiation balance (equation (1)). For fine-weather conditions during the growing season, midday sky temperatures are fairly consistent, so contrasts or changes in surface temperature account for changes in longwave radiation balance. Surface temperature, in turn, is strongly affected by the efficiency of coupling between the surface and atmosphere, stomatal conductance [Blanken et al., 1997; McGuffie et al., 1998], and the thermal characteristics of the surface [Oke, 1987]. Canopy characteristics dictate the efficiency of the surface-atmosphere coupling. Although the low albedo of black spruce forests leads to effective absorption of incoming shortwave radiation, the tapered coniferous crowns make the surface aerodynamically rough, and the cylindrical needles are very efficient at shedding heat to the atmosphere [Baldocchi et al., 2000]. Consequently, black spruce forests are thermally well coupled to the atmosphere and typically maintain modest surface-to-air temperature differences.

equation image

Here ε is the emissivity, α is the albedo, T absolute temperature (°K) and σ the Stefan-Boltzmann constant (5.67 × 10−8 W m−2 K−4).

[47] The high midday radiation efficiency (75%) and net radiation that we observed in unburned black spruce forests was a logical consequence of their low albedo (effective trapping of shortwave radiation) and low longwave emission (effective coupling to the atmosphere). Canopy architecture contributed to both the short- and long-wave radiation balances. Among our study sites, for example, radiation efficiency increased by 6% from short to tall canopies (Figure 9a). The radiation efficiency of the taller Canadian black spruce stands was even greater (78–80% [Kaminsky and Dubayah 1997]).

[48] The reduction in net radiation that we observed after fire in the Alaskan black spruce stands is similar to that observed following fire in a Canadian spruce forest [Rouse, 1976] where trees were felled and in burned tundra (S. Chambers, et al., Canopy controls of net radiation and energy partitioning following fire in Alaskan black spruce and tundra ecosystems, in preparation for submission to Agricultural and Forest Meteorology), although Amiro et al. [1999] reported no significant change in net radiation following fire where the taller burned stems remained standing. In our studies, the radiation balance was affected most strongly by elevated surface temperatures immediately after fire, due to the reduction in surface roughness and surface conductance, and by increased albedo in mid-succession. Midday radiation efficiency reached a minimum 7 years after fire. At this point the cover of successional vegetation was sufficient to yield a high surface albedo, but the developing canopy was too short (<1 m) and aerodynamically smooth to prevent high surface temperatures. However, based on daily averages radiation efficiency was greatly reduced in the older successional stands due to the combination of increased minimum albedo and greatly accentuated diurnal albedo variation.

[49] Alaskan unburned black spruce forests partitioned twice as much near-noon energy into ground heat flux (7.9% of Rn) as did Canadian black spruce stands [Jarvis et al., 1997; Pattey et al., 1997], because the short, sparse canopy of Alaskan forests allowed more radiation to penetrate to the ground surface, and permafrost created a stronger thermal gradient. The increase in ground heat flux that occurred immediately after fire resulted from several factors. The decreased albedo of the charred organic mat and reduced surface conductance after fire increased surface temperature and therefore the thermal gradient that drives ground heat flux. In addition, fire in Alaskan black spruce forests generally consumes more than half of the prefire organic mat [Kasischke et al., 1995], and increases the thermal conductivity of the remaining charred humic material threefold relative to that of the original live moss layer (data not shown). Together these changes in thermal properties reduce the resistance of heat flux into the soil. The lack of a well-developed moss and organic mat beneath the deciduous canopy [Van Cleve et al., 1991] favors ground heat flux in deciduous stands further contributing to permafrost retreat at the site. In summary, high daily averaged ground heat flux is typical early in postfire succession, but changes caused to the thermal characteristics of the surface will prevent the return of permafrost to the site for a longer period.

[50] Surface turbulent fluxes are the driving force for development of the planetary boundary layer (PBL) [Betts et al., 1996]. Over the diurnal cycle the amount of energy partitioned into turbulent fluxes (RnG), and in particular, the relative proportion of this total that is attributed to sensible heating of the atmosphere (H), dictate the extent and degree of mixing of the PBL. The reduction in net radiation at the surface (despite the reduction of albedo), and the increased ground heat flux following fire reduce the energy available to be partitioned into turbulent fluxes. Since the absolute sensible heat flux is the dominant factor in PBL development, the immediate effects of fire on PBL development are small. This is because the reduction in net radiation and simultaneous increase in H/Rn immediately postburn were of a similar magnitude (Figures 9b and 16), resulting in a negligible change in absolute sensible heating. However, as the successional vegetation becomes established, there is initially a further reduction in net radiation, followed by a reduction in Bowen ratio as a more substantial deciduous canopy develops and the stomatal conductance of the stand increases.

[51] Under clear-sky conditions the 9.3% decrease that we observed in midday radiation efficiency between the unburned black spruce stands and early successional stands corresponds to a reduction of roughly 70 W m−2 in net radiation at the surface. Based on daily totals the decrease is smaller (5.5%) corresponding to a change in net radiation of 42 W m−2. Together with the midday decrease in H/Rn from roughly 0.5 to 0.4, this corresponds to a reduction of more than 80 W m−2 in the sensible heat flux over early successional deciduous vegetation compared to the unburned stands. As the deciduous canopy becomes taller and more dense later in postfire succession, growing-season Bowen ratio (H/LE) values less than 1.0 could be expected [Blanken et al., 1997; Baldocchi et al., 2000; Eugster et al., 2000] yielding even larger differences in pre- to postfire sensible heat flux. This growing-season cooling of the PBL over burn scars is a combined effect of increased stomatal conductance, increased stand albedo, and decreased coupling between the surface and atmosphere, and agrees qualitatively with the boreal deforestation simulations of Bonan et al. [1992]. Fire should therefore act as a negative feedback to regional warming during the growing season [Chapin et al., 2000]. Fire should also contribute to regional cooling during winter due to the reduction in canopy and the increased albedo [Betts et al., 1997].

[52] More than 90% of the area burned annually in boreal forests is a result of individual fire scars that exceed 104 ha [Kasischke and Stocks, 2000]. These fire scars are large enough to modify the characteristics and extent of the planetary boundary layer and induce mesoscale circulation [Shuttleworth, 1988]. Similar changes in sensible heat flux, albedo and roughness over regions similar in size to Alaskan fire scars generate significant changes in-depth of the PBL and induce the development of mesoscale circulations [Anthes 1984; André et al., 1989; Pielke and Vidale 1995; Eltahir, 1996; Chambers, 1998]. If a mesoscale circulation were to develop across a fire scar boundary, subsidence over the fire scar where sensible heat fluxes are suppressed would result in surface flow toward the unburned forest. Given the high stomatal conductance of most successional species compared to black spruce, this scenario would result in a flow of relatively moist air toward a sharp gradient in roughness and sensible heating of the atmosphere. These are ideal conditions for the triggering of intense convective development. Lightning-induced fires account for 87% of the area burned in Alaskan boreal forests [Gotholdt, 1998]. In this way, the mesoscale circulations induced by large fires could perpetuate the importance of fire in the boreal forest, despite the net cooling effect of fire scars on regional climate.

7. Conclusions

[53] Changes in canopy architecture and albedo following fire in Alaskan boreal black spruce forests reduce the net radiation available at the surface. The reduction of above ground biomass and organic mat thickness increase the relative partitioning of net radiation into ground heat fluxes. The combination of these effects reduces the amount of energy partitioned into sensible and latent heat fluxes during the growing season compared to similar unburned regions. Later in succession the relative partitioning of the available energy (RnG) into turbulent fluxes favors latent heat flux with the establishment of an intermediate deciduous canopy. The subsequent reduction in sensible heat fluxes results in locally cooler conditions and restricted development of the planetary boundary layer. The fire-induced change in surface-atmosphere interaction and sizes typical of naturally occurring wildfires in Alaska are sufficiently large to generate mesoscale circulations that could perpetuate the fire-cycle. Due to the large area of boreal forest that burns annually and a prolonged successional process over decades, the fire-disturbed regions of Alaskan boreal forests are likely to contribute significantly to regional climate patterns. Given the importance of the boreal forest as a terrestrial biome, these effects may have global significance, although further studies that address the complete annual cycle and the mid-successional timeframe are required to verify this hypothesis.


[54] We thank Michelle Durant, Melissa Chapin, Stefan and Naoko Maurer, Malte Kreutzfeld, Mike Dang, Eric Huber, Phillip Johnson, Thomas Friborg, Jay Raymond, Kevin Davey, and Jim Randerson for their assistance in field measurements, Eric Kasischke, Kathy O'Neill, and Larry Hinzman for help in locating field sites, and the National Science Foundation for funding through the Bonanza Creek Long-Term Ecological Research program, and the FROSTFIRE experiment. We also thank Nancy French for providing the fire scar information and graphic, Jason Beringer for assistance with thermal conductivity measurements, Joe McFadden and Jim Randerson for their thought provoking discussions.