Ager, A. A., M. A. Finney, B. K. Kerns, and H. Maffei. 2007. Modeling wildfire risk to northern spotted owl (Strix occidentalis caurina) habitat in Central Oregon, USA. Forest Ecology and Management 246:45–56.
In this paper, Ager et al. demonstrate a method to quantitatively evaluate the effectiveness of different landscape fuel treatment intensities in reducing wildfire risk to spotted owl habitat. This was the first application of quantitative risk analysis for fuel treatment planning on federal land. First, a subset of forest stands were selected for treatment based on total available treatment effort and each candidate stand’s priority ranking, where a stand’s priority ranking was a function of stand density and location relative to spotted owl habitat. Next, they used Forest Vegetation Simulator (FVS) to simulate the effects of treatment on fuel conditions within each selected stand, and to identify stand-specific fire intensity thresholds for habitat loss. Finally, the authors estimated conditional probabilities for threshold fire behavior at each pixel using 1,000 randomly located ignitions and FlamMap, then converted these to areal quantities to convey the expected loss of spotted owl habitat across the landscape. Importantly, the process described in this paper provides a means for evaluating treatment effectiveness outside the treatments themselves; incorporates fire spread, intensity, and effects; and it does all this using widely accessible tools such as GIS, FVS, and FlamMap. However, the approach has several important shortcomings. First, because their wildfire simulations did not include modeling of spatiotemporal patterns in ignitions or weather, and because they restricted their wildfire simulations to their study landscape alone, they were not capable of calculating absolute burn probabilities. Second, their process for selecting stands for treatment was unlikely to produce optimal treatment designs because it relied on highly simplistic assumptions regarding ignition locations and fire spread direction, ignored potential variation in baseline risk among spotted owl habitats, and ignored the interacting effects of stands treated in combination. Third, like Finney et al. (2007) the paper provides little insight into how to identify and balance the needs of multiple types of resources at risk from wildfire. This was the first case study using Finney’s (2005) risk framework.
Ager, A. A., N. M. Vaillant, and M. A. Finney. 2010. A comparison of landscape fuel treatment strategies to mitigate wildland fire risk in the urban interface and preserve old forest structure. Forest Ecology and Management 259:1556–1570.
Ager et al. (2010) quantitatively evaluated the effectiveness of alternative landscape fuel treatment design strategies for reducing wildfire risk to large trees and wildland-urban interface (WUI) structures within a 16,343-ha landscape in northeast Oregon. Specifically, they compared, across a range of treatment intensities, two treatment design strategies – one that prioritized stands for treatment based on stand density, and the other based on residential density. The authors relied heavily on geographic information systems (GIS), used previously delineated forest stands as candidate treatment units, used Forest Vegetation Simulator (FVS) to simulate the effects of treatment on fuel conditions in each selected stand, and simulated wildfires from 10,000 randomly located ignitions using the minimum travel time (MTT) algorithm within FlamMap. This paper expanded on previous work by Ager et al. (2007) and Finney et al. (2007) by attempting to evaluate treatment designs for their efficacy in achieving multiple land management objectives simultaneously. However, it provided no strategy for designing landscape fuel treatment scenarios that would optimally address multiple objectives, nor did it provide a synthetic metric for quantifying the degree to which multiple objectives have been met.
Ager, A. A., N. M. Vaillant, and M. A. Finney. 2011. Integrating fire behavior models and geospatial analysis for wildland fire risk assessment and fuel management planning. Journal of Combustion 2011:1–19.
In this paper, Ager et al. reviewed available wildfire and vegetation modeling systems, paying particular attention to FlamMap and Forest Vegetation Simulator-Fire and Fuels Extension (FVS-FFE), then went on to describe the use of ArcFuels to integrate these applications for the purposes of fuel management planning and wildfire risk assessment. They pointed out that fuel management planning and risk analyses are increasingly important elements of modern wildfire management in the US, and that these tasks are complex and fundamentally spatial, yet computer-aided geospatial techniques had been poorly adopted by fire managers due to the lack of a clear framework and supporting tools for geospatially integrating the necessary modeling applications. This paper and the tools it described filled that critical gap. FVS-FFE can be used to model stand-level fuel treatment prescriptions, and to build loss-benefit functions that translate fire intensities to fire effects for risk analyses. FlamMap can be used develop surfaces of conditional burn probability and conditional fire intensity, and its treatment optimization model (TOM) can be used to design treatments scenarios that minimize landscape fire spread. Finally, ArcFuels allows users to conveniently integrate these functions from within ArcMap. Specifically, FVS-FFE models fuel treatment prescriptions and loss-benefit functions for stands within a landscape. FlamMap models the resulting burn probability and intensity surfaces for the landscape, and also provides a treatment optimization tool (TOM) that designs landscape treatment scenarios that minimize landscape fire spread. Finally, ArcFuels serves as a wrapper for FVS-FFE and FlamMap, providing convenient integration of their functions from within the ArcMap environment. This framework has been demonstrated by Ager et al. (2007) and Ager et al. (2010).
Ager, A. A., N. M. Vaillant, M. A. Finney, and H. K. Preisler. 2012a. Analyzing wildfire exposure and source–sink relationships on a fire prone forest landscape. Forest Ecology and Management 267:271–283.
Here, authors used wildfire simulation modeling to assess wildfire exposure and transmission properties to various land designations in a 0.6 million ha forested landscape in central Oregon. Specifically, they used landscape fuel data from LANDFIRE, fire weather data derived from recent large fire events, 50,000 randomly located ignitions, and the minimum travel time (MTT) fire spread algorithm within FlamMap to produce surfaces for conditional burn probability, conditional flame length, and source-sink ratio (calculated as the ratio of the size of fires incorporating a pixel to the conditional probability of that pixel burning) across a range of land designations ranging from bald eagle habitat, to research natural areas, to wildland-urban interface. This paper added to similar, previous work by Ager et al. (2007) and Ager et al. (2010) in that it a) used fuels data from a nationally consistent dataset, b) included a buffer around the study area for all wildfire simulations, c) included multiple weather scenarios in its wildfire simulations, and d) attempted to assess variability in how different portions of the landscape contribute to large fire occurrence.
Ager, A. A., N. M. Vaillant, and A. McMahan. 2013. Restoration of fire in managed forests: a model to prioritize landscapes and analyze tradeoffs. Ecosphere 4:art29.
In this paper, Ager et al. demonstrate a method to optimize the locations and prescriptions of fuel treatments in order to maximize ecosystem restoration benefits given constrained treatment capacity. Specifically, they used geographic information systems (GIS) to delineate forest stands, Forest Vegetation Simulator and the Fire and Fuels Extension (FVS-FFE) to model the effects of fuel treatments and wildfire for each stand, and then used Landscape Treatment Designer (LTD, Ager et al. 2012b) to identify optimal treatment designs for preserving old-growth ponderosa pine under different combinations of treatment capacity and stand-level treatment prescriptions. The authors acknowledged several drawbacks of the approach in that it did not test every possible spatial arrangement of treatments, and that it only used wildfire simulations at the stand-scale (FVS-FFE), as opposed to landscape-scale (e.g., FlamMap). However, the authors argued that these aspects of the model were justified because they conferred greater model accessibility to its intended users, and that landscape-scale fire simulations provided little benefit (although no data was presented to support this claim). This paper built on Ager et al. (2012b) by illustrating how LTD can be used to quickly explore tradeoffs among multiple treatment constraint criteria. In addition, it complemented earlier works by Finney (2007) and Wei et al. (2008) in that it broadened the scope of spatial optimization of fuel treatment designs beyond disruption of fire spread to include ecological restoration objectives.
Ager, A. A., N. M. Vaillant, D. E. Owens, S. Brittain, and J. Hamann. 2012b. Overview and example application of the Landscape Treatment Designer.
This report briefly describes and demonstrates Landscape Treatment Designer (LTD), a computer application intended to improve the efficiency of ecosystem management programs by assisting land managers to optimally allocate their treatment effort across the landscape. More specifically, LTD identifies the set of discrete areas on a landscape where treatment of vegetation would yield the greatest simultaneous progress toward multiple land management objectives while also adhering to limitations on total treatment capacity. Operationally, LTD requires 1) a geographic information system (GIS), 2) vector data delineating candidate areas for treatment, 3) the desired and existing conditions for valued attributes in those candidate areas, 4) the relative importance of those attributes, and 5) definition of the limit on total treatment capability. For example, by incorporating fire hazard ratings developed with FlamMap, managers can identify areas where a given amount of treatment effort would lead to the greatest improvement in forest stand density, forest stocking, and fire hazard, simultaneously. In addition, LTD allows land managers to easily explore how variation in treatment capacity, attribute importance, and desired attribute levels affects optimal treatment design.
Finney, M. A. 2007. A computational method for optimising fuel treatment locations. International Journal of Wildland Fire 16:702.
This paper describes and evaluates a hueristic algorithm intended to identify a limited set of locations within a heterogeneous landscape where treatment of fuels would result in the greatest reduction in fire growth. The algorithm used a GIS, a minimum-travel-time (MTT) algorithm for spatially explicit modeling of fire behavior, information on weather conditions, and data on the type of fuel present at each location in the landscape and whether or not it is considered treatable. To evaluate the algorithm’s performance, the authors simulated growth of numerous randomly ignited fires on a sample landscape containing an algorithm-derived pattern of treatments, and compared the resulting fire growth metrics with those resulting from identical simulations on a sample landscape containing randomly located treatments. The authors found that algorithm-derived patterns were more efficient at reducing fire growth than random patterns, for each level of sample landscape heterogeneity, proportion of landscape treated, individual treatment size, and simulated fire duration. However, they note that the advantage of the algorithm-derived treatment pattern declines with the abundance of untreatable areas, which are common in many of the real landscapes where reduced fire growth is desired, and also declines with proportion of the landscape treated.
Martin, A., B. Botequim, T. M. Oliveira, A. Ager, and F. Pirotti. 2016. Temporal optimization of fuel treatment design in blue gum (Eucalyptus globulus) plantations. Forest Systems 25:eRC09.
Martin et al. use FlamMap and Landscape Treatment Designer (LTD) to devise an optimal spatiotemporal fuel treatment strategy for eucalyptus plantations in Portugal, with the intent to simultaneously reduce fire hazard, maximize timber volume and total biomass as carbon storage, while staying within budgetary constraints. As with other treatment optimization efforts that have used LTD, their approach did not take into consideration spatiotemporal variation in ignition probability, burn probability, or fire intensity probability, rather only deterministic fire intensity. To incorporate a temporal component, they estimated values for all objective attributes, for each stand, at each of four time steps, although it is not clear if these estimations of future attribute values factored in the effects of future fuel treatments, harvest rotations, or fires.
Miller, C., and A. A. Ager. 2013. A review of recent advances in risk analysis for wildfire management. International Journal of Wildland Fire 22:1.
This paper summarizes the significant advances in risk analysis made by the wildfire management community since 2005, and discusses some of the major challenges that remain. After defining the term “risk” and its constituent elements (likelihood, intensity, effects), they discussed how each of these elements can be derived, how they can be integrated into a single synthetic risk metric (expected net value change [e(NVCj)]) using the Integral Risk Model (IRM, Finney 2005), and reviewed recent progress in applying these concepts to real landscapes using geographic information systems (GIS) and wildfire simulation applications. However, they also discussed several remaining challenges. First, most risk analyses at the time of publication did not incorporate temporal dynamics, making them of limited use for evaluating risk management activities over most timeframes used in land management planning. Second, at the time of publication, efforts at spatial optimization of treatments had only managed to produce solutions that were heuristic, rather than truly optimal. Given that treatment optimization algorithms increase exponentially with the size of the landscape, number of treatment options, and degree of modeled stochasticity, the authors argued that spatial optimization of treatments was likely to remain an intractable problem for some time to come.
Scott, J. H., M. P. Thompson, and D. E. Calkin. 2013. A wildfire risk assessment framework for land and resource management. General Technical Report RMRS-GTR-315, USDA Forest Service, Missoula, Montana, USA.
In this technical report, Scott et al. provide and accessible description of a framework for developing sophisticated wildfire risk assessments and applying them to fire management. Notably, the report was written for a non-academic, land manager audience. The authors begin by explaining the need for wildfire risk assessments, followed by a framework for conceptualizing wildfire risk that aligns closely with that of Finney (2005) and Miller and Ager (2013). They then describe the analytical process for performing a wildfire risk assessment, including application of geographic information systems (GIS), wildfire simulation (using FSPro, FlamMap 5, or FSim), identification and characterization of highly valued resources and assets (HVRAs), and exposure and effects analysis. An example study is provided, for the Bridger-Teton National Forest. The report concludes with a largely conceptual discussion of how wildfire risk analyses can contribute to fire management and land management planning. A glossary of important risk-related terminology is provided at the end.
Scott, J. H., M. P. Thompson, and J. W. Gilbertson-Day. 2016. Examining alternative fuel management strategies and the relative contribution of National Forest System land to wildfire risk to adjacent homes – A pilot assessment on the Sierra National Forest, California, USA. Forest Ecology and Management 362:29–37.
In this paper, Scott et al. quantified how different fuel treatment scenarios on federal land could reduce transmission of wildfire risk to the wildland-urban interface. Specifically, using LANDFIRE fuels data, WUI location data from the West-wide Wildfire Risk Assessment Where People Live database, and the FSim large-fire simulator, the authors examined how expected annual number of homes exposed to any wildfires and expected annual number of homes exposed to wildfires originating from federal land varied across five different hypothetical fuel treatment scenarios ranging from making all federal land unburnable, to treating only non-federal land immediately adjacent to WUI. Results highlighted the limited impact that even the most extreme and impractical fuel treatment scenarios on federal lands could have on reducing expected exposure of nearby homes to wildfire, and the overwhelming advantage of treating WUI areas themselves rather than federal lands in terms home exposure reduction per unit area treated. In short, this paper illustrates the importance of landscape-scale assessments of risk transmission for the development of effective and efficient fuel treatment strategies.
Thompson, M., P. Bowden, A. Brough, J. Scott, J. Gilbertson-Day, A. Taylor, J. Anderson, and J. Haas. 2016a. Application of Wildfire Risk Assessment Results to Wildfire Response Planning in the Southern Sierra Nevada, California, USA. Forests 7:64.
This paper describes a framework for risk-based response planning and demonstrates its application on a 9.7 million ha landscape in the California’s southern Sierra Nevada. The authors begin by describing a four-dimensional continuum for conceptualizing fire management, defined by 1) fire location, 2) social and ecological conditions, 3) response options, and 4) temporal fluidity of these three conditions. The authors define fire management zones, referred to as “Potential wildland fire Operational Delineations” (PODs) in collaboration with local land management staff, assess risk using wildfire simulations from FSim, and then assign each POD to one of three response categories (protect, restore, maintain) based on cumulative risk each POD contains. Importantly, they assessed two types of risk – in situ risk, and source risk. In situ risk represented the expected or conditional net value change (eNVC or cNVC, respectively) for a given pixel, while source risk represented the sum of eNVC or cNVC across all pixels falling within perimeters of fires originating from a given pixel. This paper builds on Thompson et al. (2013) by proposing a simple framework for operationalizing risk analysis results for comprehensive fire management planning purposes.
Thompson, M. P., J. W. Gilbertson-Day, and J. H. Scott. 2016b. Integrating Pixel- and Polygon-Based Approaches to Wildfire Risk Assessment: Application to a High-Value Watershed on the Pike and San Isabel National Forests, Colorado, USA. Environmental Modeling & Assessment 21:1–15.
In this paper, Thompson et al. demonstrate how to glean additional for fire management planning from pixel-based wildfire risk assessments by incorporating polygon-based metrics into the wildfire risk assessment process. Wildfire risk assessments rely heavily on geographic information systems (GIS). Although most wildfire risk assessment work to-date has been pixel-based, which yields important local-scale information on burn probability, fire intensity, and highly valued resources and asset (HVRA) susceptibility, it fails to provide fire-level and season-level metrics which could further assist fire management planning. Here, the authors integrate pixel-based and polygon-based approaches with the large fire simulator FSim so that in addition to the usual suite of pixel-based outputs such as annual burn probability and expected net value change (eNVC), their analysis also generated metrics such as fire-level conditional net value change (cNVCfire) and produce maps showing risk transmission (i.e., source risk, where each cNVCfire is mapped back to its point of origin). This integrated approach has promise for assisting with the delineation of firesheds, the basic spatial units for fire management.
Thompson, M. P., J. Scott, D. Helmbrecht, and D. E. Calkin. 2013. Integrated wildfire risk assessment: Framework development and application on the Lewis and Clark National Forest in Montana, USA. Integrated Environmental Assessment and Management 9:329–342.
In this paper, Thompson et al. describe the application of the general risk assessment framework outlined in Scott et al. (2013) to a real landscape in Montana. This paper is most notable for the degree of detail the authors provide regarding how highly valued resources and assets (HVRAs) were identified, how they were weighted for relative importance, and how response functions were developed for each. Risk assessments are highly sensitive to these inputs, and eliciting them from stakeholders can be a surprisingly challenging process.
Vogler, K., A. Ager, M. Day, M. Jennings, and J. Bailey. 2015. Prioritization of Forest Restoration Projects: Tradeoffs between Wildfire Protection, Ecological Restoration and Economic Objectives. Forests 6:4403–4420.
Vogler et al. use the Wallowa-Whitman National Forest as a case study to illustrate the use of geographic information systems (GIS), Landscape Treatment Designer (LTD, Ager et al. 2012b), and production possibility frontiers (PPF) for assisting forest managers in prioritizing areas for forest restoration treatments given limited treatment capability and multiple restoration objectives. As the first step in the process, the authors identified five management objectives for the Forest (e.g., reduction of wildfire risk to the wildland-urban interface, timber volume) and mapped forest condition in terms of these attributes across all forest stands in a GIS. They then, given an area-based treatment constraint, used LTD to identify the combination of treatable stands that would yield the greatest simultaneous progress toward all restoration objectives, and repeated this process using different combinations of priority weightings for the five objectives. PPFs were used to illustrate the pairwise tradeoffs realized by shifting priorities among objectives. This paper extends the work of Ager et al. (2013) in that it illustrates the application of LTD on a real landscape, to a spatial optimization problem involving multiple objectives.
