Impact Factor 4.106 | CiteScore 4.47
More on impact ›

Original Research ARTICLE

Front. Plant Sci., 11 September 2018 | https://doi.org/10.3389/fpls.2018.01318

Does Environment Filtering or Seed Limitation Determine Post-fire Forest Recovery Patterns in Boreal Larch Forests?

Wen H. Cai1,2,3, Zhihua Liu1*, Yuan Z. Yang2,3 and Jian Yang4*
  • 1CAS Key Laboratory of Forest Ecology and Management, Institute of Applied Ecology, Chinese Academy of Sciences, Shenyang, China
  • 2Key Laboratory of Environment Change and Resources Use in Beibu Gulf, Guangxi Teachers Education University, Ministry of Education, Nanning, China
  • 3Guangxi Key Laboratory of Earth Surface Processes and Intelligent Simulation, Guangxi Teachers Education University, Nanning, China
  • 4Department of Forestry and Natural Resources, Thomas Poe Cooper Building, University of Kentucky, Lexington, KY, United States

Wildfire is a primary natural disturbance in boreal forests, and post-fire vegetation recovery rate influences carbon, water, and energy exchange between the land and atmosphere in the region. Seed availability and environmental filtering are two important determinants in regulating post-fire vegetation recovery in boreal forests. Quantifying how these determinants change over time is helpful for understanding post-fire forest successional trajectory. Time series of remote sensing data offer considerable potential in monitoring the trajectory of post-fire vegetation recovery dynamics beyond current field surveys about structural attributes, which generally lack a temporal perspective across large burned areas. We used a time series of the normalized difference vegetation index (NDVI) and normalized difference shortwave infrared reflectance index (NDSWIR) derived from Landsat images to investigate post-fire dynamics in a Chinese boreal larch forest. An adjacent, unburned patch of a similar forest type and environmental conditions was selected as a control to separate interannual fluctuation in NDVI and NDSWIR caused by climate from changes due to wildfire. Temporal anomalies in NDVI and NDSWIR showed that more than 10 years were needed for ecosystems to recover to a pre-fire state. The boosted regression tree analysis showed that fire severity exerted a persistent, dominant influence on vegetation recovery during the early post-fire successional stage and explained more than 60% of variation in vegetation recovery, whereas distance to the nearest unburned area and environmental conditions exhibited a relatively small influence. This result indicated that the legacy effects of fire disturbance, which control seed availability for tree recruitment, would persist for decades. The influence of environmental filtering could increase with succession and could mitigate the initial heterogeneity in recovery caused by wildfire.

Introduction

Wildfire is the primary natural disturbance in boreal forest ecosystems and plays a key role in regulating the carbon, water and energy exchange between the land and atmosphere via its influence on ecosystem structure and function in the region (Harden et al., 2000; Carcaillet et al., 2001; Conard et al., 2002; Kashian et al., 2006; Flannigan et al., 2009; Young et al., 2016). Due to climate-induced increases in fire frequency, severity and size, a possible shift in forest composition from conifer (e.g., Picea spp. and Abies spp.) toward broadleaf (e.g., Betula spp. and Populus spp.) species may significantly alter the land surface radiative budget (Amiro et al., 2006; Randerson et al., 2006) and initiate a positive/negative feedback to climate (Balshi et al., 2009; Goulden et al., 2011; Shuman et al., 2011). Therefore, understanding the mechanisms that determine post-fire forest recovery is critical to predicting the potential response of forest ecosystems and its climate feedback in boreal regions.

Previous studies have identified that seed limitation and environmental filtering are two key processes that control post-fire species composition and subsequent successional dynamics in boreal ecosystems (Bond-Lamberty et al., 2014; Chambers et al., 2016), although post-fire climate conditions are also an important control in other ecosystems (Shryock et al., 2015; Hansen et al., 2016). Seed limitation is mainly controlled by in-situ seed availability and seed dispersal, which are related to fire severity and size (Tautenhahn et al., 2016). Environmental filter often refers to local biophysical conditions (e.g., topographic positions, soil moisture, and solar radiation) and microclimate conditions that control seed germination and seedling establishment once seeds arrive. Post-fire seedling establishment is also vulnerable to drought and light conditions (Clark and St Clair, 2011; Walker et al., 2015; Harvey et al., 2016). The relative importance of seed limitation and environmental filtering on post-fire recovery varies depending on the interaction between species’ life-history traits and fire characteristics in the boreal forests. For example, in Alaskan black spruce (Picea mariana) forests, serotinious cones (the retention of seeds in cones that open after a fire) release abundant seeds after a fire, and environmental filtering is often the major factor that determines the post-fire successional trajectory due to the different regeneration abilities of deciduous hardwoods and conifers (Johnstone et al., 2010). In Central Siberian dark taiga forests dominated by Abies sibirica, Picea obovata, and Pinus sibirica, seed limitation plays a major role in determining post-fire successional trajectories via the different dispersal abilities of deciduous hardwoods and conifers (Tautenhahn et al., 2016). These studies have offered insightful implications for how an intensified fire regime will affect future ecosystem structures and functions in similar boreal forest ecosystems. However, relatively little is known about the mechanisms that regulate post-fire community assembly in the Siberian light taiga dominated by Pinus sylvestris and Larix spp, because the results from North American boreal forests may not be applicable to Siberian larch forests due to differences in species’ life history traits, fire regimes, climate, and environmental conditions. Furthermore, most of these studies are based on local observations and therefore are spatially and temporally limited (Johnstone et al., 2004; Lecomte et al., 2006; Cai et al., 2013). Consequently, its application over multiple wildfires and large regions remains uncertain.

Remote sensing offers a unique opportunity to explore burned patterns and the spatial controls on post-fire recovery over large spatial and temporal scales. Fire reduces the reflectance in the visible-to-near-infrared wavelengths (∼0.4–∼0.9 μm) due to the loss of photosynthetic vegetation and an accompanied increase in reflectance in short and middle infrared wavelengths (∼1.5–∼2.3 μm) due to char/ash accumulation on the ground (Lentile et al., 2006; Miller and Thode, 2007). Post-fire recovery has the opposite effect on surface reflectance due to the reestablishment of photosynthetic active vegetation. Such changes in spectral signatures, if well-calibrated with field survey data, can be valuable to characterize fire severity and vegetation recovery (Chu and Guo, 2014; Mouillot et al., 2014). For example, the normalized burn ratio (NBR) is often used as a fire severity indicator, due to its sensitivity to changes in vegetation and soil conditions (Miller et al., 2009). The normalized difference vegetation index (NDVI) and normalized difference shortwave infrared reflectance index (NDSWIR) are commonly used as indicators to monitor post-fire vegetation recovery in boreal forests (Cuevas-González et al., 2009) because they are associated with vegetation cover and canopy water content, respectively.

Siberian Larch forests constitute approximately 20% of global boreal forests and are associated with relatively low-severity fires due to their sparse stand structure in a permafrost environment (Conard et al., 2002). Chinese boreal larch forests are located on the southern border of Siberian larch forest and are sensitive to climate warming. Understanding the mechanisms that drive post-fire vegetation recovery could help to better predict potential responses of boreal biomes to climate warming and intensified wildfire regimes. In this research, we focused on a burned area that was disturbed in 2000, in the Great Xing’ an Mountains, China. We examined the effect of fire severity on an early post-fire vegetation recovery pattern using Landsat-derived time-series NDVI and NDSWIR as proxies of vegetation recovery. We also detected the roles of seed limitation and environmental filtering and their temporal dynamics during the post-fire forest successional stage. This study helps to understand the extended legacy effects of fire on Siberian larch forests, a globally important yet understudied ecosystem, and contributes to understanding the vulnerability of high-latitude ecosystems to environmental change by complementing results from North America (Johnstone et al., 2010; Kasischke et al., 2010).

Materials and Methods

Study Area

The Huzhong National Natural Reserve is located in the Great Xing’ an Mountains of northeastern China (51°56′31″N 122°42′14″E to 51°17′42″N 123°18′05″E) and has a total area of 190,000 ha (Figure 1). The region is characterized by a continental monsoon climate with a mean annual precipitation and temperature of 500 mm and -4.7°C, respectively. The terrain is characterized by gently sloping uplands, and the valley bottom is frequently underlain by permafrost or seasonally frozen ground (Xu, 1998). Wildfire is a primary natural disturbance in this ecosystem, with a historical fire return interval of approximately 120 years (Xu, 1998). We selected an 8700-ha burned area in the Huzhong Natural Reserve, burned in 2000, to examine the relative influence of seed limitation and environmental filtering and their temporal dynamics in regulating post-fire forest recovery (Figure 1). This burned patch was selected because of its representativeness of Siberian larch forests and fire regime, minimum human intervention, and existence of field investigation of post-fire stand structure and function.

FIGURE 1
www.frontiersin.org

FIGURE 1. The location of the study area in circumpolar boreal forest ecosystem (a), digitial elevation model (b), and the burned scar shown in false color composite on the 2001 Landsat image (R:G:B = B7:B4:B1) (c). Field survey sites were shown as solid black points in c.

The area is representative of the Siberian light taiga—a closed canopy boreal forest dominated by Betula platyphylla, Populus davidiana, and Populus suaveolens during early-successional stages; and Larix gmelinii, and P. sylvestris var. mongolica in late-successional stages (Zhou, 1991). The fire regimes exhibit great spatial heterogeneity in occurrence, size and severity, primarily due to biophysical environments and associated fuel characteristics (Xu, 1998; Wu et al., 2013). Fires are relatively higher frequency and lower severity in well-drained, south-facing slopes where microclimates are relatively warm and dry and herbaceous species are dominant. Fires are relatively infrequent and more intensive on north slopes and terraces that are wet and cool and where two shrub species, Ledum palustre and Vaccinium uliginosum, are dominant. Post-fire tree seedlings mainly consist of larch and birch, and their density and composition are controlled by both fire severity and topographic positions (Cai et al., 2013).

Overall Study Design

The aim of this study is to quantify the relative influence of seed limitation or environmental filtering in determining post-fire ecosystem recovery dynamics using field surveys and time series of remote sensing data. To do so, we used fire severity and distance to the nearest unburned forest, due to their relationship with in-situ residual mature trees and seed dispersal, as a proxy for seed availability. We used the topographic wetness index (TW) and soil radiation, due to their relationship with soil moisture and microclimate conditions, as proxies for environmental filters. We used time series of Landsat-derived vegetation indices as a proxy for post-fire forest recovery. To evaluate the suitability of each vegetation index in representing vegetation conditions, we compared field survey data with vegetation indices using the Pearson correlation. To evaluate whether fire severity and topographic conditions affect post-fire recovery dynamics, we compared the time series of vegetation indices under different fire severities and aspects using the paired-t test. Finally, to assess the relative influences of spatial controls on post-fire forest recovery, we used a machine learning approach to account for the potential complex relationship among post-fire vegetation recovery and a suite of variables.

Datasets

Satellite Data and Preprocessing

Eleven Landsat 5 and 7 images (path 122, row 241) acquired in peak growing seasons between 1999 and 2015 were used (Table 1). Most of the selected images were acquired in peak-summer months (June–September) with minimum cloud cover to minimize the influences of forest phenology on quantifying post-fire forest recovery dynamics (i.e., time-series NDVI and NDSWIR). The Landsat TM images acquired in 1999 and 2001 (Table 1) were chosen to calculate spectral fire severity (dNBR = NBR2001–NBR1999) (Miller and Thode, 2007; Miller et al., 2009).

TABLE 1
www.frontiersin.org

TABLE 1. Time-series Landsat images for assessing post-fire recovery of boreal larch forest in the 2000 burned area.

To ensure that the time series of vegetation indices reflected real changes in the state of the ecosystem, rather than phenology and senor-sun angle artifacts (the bidirectional reflectance distribution function), a radiometric correction was conducted to the selected Landsat images. The Landsat digital number (DN) was converted to top-of-atmosphere reflectance following a two-step process. First, the DNs were converted to radiance values using the bias and gain values specific for each scene. Second, the calculated radiance values were converted to top-of-atmosphere reflectance using the solar zenith angle and astronomical unit distance of the sun. We did not conduct a geometric correction because it had already been done by the USGS2.

Field Survey Data

We selected 99 sites in the burned areas to sample the post-fire regeneration pattern of Chinese boreal larch forests (Figure 1) from 2010 to 2012. These sites were carefully selected to represent the variability of fire severity and environmental conditions. Within each site, a 50 m × 50 m plot was established based on its representativeness of fire severity, topographical position, and post-fire vegetation recovery pattern. Within each plot, the remaining unburned adult trees were counted for density and diameter in two subplots (10 m × 10 m), and post-fire recruited seedling/saplings for each species were surveyed for density in four subplots (5 m × 5 m). Detailed information regarding the field survey and forest recruitment conditions can be found in Cai et al. (2013) and Cai and Yang (2016). These field data were used for qualitative analysis of the NDVI and NDSWIR values.

Proxy for Post-fire Forest Recovery: NDVI and NDSWIR

The images used for characterizing post-fire forest recovery were based on Landsat images acquired from 2001 to 2016. NDVI and NDSWIR were calculated using reflectance (ρ) values from the red (band 3), NIR (band 4), and SWIR (band 5) bands, as shown below:

NDVI=(ρNIRρred)/(ρNIR+ρred)
NDSWIR=(ρNIRρSWIR)/(ρNIR+ρSWIR)

Proxy for Seed Limitation

Post-fire seed limitation for forest recruitment included in-situ residual mature trees and seed dispersal from around unburned areas. We used fire severity as a proxy for in-situ residual mature trees. The Landsat-derived differenced NBR (dNBR) is widely used for fire severity assessments in boreal forests (French et al., 2008; Soverel et al., 2011) and has been proven to be applicable in Chinese boreal forests (Fang and Yang, 2014). The dNBR has a high correlation with a field-based assessment of fire severity called the composition burn index (CBI) and can show variations in canopy tree mortality rate (Cocke et al., 2005). The dNBR was computed from pre-fire (09/05/1999) and post-fire (09/14/2000) images using the algorithm described by Lentile et al. (2006). As the effective seed dispersal distance of the dominant species Gmelin’s larch is approximately 100 m, we used a moving window (5 × 5 pixels) resampling approach to minimize the neighborhood effect of nearby surviving trees on post-fire forest recovery.

In this analysis, pre-fire species composition and age retrieved from forest inventory data were employed as proxy data for seed availability. Information regarding pre-fire vegetation type was needed for comparing the responses of different forest types to fire disturbance. The Forest Inventory Data conducted in 1990 is the available land-cover map before the 2000 fire. These data were used to identify the pre-fire forest type of the burned area. We also calculated the proportion of broadleaf species, which was used as an independent factor in regulating post-fire forest recovery. We used ArcGIS to calculate the distances to the nearest unburned area as a proxy for seed availability and delivery from around the unburned area (Greene et al., 1999).

Proxy for Environmental Filters

We calculated a TW from a digital elevation model using the Hydrology module in ArcGIS to indicate soil moisture availability due to its strong control on tree regeneration and establishment ability (Zinko et al., 2005). We also calculated a solar radiation index to indicate the potential atmospheric water demand (Kumar et al., 1997). The solar radiation index was estimated using the Solar Radiation analysis toolset in ArcGIS. The toolset used variability in the orientation (slope and aspect) to calculate direct and diffuse radiation for each pixel of the elevation model using viewshed algorithms. We report total insolation during the growing season for each pixel. Slope and aspect were also included. All of the indices were calculated from a 30 m resolution digital elevation model.

Data Analysis

To explore the suitability of each vegetation index in representing post-fire forest recovery, we compared field survey data (tree recruitment density and species composition) with NDVI and NDSWIR. The Pearson correlation coefficient was calculated in addition to NDVI, NDSWIR, log-transformed tree recruitment densities and tree species composition. These vegetation indices were calculated from TM images that were acquired on August 26, 2010.

To explore post-fire forest recovery trajectories, time series of NDVI and NDSWIR were sampled. First, the minimum distance between each sampling point was set as 1000 m to remove a spatial autocorrelation issue. Second, a normalized time series vegetation index was calculated to remove the potential bias caused by phonology, the senor artifact, or topographic influence among years. To do so, we created a 2000 m buffer zone around the burned patch, and classified the landscape (burned patch + buffer zone) into three land types based on topographic position because of its influence on ecosystem and fire characteristics (flat valley bottom, south-facing, and north-facing slopes; see study area description). In each land type, the mean value of the vegetation index of buffer zone (termed as unburned area, e.g., NDVIunburned) and burned patches (termed as burned area, e.g., NDVIburned) was calculated. The normalized NDVI and NDSWIR in each land type were calculated using the following formulas:

NormalizedNDVI=(NDVIburnedNDVIunburned)/NDVIunburned
NormalizedNDSWIR=(NDSWIRburnedNDSWIRunburned)/NDSWIRunburned

The mean and standardized values of vegetation indices were plotted against time since last fire to show the post-fire forest recovery trajectory.

A paired t-test was applied to examine the variations in NDVI and NDSWIR between high severity and low severity burned areas, as well as among different topographic positions (flat valley bottoms, south-facing, and north-facing slopes). Statistical significance in these analyses was assessed at a = 0.05.

We also used a boosted regression tree (BRT) analysis to examine the influence of a suite of explanatory variables, including fire severity, TW, solar radiation, distance to the nearest unburned area and pre-fire forest type, on post-fire forest recovery. BRT is a machine-learning approach based on classification and regression trees (Elith et al., 2008). It is capable of handling complex nonlinear relationships and missing values without the restrictive assumptions of parametric statistics. The relationship between dependent and independent variables could visually be shown using a partial dependency plot, and it could calculate the relative influence of each independent variable on the dependent variable. The BRT method combines the advantages of ensemble regression trees (e.g., random forest), which relate a response to their predictors by recursive binary splits and boosting algorithms, which combine many simple models to give improved predictive performance (Elith et al., 2008). In the fitted BRT models, both NDVI and NDSWIR were selected as response variables. Biotic and abiotic explanatory variables for the two analytical approaches are listed in Table 2. The BRT models were fitted with the following parameters: Gaussian error distribution, a learning rate of 0.005, and a bag fraction of 0.5. Others used default parameters. The BRT analysis was run in the gbm module of R software (R Core Team, 2013).

TABLE 2
www.frontiersin.org

TABLE 2. List of dependent and predictor variables considered in Boosted regression tree analysis.

Results

The normalized NDVI and NDSWIR removed the influence of interannual variability in climate, and its dynamics reflected the post-fire vegetation recovery dynamics to fire severity and environmental conditions. A negative value of these indices indicated departure from pre-fire vegetation conditions. Figure 2 shows reduced values 1 year after the wildfire, followed by a gradual recovery period. It took approximately 10 years for NDVI to recover to the pre-fire state, and at least 15 years for NDSWIR (Figure 2). Our results showed that different vegetation indices followed special recovery trajectories, reflecting the fact that they may represent different aspects of ecosystem characteristics.

FIGURE 2
www.frontiersin.org

FIGURE 2. The normalized difference of Normalized Difference Vegetation Index (red line) and Normalized Difference Shortwave Infrared reflectance Index (blue line) between burned areas and unburned control sites plotted as a function of time since last fire.

Post-fire recovery rates also depended on the degree to which the larch forest ecosystem had changed due to the fire disturbance. Figure 3 shows the variation in vegetation recovery between high severity and low severity burned areas. The recovery rate was higher in the low severity burned area (Figure 3). A high fire severity burn could exert a long-term negative effect on vegetation recovery. Our field observations showed that post-fire tree recruitment was different among topographic positions, with south-facing slopes favoring broadleaf tree recruitment. Our results also showed that the recovery rate of NDVI and NDSWIR did not differ significantly (p > 0.05) for different topographic conditions (Figure 4).

FIGURE 3
www.frontiersin.org

FIGURE 3. Post-fire dynamics of NDVI and NDSWIR values (mean ± SD) in different fire severity levels.

FIGURE 4
www.frontiersin.org

FIGURE 4. Post-fire dynamics of NDVI and NDSWIR values (mean ± SD) in different topographic positions.

To analyze whether NDVI and NDSWIR could reflect post-fire vegetation recovery conditions, we compared the vegetation indices with field sampling data. The results showed that NDVI and NDSWIR had a positive relationship with post-fire stand density (r = 0.32 and 0.25 for stand density, respectively, p < 0.05) but had no relationship with post-fire forest composition (p > 0.05) (Figure 5).

FIGURE 5
www.frontiersin.org

FIGURE 5. The relationship between field surveyed post-fire tree recruitment and NDVI and NDSWIR values (10 years after fire), respectively.

The BRT analysis showed the relative contributions of fire severity and other independent variables (Table 2) to vegetation recovery. The results showed that fire severity was the most important factor in determining post-fire NDVI and NDSWIR recovery (Figure 6). It explained more than 60% variation of post-fire vegetation recovery. Pre-fire forest structure and distance to the nearest unburned area had little influence on forest recovery. Although solar radiation and TW, which depict environmental conditions, had relatively small effects on NDVI and NDSWIR, their effects increased with forest succession. Because of the similar trend between NDVI and NDSWIR, only the relative influence of each predictor on NDSWIR is shown in Figure 6.

FIGURE 6
www.frontiersin.org

FIGURE 6. Relative importance of predictor variables on NDSWIR. dNBR stands for differenced Normalized Burn Ratio; solar stands for solar radiation; TW stands for topographic wetness index; distance stands for distance away from unburned area; BP stands for pre-fire proportion of broadleaf species.

Discussion

The Suitability of NDVI/NDSWIR in Monitoring Post-fire Forest Recovery

Our results, consistent with other studies (Cuevas-González et al., 2009; Jones et al., 2013; Meng et al., 2015; Chu et al., 2016, 2017), suggested that time series of spectral indices are an effective way to monitor post-fire vegetation recovery dynamics over large spatiotemporal scales. The NDVI, a commonly used vegetation greenness index, is often used to assess chlorophyll abundance and the vigor of green vegetation. The NDSWIR, a shortwave infrared-based index, is sensitive to canopy water content and stand structure. Therefore, these two spectral indices indicate different aspects of vegetation characteristics. In this analysis, NDVI reached its maximum quickly after the fire and resulted in a faster vegetation recovery rate, suggestive of rapid recruitment of shrubs and herbaceous vegetation after fire (Hart and Chen, 2006). However, NDSWIR recovered at slower rate, indicative of much slower establishment and growth of forest stands. Although NDSWIR and NDVI are good indicators of vegetation cover and structure, they offer little information regarding species composition (Figure 5). Thus, they should be used with caution when applying spectral indices for detecting the legacy effects of disturbance on post-fire forest successional trajectories (i.e., dynamics of vegetation communities). Nevertheless, spatially and temporally consistent time series of spectral indices remain an effective way to assess post-fire recovery and forest resilience in a warmer climate and intensified fire regime.

Time Needed for Forest Recovery

Quantifying the time needed for forest recovery is useful for evaluating the long-term legacy effects of disturbances to ecosystem properties. However, this analysis also suggested that rate of forest recovery is dependent on vegetation indices (i.e., NDVI and NDSWIR) and ecosystem types. It took at least 10 years for NDVI to return to control sites, and it would be longer for NDSWIR (Figure 2). This recovery time is similar to the recovery rate derived by Cuevas-González et al. (2009) in central Siberia. However, it is longer than that in North America and Canada, which are less than 8 years (e.g., Hicke et al., 2003; Goetz et al., 2006). These variations might be attributed to many aspects such as differences in fire regime, community composition and tree traits (e.g., fire adaptive trait) between North American and Siberian boreal forests. For example, black spruce [Picea mariana (Mill.) B.S.P.], a dominant species in many North American boreal forests, is serotinous and could release abundant seeds immediately after a fire (Johnstone et al., 2008) and therefore experience a faster recovery than boreal larch forests.

The rapid increase in NDVI following fire was considered to reflect the restoration of vegetation cover (including tree, shrub, and herbaceous vegetation). Forest regeneration and the quick recovery of vegetation cover are assumed to occur in the first few years after a fire, and this community structure could persist for a long time (Johnstone et al., 2004; Alexander et al., 2012). The rapid recovery in NDVI showed that vegetation cover can likely reach a pre-fire state within 10 years in the Great Xing’ an Mountains (Figure 2). However, NDVI could not identify species composition. The relatively slow recovery of NDSWIR indicated that the post-fire forest structure could not recover in a short time. As NDVI is assumed to have a positive relationship with forest NPP (Turner et al., 2004), the rapid increase in NDVI indicated a high resilience of NPP to wildfire. Our result indicated that ecosystem function is less sensitive to disturbance than forest structure (Cai and Yang, 2016). Boreal larch forests are quite resilient to fire disturbances, showing quick recovery, at least in terms of rapid return to previous states of vegetation cover not only in low-severity burned stands but also in high-severity burned stands.

Seed Availability vs. Environmental Filtering

The dNBR was frequently used for quantifying the conditions of affected vegetation after fire (Goetz et al., 2006; Meng et al., 2015; Bartels et al., 2016). Fire severity (dNBR) could reflect information about post-fire survival trees and residual organic layer depth, which control seed availability and seed bed quality for tree regeneration. In the Great Xing’ an Mountains, the organic layer in unburned areas is relatively shallow compared to that in North American boreal forests, and it is consumed by relatively low severity fires (based on field observations). The difference in fire severity is often considered as the variation in in-situ survival trees. Severe fires significantly reduced forest recovery rates, and this negative effect on post-fire forest recovery would persistently dominate for at least 10 years. This result is convincing evidence that fire severity has a legacy effect on forest recovery, confirming previous studies (Pan et al., 2011; Fletcher et al., 2014; Brown et al., 2015). Seed availability is the most important factor in regulating post-fire recovery trends. However, distance to the nearest unburned area had a relatively small effect. Seed dispersal from unburned area contributed little to post-fire forest recovery. This might connect with the fact that the effective seed dispersal distance of the dominant conifer species (L. gmelinii) is less than 100 m. Mature trees within the burned patch have similar influences on tree regeneration as the edge of the burned patch. Over the expanse of the burned area, post-fire tree regeneration is dependent on in-situ residual seed banks or survival propagules.

Environmental factors had a relatively small influence on post-fire forest recovery. However, their effect increased with succession. This result indicated that environmental filtering would mitigate the legacy effect of fire disturbance on forest ecosystems. The legacy effects caused by spatial heterogeneities of fire severity tend to become mitigated with succession in terms of NDVI and NDSWIR (Figure 3). However, compared to upland sites, in valley bottoms, differences in NDVI/NDSWIR between fire severity patches are more persistent. This result is likely due to the existence of seasonal permafrost and poor nutrient conditions in valley bottoms, where Gemlin’s larch is adaptive.

Our results showed that elevation changes also influence post-fire vegetation recovery, likely due to its relationship to climatic conditions. This result indicated that climatic conditions are also the drivers of vegetation response after disturbances. Post-fire climate might control tree regeneration and interact with factors such as seed availability and environmental conditions to alter post-fire recovery (Meng et al., 2015). Post-fire climate in the first growing seasons following fire may be highly important for forest recruitment (Shryock et al., 2015). As we only assessed post-fire forest recovery in a single burned patch, climate factors were not considered comprehensively in this study. Additionally, climate should be considered in studies with multiple fires.

The BRT models could significantly explain the temporal and spatial variation in NDVI and NDSWIR as a function of independents indicating seed availability and environment properties. Our field sampling data showed that NDVI, especially NDSWIR, has a good positive relationship with tree recruitment density. Hence, the influence of selected independent variables on NDVI/NDSWIR could reflect underlying ecological processes that affect post-fire vegetation recovery. Seed availability and environmental filtering are two important processes that control post-fire forest dynamics. Seed availability had a dominant effect during the early successional stage, but its effect was mitigated by environmental filtering.

Limitations

First, we analyzed the relative influence of seed limitation and environmental filtering on post-fire vegetation recovery for only one fire, which may not fully represent the full spectrum in terms of fire severity and environmental gradients for larch forest. However, this analysis also provided us to minimize other confounding factors such as past disturbance history, climate variability, and vegetation conditions, and therefore allowed us to confidently assess the effects of seed limitation and environmental filtering on post-fire recovery. Second, although time series analyses of NDVI and NDSWIR are good indicators of photosynthetically active vegetation or “greenness,” they did not necessarily reflect recovery in other important ecosystem variables such as biogeochemistry, radiation budgets, or water balance. In the future, we will utilize multiple recovery indicators to provide complementary measurements of vegetation dynamics across many fires to study the interactions among ecosystem vulnerability, disturbances, and climate in high-latitude ecosystems.

Implications

The post-fire forest recovery rate is strongly related to ecosystem carbon storage and its resultant effects on climate feedback. Slow vegetation recovery rates likely induce low carbon storage. Post-fire ecosystem carbon storage shows a sigmoid pattern that initially decreases and then increases, and eventually balances (Yang et al., unpublished). As fire disturbances in boreal forests are predicted to be larger and more severe, these changes imply an increasing risk of seed limitation for regulating post-fire forest recovery, loss of ecosystem resilience, and increases in carbon release. The coupled effect of vegetation destruction and low forest recovery rates due to high-severity wildfires inevitably leads to a rapid increase in carbon loss.

Using spatially and temporal consistent spectral indices, we showed that fire severity is a dominant factor in post-fire forest recovery, and its effect could last many decades. The effect of environmental filtering on forest recovery would increase with succession, reflecting the convergence of ecosystem to climax regardless of the initial conditions. In-situ seed availability remains the primary factor in regulating vegetation recovery. Understanding the mechanisms that underlie the interaction of fire disturbance and environmental conditions on vegetation recovery will assist post-fire forest management efforts to enhance the resilience of ecosystems in Siberian larch forests.

Author Contributions

ZL, JY, and WC conceived and designed research. WC performed the experiments, analyzed the data and wrote the manuscript. ZL and JY revised the manuscript. YY helped in conducting the experiments, analyzing the data and editing the manuscript.

Funding

This research was funded by the National Key Research and Development Program of China (2017YFA0604403), CAS Pioneer Hundred Talents Program, and National Natural Science Foundation of China (Project Nos. 31470517 and 31500387).

Conflict of Interest Statement

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Acknowledgments

We thank Huzhong Forestry Bureau and Huzhong Natural Reserve for the help in field investigations. We also thank three reviewers and academic editor for comments that improved this manuscript.

Footnotes

  1. ^http://glovis.usgs.gov/
  2. ^http://landsat.usgs.gov/

References

Alexander, H. D., Mack, M. C., Goetz, S., Loranty, M. M., Beck, P. S. A., Earl, K., et al. (2012). Carbon accumulation patterns during post-fire succession in Cajander Larch (Larix cajanderi) forests of Siberia. Ecosystems 15, 1065–1082. doi: 10.1007/s10021-012-9567-6

CrossRef Full Text | Google Scholar

Amiro, B. D., Orchansky, A. L., Barr, A. G., Black, T. A., Chambers, S. D., Chapin Iii, F. S., et al. (2006). The effect of post-fire stand age on the boreal forest energy balance. Agric. For. Meteorol. 140, 41–50. doi: 10.1016/j.agrformet.2006.02.014

CrossRef Full Text | Google Scholar

Balshi, M. S., McGuire, A. D., Duffy, P., Flannigan, M., Kicklighter, D. W., and Melillo, J. (2009). Vulnerability of carbon storage in North American boreal forests to wildfires during the 21st century. Glob. Change Biol. 15, 1491–1510. doi: 10.1111/j.1365-2486.2009.01877.x

CrossRef Full Text | Google Scholar

Bartels, S. F., Chen, H. Y. H., Wulder, M. A., and White, J. C. (2016). Trends in post-disturbance recovery rates of Canada’s forests following wildfire and harvest. For. Ecol. Manag. 361, 194–207. doi: 10.1016/j.foreco.2015.11.015

PubMed Abstract | CrossRef Full Text | Google Scholar

Bond-Lamberty, B., Rocha, A. V., Calvin, K., Holmes, B., Wang, C., and Goulden, M. L. (2014). Disturbance legacies and climate jointly drive tree growth and mortality in an intensively studied boreal forest. Glob. Change Biol. 20, 216–227. doi: 10.1111/gcb.12404

PubMed Abstract | CrossRef Full Text | Google Scholar

Brown, C. D., Liu, J., Yan, G., and Johnstone, J. F. (2015). Disentangling legacy effects from environmental filters of post-fire assembly of boreal tree assemblages. Ecology 96, 3023–3032. doi: 10.1890/14-2302.1

CrossRef Full Text | Google Scholar

Cai, W., Yang, J., Liu, Z., Hu, Y., and Weisberg, P. J. (2013). Post-fire tree recruitment of a boreal larch forest in Northeast China. For. Ecol. Manage. 307, 20–29. doi: 10.1016/j.foreco.2013.06.056

CrossRef Full Text | Google Scholar

Cai, W. H., and Yang, J. (2016). High-severity fire reduces early successional boreal larch forest aboveground productivity by shifting stand density in north-eastern China. Int. J. Wildland Fire 25, 861–875. doi: 10.1071/WF15026

CrossRef Full Text | Google Scholar

Carcaillet, C., Bergeron, Y., Richard, P. J. H., Frechette, B., Gauthier, S., and Prairie, Y. T. (2001). Change of fire frequency in the eastern Canadian boreal forests during the Holocene: does vegetation composition or climate trigger the fire regime? J. Ecol. 89, 930–946. doi: 10.1111/j.1365-2745.2001.00614.x

CrossRef Full Text | Google Scholar

Chambers, M. E., Fornwalt, P. J., Malone, S. L., and Battaglia, M. A. (2016). Patterns of conifer regeneration following high severity wildfire in ponderosa pine – dominated forests of the Colorado Front Range. For. Ecol. Manage. 378, 57–67. doi: 10.1016/j.foreco.2016.07.001

CrossRef Full Text | Google Scholar

Chu, T., and Guo, X. (2014). An assessment of fire occurrence regime and performance of Canadian fire weather index in south central Siberian boreal region. Nat. Hazards Earth Syst. Sci. Discuss. 2, 4711–4742. doi: 10.5194/nhessd-2-4711-2014

CrossRef Full Text | Google Scholar

Chu, T., Guo, X., and Takeda, K. (2016). Remote sensing approach to detect post-fire vegetation regrowth in Siberian boreal larch forest. Ecol. Indic. 62, 32–46. doi: 10.1016/j.ecolind.2015.11.026

CrossRef Full Text | Google Scholar

Chu, T., Guo, X., and Takeda, K. (2017). Effects of burn severity and environmental conditions on post-fire regeneration in Siberian larch forest. Forests 8:76. doi: 10.3390/f8030076

CrossRef Full Text | Google Scholar

Clark, A. L., and St Clair, S. B. (2011). Mycorrhizas and secondary succession in aspen-conifer forests: light limitation differentially affects a dominant early and late successional species. For. Ecol. Manage. 262, 203–207. doi: 10.1016/j.foreco.2011.03.024

CrossRef Full Text | Google Scholar

Cocke, A. E., Fule, P. Z., and Crouse, J. E. (2005). Comparison of burn severity assessments using differenced normalized burn ratio and ground data. Int. J. Wildland Fire 14, 189–198. doi: 10.1071/WF04010

CrossRef Full Text | Google Scholar

Conard, S. G., Sukhinin, A. I., Stocks, B. J., Cahoon, D. R., Davidenko, E. P., and Ivanova, G. A. (2002). Determining effects of area burned and fire severity on carbon cycling and emissions in Siberia. Clim. Change 55, 197–211. doi: 10.1023/A:1020207710195

CrossRef Full Text | Google Scholar

R Core Team (2013). R: A Language and Environment for Statistical Computing. Vienna: R Foundation for Statistical Computing.

Google Scholar

Cuevas-González, M., Gerard, F., Balzter, H., and Riano, D. (2009). Analysing forest recovery after wildfire disturbance in boreal Siberia using remotely sensed vegetation indices. Glob. Change Biol. 15, 561–577. doi: 10.1111/j.1365-2486.2008.01784.x

CrossRef Full Text | Google Scholar

Elith, J., Leathwick, J., and Hastie, T. (2008). A working guide to boosted regression trees. J. Anim. Ecol. 77, 802–813. doi: 10.1111/j.1365-2656.2008.01390.x

PubMed Abstract | CrossRef Full Text | Google Scholar

Fang, L., and Yang, J. (2014). Atmospheric effects on the performance and threshold extrapolation of multi-temporal Landsat derived dNBR for burn severity assessment. Int. J. Appl. Earth Obs. Geoinf. 33, 10–20. doi: 10.1016/j.jag.2014.04.017

CrossRef Full Text | Google Scholar

Flannigan, M., Stocks, B., Turetsky, M., and Wotton, M. (2009). Impacts of climate change on fire activity and fire management in the circumboreal forest. Glob. Change Biol. 15, 549–560. doi: 10.1111/j.1365-2486.2008.01660.x

CrossRef Full Text | Google Scholar

Fletcher, M.-S., Wood, S. W., and Haberle, S. G. (2014). A fire-driven shift from forest to non-forest: evidence for alternative stable states? Ecology 95, 2504–2513.

Google Scholar

French, N. H. F., Kasischke, E. S., Hall, R. J., Murphy, K. A., Verbyla, D. L., Hoy, E. E., et al. (2008). Using Landsat data to assess fire and burn severity in the North American boreal forest region: an overview and summary of results. Int. J. Wildland Fire 17, 443–462. doi: 10.1071/WF08007

CrossRef Full Text | Google Scholar

Goetz, S. J., Fiske, G. J., and Bunn, A. G. (2006). Using satellite time-series data sets to analyze fire disturbance and forest recovery across Canada. Remote Sens. Environ. 101, 352–365. doi: 10.1016/j.rse.2006.01.011

CrossRef Full Text | Google Scholar

Goulden, M. L., McMillan, A., Winston, G., Rocha, A., Manies, K., Harden, J. W., et al. (2011). Patterns of NPP, GPP, respiration, and NEP during boreal forest succession. Glob. Change Biol. 17, 855–871. doi: 10.1111/j.1365-2486.2010.02274.x

CrossRef Full Text | Google Scholar

Greene, D. F., Zasada, J. C., Sirois, L., Kneeshaw, D., Morin, H., Charron, I., et al. (1999). A review of the regeneration dynamics of North American boreal forest tree species. Can. J. For. Res. 29, 824–839. doi: 10.1139/x98-112

CrossRef Full Text | Google Scholar

Hansen, W. D., Romme, W. H., Ba, A., and Turner, M. G. (2016). Shifting ecological filters mediate postfire expansion of seedling aspen (Populus tremuloides) in Yellowstone. For. Ecol. Manage. 362, 218–230. doi: 10.1016/j.foreco.2015.12.012

CrossRef Full Text | Google Scholar

Harden, J., Trumbore, S., Stocks, B., Hirsch, A., Gower, S., O’neill, K., et al. (2000). The role of fire in the boreal carbon budget. Glob. Change Biol. 6, 174–184. doi: 10.1046/j.1365-2486.2000.06019.x

CrossRef Full Text | Google Scholar

Hart, S. A., and Chen, H. Y. H. (2006). Understory vegetation dynamics of North American boreal forests. Crit. Rev. Plant Sci. 25, 381–397. doi: 10.1093/biosci/biw080

PubMed Abstract | CrossRef Full Text | Google Scholar

Harvey, B. J., Donato, D. C., and Turner, M. G. (2016). High and dry: post-fire tree seedling establishment in subalpine forests decreases with post-fire drought and large stand-replacing burn patches. Glob. Ecol. Biogeogr. 25, 655–669. doi: 10.1111/geb.12443

CrossRef Full Text | Google Scholar

Hicke, J. A., Asner, G. P., Kasischke, E. S., French, N. H. F., Randerson, J. T., James C. G., et al. (2003). Postfire response of North American boreal forest net primary productivity analyzed with satellite observations. Glob. Change Biol. 9, 1145–1157. doi: 10.1046/j.1365-2486.2003.00658.x

CrossRef Full Text | Google Scholar

Johnstone, J. F., Chapin Iii, F., Foote, J., Kemmett, S., Price, K., and Viereck, L. (2004). Decadal observations of tree regeneration following fire in boreal forests. Can. J. For. Res. 34, 267–273. doi: 10.1139/x03-183

CrossRef Full Text | Google Scholar

Johnstone, J. F., Hollingsworth, T. N., and Chapin, F. S. (2008). A Key for Predicting Postfire Successional Trajectories in Black Spruce Stands of Interior. Juneau, AK: US Department of Agriculture, Forest Service, Pacific Northwest Research Station. doi: 10.2737/PNW-GTR-767

CrossRef Full Text | Google Scholar

Johnstone, J. F., Hollingsworth, T. N., Chapin, F. S., and Mack, M. C. (2010). Changes in fire regime break the legacy lock on successional trajectories in Alaskan boreal forest. Glob. Change Biol. 16, 1281–1295. doi: 10.1111/j.1365-2486.2009.02051.x

CrossRef Full Text | Google Scholar

Jones, M. O., Kimball, J. S., and Jones, L. A. (2013). Satellite microwave detection of boreal forest recovery from the extreme 2004 wildfires in Alaska and Canada. Glob. Change Biol. 19, 3111–3122. doi: 10.1111/gcb.12288

PubMed Abstract | CrossRef Full Text | Google Scholar

Kashian, D. M., Romme, W. H., Tinker, D. B., Turner, M. G., and Ryan, M. G. (2006). Carbon storage on landscapes with stand-replacing fires. Bioscience 56, 598–606. doi: 10.1641/0006-3568(2006)56[598:CSOLWS]2.0.CO;2

CrossRef Full Text

Kasischke, E. S., Verbyla, D. L., Rupp, S., McGuire, A. D., Murphy, K. A., Jandt, R., et al. (2010). Alaska’s changing fire regime – implications for the vulnerability of its boreal forests. Can. J. For. Res. 40, 1313–1324. doi: 10.1139/X10-098

CrossRef Full Text | Google Scholar

Kumar, L., Skidmore, A. K., and Knowles, E. (1997). Modelling topographic variation in solar radiation in a GIS environment. Int. J. Geogr. Inform. Sci. 11, 475–497. doi: 10.1080/136588197242266

CrossRef Full Text | Google Scholar

Lecomte, N., Simard, M., and Bergeron, Y. (2006). Effects of fire severity and initial tree composition on stand structural development in the coniferous boreal forest of northwestern Québec, Canada. Ecoscience 13, 152–163. doi: 10.2980/i1195-6860-13-2-152.1

CrossRef Full Text | Google Scholar

Lentile, L. B., Holden, Z. A., Smith, A. M. S., Falkowski, M. J., Hudak, A. T., Morgan, P., et al. (2006). Remote sensing techniques to assess active fire characteristics and post-fire effects. Int. J. Wildland Fire 15, 319–345. doi: 10.1071/WF05097

CrossRef Full Text | Google Scholar

Meng, R., Dennison, P. E., Huang, C., Moritz, M. A., and D’Antonio, C. (2015). Effects of fire severity and post-fire climate on short-term vegetation recovery of mixed-conifer and red fir forests in the Sierra Nevada Mountains of California. Remote Sens. Environ. 171, 311–325. doi: 10.1016/j.rse.2015.10.024

CrossRef Full Text | Google Scholar

Miller, J., Knapp, E., Key, C., Skinner, C., Isbell, C., Creasy, R., et al. (2009). Calibration and validation of the relative differenced Normalized Burn Ratio (RdNBR) to three measures of fire severity in the Sierra Nevada and Klamath Mountains, California, USA. Remote Sens. Environ. 113, 645–656. doi: 10.1016/j.rse.2008.11.009

CrossRef Full Text | Google Scholar

Miller, J. D., and Thode, A. E. (2007). Quantifying burn severity in a heterogeneous landscape with a relative version of the delta Normalized Burn Ratio (dNBR). Remote Sens. Environ. 109, 66–80. doi: 10.1016/j.rse.2006.12.006

CrossRef Full Text | Google Scholar

Mouillot, F., Schultz, M. G., Yue, C., Cadule, P., Tansey, K., Ciais, P., et al. (2014). Ten years of global burned area products from spaceborne remote sensing – A review: analysis of user needs and recommendations for future developments. Int. J. Appl. Earth Obs. Geoinf. 26, 64–79. doi: 10.1016/j.jag.2013.05.014

CrossRef Full Text | Google Scholar

Pan, Y., Chen, J. M., Birdsey, R., McCullough, K., He, L., and Deng, F. (2011). Age structure and disturbance legacy of North American forests. Biogeosciences 8, 715–732. doi: 10.5194/bg-8-715-2011

CrossRef Full Text | Google Scholar

Randerson, J., Liu, H., Flanner, M., Chambers, S., Jin, Y., Hess, P., et al. (2006). The impact of boreal forest fire on climate warming. Science 314, 1130–1132. doi: 10.1126/science.1132075

PubMed Abstract | CrossRef Full Text | Google Scholar

Shryock, D. F., Esque, T. C., and Chen, F. C. (2015). Topography and climate are more important drivers of long-term, post-fire vegetation assembly than time-since-fire in the Sonoran Desert, US. J. Veg. Sci. 26, 1134–1147. doi: 10.1111/jvs.12324

CrossRef Full Text | Google Scholar

Shuman, J. K., Shugart, H. H., and O’Halloran, T. L. (2011). Sensitivity of Siberian larch forests to climate change. Glob. Change Biol. 17, 2370–2384. doi: 10.1111/j.1365-2486.2011.02417.x

CrossRef Full Text | Google Scholar

Soverel, N. O., Coops, N. C., Perrakis, D. D. B., Daniels, L. D., and Gergel, S. E. (2011). The transferability of a dNBR-derived model to predict burn severity across 10 wildland fires in western Canada. Int. J. Wildland Fire 20, 518–531. doi: 10.1071/WF10081

CrossRef Full Text | Google Scholar

Tautenhahn, S., Lichstein, J. W., Jung, M., Kattge, J., Bohlman, S. A., Heilmeier, H., et al. (2016). Dispersal limitation drives successional pathways in Central Siberian forests under current and intensified fire regimes. Glob. Chang Biol. 22, 2178–2197. doi: 10.1111/gcb.13181

PubMed Abstract | CrossRef Full Text | Google Scholar

Turner, M. G., Tinker, D. B., Romme, W. H., Kashian, D. M., and Litton, C. M. (2004). Landscape patterns of sapling density, leaf area, and aboveground net primary production in postfire lodgepole pine forests, Yellowstone National Park (USA). Ecosystems 7, 751–775. doi: 10.1007/s10021-004-0011-4

CrossRef Full Text | Google Scholar

Walker, X. J., Mack, M. C., and Johnstone, J. F. (2015). Stable carbon isotope analysis reveals widespread drought stress in boreal black spruce forests. Glob. Chang Biol. 21, 3102–3113. doi: 10.1111/gcb.12893

PubMed Abstract | CrossRef Full Text | Google Scholar

Wu, Z., He, H. S., Liang, Y., Cai, L., and Lewis, B. J. (2013). Determining relative contributions of vegetation and topography to burn severity from LANDSAT imagery. Environ. Manage. 52, 821–836. doi: 10.1007/s00267-013-0128-3

PubMed Abstract | CrossRef Full Text | Google Scholar

Xu, H. C. (1998). Forests in the Great Xing’an Mountains of China. Beijing: Science Press.

Young, A. M., Higuera, P. E., Duffy, P. A., and Hu, F. S. (2016). Climatic thresholds shape northern high-latitude fire regimes and imply vulnerability to future climate change. Ecography 39, 1–12.

Google Scholar

Zhou, Y. (1991). Vegetation of Da Hinggan Ling in China. Beijing: Science Press.

Google Scholar

Zinko, U., Seibert, J., Dynesius, M., and Nilsson, C. (2005). Plant species numbers predicted by a topography-based groundwater flow index. Ecosystems 8, 430–441. doi: 10.1007/s10021-003-0125-0

CrossRef Full Text | Google Scholar

Keywords: fire disturbance, climate change, environmental filtering, seed availability, forest recovery, boosted regression tree analysis

Citation: Cai WH, Liu Z, Yang YZ and Yang J (2018) Does Environment Filtering or Seed Limitation Determine Post-fire Forest Recovery Patterns in Boreal Larch Forests? Front. Plant Sci. 9:1318. doi: 10.3389/fpls.2018.01318

Received: 15 January 2018; Accepted: 21 August 2018;
Published: 11 September 2018.

Edited by:

Pan Kaiwen, Chengdu Institute of Biology (CAS), China

Reviewed by:

Andrey S. Zaitsev, Justus-Liebig-Universität Gießen, Germany
Fei-Hai Yu, Taizhou University, China

Copyright © 2018 Cai, Liu, Yang and Yang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

*Correspondence: Zhihua Liu, liuzh@iae.ac.cn Jian Yang, jian.yang@uky.edu