Remote Sensing of Environment 130 (2013) 171–181
Contents lists available at SciVerse ScienceDirect
Remote Sensing of Environment journal homepage: www.elsevier.com/locate/rse
Controls on variations in MODIS fire radiative power in Alaskan boreal forests: Implications for fire severity conditions Kirsten Barrett a,⁎, Eric S. Kasischke b a b
U.S.G.S. Alaska Science Center, 4230 University Drive, Anchorage, AK, United States Department of Geographical Sciences, University of Maryland, College Park, MD, United States
a r t i c l e
i n f o
Article history: Received 29 February 2012 Received in revised form 14 November 2012 Accepted 17 November 2012 Available online 21 December 2012 Keywords: Fire Boreal Fire radiative power Fire intensity Fire severity
a b s t r a c t Fire activity in the Alaskan boreal forest, though episodic at annual and intra-annual time scales, has experienced an increase over the last several decades. Increases in burned area and fire severity are not only releasing more carbon to the atmosphere, but likely shifting vegetation composition in the region towards greater deciduous dominance and a reduction in coniferous stands. While some recent studies have addressed qualitative differences between large and small fire years in the Alaskan boreal forest, the ecological effects of a greater proportion of burning occurring during large fire years and during late season fires have not yet been examined. Some characteristics of wildfires that can be detected remotely are related to fire severity and can provide new information on spatial and temporal patterns of burning. This analysis focused on boreal wildfire intensity (fire radiative power, or FRP) contained in the Moderate Resolution Imaging Spectroradiometer (MODIS) daily active fire product from 2003 to 2010. We found that differences in FRP resulted from seasonality and intra-annual variability in fire activity levels, vegetation composition, latitudinal variation, and fire spread behavior. Our studies determined two general categories of active fire detections: new detections associated with the spread of the fire front and residual pixels in areas that had already experienced front burning. Residual pixels had a lower average FRP than front pixels, but represented a high percentage of all pixels during periods of high fire activity (large fire years, late season burning, and seasonal periods of high fire activity). As a result, the FRP from periods of high fire activity was less intense than those from periods of low fire activity. Differences related to latitude were greater than expected, with higher latitudes burning later in the season and at a higher intensity than lower latitudes. Differences in vegetation type indicate that coniferous vegetation is the most fire prone, but deciduous vegetation is not particularly fire resistant, as the proportion of active fire detections in deciduous stands is roughly the same as the fraction of deciduous vegetation in the region. Qualitative differences between periods of high and low fire activity are likely to reflect important differences in fire severity. Large fire years are likely to be more severe, characterized by more late season fires and a greater proportion of residual burning. Given the potential for severe fires to effect changes in vegetation cover, the shift toward a greater proportion of area burning during large fire years may influence vegetation patterns in the region over the medium to long term. Published by Elsevier Inc.
1. Introduction Fire, in concert with other abiotic components such as climate, topography, soil, substrate, and permafrost condition, is an integral part of boreal forest ecosystems. Fire regime characteristics in the Alaskan boreal forest have responded to recent climate changes through an increase in area burned and average fire size resulting in a decrease in fire return interval over the last several decades (Kasischke & Turetsky, 2006; Kasischke et al., 2010). Black spruce stands, which account for 70% of the mature and regenerating forest cover in the region (source: 2001 NLCD), have experienced a marked increase in fire
⁎ Corresponding author. Tel.: +1 907 786 7419; fax: +1 907 786 7150. E-mail address:
[email protected] (K. Barrett). 0034-4257/$ – see front matter. Published by Elsevier Inc. http://dx.doi.org/10.1016/j.rse.2012.11.017
severity (i.e., the removal of surface organic material through combustion) over the last decade (Turetsky et al., 2011). Large fire years, defined as those years when > 1% of the boreal forest area burns, have increased in frequency, with the 2000s having the greatest number of large fire years in the period since large fires have been recorded on a yearly basis. Area burned in the Alaskan boreal forest responds both to short-term weather patterns and longer term variations in climate. Large fire years typically occur in years with high early season temperatures (Duffy et al., 2005) such as 2004, the largest fire year in 60 years of systematic data collection (Alaska Large Fire Database [ALFD], http://fire.ak.blm.gov/incinfo/ aklgfire.php). On a seasonal time scale, negative precipitation anomalies are linked to greater burned area (Abatzoglou & Kolden, 2011). The warm, dry conditions that cause large fire years or intra-annual periods of high fire activity are likely to exacerbate severity conditions
172
K. Barrett, E.S. Kasischke / Remote Sensing of Environment 130 (2013) 171–181
as well, and there are likely to be important qualitative differences between large and small fire years. Variation in fire severity is critical in the boreal forest because of the potential for high-severity fires to precipitate shifts in forest vegetation communities (Johnstone & Chapin, 2006; Johnstone & Kasischke, 2005), cause losses of permafrost (Yoshikawa et al., 2003), and possibly contribute to recruitment failure on previously forested land (Johnstone et al., 2010a; Kasischke et al., 2007). Low severity wildfires in the region generally favor self-replacement in coniferous and deciduous stands through feedbacks that reinforce thermal and moisture conditions in the surface organic layer (Johnstone et al., 2010b; Shenoy et al., 2011). High severity fires in the Alaskan boreal forest, however, reduce the surface organic layer in black spruce stands to a critical depth, releasing as much as 65 Tg C to the atmosphere in a single large-fire year (Kasischke & Hoy, 2012) and creating an opportunity for deciduous seedlings to proliferate and dominate post-fire succession (Johnstone & Chapin, 2006; Johnstone & Kasischke, 2005; Johnstone et al., 2010; Shenoy et al., 2011). The combination of increased fire frequency and severity is likely to lead to a decrease in average stand age (Turner et al., 1993), greater deciduous dominance in the region (Barrett et al., 2011), and possibly a loss of forested land cover. Active fire detections from thermal IR remote sensors permit researchers to examine some of the spatial and temporal patterns of wildfires that reflect differences in fire intensity, which in turn control wildfire emissions and severity. Daily fire registration data are available from the Advanced Very High Resolution Radiometer (AVHRR) since 1978. These data have been used to study temporal patterns of burning at regional to global scales (Pu et al., 2007; Stroppiana et al., 2000; Sukhinin et al., 2004). The AVHRR sensor records the presence of fires, but because of saturation at relatively low temperatures, data processing requires an algorithm to systematically remove false positives, and no information on fire intensity can be derived from this data source. Since the launch of the Moderate Resolution Imaging Spectroradiometer (MODIS) sensor aboard the Terra platform in December 1999, the global product of daily fire activity has recorded fire intensity, or Fire Radiative Power (FRP) (Giglio et al., 2003; Justice et al., 2002). The MODIS mid-infrared band is less susceptible to saturation than the similar band from AVHRR, but estimates of fire radiative power for MODIS active fire detections are available only since 2000. Comparisons of FRP values in North America and Eurasia for the first two years of MODIS operation revealed important differences between these two continental boreal fire regimes. Fires in Eurasia were of lower intensity than those in North America (Mottram et al., 2005; Wooster & Zhang, 2004), likely due to differences in species composition that favor a higher fraction of crown fires in North America (Wooster & Zhang, 2004). Integrating the observed FRP values (measured in MW) over time yields Fire Radiative Energy (FRE), or the total amount of energy emitted over a bounded spatio-temporal unit (Kumar et al., 2011; Wooster, 2002). FRE is closely related to emissions of trace gases and particulate matter during wildfire, and the MODIS active fire product has been used to infer total emissions from biomass burning in various ecosystems, including boreal forests, with promising results (Ellicott et al., 2009; Ichoku & Kaufman, 2005; Kaiser et al., 2012; Vermote et al., 2009). Here, we present the results from a study that analyzed the MODIS active fire data for the Alaskan boreal forest for the period of 2003 to 2010 to determine the spatial and temporal characteristics of the fire regime related to fire intensity. We focused on variations in fire activity over annual and intra-annual time scales, and spatial variations in burning based on vegetation type, latitudinal variation, and fire spread characteristics. 2. Hypotheses For this study, the MODIS active fire detections were characterized by intensity according to spatial and temporal factors that were thought to influence fire activity. Fig. 1 presents the general hypothetical
relationships between factors that control fire intensity information that is provided by the MODIS active fire product. Intensity of burning is expected to increase during periods of high fire activity and towards the later part of the season as seasonal thawing of frozen soil leads to better drainage and drier site conditions (Fig. 1A). Intensity also increases with crowning fires, which are more common in coniferous vegetation (Johnson, 1992) (Fig. 1B). The following hypotheses were developed to test the effect of fire seasonality, frequency, fuel type, latitude, and differences between front burning and residual burning on intensity. To elaborate on the findings, additional analysis of factor combinations and their impact on intensity were evaluated for significance. 2.1. Hypotheses related to fire seasonality
H1. Periods of high fire activity are more common in the later part of the fire season. Active fire detections in the boreal forest tend to cluster in periods of amplified activity that persist for several days. As much as half of all active fire detections occur in the 10 days of highest seasonal fire activity (K. Barrett, unpublished data). We hypothesized that these periods of high fire activity would be most common toward the end of the fire season, when fuels have been subjected to prolonged dry and warm conditions. H2. Late season fires are more intense than early season fires. Many ecosystems experience more intense fires during the later part of the fire season, when warmer temperatures and drier conditions may intensify (e.g. Cooke et al., 1996; Jones et al., 2009; Williams et al., 1999). In the boreal forest, the most severe burning in terms of ground-layer biomass consumed by burning occurs during large fire years in the late season (Kasischke & Hoy, 2012; Turetsky et al., 2011) when the ground has thawed and surface organic materials become desiccated and vulnerable to deep burning. We anticipated that intensity would similarly increase toward the end of the season, consistent with results from studies of other fire-affected ecosystems. H3. The start of the fire season is later at higher latitudes. The start of the fire season is associated with the beginning of the snow-free period in the boreal forest. While there is snow on the ground there is little exposed surface organic material that aids the spread of fire, particularly in areas of low tree density (Viereck, 1983). Because the onset of the snow-free period occurs later at higher latitudes, we anticipated that the beginning of the fire season would be similarly delayed. The end of the fire season is typically associated with precipitation events that do not show a latitudinal gradient, so we anticipated that the end of the season would not vary as a function of latitude. 2.2. Hypotheses related to spatial patterns of burning
H4. Residual burning is lower in intensity than front burning. The active fires detected by satellite-borne thermal radiometers can be divided into two categories (Fig. 2). The first category (front burning) applies to those detections associated with the outermost edge of the fire perimeter, which expands over time as the fire event size increases. The second category of detections (residual burning) occurs within the perimeter of burned area, typically located >1 km from the edge of the fire event. Differences between front burning and residual burning reflect fuels and burning conditions. We hypothesize that front burning is composed largely of flaming components that consume canopy
K. Barrett, E.S. Kasischke / Remote Sensing of Environment 130 (2013) 171–181
173
Fig. 1. Hypothetical relationships between temporal (A) and spatial (B) characteristics of burning and intensity.
vegetation and other above ground biomass (French et al., 2003; Lobert & Warnatz, 1993). After a front moves through a stand, residual burning may continue as smoldering combustion with intermittent flaming of unburned islands of vegetation within the fire, and consumption of surface organic materials and coarse woody debris. H5. Burning in coniferous vegetation (primarily black spruce) produces higher intensity than other vegetation types in the region (such as deciduous vegetation, mixed deciduous and coniferous stands, and shrubs). Black spruce trees are adapted to fire disturbance, characterized by semi-serotinous cones and low-level branches that serve as “ladder fuels” to promote fire (Johnson, 1992; Viereck, 1983). Low canopy moisture and the high frequency of crowning in black spruce stands mean that fires in these stands are likely to be more intense than in deciduous vegetation. The intensity of burning in shrub vegetation is likely constrained by fuels, i.e., less biomass and a shallower organic layer than is found in coniferous stands. 3. Data and methods Variations in fire intensity were determined using the MODIS daily active fire detection data product (MCD14ML, Collection 5, scene h12v02). The data were downloaded from the Land Processes Distributed Active Archive Center (LP-DAAC) using the EOS Data Gateway (http://wist.echo.nasa.gov/) on April 28, 2011. The combined dataset included data from both Terra and Aqua satellites, and spans the period from 2001 to 2010. Data were analyzed during the period for which data are available for both platforms (2003 and later). The study area was defined by the extent of the boreal ecoregion delineated by Nowacki et al. (2001). The active fire detection product is likely a conservative estimate of the actual amount of burning in the study region because of omissions due to cloud cover, thick smoke, and surface fires of lower intensity that escape detection (Sukhinin et al., 2004). Because we use the data only to compare relative intensity between different types of burning, the absolute amount of burning is not a factor in the analysis. Each active fire detection was categorized by vegetation type, fire seasonality, high versus low latitude, fire type (front or residual burning), and level of intra- and inter-annual fire activity across the region. Vegetation cover was derived from the 2001 National Land Cover Database (Homer et al., 2004) with expanded wetland categories. Fires were characterized as early or late season using a threshold of Julian day 200 (July 19) to be consistent with previous studies (Barrett et al., 2010,
2011). The start of the fire season for each year was defined as the average of the three earliest days when active fires were recorded. The threshold for high latitude detections was 65.5°N, which accounted for 30% of the boreal area of Alaska being studied and 55% of MODIS active fire detections in the region. The threshold was determined based on the histogram of active fire detections, which show a bi-modal distribution with peaks at lower and higher latitudes with an apparent break at 65.5°N (Fig. 3). To determine whether or not an active fire detection was part of a fire front, the data were processed daily over the entire fire season in a given year. First, the active fire detections on the first day of each year of burning were used to create a polygon (Fig. 2A). For the following day, active fire detections within the polygon were considered to be residual burning, and detections outside the polygon were considered front burning (Fig. 2B). This process was repeated iteratively until all active fire detections were categorized. At the scale of the MODIS pixel (1 km2), the type of burning is unlikely to be homogeneous (Wooster et al., 2005). The pixels classified as “front burning” according to this analysis approach are those that likely contain the fire front, and may contain residual burning as well. The level of intra-annual regional fire activity was quantified using a daily running count of all detections in the study area. The relative level of fire activity was characterized for each day by the number of active fire detections within five days prior to five days later over the entire study region. The threshold for high frequency burning periods was 10,000 active fire detections within a 10 day period of burning. Large fire years were those during which more than 1% of the region (5500 km 2) burned, according to the Alaska Large Fire Database, maintained by the Bureau of Land Management Alaska Fire Service (Kasischke et al., 2002). Large fire years occurred during the study period in 2004, 2005, and 2009. Fire intensity was based on the Fire Radiative Power (FRP) associated with the active fire registration. FRP is calculated from the MODIS mid-infrared band 21 (3.93–3.99 μm) by comparing individual pixels with surrounding observations (i.e., background) to detect thermal anomalies (Kaufman et al., 1998). FRP values in this study followed a power law distribution owing to the tendency for values to be generally low, with a long ‘tail’ of registrations of higher intensity. FRP values were log transformed to produce a normal distribution to allow comparison of means. Although signals from the Terra-based sensor have shown signs of faster signal degradation than the signal from the Aqua-based sensor (Wang et al., 2012), the fire detections were not analyzed separately because there was no significant difference in the proportion of detections from Terra versus Aqua for any subgroupings used in the analysis. Giglio et al. (2006) found a high level of agreement between Aqua and Terra
174
K. Barrett, E.S. Kasischke / Remote Sensing of Environment 130 (2013) 171–181
Fig. 2. The process for differentiating front burning and residual burning. In (a), the active fire detections on the first day of burning are used to create a polygon. The next day (b), active fire detections within the polygon are considered to be residual burning (triangles) and detections outside the polygon are considered front burning (grey circles).
data in terms of seasonality and length of season, and that FRP from the Aqua sensor is about 10% higher than Terra, probably because the overpass occurs closer to the time of peak burning. There were differences in the diel distribution (i.e., time of day) that fires were registered between large and small fire years. This is perhaps owing to the presence of clouds in the afternoon, particularly during cooler, wetter small fire years, that obscure active fires from detection. Because fires show a diel cycle of intensity (Ichoku et al., 2008; Kumar et al., 2011), such a discrepancy may bias results. To address this issue we resampled the active fire detections (with replacement) from small fire years using the daily distribution of active fire registrations during the same Julian day across all large fire years. The resampling was done by calculating the fraction of active fire registrations during each time period for a given day across all large fire years. The observations that occurred on the same Julian day in a small fire year were then resampled to match the distribution for large fire years. If there were insufficient observations during a given time period to reproduce the pattern observed
in large fire years, single observations could be used more than once. While there may in fact be differences in the diel patterns of wildfire activity between large and small fire years, this seemed the most effective way to remove known biases in the data. The data were then used to test the hypotheses related to fire seasonality, intensity, fuel type, and differences between front burning and residual burning. Differences in fire intensity were determined by comparing the distribution of log transformed FRP values, and significance was assessed using the Wilcoxon rank sum test and chi-squared test. 4. Results The differences in intensity based on temporal and spatial variation in active fire detections are displayed in Fig. 4. The detections were grouped into 50 MW bins and the proportion of each bin of the total number of detections is compared between the categories used in the
K. Barrett, E.S. Kasischke / Remote Sensing of Environment 130 (2013) 171–181
Fig. 3. Histogram of the latitude of all active fire detections.
analysis. The data showed a negative exponential distribution, with very few observations having an FRP of 100 or higher. The FRP values were log transformed to compare means and assess between-group differences using the Wilcoxon rank sum test. The back-transformed mean FRP values are shown for each category of burning in Fig. 5. The effect of changing buffer size in discriminating between continued and residual burning is shown in Fig. 4 (panels G and H) and Table 1. A smaller buffer causes greater differences between the intensity of front burning and continued burning, while a bigger buffer likely leads to more front
175
burning being classified as continued burning, and the differences in intensity are lower (though still significant). There were significant differences in FRP (pb 0.001) between large and small fire years, early and late season fires, residual versus front burning and periods of high and low fire activity. Differences in FRP among vegetation types were also significant (pb 0.001), with coniferous vegetation burning most intensely and shrubs (not deciduous forest, as hypothesized) burning with the least intensity. The differences in intensity based on location (vegetation type, front versus residual burning) were generally consistent with our hypotheses. Differences based on temporal factors (seasonality, intra- and inter-annual fire activity) conflicted with our hypotheses that late season fires and high fire activity periods would be more intense. The differences between the different categories of burning were generally small (average delta= 6.6 MW or 10%). There is likely some temporal and spatial autocorrelation between samples that renders the effective sample size smaller than the total number of observations. Periods of high fire activity (10 day total active fire detections > 10,000) occurred within a fairly brief time frame between Julian day 220 (August 8) to 240 (August 28), and only during large fire years (Fig. 6). In contrast, periods of low fire activity (10 day total active fire detections b 10,000) occur throughout the entire fire season. About 35% of active fire detections occurred during periods of seasonally high fire activity. The highest fire activity seen during a small fire year was 1587 registrations within a 10 day period. Over the fire season, average Julian date of burning was four days later in higher latitudes (Table 2). Latitudinal differences in date of burn are more pronounced in small fire years, when average Julian date of burning is 27 days later in high latitudes. During large fire years average Julian date of burning is only three days later. There was no significant difference in the earliest active fire detection between large and small fire
Fig. 4. Differences in intensity based on (A) annual variability in area burned, (B) fire seasonality, (C) intra-seasonal variation, (D) vegetation type, (E) fire spread, and (F) latitudinal zone. Panels G and H show FRP for front and residual burning based on 1 km and 3 km thresholds, respectively, for comparison with panel E which shows a 2 km threshold (as used in the analysis).
176
K. Barrett, E.S. Kasischke / Remote Sensing of Environment 130 (2013) 171–181
Intensity of Fire by Category
Fire Radiative Power (MW)
80 70 60 50 40 30 20 10 0
Large Small Fire Year Fire Year
Early Season
Late Season
Residual Burning
Front Burning
Low Latitude
High Latitude
Fig. 5. Comparison of back transformed (from log) mean values of FRP for each active fire detection category. Error bars are one sigma.
years or between higher and lower latitudes. Year to year variation in the earliest active fire detection was high, and in two years the first active fire detection occurred earlier in the northern half of the study area. Higher latitudes experience higher intensity burning (Figs. 4 and 5) although they have a greater proportion of residual burning (47% versus 39%). Although the proportion of residual and continued burning was not the subject of any of our hypotheses, we investigated these patterns to help resolve some of the unanticipated findings reported above. We found that the fraction of burning that continues after a front has moved through an area increases with the total number of active fire detections (Fig. 7). Large fire years have a greater fraction of residual burning, particularly in the latter half of the fire season (Fig. 8). Another interesting difference between large and small fire years was the seasonality of burning. In large fire years, about 60% of the active fire detections occurred in the latter half of the fire season, while in small fire years, the majority of detections occur in the earlier half. Table 3 shows the differences in the frequency of four vegetation types for the spatial and temporal factors used in the analysis. The four categories: conifer, deciduous, mixed (i.e., neither conifer nor deciduous vegetation account for > 75% of the area), and shrub account for 95% of all active fire detections in the region. There are significant differences (chi-squared test) in the vegetation composition across all factors (annual and intra-annual variation, seasonality, front versus residual burning, and latitude), but the differences are minor (generally ≤0.03%). According to the analysis, mature conifer vegetation is more prevalent in the active fire detections (56% of detections) than in the region (43% of land cover), indicating that these stands are more fire prone than other vegetation types (a result consistent with Kasischke et al., 2010). Shrub vegetation is apparently less susceptible to burning, comprising about 40% of the land cover, but only 28% of the active fire detections.
5. Discussion In general the data contradicted the hypotheses related to the relationship between temporal factors and intensity (H1 and H2), but were consistent with those hypotheses related to the influence of spatial pattern and burning (H4 and H5). The data did not support the hypothesis regarding a later start of the fire season in higher latitudes, but average burn date is somewhat later for higher latitudes (H3). 5.1. Annual and intra-annual variation The spatial and temporal patterns of active fire detections in the region indicate qualitative differences between periods of high and low fire activity for the nine years used in this study. Fig. 8 shows that the seasonal pattern of burning is different between large and small fire years. In small fire years there is initially a high rate of spread, reflected in the steepness of the curve at the beginning of the season. Burning levels off around Julian day 230 (August 18) and the curve flattens. Cumulative fire activity during large fire years follows a sigmoidal pattern, starting out at a relatively low rate of spread, and increasing at a
Table 1 Comparison of back transformed (from log) mean values of FRP (MW) for residual and front hot pixel detections using thresholds of 1 km, 2 km (used in the analysis), and 3 km buffers to distinguish fronts from continued burning. Uncertainty estimates are one sigma. Buffer distance
Residual FRP
Front FRP
1 km 2 km 3 km
56.3 ± 2.2 59.71 ± 2.4 60.72 ± 2.4
64.22 ± 2.6 66.39 ± 2.7 66.66 ± 2.8
Fig. 6. Distribution of mean date of burn of periods of low fire activity and high fire activity. While periods of low fire activity occur throughout the growing season, high fire activity only occurred during the late season in large fire years.
K. Barrett, E.S. Kasischke / Remote Sensing of Environment 130 (2013) 171–181 Table 2 Differences in mean burn date and fraction of residual burning between latitudinal zones (threshold = 65.5 N).
Mean Julian date • Large fire years • Small fires years Fraction of residual burning
Low latitudes
High latitudes
(n = 50,699)
(n = 62,542)
207 210 171 0.39
211 213 198 0.47
progressively faster rate. The curve during large fire years does not plateau until around Julian day 240 (August 28). Periods of high fire activity tend to occur later in the season, a result consistent with previous studies of wildfires in North American (Kasischke & Turetsky, 2006; Kasischke et al., 2010) and Eurasian boreal forests (Sukhinin et al., 2004). Late season burning is characterized by a greater proportion of residual burning and subsequently lower fire intensity. Late season burning and a longer “residence time” mean that periods of high fire activity are likely to consume more surface organic materials and to have a more severe impact on ecosystem recovery. Periods of low fire activity (i.e., small fire years and seasonal periods of low fire activity) are dominated by front burning and high fire intensity. Small fire years account for just 10% of the active fire detections in the region, a figure consistent with area burned as determined from the Alaska Large Fire Database (Kasischke et al., 2002). Seasonal periods of low fire activity, however, account for the majority of active fire detections (56% in large fire years, 67% in small fire years). This is because there is no period of high fire activity (10 day active fire detection n>10,000) that occurs in small fire years, and periods of relatively low fire activity can occur throughout the fire season during large fire years (Fig. 6). Seasonal or inter-annual periods of low fire activity appear to be composed mainly of flaming fronts that move relatively quickly through an area. It is unlikely that these flaming fronts burn deeper organic layers because duff layers have higher fuel moisture and are too dense to allow sufficient oxygen to sustain flaming combustion, (Kasischke et al., 2005). Severe fires are less likely during small fire years because small fire years are dominated by early season fires and front activity. According to meteorological data from Remote Automated Weather Stations (RAWS) in the region, late season increases in precipitation are greater in small fire years, which are responsible for the reduction in the duration of fire activity (Abatzoglou & Kolden, 2011).
The moderately higher intensity of seasonal or inter-annual periods of low fire activity can be explained in part by more front burning during these periods. The hypothetical positive relationship between late season fires, high fire activity, and high fire intensity was not supported. To understand the temporal patterns of fire severity, it is necessary to include a consideration of spatial variability and the nature of the fire spread behavior. Fig. 9 shows the revised conceptual relationships among residual burning, fire activity, and intensity, wherein the intensity of seasonal or inter-annual high fire activity is moderated by an increased fraction of residual burning. Even considered in isolation, however, front burning is more intense during periods of low fire activity by about 8 MW or 11% compared to high activity periods. It is possible that this is due to misclassification caused by different rates of spread in periods of high and low fire activity. We used a threshold of 2 km from previous detections to separate front burning from residual burning, but this method assumes a consistent rate of spread. If a fire moves quickly through an area, some active fire detections that are actually residual burning may be included in the “front” burning because they are far enough away (i.e., >2 km) from previous detections. This type of misclassification would reduce the estimated intensity of front burning during periods of high fire activity. 5.2. Front burning versus residual burning The difference in intensity between residual and front burning is important with respect to fire severity and emissions. The residual burning that occurs more frequently during periods of high fire activity, particularly toward the end of the fire season, is likely responsible for consuming more of the surface organic layer than front burning. Surface fires are typically smoldering fires (Johnson, 1992), and although they are less intense than crown fires, they are responsible for most of the emissions from boreal wildfires in Alaska (Kasischke & Hoy, 2012). Overall emissions that occur during the residual burning period are therefore likely to be higher than during the front burning period, even if the instantaneous amount of energy released by the fire (FRP) is lower. It is also likely that smoldering ground fires more easily escape detection through the MODIS algorithm used to detect active fires (Kasischke & Bruhwiler, 2003), particularly in areas of thicker overstory vegetation, heavy smoke or cloud cover (Giglio et al., 2009). While CO2 emissions from flaming combustion are about 20% higher than smoldering, CO and CH4 are two to three times higher from smoldering combustion (French et al., 2003). The underestimation of residual burning, and the lower intensity of these fires are likely to bias estimates of emissions based on Fire Radiative Energy and may result in substantial underestimation of emissions. 5.3. Latitudinal variation
0.8
R²=0.7781
0.7
Fraction of Residual Burning
177
0.6 0.5 0.4 0.3 0.2 0.1 0 2
4
6
8
10
12
14
16
18
20
Frequency of Active Fire Detections/1000 Fig. 7. The relationship between continued burning and total number of active fire detections. Ten-day periods were binned in increments of 2000 detections. As more burning is detected, the fraction of residual burning for that 10-day period increases.
While other studies have found a significant effect of latitude on FRP (e.g., Mottram et al., 2005), we did not expect to see a dramatic effect in Alaska because the region spans a small latitudinal gradient (~6.5°). At a global scale, fires at higher latitudes are sampled more often because of the greater overlap between MODIS scans at the poles. Globally, MODIS active fire registrations tend to decrease with latitude, while FRP tends to increase because high latitude fires are more episodic and characterized by greater fuel loads (Giglio et al., 2006). Previous studies with a greater latitudinal range in the boreal forest have observed that the number of detections decreases with increasing latitude (Mottram et al., 2005). Over the time period used in our study region, the density of active fire detections over the period from 2003 to 2010 was greater at higher latitudes (0.44 detections per km2) than lower latitudes (0.16 detections per km2). The lower latitude portion of the study area is characterized by a higher mean elevation, which is presumably more likely to reduce the effect of latitudinal differences because of lower surface and soil temperatures at higher altitudes. Given the small latitudinal range and the mitigating factor of elevation, the lack of a significant difference in the
178
K. Barrett, E.S. Kasischke / Remote Sensing of Environment 130 (2013) 171–181
Fig. 8. The cumulative amount of front burning (solid line) and residual burning (dashed line) for small (A) and large (B) fire years, 2003–2010. Residual burning is defined as burning that continues after a front has moved through a region, and likely contains a greater area of surface fires than front burning.
start of the fire season might have been anticipated. It was therefore somewhat surprising to see significant differences between high and low latitude areas in terms of intensity and proportion of residual burning. Higher latitudes are characterized by higher intensity and greater proportion of residual burning, a likely indication that fires in these areas are more severe than fires at lower latitudes. Severe fires in higher latitude boreal forests may act as a control on the ability of tree populations such as black spruce to extend their northern border in response to climate change (Brown & Johnstone, 2012). It is possible that regional patterns of precipitation may influence fire activity. Based on an analysis of RAWS data, higher latitudes in the region receive less precipitation than lower latitudes, a difference that is amplified during large fire years. This may have contributed to the earlier start of the fire season at high latitudes during large fire years.
in the region with increases in fire severity. Some (e.g., Chapin et al., 2000; McGuire et al., 2006) have suggested that more deciduous vegetation represents a possible negative feedback on fire occurrence, but such a feedback has not been shown to counteract increased fire activity in the region (Johnstone et al., 2011). The results of this analysis suggest that deciduous vegetation is not particularly resistant to burning, although it is less fire prone than coniferous vegetation. The shrub land cover includes areas that are early succession stands and the resistance of these areas to fire may indicate a biomass limitation on burning. Given that fire return intervals in the region are decreasing (Kasischke et al., 2010), the proportion of the landscape occupied by early succession stands is likely to increase. Vegetation shifts from increased fire activity may therefore constrain fire activity through reduced biomass.
5.5. Implications for post-fire vegetation shifts 5.4. Vegetation composition The hypothesized relationship among vegetation type, crowning, and intensity appears to be correct according to the results of this analysis. While we did not study crowning events directly, the active fires classified as conifer fuels are those most likely to contain flaming fires, which are the highest intensity (mean FRP of conifer vegetation=66.1, compared to deciduous=59.9, mixed=61.5, and shrub=60.26). There were minor, though significant differences in vegetation cover between different types of burning (Table 3). In some cases a greater proportion of conifer fuel and reduced shrub fuel may have led to more intense burning and subsequently higher FRP. A previous study of North American wildfires found that deciduous and mixed stands are resistant to burning (Pu et al., 2007), a finding inconsistent with our analysis. This discrepancy suggests that the findings reported here pertain to the study region and cannot be extrapolated to the continental scale without additional analysis. Projected changes in vegetation may influence future forest fire regimes through variation in fuel type and availability (Krawchuk et al., 2006). It is likely that deciduous vegetation will become more dominant
The study of fire fronts and residual burning highlights the potential for shifts in vegetation communities during post-fire succession. Fire fronts are defined as being characterized by flaming combustion (National Wildfire Coordinating Group Glossary of Wildland Fire Terminology [www.nwcg.gov/pms/pubs/glossary], Glossary of Fire Science Terminology [www.firewords.net]). It is reasonable, therefore, to assume that the front burning active fire detections represent more flaming combustion, because this is how fire spreads to new areas. Smoldering is the phase of burning that follows flaming, and although we cannot know exactly when an area transitions from flaming to smoldering, it seems likely that the flaming phase does not continue for an extended time. This is because the majority of the fuels in boreal forest fires are from surface organic material (Kasischke & Hoy, 2012), which tends to burn more through smoldering combustion (Kasischke et al., 2005). Areas of residual burning are associated with smoldering combustion, and are more likely to burn deeply enough to precipitate a shift toward greater deciduous dominance in the boreal forest. Periods of low fire activity show comparatively low residual burning, and therefore are likely to experience less organic layer depth reduction.
K. Barrett, E.S. Kasischke / Remote Sensing of Environment 130 (2013) 171–181
179
Table 3 Number of fire hot spots detected as a function of vegetation type for the various categories of burning examined in the analysis. The differences for each factor are statistically significant, although minimal. The numbers in parentheses represent the fraction of pixels in a specific category. Conifer forest (unburned fraction = 0.43)
Deciduous forest (unburned fraction = 0.09)
Mixed forest (unburned fraction = 0.09)
Shrub forest (unburned fraction = 0.39)
Annual variability (area burned threshold = 5500 km2) χ2 = 191.32, p b 0.00001 Large fire year 54227 (0.56) Small fire year 6537 (0.59)
7783 (0.08) 1115 (0.10)
7575 (0.08) 908 (0.08)
27227 (0.28) 2467 (0.22)
Intra-annual variability (10 day fire frequency threshold = 10 K) High fire activity 21643 (0.56) χ 2 = 143.25, p b 0.00001 Low fire activity 39121 (0.56)
3284 (0.09) 5614 (0.08)
3485 (0.09) 4998 (0.07)
10144 (0.26) 19550 (0.28)
Seasonality (DOY threshold = 200) χ 2 =113.80, p b 0.00001 Late season Early season
5610 (0.08) 3241 (0.09)
5802 (0.08) 2626 (0.08)
20277 (0.28) 9117 (0.26)
3962 (0.09) 4936 (0.08)
3754 (0.08) 4729 (0.08)
12188 (0.26) 17506 (0.29)
1981 (0.04) 5654 (0.12)
2663 (0.06) 3576 (0.10)
14636 (0.31) 11030 (0.25)
40296 (0.56) 19966 (0.57)
Front vs. residual burning (front threshold > 2 km from previous detections) χ 2 = 88.60, p b 0.00001 Residual 26838 (0.57) Front 33926 (0.56) Latitude (threshold = 65.5°N) χ 2 = 2742.536, p b 0.00001
Low latitude High latitude
28606 (0.60) 25562 (0.54)
Given the strong control of post-fire organic layer depth on post-fire regeneration (Johnstone & Chapin, 2006; Johnstone & Kasischke, 2005), the areas that burn less severely during periods of low fire activity may represent a kind of check on the ability of the boreal system to experience a shift in vegetation composition caused by deep-burning fires. The majority of active fire detections occur during periods of low fire frequency, which represents a large area that is less vulnerable to severe burning and subsequent shifts in post-fire regeneration. Periods of low fire activity (including small fire years), then, are likely to bolster the resilience of the boreal ecosystem by reducing the area available for severe burning during periods of high fire activity. The proportion of North American boreal forest that burns during large fire years has increased over the last several decades (Kasischke & Turetsky, 2006), and there has been a marked shift toward late season burning in Alaska over the last decade (Kasischke
et al., 2010). Whether there has been a similar shift in the proportion of area that burns during periods of high fire activity is unknown. The reduction of conifer dominated stands in severely burned areas would influence the net ecosystem carbon balance in the region directly through combustion and indirectly through increased soil respiration. The increase in above ground biomass resulting from a greater fraction of deciduous stands may not offset these losses. Long-term fire regime and vegetation dynamics are complex (Johnstone et al., 2011), but the increase in albedo from more deciduous vegetation may effectively counteract the radiative forcing from increased carbon emissions (Randerson et al., 2006). The dynamics of a system de-stabilized through a changing disturbance regime are difficult to predict (Buma & Wessman, 2011; Fraterrigo & Rusak, 2008; Turner et al., 1993), and the ultimate effect of increased fire disturbance in the region requires more concerted study (Chapin et al., 2010).
6. Conclusions
Fig. 9. A re-conceptualization of the relationships among seasonal or inter-annual high fire activity, residual burning, and fire intensity.
Based on initial and revised hypotheses, the spatial and temporal factors that control fire intensity are inextricably linked, and the relationship between these influences is important to understand the dynamics of boreal wildfire activity that are likely to affect post-fire vegetation. Seasonal or inter-annual periods of high fire activity in the Alaskan boreal forest are characterized by a higher proportion of residual burning, which are more likely to lead to a shift toward higher deciduous dominance in stands dominated by spruce. This finding is significant because inter-annual periods of high fire activity are increasing, though data for seasonal fire activity are lacking. The proportion of areas that burn during small fire years has decreased in the most recent decade to 19% from 32% during the preceding 50 years (source: Alaska Large Fire Database [fire.ak.blm.gov/incinfo/ aklgfire.php]). It is possible that older fire scars, mapped from sketches of observed fire affected areas, are less accurate than areas mapped more recently using remotely sensed data. However, even when the period is restricted to more recent burns (1980 to 2000) the region experienced a greater fraction of area burned during small fire years (36%) than the most recent decade. The reduction in areas that burn in small fire years leaves a greater area available to be burned severely. As the remote sensing data record grows, our ability to characterize long-term trends in ecosystem dynamics and detect real shifts in structure and function improves. Future work related to the study of
180
K. Barrett, E.S. Kasischke / Remote Sensing of Environment 130 (2013) 171–181
intra-annual patterns of boreal wildfires will include a descriptive model of periods of elevated fire activity. The explicit inclusion of meteorological data, in combination with the information used in this analysis, is likely to yield insights into the factors that control fire intensity. Acknowledgments Funding for this research was provided to KB by the U.S. Geological Survey and to EK by NASA through grant number NNG04GD25G. We also thank A. D. McGuire, Zachary Christman, David Selkowitz, and three anonymous reviewers for helpful suggestions and comments on the manuscript. References Abatzoglou, J. T., & Kolden, C. A. (2011). Relative importance of weather and climate on wildfire growth in interior Alaska. International Journal of Wildland Fire, 20, 479–486. Barrett, K., Kasischke, E. S., McGuire, A. D., Turetsky, M. R., & Kane, E. S. (2010). Modeling fire severity in black spruce stands in the Alaskan boreal forest using spectral and non- spectral geospatial data. Remote Sensing of Environment, 114, 1494–1503. http://dx.doi.org/10.1016/j.rse.2010.02.001. Barrett, K., McGuire, A. D., Hoy, E. E., & Kasischke, E. S. (2011). Potential shifts in dominant forest cover in interior Alaska driven by variations in fire severity. Ecological Applications, 21, 2380–2396. Brown, C. D., & Johnstone, J. F. (2012). Once burned, twice shy: Repeat fires reduce seed availability and alter substrate constraints on Picea mariana regeneration. Forest Ecology and Management, 266, 34–41. Buma, B., & Wessman, C. A. (2011). Disturbance interactions can impact resilience mechanisms of forests. Ecosphere, 2(5). http://dx.doi.org/10.1890/ES11-00038.1. Chapin, F. S., McGuire, A. D., Randerson, J., Pielke, R., Sr., Baldocchi, D., Hobbie, S. E., et al. (2000). Arctic and boreal ecosystems of western North America as components of the climate system. Global Change Biology, 6, 211–223. Chapin, F. S., McGuire, A. D., Ruess, R. W., Hollingsworth, T. N., Mack, M. C., Johnstone, J. F., et al. (2010). Resilience of Alaska's boreal forest to climatic change. Canadian Journal of Forest Research, 40, 1360–1370. http://dx.doi.org/10.1139/X10-074. Cooke, W. F., Kotti, B., & Gregoire, J. -M. (1996). Seasonality of vegetation fires in Africa from remote sensing data and application to a global chemistry model. Journal of Geophysical Research, 101, 21,051–21,065. Duffy, P. A., Walsh, J. E., Graham, J. M., Mann, D. H., & Rupp, T. S. (2005). Impacts of large-scale atmospheric-ocean variability on Alaskan fire season severity. Ecological Applications, 15, 1317–1330. http://dx.doi.org/10.1890/04-0739. Ellicott, E., Vermote, E., Giglio, L., & Roberts, G. (2009). Estimating biomass consumed from fire using MODIS FRE. Geophysical Research Letters, 36. http: //dx.doi.org/10.1029/2009GL038581. Fraterrigo, J. M., & Rusak, J. A. (2008). Disturbance-driven changes in the variability of ecological patterns and processes. Ecology Letters, 11, 756–770. http://dx.doi.org/ 10.1111/j.1461-0248.2008.01191.x. French, N. H. F., Kasischke, E. S., & Williams, D. G. (2003). Variability in the emission of carbon-based trace gases from wildfire in the Alaskan boreal forest. Journal of Geophysical Research, 108(D1). http://dx.doi.org/10.1029/2001JD000480. Giglio, L., Csiszar, I., & Justice, C. O. (2006). Global distribution and seasonality of active fires as observed with the Terra and Aqua Moderate Resolution Imaging Spectroradiometer (MODIS) sensors. Journal of Geophysical Research, 111, G02016. http://dx.doi.org/ 10.1029/2005JG000142. Giglio, L., Descloitres, J., Justice, C. O., & Kaufman, Y. J. (2003). An enhanced contextual fire detection algorithm for MODIS. Remote Sensing of Environment, 87, 273–282. http://dx.doi.org/10.1016/S0034-4257(03)00184-6. Giglio, L., Loboda, T., Roy, D. P., Quayle, B., & Justice, C. O. (2009). An active-fire based burned area mapping algorithm for the MODIS sensor. Remote Sensing of Environment, 113, 408–420. Homer, C., Huang, C. Q., Yang, L. M., Wylie, B., & Coan, M. (2004). Development of a 2001 national land-cover database for the United States. Photogrammetric Engineering & Remote Sensing, 70, 829–840. Ichoku, C., Giglio, L., Wooster, M. J., & Remer, L. A. (2008). Global characterization of biomass-burning patterns using satellite measurements of fire radiative energy. Remote Sensing of Environment, 112, 2950–2962. http://dx.doi.org/10.1016/j.rse.2008.02.009. Ichoku, C., & Kaufman, Y. J. (2005). A method to derive smoke emission rates from MODIS fire radiative energy measurements. Geoscience and Remote Sensing, IEEE Transactions on, 43(11), 2636–2649 (IEEE). Johnson, E. A. (1992). Fire and vegetation dynamics: Studies from the North American boreal forest. New York: Cambridge University Press. Johnstone, J. F., & Chapin, F. S. (2006). Effects of soil burn severity on post-fire tree recruitment in boreal forest. Ecosystems, 9, 14–31. http://dx.doi.org/10.1007/s10021-0040042-x. Johnstone, J. F., Chapin, F. S., Hollingsworth, T. N., Mack, M. C., Romanovsky, V., Turetsky, M., et al. (2010a). Fire, climate change, and forest resilience in interior Alaska. Canadian Journal of Forest Research, 40, 1302–1312. Johnstone, J. F., Hollingsworth, T. N., Chapin, F. S., & Mack, M. C. (2010). Changes in fire regime break the legacy lock on successional trajectories in Alaskan boreal forest. Global Change Biology, 16, 1281–1295.
Johnstone, J. F., & Kasischke, E. S. (2005). Stand-level effects of soil burn severity on postfire regeneration in a recently burned black spruce. Canadian Journal of Forest Research, 2163, 2151–2163. http://dx.doi.org/10.1139/X05-087. Johnstone, J. F., Rupp, T. S., Olson, M., & Verbyla, D. (2011). Modeling impacts of fire severity on successional trajectories and future fire behavior in Alaskan boreal forests. Landscape Ecology, 26, 487–500. http://dx.doi.org/10.1007/s10980-011-9574-6. Jones, B. M., Kolden, C. A., Jandt, R., Abatzoglou, J. T., Urban, F., & Arp, C. D. (2009). Fire behavior, weather, and burn severity of the 2007 Anaktuvuk River Tundra Fire, North Slope, Alaska. Arctic, Antarctic, and Alpine Research, 41, 309–316. http://dx.doi.org/ 10.1657/1938-4246-41.3.309. Justice, C. O., Giglio, L., Korontzi, S., Owens, J., Morisette, J. T., Roy, D., et al. (2002). The MODIS fire products. Remote Sensing of Environment, 83, 244–262. Kaiser, J. W., Heil, A., Andreae, M. O., Benedetti, A., Chubarova, N., Jones, L., et al. (2012). Biomass burning emissions estimated with a global fire assimilation system based on observed fire radiative power. Biogeosciences, 9, 527–554. http://dx.doi.org/ 10.5194/bg-9-527-2012. Kasischke, E. S., Bourgeau-Chavez, L. L., & Johnstone, J. F. (2007). Assessing spatial and temporal variations in surface soil moisture in fire-disturbed black spruce forests using space- borne synthetic aperture radar imagery: implications for post-fire tree recruitment. Remote Sensing of Environment, 108, 42–58. Kasischke, E. S., & Hoy, E. E. (2012). Controls on carbon consumption during Alaskan wildland fires. Global Change Biology, 18, 685–699. http://dx.doi.org/10.1111/ j.1365-2486.2011.02573.x. Kasischke, E. S., Hyer, E. J., Novelli, P. C., Bruhwiler, L. P., French, N. H. F., Sukhinin, A. I., et al. (2005). Influences of boreal fire emissions on Northern Hemisphere atmospheric carbon and carbon monoxide. Global Biogeochemical Cycles, 19, GB1012. Kasischke, E. S., & Bruhwiler, L. P. (2003). Emissions of carbon dioxide, carbon monoxide, and methane from boreal forest fires in 1998. Journal of Geophysical Research, 108(D1), 8146. http://dx.doi.org/10.1029/2001JD000461. Kasischke, E., & Turetsky, M. (2006). Recent changes in the fire regime across the North American borealregion—Spatial and temporal patterns of burning across Canada and Alaska. Geophysical Research Letters, 33, L09703. Kasischke, E. S., Verbyla, D., Rupp, T. S., McGuire, A. D., Murphy, K. A., Allen, J. L., et al. (2010). Alaska's changing fire regime—Implications for the vulnerability of its boreal forests. Canadian Journal of Forest Research, 40, 1313–1324. Kasischke, E. S., Williams, D., & Barry, D. (2002). Analysis of the patterns of large fires in the boreal forest region of Alaska. International Journal of Wildland Fire, 11, 131–144. Kaufman, Y. J., Justice, C. O., Flynn, L. P., Kendall, J. D., Prins, E. M., Giglio, L., et al. (1998). Potential global fire monitoring from EOS-MODIS. Journal of Geophysical Research, 103, 32,215–32,238. Krawchuk, M. A., Cumming, S. G., Flannigan, M. D., & Wein, R. W. (2006). Biotic and abiotic regulation of lightning fire initiation in the mixedwood boreal forest. Ecology, 87, 458–468. Kumar, S. S., Roy, D. P., Boschetti, L., & Kremens, R. (2011). Exploiting the power law distribution properties of satellite fire radiative power retrievals: A method to estimate fire radiative energy and biomass burned from sparse satellite observations. Journal of Geophysical Research, 116, 1–18. http://dx.doi.org/10.1029/2011JD015676. Lobert, J. M., & Warnatz, J. (1993). Emissions from the combustion process in vegetation. In P. J. Crutzen, & J. G. Goldammer (Eds.), Fire in the environment: The ecological, atmospheric, and climatic importance of vegetation fires (pp. 15–37). Chichester: John Wiley & Sons. McGuire, A. D., Chapin, F. S., Walsh, J. E., & Wirth, C. (2006). Integrated regional changes in arctic climate feedbacks: Implications for the global climate system. Annual Review of Environment and Resources, 31, 61–91. http://dx.doi.org/10.1146/annurev.energy. 31.020105.100253. Mottram, G. N., Wooster, M., Balzter, H., George, C., Gerrard, F., & Beisley, J. (2005). The use of MODIS-derived fire radiative power to characterise Siberian boreal forest fires. Proceedings of the 31st International Symposium on Remote Sensing of Environment, 19. (pp. 1–4). Nowacki, G., Spencer, P., Fleming, M., Brock, T., & Jorgenson, T. (2001). Unified ecoregions of Alaska: 2001. US Geological Survey, Open-File Report 020–297, Reston, Va. Pu, R., Li, Z., Gong, P., Csiszar, I., Fraser, R., Hao, W., et al. (2007). Development and analysis of a 12-year daily 1-km forest fire dataset across North America from NOAA/AVHRR data. Remote Sensing of Environment, 108, 198–208. http://dx.doi.org/10.1016/j.rse.2006.02.027. Randerson, J. T., Liu, H., Flanner, M. G., Chambers, S. D., Jin, Y., Hess, P. G., et al. (2006). Impact of Boreal Forest Fire on Climate Warming. Science, 314(5802), 1130–1132. http://dx.doi.org/10.1126/science.1132075. Shenoy, A., Johnstone, J. F., Kasischke, E. S., & Kielland, K. (2011). Persistent effects of fire severity on early successional forests in interior Alaska. Forest Ecology and Management, 261, 381–390. http://dx.doi.org/10.1016/j.foreco.2010.10.021. Stroppiana, D., Pinnock, S., & Gregoire, J. M. (2000). The Global Fire Product: Daily fire occurrence from April 1992 to December 1993 derived from NOAA AVHRR data. International Journal of Remote Sensing, 21, 1279–1288. Sukhinin, A. I., French, N. H. F., Kasischke, E. S., Hewson, J. H., Soja, A. J., Csiszar, I. A., et al. (2004). AVHRR-based mapping of fires in Russia: New products for fire management and carbon cycle studies. Remote Sensing of Environment, 93, 546–564. Turetsky, M. R., Kane, E. S., Harden, J. W., Ottmar, R. D., Manies, K. L., Hoy, E., et al. (2011). Recent acceleration of biomass burning and carbon losses in Alaskan forests and peatlands. Nature Geoscience, 4, 27–31. http://dx.doi.org/10.1038/ngeo1027 (Nature Publishing Group). Turner, M. G., Romme, W. H., Gardner, R. H., O'Neill, R. V., & Kratz, T. K. (1993). A revised concept of landscape equilibrium: Disturbance and stability on scaled landscapes. Landscape Ecology, 8, 213–227. http://dx.doi.org/10.1007/BF00125352. Vermote, E., Ellicott, E., Dubovik, O., Lapyonok, T., Chin, M., Giglio, L., et al. (2009). An approach to estimate global biomass burning emissions of organic and black carbon from MODIS fire radiative power. Journal of Geophysical Research, 114, 1–22. http://dx.doi.org/10.1029/2008JD011188.
K. Barrett, E.S. Kasischke / Remote Sensing of Environment 130 (2013) 171–181 Viereck, L. A. (1983). The effects of fire in black spruce ecosystems of Alaska and northern Canada. In Ross W. Wein, & D. A. MacLean (Eds.), The role of fire in northern circumpolar ecosystems (pp. 201–220). Chichester: John Wiley & Sons. Wang, D., Morton, D., Masek, J., Wu, A., Nagol, J., Xiong, X., et al. (2012). Impact of sensor degradation on the MODIS NDVI time series. Remote Sensing of Environment, 119, 55–61. Williams, R. J., Cook, G. D., Gill, A. M., & Moore, P. H. R. (1999). Fire regime, fire intensity and tree survival in a tropical savanna in northern Australia. Austral Ecology, 24, 54–59. Wooster, M. J. (2002). Small-scale experimental testing of fire radiative energy for quantifying mass combusted in natural vegetation fires. Geophysical Research Letters, 29, 21–24. http://dx.doi.org/10.1029/2002GL015487.
181
Wooster, M. J., Roberts, G., Perry, G. L. W., & Kaufman, Y. J. (2005). Retrieval of biomass combustion rates and totals from fire radiative power observations: FRP derivation and calibration relationships between biomass consumption and fire radiative energy release. Journal of Geophysical Research, 110(D24), 1–24. http://dx.doi.org/ 10.1029/2005JD006318. Wooster, M. J., & Zhang, Y. H. (2004). Boreal forest fires burn less intensely in Russia than in North America. Geophysical Research Letters, 31, 2–4. http://dx.doi.org/ 10.1029/2004GL020805. Yoshikawa, K., Bolton, W. R., Romanovsky, V. E., Fukuda, M., & Hinzman, L. D. (2003). Impacts of wildfire on the permafrost in the boreal forests of Interior Alaska. Journal of Geophysical Research, 107(D1), 8148.