Fire effects on seedling establishment success across treeline: implications for future tree migration and flammability in a changing climate

Understanding the complex mechanisms controlling treeline advance or retreat in the arctic and subarctic has important implications for projecting ecosystem response to changes in climate. Changes in landcover due to a treeline biome shift would alter climate feedbacks (carbon storage and energy exchange), ecosystem services such as wildlife and berry habitat, and landscape flammability. Wildfire frequency and extent has increased in the last half-century in the boreal forest and tundra in response to warmer weather and lower precipitation. Invasion of tundra by trees may be facilitated by wildfire disturbance, which exposes new seedbeds, increases nutrient availability immediately post-fire, and creates opportunities for establishment in an ecosystem where tree recruitment is otherwise rare. Coupled with projects specifically investigating biotic factors influencing tree seedling establishment, we evaluated the regional abiotic factors governing seedling performance and establishment success across treeline after fire. Addressing the Joint Fire Science Program (JFSP) Graduate Research INnovation (GRIN) topic of climate change and fire effects, we investigated regional controls on seedling growth across a latitudinal treeline gradient post-fire in Alaska. We used samples from a tree seedling out-plant experiment and an observational study of naturally established tree seedlings to investigate how establishment success and physiological performance is limited by drought stress and nutrient acquisition across the treeline ecotone. We developed a conceptual model of the abiotic and biotic factors that govern seedling performance at treeline and tundra. This conceptual model has been implemented in ALFRESCO, a landscape-level model of vegetation-fire-climate dynamics. Using ALFRESCO we are investigating the potential for fire-initiated tree migration. Background and purpose The location of arctic treeline has important effects on the climate system through changes in carbon storage (McGuire et al. 2001), regional albedo (Chapin et al. 2000), and the position of the Arctic Front (Pielke & Vidale 1995). Expansion of Alaska treeline is a climatically relevant event due to positive feedbacks to climate warming that arise with conversion of tundra to forest (Chapin et al. 2005). Advance of arctic treeline is usually modeled in Dynamic Global Vegetation Models as a function of temperature, where increases in temperature lead to more trees. However there is an increasing body of literature to suggest that, at treeline, warming may negatively affect growth, reproduction, and establishment of trees due to drought stress (Wilmking & Juday 2005). Ecosystem resilience to vegetation change at treeline can be understood through exposure to climate warming and through sensitivity to climate-induced changes in the fire regime. A warming climate promotes disturbances such as fire (Kasischke & Turetsky 2006), which facilitate seedling establishment (Lloyd & Fastie 2003), growth (Hobbie & Chapin 1998), and affects treeline resilience. The boreal forest is North America’s largest biome (Johnson et al. 1995), and wildfires are the dominant large-scale disturbance in Alaskan boreal forests. In Alaska, changes in climate have directly affected the fire regime in the boreal forest (Rupp et al. 2007). From 1959–99 there has been a doubling of both the annual area burned and frequency of large fire years in North American boreal forests (Kasischke & Turetsky 2006). As climate changes, fires at treeline and tundra are expected to become more frequent due to drier fuel loads and more frequent thunderstorms. In the boreal forest, fire is a major driver of tree recruitment as well as tree migration (Johnstone et al. 2004). Post-fire successional trajectories are closely related to fire severity, which affects density and composition of tree seedlings that establish after fire. High-severity fires expose mineral soils, which are favorable to seedling germination, net seedling establishment, and growth of transplanted seedlings(Johnstone & Chapin 2006). In addition, northern migration of lodgepole pine into Alaska appears to be tightly linked to high-severity fires (Johnstone & Chapin 2003). Because fire disturbance is critical to boreal forest vegetation patterns, it seems likely that a changing fire regime will dramatically affect where on the landscape trees will occur in the future. In Alaska recent studies have focused on fire effects on post-fire successional trajectories in black spruce forests; however, impacts of fire on successional trajectories and future flammability at treeline forest sites are unknown. In order to make accurate predictions of the rate and magnitude of future changes in boreal-tundra landcover in the next decades, it is important to understand the interacting factors that can influence tree seedling recruitment and performance at treeline. The Joint Fire Science Program (JFSP) Graduate Research INnovation (GRIN) project allowed us to investigate the relative importance of abiotic factors pertaining to fire severity, drought stress, and nutrient availability. The research was in relation to Rebecca Hewitt’s dissertation projects focused on fire-severity effects on symbiotic fungal communities and the impacts of plant-fungal interactions of tree seedling growth and establishment at treeline after fire. With the completion of the JFSP GRIN data collect, we were able to develop a hierarchy of biotic and abiotic filters that determine seedling establishment and performance at treeline after fire. In order to assess treeline ecosystem vulnerability to climate-induced changes in fire activity, we developed a conceptual model of abiotic and biotic controls over seedling establishment based on our field observations and the literature. This conceptual framework has been implemented in the spatially explicit fire-climatevegetation landscape model, Alaska Frame-Based Ecosystem Code (ALFRESCO)(Rupp et al. 2000). This project links mechanistic investigation of regional patterns of seedling establishment to model projects of continental biome shifts after fire using a novel suite of analytical tools in order to address fire effects on treeline successional trajectories and future flammability. Study description and location Study sites: The study area includes a burned arctic tundra site north of latitudinal treeline and two burned treeline sites in the upland boreal forest of interior Alaska bounded by the Brooks Range and latitudinal treeline to the north (67 ̊N) and the Alaska Range to the south (63 ̊N) (Fig. 1). Tundra fire sampling occurred in the burn scar of the 2007 Anaktuvuk River Fire. This fire burned 1000 km on the North Slope of Alaska at mostly moderate to high severities. Latitudinal treeline sampling was focused in the uplands of the southern foothills of the Brooks Range at Finger Mountain. Alpine treeline sampling was focused in the White Mountains. Forest cover is dominated by black spruce (Picea mariana) with patches of trembling aspen (Populus tremuloides) and Alaskan paper birch (Betula neo-alaksana), and treeline cover is primarily white spruce (Picea glauca). The treeline sites burned in the widespread fires of 2004 that burned over 2.7 million ha of forest across the interior of Alaska. Figure 1: Map of study sites for outplanted and naturally established seedlings. Orange polygons represent burn scars. Seedlings were outplanted at the two northern-most sites in tundra and at latitudinal treeline. Naturally established seedlings were sampled at the latitudinal treeline site and the alpine treeline to the south. The sites are affiliated with the two NSF LongTerm Ecological Research Sites in Alaska. Sampling design and measurements: In 2009 we sampled naturally established seedlings at alpine and latitudinal treeline sites that had burned in 2004. We excavated root systems to sample for fungal symbionts, ectomycorrhizal fungi (EMF). We noted the substrate, dominant vegetation within one m of the focal seedling, the depth of the organic layer (fire severity metric), and the distance between the seedling and the closest shrub. These samples of naturally established seedlings were used to assess the relative importance of abiotic and biotic factors for seedling biomass. In 2009 we out-planted seedlings (white spruce, black spruce, birch and alder) inoculated with different fungal communities at a latitudinal treeline site and an arctic tundra site, both of which burned recently. The treatments were (1) sterile autoclaved inoculum, (2) unburned boreal forest inoculum (3) burned boreal forest inoculum. We harvested seedlings in 2010 and 2011, and preliminary results show that there was significantly lower survivorship at the latitudinal treeline site than the arctic tundra site, indicating that abiotic factors may impact seedling establishment success more than fungal symbiont availability at some sites. For the JFSP GRIN project we collected data on foliar and soil isotopes to determine the relative importance of abiotic factors on seedling attributes. Foliar isotope signatures, which provide an integrative measurement of seedling performance, were related to soil isotopes and previously developed fungal community profiles for each seedling. We measured the natural abundance of δN and δC in leaves and soil, as well as C and N concentrations using Isotope Ratio Mass Spectrometry. During harvest of the out-planted seedlings in 2010 and 2011, we separated current years growth, dried the material, and ground the foliar tissue. With funding from JFSP GRIN, we ran the foliar and soil samples on an Elemental Analyzer (ECS 4010, Costech Analytical, Valencia, California, USA) coupled to a Delta Plus XL Stable Isotope Ratio Mass Spectrometer (Delta Plus XL, Finnigan, Bremen, Germany) via a Conflo III (Finnigan, Bremen, Germany). Foliar δC signatures provide an integrated indicator of photosynthetic activity and water relations with enriched δC signatures indicating greater drought stress. Foliar δN signatures in comparison with soil δN signatures indicate soil N pools accessed by seedlings, ecosystem N availability, and N uptake by fungal symbionts. The use of dual isotope analysis in relation to abiotic and biotic drivers of seedling performance is a novel approach to address regional seedling establishment success. Random Forest analysis of abiotic and biotic variable importance: We used Random Forest classification and regression tree method to identify important associations between seedling biomass and explanatory variables: fire severity, fungal and vegetation variables. Random Forest is an ensemble decision tree method that uses an algorithmic approach to make predictions based on the input variables (Breiman 2001b). Multiple decision trees are constructed through a process known as bagging (Breiman 1996), which utilizes a bootstrap sample from a subset of the data (~63% of the data) and the random selection of a subset of the predictor variables. The most important variable from the random subset of predictor variables is then used to produce each node split (Breiman 2001a). The fully-grown (statistical) trees are used to predict the observations excluded from the bootstrap sample, i.e. out-of bag-data (~37% of the full dataset). Predictions are calculated by combining the predictions across the forest of regression trees by averaging. The importance of each explanatory variable is determined when permutations of the observed values for a predictor variable in the out-of-bag data are run down the tree to predict the response variable. If an explanatory variable is important, than the permutation of that variable will have a large effect on the prediction of the response variable (Breiman 2001b). The variable importance is reported as the percent increase in the miscalculation rate, i.e., decrease in accuracy, for each predictor when the predictor is permuted (Breiman 2001b). We used the R package (R Development Core Team, 2013, R Foundation for Statistical Computing, Vienna, Austria), Random Forest to determine which variables are important to predictions of seedling biomass. Using Random Forest we built 500 regression trees using a random sample of the 70 observations of seedling biomass. At each node in the trees, two predictor variables were chosen at random from the nine explanatory variables. Analysis of natural abundance isotopes: We used linear regression to investigate relationships between δC and δN isotopes and seedling taxon, treatment with EMF, and site. We used ANOVA to initially investigate isotope signatures in relation to a site and soil factors. We also investigated the effect of fungal colonization on drought stress and seedling N acquisition. Enriched δC values indicate drought stress, while relative enrichment of δN values will indicate plant uptake of different sources of soil nitrogen and the role of mycorrhizae in nitrogen uptake. Correlations of δN values with particular biotic drivers of site level soil parameters may indicate fire effects on soil nitrogen availability and plant N acquisition. This research produced isotopic signature data for foliar material and soils. As with all Long Term Ecological Research (LTER) projects data from this research will be archived with the Bonanza Creek LTER. Data will be available to those that adhere to the LTER data access policy. Model outputs will be archived by the Scenarios Network for Alaska Planning (SNAP). Final model outputs are accessible via web portals. Documentation and presentations of my findings will be available to fire managers through the Alaska Fire Science Consortium. Synthesizing Results: Modeling the potential for fire-initiated tree migration into previously unforested alpine and arctic tundra. We collaborated with Dr. T. Scott Rupp, director of the SNAP, to develop the conceptual model from our empirical data on seedling establishment success and seedling physiology. We parameterized the model based on our findings from the tree out-planting experiment (drought tolerance and nutrient limitation impacts on performance and survival) and previous investigations of biotic parameters. ALFRESCO simulations now include more biologically accurate mechanisms governing seedling establishment and vegetation transitions from tundra to spruce.