LD1-1: Fire enhances regional resilience by maintaining a mosaic of flammable and non-flammable vegetation. Climatic or human changes in fire regime will therefore likely alter the landscape mosaic

1.1 Field sampling

The primary goal of data collection activities is characterization of vegetation age structure for different vegetation (i.e., fuel) types. Estimates of mean fire intervals (MFIs) derived from the age structure data will be checked using estimates of MFIs derived from fire scarred trees (see below). Within each of the three study areas, we will visit 250 randomly chosen plots. A combination of fixed-wing aircraft, helicopter, waterway, and ground transportation will be used for field transport. Each plot will encompass 1 km², the spatial resolution used in Boreal ALFRESCO. Plots will be stratified by fuel type, with 50 plots in each of the 5 vegetation/ fuel types. In each plot, we will ground truth the vegetation and fuel type, and describe organic soil horizons (an important fuel component in the boreal forest). We will determine time since last fire (TSLF) in each plot by felling 10 randomly chosen trees and retrieving a full-cross section from each near the root crown. This sampling density will equal or exceed that of similar studies in the boreal forest. We will also search out and collect cross sections from any fire-scarred trees within a kilometer of our landing zone.

1.1.1. Laboratory Work: Ring Counts and Statistical Analyses

Cross sections will be trimmed and polished, then their rings counted and ring widths measured on a Bannister Bench under a binocular dissecting microscope. All sections will be cross dated to correct for missing rings. Initial cross dating will be done visually using marker rings. These results will be checked using the computer program COFECHA.

We will derive estimates of the hazard-of-burning for different fuel/vegetation types from data describing times since last fire (TSLF). In the boreal forest of interior Alaska, fire is the dominant mortality agent for trees, and tree ages describe TSLF and fire-interval distributions. We will use likelihood techniques to estimate parameters in statistical models of fire-interval distribution. Using methods from the statistical field of survival analysis, the probability of fire (mortality) during a given time interval can be expressed analytically according to stand age. Parameter estimates from the maximum likelihood estimation will be used to specify an analytic hazard function (based on the data) for each vegetation/fuel type within the fire subroutine in Boreal ALFRESCO. Through the use of exploratory data analysis, a statistical model will be constructed for the TSLF data. The distribution chosen for the statistical model (i.e. exponential, Weibull) will determine the explicit form of the corresponding hazard-of-burning as a function of tree age. The use of the negative exponential distribution for fire intervals assumes that the hazard-of-burning is constant through time. If the data suggest this is not the case, we will use another distribution such as the Weibull or Gamma. These distributions can accommodate hazards that change with tree age. We will then use the Chi-square test and the Anderson-Darling goodness of fit tests to evaluate the statistical significance of differences between empirical data and the statistical model for fire frequency. For each vegetation/fuel type and study area (different climate regimes) a different hazard function will be utilized to depict fire-vegetation-climate interactions in Boreal ALFRESCO.

We will use fire scars in tree cross sections to check the estimates of fire frequency made from tree-age distributions. Though fire scars are not common in interior Alaska compared to pine forests in the Rockies, they do occur and are common in some firebreak areas. Fire intervals are the periods between two consecutive fires or between tree establishment and the first fire scar. From fire-interval distributions for different fuel/vegetation types within each study area, we will derive MFIs, the average number of years between consecutive fire dates in a composite chronology. The construction of a composite (or master) fire chronology implicitly results in a measure of fire frequency that is dependent on the size of the study region, unless one can be sure that each fire recorded in the chronology burned uniformly in the area of interest. When this assumption is not satisfied, the estimate of fire frequency increases as the size of the study area increases. For this reason, MFI data can be used only as a lower bound on estimates of fire frequency that are obtained from the stand age analysis. Nonetheless, scar-derived estimates of MFI will provide a valuable check on the results of fire frequency estimated by stand age analysis.

1.2. Modeling Component

The changes we plan to make in existing versions of ALFRESCO emphasize improvements in how the model depicts feedbacks between vegetation, fuel, and fire regime. The current version of Boreal ALFRESCO simulates fire ignition and spread using coarse-scale rules relating ignition and fire-spread hazard to climate, vegetation type, and stand age. The model originally was parameterized to simulate general fire frequency trends in Alaska as reported in the literature. Our proposed research aims to improve the precision of fire hazard-fuel load relationships by using results from survival analysis. This will result in a more accurate simulation of fire regimes. We will use the hazard-of-burning functions developed from the analysis of field data (see above) to develop algorithms for the five major boreal forest fuel types. In addition, the climatic gradient of the study region will provide empirical data for improved representation of fire hazard-climate relationships within and among fuel types in interior Alaska. Currently, the successional dynamics component of Boreal ALFRESCO simulates five fuel types, including a dry grassland type. In this project, we will develop new frames (i.e., fuel types) that specifically represent the seven major fuel types that are currently utilized in fire management. The major fuel types consist of five forest types (deciduous forest, open black spruce forest, closed black spruce forest, upland white spruce forest, and riparian white spruce forest) and two non-forest types (tussock tundra and upland tundra). We will utilize the current flammability functions in Boreal ALFRESCO for the tundra fuel types, which were developed from the Alaska fire literature.

The new model version will function at a 1 x 1 km grid cell resolution and will have interior Alaska as its spatial domain. We are using a 1 x 1 km spatial resolution for three reasons. First, fires in Interior Alaska tend to be very large, and the spatial distribution of vegetation is homogeneous relative to that of the Intermountain West. Second, although 30 x 30 m data exist for our three study areas, only 1 x 1 km data are available for much of Interior Alaska. Third, computational efficiency inhibits finer resolution model simulations of the entire spatial domain. This is important because the objective is to provide an efficient decision support tool that can be easily used by managers. The current model structure runs on a 10-yr-time step and simulates system dynamics on century to millennia time scales. For this project, we will modify Boreal ALFRESCO to run on an annual time step to simulate inter-annual variability of the fire regime over decadal to century time scales in response to changing climate and management policies.

1.2.1 Testing the Model

In addition to informing model development, the field data and hazard-of-burning functions will provide tests of model performance in the three study areas. We will test the model by showing that the variability and measures of central tendency for the output variables of interest from the simulation results (e.g., number of fires and area burned) are reasonable when compared with the field data. In other words, we will test the model by seeing how closely it can predict the vegetation, fuel, and fire regime that we actually find on the present landscape. We will accomplish this by running the model to equilibrium (i.e., until vegetation and climate stabilize) over a period of several thousand years and under current climate conditions. Once it passes this test, Boreal ALFRESCO will be used to explore the interactions between climate, fire regime, vegetation, and fire management under different scenarios of future change.

1.2.2 Model Application to Management Issues

Following model calibration and testing, we will work to develop it into an useful management tool that is capable of simulating fire risk in interior Alaska and predicting the impacts of fires and fire management on ecosystem dynamics. The first step will be to develop georeferenced input data on fuel types across the model domain. This work already has begun by the Alaskan Wildland Fire Coordinating Group. The model will use wildland fuel maps to estimate current fire risk throughout Interior Alaska based on the empirically derived hazard-of-burning functions. Boreal ALFRESCO dynamically simulates climate, the timing and location of fires, and the response of the individual fuel types to environmental change. Once the model is calibrated using field data, the only input data required for model simulations are the initial vegetation conditions and the user defined management options and climate scenario. This initial modeling exercise will allow fire managers to conduct assessments of present day fire risk in communities adjacent to Federal wildlands.

The second collaborative task will be to develop predictions of fire-risk under different fire-management and climate scenarios. Alaska’s interagency fire-management plans designate several fire-management options, which range from aggressive action on any new fire start to monitoring and structure protection with no interference with natural fire spread. Options are reviewed annually by land-management agencies comprising the Alaska Wildland Fire Coordinating Group. Boreal ALFRESCO will be able to analyze the long-term effects of different fire-management options simultaneously with different scenarios of human development, land-use, and wildlife populations. In addition, we will be able to input future climate-change scenarios covering the expected range of variability associated with anthropogenic climate change. The model will output short-term, long-term, and cumulative impacts of different management scenarios under different regimes of climate change. Impacts of using different scales will be assessed and output as well. Comparison of the results of model runs with both increased and decreased spatial resolution will be performed to accommodate differing management needs. Mapped time series of vegetation, fuel, and fire regime changes will provide Federal land managers with a useful planning tool for assessing management options.

The third application of Boreal ALFRESCO will be to model how changes in fire regime impact natural ecosystems. Rupp has initiated a pilot study complementing a research project funded by the Department of the Interior Fire Research Committee, led by USGS–BRD and ADF&G, to investigate how changes in fire regime affect caribou winter foraging habitat. Caribou are an important subsistence resource and game animal in Alaska. In winter, they rely heavily on lichens, which are abundant only during certain stages of post-fire succession. Wildlife managers want to know how changes in fire regime occurring under different fire-management and climate-change scenarios could impact caribou winter forage. Boreal ALFRESCO will predict future fire regimes and resultant vegetation distribution. From model output, we can make maps depicting how the quality of caribou winter range might change over decades to centuries.