Datasets, code, and metadata associated with,
Hansen et al, (in press). An alternate vegetation type proves resilient and persists for decades following forest conversion in the North American boreal biome. Journal of Ecology
1. Changing climate and natural disturbances are increasingly causing forests to transition to alternate vegetation types (e.g., new tree species assemblages, grasslands). Determining whether and how these new vegetation types will persist is essential for forecasting earth system function during this century. However, long-term studies of past disturbance-induced forest conversion are rare, and future climates may not have a present day analog. Both limit the utility of empirical approaches for evaluating the fate of alternate vegetation types.
2. We conducted individual-based simulations to test how changing climate, disturbance, and biotic interactions shape the resilience of deciduous broadleaf forest, which has begun to replace spruce after severe wildfires in interior Alaska, USA.
3. Deciduous forest persisted in 86% of simulated stands and was especially resilient with fire return intervals of 50 years or shorter. However, when transitions to another vegetation type did occur, mixed forest was most common, particularly when fire return intervals were longer than 50 years and when seed source was distant. Recovery to spruce forest almost never occurred. Moose browsing and postfire drought also influenced outcomes, but effects were contingent on fire-regime characteristics. When fire return intervals were long and postfire seed sources were 500 m away or farther, moose browsing reduced deciduous sapling growth and survival, helping spruce better compete. Late 21st-century drought following short-interval fire was sufficient to occasionally cause conversion to nonforest.
4. Synthesis. Our analyses indicate that emerging postfire deciduous forest will almost certainly prove resilient for decades to centuries, which will shape biophysical and biogeochemical feedbacks to climate and alter subsequent disturbance. This paper offers a framework for quantifying the long-term resilience of alternate vegetation types following forest conversion and lends critical insights into the biotic and abiotic agents that are likely to underpin similar vegetation transitions across the North American boreal biome.
Please see attached file, table_descriptions_stability_project.csv for descriptions of each of the data tables in this project.
Earth Institute at Columbia University
Center for Climate and Life at Columbia University
Alaska Climate Science Center
National Park Service Task Agreement P17AX01470, National Park Service Fuels Reserve Fund
Holstrom Undergraduate Research Fellowship, University of Wisconsin Madison
National Science Foundation, AGS-1243204
Geographic descriptionGeographic Coordinates:
Lower right: 63.0998°, -142.5507°
Lower left: 63.0692°, -158.2863°
Upper left: 63.0692°, -158.2863°
Upper right: 66.8758°, -159.3156°
Time periodBegin date: January, 1 2018
End date: February, 1 2020
We conducted a simulation experiment to test the resilience of an alternate forest state using the process-based forest model iLand (Online documentation: http://iland.boku.ac.at).
Initial forest structure and topoedaphic conditions
We initialized iLand based on field surveys of forests that burned across interior Alaska during the large fire season of 2004. Before the 2004 fires, all stands were dominated by black spruce. We selected stands for this experiment that burned severely and then transitioned to deciduous forest (trembling aspen or Alaskan birch comprised ≥ 50% of postfire stand density). This yielded 31 independent early postfire deciduous stands (1 ha size) scattered along a ~500 km transect in interior Alaska for the experiment.
For each of the stands selected from the field data, we extracted geographically corresponding soil characteristics including percent sand, silt, clay, and effective soil depth, derived from the Alaska STATSGO database. Relative soil fertility (expressed as plant available nitrogen) was set at 45 kg ha -1 yr -1, which aligns with field based measurements of plant available nitrogen in mixed forest and deciduous forest of interior Alaska. Remote sensing indicated that most of the selected stands (~75%) are already not underlain by near-surface permafrost, and the extent of permafrost is expected to decrease in coming decades. Thus, we assumed permafrost was not present in simulations.
The model was initialized so that tree species composition, stand age (11-years postfire), and stem densities matched field measurements. Seedling heights were not available from field observations. Thus, we fit species-specific distributions (gamma for aspen and spruce, Weibull for birch) to seedling-height data collected in a subset of the sampled fires and assigned initial heights to simulated seedlings and saplings by drawing randomly from these distributions.
Beginning 11 years after the 2004 fires, deciduous stands were simulated independently from one another under contemporary and future climate conditions for varying lengths of time until they were scheduled to burn in a second severe fire. After the second fire, stands were simulated for an additional 50 years with the amount of seed arriving at each stand varying as if spruce and deciduous seed sources were varying distances away. Throughout the simulations, both spruce and deciduous trees were subjected to different browsing intensities.
We represented effects of current and future climate on simulated stands by forcing the model under three sets of climate conditions: a contemporary period (1971-2000), a mid-21st century period (2031-2060), and a late-21st century (2071-2100) period. iLand ingests daily climate (minimum and maximum temperature, sum of precipitation, sum of shortwave radiation, and mean vapor pressure deficit). For each climate period, we used daily meteorological data simulated by three different GCMs that have been shown to represent historical climate well in Alaska and span the range of projected 21st-century conditions. The models were the MRI-CGCM3, IPSL-CM5A-LR, and CCSM4 GCMs. All future simulations were forced with the RCP 8.5 emissions scenario. Modeled daily meteorological grids were statistically downscaled to a 1-km spatial resolution using a quantile matching approach. Annually, a random year of climate data was chosen with replacement from the appropriate 30-yr climate period to ensure the order of the climate record did not influence simulation results.
Fire return interval
Each stand was simulated until scheduled to burn at 11, 50, 125, and 250 years after the initial 2004 fire. This spans the observed range of FRIs from very short-interval fires to long FRIs commonly observed during the Holocene. In this application, fire ignition and spread were not simulated explicitly. Instead, we represented the effects of stand-replacing fires by prescribing 100% mortality to all the prefire trees, saplings, and seedlings in each burned stand.
Distance to seed source
In iLand, the probability of seeds distributed to a given distance from the source is simulated using a two-part exponential equation that jointly accounts for short (e.g., wind) and long distance (e.g., animals) dispersal agents. Dispersal probability declines with distance from seed source. Following the scheduled fire, we simulated each stand as if the nearest spruce and deciduous seed source was 50 m, 500 m, or 1 km away. Treatments were chosen to reflect distances to seed source commonly measured in the field surveys that we used to initialize the simulation experiment. Black spruce trees could also serve as an in situ seed source following simulated fires because of their semi-serotinous cones. Simulated aspen and birch trees that burned could also resprout asexually.
Deciduous and black spruce seedlings and saplings shorter than two meters are commonly browsed by moose and snowshoe hares, respectively, in interior Alaska. Thus, we accounted for effects of browsing on both species. In the model, a user defined species-specific probability of browsing can be applied. This determines the annual percentage of seedlings and saplings that are randomly selected to be browsed. If a seedling or sapling is browsed in a given year, annual biomass production does not occur (i.e., stems do not grow and leaves are not produced). This causes stress to build, as carbon reserves are reduced, which may eventually kill the simulated seedling or sapling.
Background probabilities of browsing for deciduous (20 %) and black spruce (1.5 %) seedlings and saplings were based on field surveys. To evaluate effects of more intense browsing, we compared simulated stands with the background level of browsing to stands that experienced medium intensity browsing (45% and 15% of deciduous and black spruce < 2 m tall) and chronic browsing (60% and 25% of deciduous and black spruce < 2 m tall).
Because annual climate conditions were randomly selected from respective 30-yr climate periods, five replicates of each of the 31 independent stands were run for all combinations of climate period (three levels), GCM (three levels), FRI (four levels), distance to seed source (three levels), deciduous browsing intensity (three levels), and black spruce browsing intensity (three levels) (n = 150,660 total simulated stands).
Model outputs and data analysis
We quantified the mean percentage of simulated stands that remained deciduous forest across replicates, and the mean percentage that transitioned to mixed forest (co-dominance of deciduous species and spruce), black spruce forest, or nonforest by 50 years after simulated fire. Based on all trees with a DBH >= 2.5 cm, stands were considered deciduous forest if the importance value (IV, deciduous proportion of density plus deciduous proportion of basal area; values range from zero to two) was above 1.5. They were considered mixed forest when deciduous IV ranged from 0.5 to 1.5, and were considered black spruce forest if deciduous IV was below 0.5. Stands were classified as nonforest if there were fewer than 50 stems ha-1.
Random forest classification was used to quantify how the dominant vegetation type (deciduous forest, mixed forest, black spruce forest, nonforest) at the end of simulations differed across combinations of the experimental treatments and varied by simulation replicate.
Taxonomic species or groupsPicea mariana, Populus tremuloides, Betula neoalaskana
SoftwareR code files:
Analysis of the condensed simulation output for results in the paper
Analysis of the climate data for results in the paper
Data provenance1. Downscaled daily GCM data for Alaska. A. Park Williams (firstname.lastname@example.org)
2. Alaska 2004 Burns: Densities of tree seedlings after fire; measured in 2006, 2008, 2011, and 2017 by Joint Fire Science Program. https://www.lter.uaf.edu/data/data-detail/id/398. Jill Johnstone (email@example.com)
Investigator 1Winslow D. Hansen
Investigator 2Ryan Fitzsimmons
University of Wisconsin Madison
Investigator 3Justin Olnes
Alaska Department of Fish and Game
Investigator 4A. Park Williams