Chapter 6: Forest Suitability for Large Wildfires

Dataset Overview

This dataset represents projections of environmental suitability for large wildfires in Pacific Northwest forests (figure 6.1). Wildfire occurrence was modeled using climate and terrain data (over the climate normal period 1971-2000), which was validated with area burned from 2001 to 2015 (see chapter glossary for definitions of terms). This model was projected into the future, producing maps of future suitability for large wildfires. Results from this modeling effort show predictions of increasing area conducive to occurrence of large forest wildfires under climate-change scenarios, with increases varying by ecoregion.

Data Access: The 2-state dataset (Washington and Oregon) can be viewed online and downloaded from:

Figure 6.1


Conservation Applications

Potential conservation applications of this dataset could include the following:

  • This dataset offers forest managers a way of visualizing geographic patterns of how climate change may affect geographic occurrence patterns of large forest wildfires. These projections could help inform forest management including fuels treatment, forest reserve network design, and forest carbon management. Projected changes in suitability for large wildfires could also be considered in urban planning and land-use decisions in the wildland-urban interface.

Applicable scales for detailed spatial assessments:

  • For conservation applications requiring detailed assessment of spatial variation of the dataset within a geographic boundary (such as a protected area), the following geographic scales may be most appropriate (see appendix 3 for more information).
  • At the scale of: a Bureau of Land Management (BLM) district, a river watershed (8-digit hydrologic unit code [HUC-8]), an individual county, a national forest, a level-3 ecoregion (e.g. the North Cascades), a single state (Washington, Oregon, or Idaho), a region (multiple states in the Pacific Northwest).

Applicable scales for assessing general patterns:

  • Due to spatial resolution, the dataset may not show detailed spatial variation at the following geographic scales, however, the dataset may be useful to assess general patterns or for comparison to other locations (see appendix 3 for more information).
  • At the scale of: a small (<1 km2) nature preserve, a state park or state wildlife area, a local watershed (12-digit hydrologic unit code [HUC-12]).

Use of the dataset in conservation applications may be limited by the following considerations:

  • Model projections represented in this dataset did not incorporate some of the management considerations and processes that may be important in shaping wildfire patterns into the future. These include fuels management programs (e.g., controlled burns, forest thinning), as well as possible transitions from forest to non-forest ecosystems in some areas as a result of climate change. Furthermore, the dataset represents only relative suitability for large wildfires over 30-year periods and cannot generally be used to make short-term predictions for fire behavior. In addition, the dataset does not address fire severity, which may be an important management consideration in addition to fire occurrence.

Past or current conservation applications:

Chapter 6 Case Study 1

Chapter 6 Case Study 2

Dataset citation:

Davis, R., Z. Yang, A. Yost, C. Belongie, and W. Cohen. 2017. The normal fire environment — modeling environmental suitability for large forest wildfires using past, present, and future climate normals. Forest Ecology and Management 390:173–186.

Note that this publication presents a version of the dataset for the states of Oregon and Washington (2-state dataset). The model algorithm was later used to expand the dataset to include the State of Idaho to produce the 3-state dataset depicted in figure 6.1.

Dataset documentation link: (open access)

Data access:

The 2-state dataset (Washington and Oregon) can be viewed online and downloaded from:

The 3-state dataset (which also includes Idaho) may be obtained by contacting the corresponding authors.

Metadata access: Formal metadata is not available for this dataset.

Dataset corresponding authors:
Raymond Davis
USDA Forest Service, Pacific Northwest Region

Zhiqiang Yang
USDA Forest Service, Rocky Mountain Research Station

Data type category (as defined in the Introduction to this guidebook): Fire, climate

Species or ecosystems represented: This dataset represents forested ecosystems of the Pacific Northwest.

Units of mapped values:  Continuous maps are based on an index of environmental suitability for large wildfire occurrence. Classified maps divide this index into classes of low, moderate, or high suitability.

Range of mapped values:

Continuous maps - 0 to 1
Classified maps – 1 (low), 2 (moderate), 3 (high) environmental suitability

Spatial data type: a raster dataset (grid)

Data file format(s): GeoTiff

Spatial resolution: 800 m

Geographic coordinate system: North American Datum of 1983

Projected coordinate system: North American Datum of 1983 Albers

Spatial extent: Regional (Oregon, Washington, and Idaho)

Dataset truncation: The dataset is truncated along the borders of the States of Washington, Oregon, and Idaho.

Time period represented: Future

Future time period(s) represented: Mid-century (2031 to 2060), end-of-century (2071 to 2100)

Baseline time period (against which future conditions were compared): 1971 to 2000.

Methods overview:

A MaxEnt model (Phillips et al., 2006) was constructed to predict occurrence of large wildfires (fires at least 40 hectares in size) using climate (temperature and precipitation) and topographic variables (elevation and slope). The model was trained using wildfire data from the period 1971 to 2000. Model performance was evaluated by predicting wildfire occurrence during the period 2001 to 2015 and comparing that prediction to real fire data from that same time period. This model validation confirmed that the model was capable of geographically predicting occurrence of large wildfires under normal environmental conditions (Davis et al., 2017). The model was then used to project future geographic patterns of large wildfire suitability using climate data for future climate normal time periods under different climate-change scenarios. For more information, please consult the dataset citation listed in section 2 of this chapter.

This dataset relied on the following general types of models: Machine-learning models based on wildfire locations

This dataset employed the following specific models: MaxEnt version 3.3 (Phillips et al., 2006)

Major input data sources for this dataset included:  Historical climate observations or models, future climate projections, digital elevation models (DEMs), and wildfire perimeter data.

This dataset used the following general circulation models (GCMs):  Downscaled climate projections from NEX-DCP30 used 33 GCMs, which are described in documentation available from:

More information about climate models is available in appendix 1. Detailed information about climate models, including model evaluation and comparison among models, is available from Randall et al. (2007) and Rupp et al. (2013).

This dataset used the following greenhouse-gas scenarios: RCP 4.5, RCP 8.5

More information about greenhouse-gas scenarios is available in appendix 2 and from Knutti and Sedláček (2013).

Creation of this dataset involved the following methods to change the spatial resolution of climate models (e.g. to downscale or resample climate models):

Downscaled climate projections from NEX-DCP30 are described in Thrasher et al. (2013) and in documentation available from:

The mapped values of the dataset may be interpreted as follows:

In the context of this dataset, environmental suitability for large wildfires is relative such that continuous map values can be compared from one location to another but do not have an absolute interpretation. Higher values for large wildfire suitability indicate landscapes that, based on climate and terrain, are more typically associated with large wildfire occurrence under normal climate conditions (3-decade averages). A classified map value of “low” indicates the likelihood of large wildfire occurrence under normal climate conditions is less than would be expected by random chance. The “moderate” and “high” classes indicate higher than random likelihoods of occurrence.

Representations of key concepts in climate-change ecology:

Climate change is expected to produce many varied indirect effects on forest ecosystems, including changes in disturbance regimes such as wildfires. Forested areas that have greater projected changes in wildfire suitability may be more vulnerable to this manifestation of climate change, either in terms of climate-change exposure (greater change in temperature or precipitation patterns that influence fires) or in terms of climate sensitivity (some landscape locations may be more or less sensitive to changing fire regimes based on terrain and topographic position).

Projected changes in forest wildfire dynamics resulting from climate change have a range of potential consequences for forest ecosystems and species, as well as for people. For example, some threatened and endangered species in the Pacific Northwest—including the northern spotted owl and the marbled murrelet—depend on mature, old-growth forests for habitat and may be negatively affected by stand-replacing fires. Changing fire dynamics also have potential implications for watershed processes such as erosion and influencing water quality in receiving stream reaches. Additionally, because fire suppression efforts represent substantial annual expenditures, projected changes in forest wildfire suitability over time represent a potentially important link between climate change and social, economic, and human health (e.g., air quality) considerations.

This dataset involves the following assumptions, simplifications, and caveats:

The modeling efforts that produced this dataset used only climate and terrain variables as inputs, however, fire behavior can be influenced by a range of other landscape characteristics. Possible changes to forest fuels were not considered, such as those that could arise from previous fires, fuel-reduction programs, carbon dioxide fertilization, or transitions from forest to non-forest ecosystems resulting from climate change. Furthermore, representation of climate variables as 30-year normals does not account for interannual variability between cool/moist years and hot/dry years. It should also be noted that the models represented likelihood of large wildfire occurrence but not fire severity.

Quantification of uncertainty:

Models of fire suitability were constructed independently for each of 33 GCMs, and results were presented as median and standard deviation across models. Thus, the standard deviation of model outputs provides information about uncertainty that derives from climate-change projections.

Field verification:

Large wildfire suitability that was modeled using climate and fire data from 1971-2000 was used to predict large wildfire suitability from 2001 to 2015. These predictions were then validated with large wildfires that occurred from 2001 to 2015. Forests mapped as having low wildfire suitability burned an average of 5 times less than would be expected by chance, whereas forests with moderate and high suitability burned on average 1.5 times and 2 to 3 times more than would be expected by chance, respectively.

Prior to dataset publication, peer review of the journal paper presenting the 2-state dataset for Oregon and Washington was conducted by external review (at least two anonymous reviewers, each from a different institution).

Davis, R., L. Evers, Y. Gallimore, J. Volpe, and C. Belongie. 2016. Appendix D – Modeling large stochastic wildfires and fire severity within the range of the northern spotted owl. Pages 1229–1244 in the Resource Management Plan for Western Oregon. Department of the Interior, Bureau of Land Management, Portland, OR.

Davis, R., Z. Yang, A. Yost, C. Belongie, and W. Cohen. 2017. The normal fire environment — modeling environmental suitability for large forest wildfires using past, present, and future climate normals. Forest Ecology and Management 390:173–186.

Knutti, R., and J. Sedláček. 2013. Robustness and uncertainties in the new CMIP5 climate model projections. Nature Climate Change 3:369–373.

Phillips, S. J., R. P. Anderson, and R. E. Schapire. 2006. Maximum entropy modeling of species geographic distributions. Ecological Modelling 190:231–259.

Randall, D., R. Wood, S. Bony, R. Colman, T. Fichefet, J. Fyfe, V. Kattsov, and et al. 2007. Cilmate models and their evaluation. in S. Solomon, D. Qin, M. Manning, Z. Chen, M. Marquis, K. Averyt, M. Tignor, and H. Miller, editors. Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, New York, NY.

Rupp, D. E., J. T. Abatzoglou, K. C. Hegewisch, and P. W. Mote. 2013. Evaluation of CMIP5 20th century climate simulations for the Pacific Northwest USA. Journal of Geophysical Research 118: 884–907.

Thrasher, B., J. Xiong, W. Wang, F. Melton, A. Michaelis, and R. Nemani. 2013. Downscaled climate projections suitable for resource management. Eos, Transactions American Geophysical Union 94:321–323.