2010
National Wildfire Risk and
Hazard - Draft Report
Summary of the First Approximation for the
Continental United States
A first approximation considers how fire likelihood and intensity influence social,
economic, and ecological values. This First Approximation of Fire Hazard and Fire
Risk was designed to meet three broad goals: 1) evaluate wild fire hazard on
federal lands; 2) develop information useful in prioritizing where fuels treatments
and mitigation measures might be proposed to address significant fire hazard and
risk; 3) develop risk-based performance measures to document the effectiveness
of fire management programs. The research effort described in this report is
designed to develop, from a strategic view, a first approximation of how both fire
likelihood and fire intensity influence risk to social, economic, and ecological
values at the national scale. The approach uses a quantitative risk framework that
approximates expected losses and benefits from wildland fire to highly valued
resources. Specifically, burn probabilities and intensities are estimated with a fire
simulation model and coupled with spatially explicit data on human and
ecological values and fire-effects response functions to estimate the percent loss
or benefit. This report summarizes risk to highly valued resources by the eight
continental US geographic areas and ranks Fire Planning Units (FPUs) according to
national relative risk scores.
David Calkin*, Alan Ager, Julie Gilbertson-Day, Joe Scott, Tom Quigley, Matthew
Thompson, Charlie Schrader-Patton, and Allison Reger
* Corresponding Author: decalkin@fs.fed.us, 406-329-3424
1
Photo credit: California BLM Office
Introduction
Reviews have been conducted by Federal oversight agencies and blue ribbon panels to
identify causal factors of the unprecedented fire suppression costs and to suggest possible
modifications to Federal fire management policy and strategies (Strategic Issues Panel 2004;
USDA Office of Inspector General 2006; Government Accountability Office 2007, 2009).
Agency and panel member reviews have found that Federal agencies with wildland fire
responsibilities are not able to quantify the value of fire management activities in terms of
reducing wildfire risk to social, economic, and ecological values. In response, the Wildland Fire
Leadership Council’s (WFLC) monitoring strategy asked: What are the trends and changes in
fire hazard on Federal lands? Fire risk assessment requires an understanding of the likelihood of
wildfire by intensity level and the potential beneficial and negative effects to valued resources
from fire at different intensity levels.
This national hazard and risk assessment was conducted to meet three broad goals: (1)
address the WFLC monitoring question regarding fire hazard on Federal lands; (2) develop
information useful in prioritizing where fuels treatments and mitigation measures might be
proposed to address significant fire hazard and risk; and (3) respond to critiques by Office of
Management and Budget, General Accounting Office, and Congress that call for risk-based
performance measures to document the effectiveness of fire management programs. The results
of this monitoring study are useful for project planning to quantify the potential effects of
proposed actions in terms of reducing risk to specific resources of concern.
Developing decision support tools that utilize an appropriate risk management framework
should address many of the issues identified within government oversight reports. Specifically,
the Office of Inspector General (USDA Office of Inspector General 2006) reviewed USDA
Forest Service (FS) large fire costs and directed that the “FS must determine what types of data it
needs to track in order to evaluate its cost effectiveness in relationship to its accomplishments. At
a minimum, FS needs to quantify and track the number and type of isolated residences and other
privately-owned structures affected by the fire, the number and type of natural/cultural resources
threatened, and the communities and critical infrastructure placed at risk.”
1
The application of fire risk and fire hazard analyses has been demonstrated at the
watershed and National Forest scales (Ager and others 2007). There, specific details regarding
probabilities of fire and fire intensity are linked with specific resource benefit and loss functions
(Ager and others 2007). Expanding these detailed analyses to regional and national scales to
provide consistent risk assessment processes is complicated by the required data specificity and
difficulty in developing loss-benefit functions for the range of human and ecological values. The
research effort described in this report is designed to develop, from a strategic view, a first
approximation of how both fire likelihood and intensity influence risk to social, economic, and
ecological values at the national scale. The approach uses a quantitative risk framework that
approximates expected losses and benefits from wildfire to highly valued resources (HVR).
The information gathered in this study can be summarized in tabular and map formats at
different scales using administrative boundaries or delineations of HVR such as built structure
density. The overall purpose of the analysis is to provide a base line of current conditions for
monitoring trends in wildfire risk over time. Future analyses would be used to determine trends
and changes in response to fuel reduction investments, climate shifts, and natural disturbance
events (e.g., bark beetles) between the timeframes analyzed. Monitoring data could be used to
address national and regional questions regarding changes in fire risk and hazard as a result of
investment strategies or changing conditions. While similar analyses could be conducted for
alternative scenarios, this work is designed to develop the base line hazard and risk situation.
Three main components were combined to generate wildfire risk outputs, namely (1)
burn probability generated from wildfire simulations, (2) spatially identified HVR, and (3)
response functions that describe the impact of fire on the HVR. Risk is calculated as the product
of the burn probability at a given fire intensity and the resulting change in value summed over all
possible fire intensities and values (Finney, 2005). Calculating risk at a given location requires
spatially defined estimates of the likelihood and intensity of fire interacted with identified values.
This interaction is quantified through the use of response functions that estimate expected
benefits and losses to resources at the specified intensities.
This report summarizes wildfire hazard and risk at the Fire Planning Unit (FPU) defined
by the Fire Program Analysis project. A detailed discussion of how the methodology to estimate
2
hazard and risk was developed is described in Calkin et al. (2010)1, with application
demonstrated for the state of Oregon.
Results
Results are summarized by the following geographic areas: California (CA), Eastern Area (EA),
Great Basin (GB), Northern Rockies (NR), Northwest (NW), Rocky Mountain (RM), Southern
Area (SA), and Southwest (SW). Results are presented in two distinct sections. Within the first
section, wildfire hazard and risk to individual HVR layers is graphically presented. Within the
second section, a system of weights and adjustment factors are developed that allow ranking of
aggregate risk at the FPU level, evaluation of the contribution of individual geographic areas to
national risk, and the relative contribution of each HVR to national risk.
Wildfire hazard and risk output
Description of the graphical output organized by geographic area for the individual FPUs
is presented below.
1. Burnable Area
The first figure, Burnable Area by FPU, illustrates the portion of burnable versus non-burnable
acres present in each FPU. Non-burnable land types (pixels) are represented by zero burn
probabilities and these pixels are excluded from the model. This figure demonstrates the area
where fire may occur in a given FPU, but does not provide an understanding of the spatial
distribution, likelihood, or relative intensity of fire in the FPU.
2. Burn Probability Distribution
Burn probabilities within each FPU are displayed using box whisker graphs. Box whisker
graphs (or box plots) summarize the range of burn probabilities found between the 10th and 90th
percent of the data points and provide a snapshot of the variability of burn probabilities in the
FPU. The box portion of the plot contains the middle fifty-percent of the data points from the
1
Calkin, David E.; Ager, Alan A.; Gilbertson-Day, Julie 2010. Wildfire risk and hazard: procedures for the first
approximation. Gen. Tech. Rep. RMRS-GTR-235. Fort Collins, CO: U.S. Department of Agriculture, Forest
Service, Rocky Mountain Research Station. 62 p. Available at: http://www.fs.fed.us/rm/pubs/rmrs_gtr235.html
3
25th percentile to the 75th percentile, with the horizontal bar representing the median. At a glance
one can determine the most frequently occurring burn probability values present in the FPU and
the variability around the center of the dataset. The location of the horizontal line within the box
(relative to the top 75th percent and 25th percent bottom of the box) is an indicator of how the
data are distributed around the median. The whiskers extending from the bottom and top of the
box represent the portion of data that lies within the 10th and 90th percentiles respectively.
3. Conditional Flame Length Distribution
Flame length distributions for each FPU are represented using box whisker plots. The values
plotted represent mean conditional flame length (CFL) for each pixel in the FPU. CFL is a
measure of the probability that fire will burn at a given intensity, conditioned on fire occurrence
within the pixel. For each pixel in an FPU, values exist for four CFL categories – Low, Medium,
High, and Very High. These four categories correspond to resource response functions used to
calculate HVR response to wildfire. It is necessary however, to condense these four categories
into one mean value in order to map/plot one flame length value per pixel. The box represents
the mean flame length for all pixels in the FPU that fall between the 25th and 75th percentiles (the
middle fifty-percent of the data), with the horizontal bar displaying the median. The lines
extending from the bottom and top of the box show the 10th and 90th percentiles respectively.
4
4. Area in HVR Value Categories
Area of each highly valued resource is calculated for all FPUs. Acres represent total HVR area
including burnable and non-burnable land types, and should be used to determine relative
abundance of each HVR category within the FPU.
5. Total Change Equivalent (TCE)
This table summarizes net change in Total Change Equivalent (TCE) measured in acres for
each value category (M, H, VH) and FPU. TCE is the equivalent area lost (or gained) assuming
100 percent loss (or gain) for a particular HVR category as measured in acres. For example,
assume that a given 18 acre pixel of critical habitat (RF 13) had a probability of burning of 0.5
percent, with a high fire intensity (70 percent loss in value), and a 1 percent chance of burning
with a very high fire intensity (80 percent loss in value). The contribution of this pixel to TCE
would equal 0.207 acres [(0.005 *0.7 * 18 acres) + (0.01 * 0.8 * 18 acres)].
5
RESULTS FOR CALIFORNIA
CA Figure 1-a. Burnable Area by FPU
CA Figure 1-b. Burn Probability Distribution by FPU
Burn probability
0.050
0.040
0.035
0.030
0.025
0.020
0.015
0.010
0.005
CA_CA_015
CA_CA_014
CA_CA_012
CA_CA_011
CA_CA_010
CA_CA_009
CA_CA_008
CA_CA_007
CA_CA_006
CA_CA_005
CA_CA_004
CA_CA_003
CA_CA_002
0.000
CA_CA_001
Burn Probability (fraction)
0.045
6
CA_CA_015
CA_CA_014
CA_CA_012
CA_CA_011
CA_CA_010
CA_CA_009
CA_CA_008
CA_CA_007
CA_CA_006
CA_CA_005
CA_CA_004
CA_CA_003
CA_CA_002
CA_CA_001
Mean Conditional Flame Length (ft)
RESULTS FOR CALIFORNIA
CA Figure 1-c. Conditional Flame Length Distribution by FPU
Flame length
12
10
8
6
4
2
0
CA Figure 1-d. Area of Highly Value Resources by FPU
7
RESULTS FOR CALIFORNIA
CA Figure 1-e. Wildfire Risk—TCE by HVR Value Category and FPU
TCE (Acres)
8
9
RESULTS FOR EASTERN AREA
EA Figure 2-a. Burnable Area by FPU
EA Figure 2-b. Burn Probability Distribution by FPU
Burn probability
0.050
0.040
0.035
0.030
0.025
0.020
0.015
0.010
0.005
EA_WV_001
EA_WI_002
EA_WI_001
EA_PA_001
EA_OH_001
EA_NJ_001
EA_NH_001
EA_MO_001
EA_MN_002
EA_MN_001
EA_MI_002
EA_MI_001
EA_IN_001
EA_IL_001
0.000
EA_IA_001
Burn Probability (fraction)
0.045
10
EA_WV_001
EA_WI_002
EA_WI_001
EA_PA_001
EA_OH_001
EA_NJ_001
EA_NH_001
EA_MO_001
EA_MN_002
EA_MN_001
EA_MI_002
EA_MI_001
EA_IN_001
EA_IL_001
EA_IA_001
Mean Conditional Flame Length (ft)
RESULTS FOR EASTERN AREA
EA Figure 2-c. Conditional Flame Length Distribution by FPU
Flame length
12
10
8
6
4
2
0
EA Figure 2-d. Area of Highly Value Resources by FPU
11
RESULTS FOR EASTERN AREA
EA Figure 2-e. Wildfire Risk—TCE by HVR Value Category
TCE (Acres)
12
13
RESULTS FOR GREAT BASIN AREA
GB Figure 3-a. Burnable Area by FPU
GB Figure 3-b. Burn Probability Distribution by FPU
Burn probability
0.050
0.040
0.035
0.030
0.025
0.020
0.015
0.010
0.005
GB_WY_001
GB_UT_005
GB_UT_004
GB_UT_003
GB_UT_002
GB_UT_001
GB_NV_006
GB_NV_005
GB_NV_004
GB_NV_003
GB_NV_002
GB_NV_001
GB_ID_004
GB_ID_003
GB_ID_002
0.000
GB_ID_001
Burn Probability (fraction)
0.045
14
GB_WY_001
GB_UT_005
GB_UT_004
GB_UT_003
GB_UT_002
GB_UT_001
GB_NV_006
GB_NV_005
GB_NV_004
GB_NV_003
GB_NV_002
GB_NV_001
GB_ID_004
GB_ID_003
GB_ID_002
GB_ID_001
Mean Conditional Flame Length (ft)
RESULTS FOR GREAT BASIN AREA
GB Figure 3-c. Conditional Flame Length Distribution by FPU
Flame length
12
10
8
6
4
2
0
GB Figure 3-d. Area of Highly Value Resources by FPU
15
RESULTS FOR GREAT BASIN AREA
GB Figure 3-e. Wildfire Risk—TCE by HVR Value Category and FPU
TCE (Acres)
16
17
RESULTS FOR NORTHERN ROCKIES AREA
NR Figure 4-a. Burnable Area by FPU
NR Figure 4-b. Burn Probability Distribution by FPU
Burn probability
0.050
0.040
0.035
0.030
0.025
0.020
0.015
0.010
0.005
NR_ND_001
NR_MT_010
NR_MT_009
NR_MT_008
NR_MT_007
NR_MT_006
NR_MT_005
NR_MT_004
NR_MT_003
NR_MT_002
NR_MT_001
0.000
NR_ID_001
Burn Probability (fraction)
0.045
18
RESULTS FOR NORTHERN ROCKIES AREA
NR Figure 4-c. Conditional Flame Length Distribution by FPU
Flame length
Mean Conditional Flame Length (ft)
12
10
8
6
4
2
NR_ND_001
NR_MT_010
NR_MT_009
NR_MT_008
NR_MT_007
NR_MT_006
NR_MT_005
NR_MT_004
NR_MT_003
NR_MT_002
NR_MT_001
NR_ID_001
0
NR Figure 4-e. Area of Highly Value Resources by FPU
19
RESULTS FOR NORTHERN ROCKIES AREA
NR Figure 4-e. Wildfire Risk—TCE by HVR Value Category and FPU
TCE (Acres)
20
21
RESULTS FOR NORTHWEST AREA
NW Figure 5-a. Burnable Area by FPU
NW Figure 5-b. Burn Probability Distribution by FPU
Burn probability
0.050
0.040
0.035
0.030
0.025
0.020
0.015
0.010
0.005
NW_WA_007
NW_WA_005
NW_WA_004
NW_WA_003
NW_WA_002
NW_WA_001
NW_OR_011
NW_OR_010
NW_OR_009
NW_OR_008
NW_OR_007
NW_OR_006
NW_OR_005
NW_OR_004
NW_OR_003
NW_OR_002
0.000
NW_OR_001
Burn Probability (fraction)
0.045
22
NW_WA_007
NW_WA_005
NW_WA_004
NW_WA_003
NW_WA_002
NW_WA_001
NW_OR_011
NW_OR_010
NW_OR_009
NW_OR_008
NW_OR_007
NW_OR_006
NW_OR_005
NW_OR_004
NW_OR_003
NW_OR_002
NW_OR_001
Mean Conditional Flame Length (ft)
RESULTS FOR NORTHWEST AREA
NW Figure 5-c. Conditional Flame Length Distribution by FPU
Flame length
12
10
8
6
4
2
0
NW Figure 5-d. Area of Highly Value Resources by FPU
23
RESULTS FOR NORTHWEST AREA
NW Figure 5-e. Wildfire Risk—TCE by HVR Value Category and FPU
TCE (Acres)
24
25
RESULTS FOR ROCKY MOUNTAIN AREA
RM Figure 6-a. Burnable Area by FPU
RM Figure 6-b. Burn Probability Distribution by FPU
Burn probability
0.050
0.040
0.035
0.030
0.025
0.020
0.015
0.010
0.005
RM_WY_003
RM_WY_002
RM_SD_003
RM_SD_002
RM_SD_001
RM_NE_001
RM_KS_001
RM_CO_008
RM_CO_007
RM_CO_006
RM_CO_005
RM_CO_004
RM_CO_003
RM_CO_002
0.000
RM_CO_001
Burn Probability (fraction)
0.045
26
RM_WY_003
RM_WY_002
RM_SD_003
RM_SD_002
RM_SD_001
RM_NE_001
RM_KS_001
RM_CO_008
RM_CO_007
RM_CO_006
RM_CO_005
RM_CO_004
RM_CO_003
RM_CO_002
RM_CO_001
Mean Conditional Flame Length (ft)
RESULTS FOR ROCKY MOUNTAIN AREA
RM Figure 6-c. Conditional Flame Length Distribution by FPU
Flame length
12
10
8
6
4
2
0
RM Figure 6-d. Area of Highly Value Resources by FPU
27
RESULTS FOR ROCKY MOUNTAIN AREA
RM Figure 6-e. Wildfire Risk—TCE by HVR Value Category and FPU
TCE (Acres)
28
29
RESULTS FOR SOUTHERN AREA
SA Figure 7-a. Burnable Area by FPU
30
0.000
SA_KY_003
SA_KY_001
SA_GA_001
SA_FL_004
SA_FL_003
SA_NC_001
0.005
SA_VA_001
0.010
SA_MS_002
0.015
SA_TX_006
0.020
SA_MS_001
0.025
SA_TX_005
0.030
SA_MD_001
0.035
SA_TX_004
0.040
SA_LA_003
0.045
SA_TX_003
0.050
SA_LA_001
Burn probability
SA_TX_002
SA_TX_001
SA_TN_001
SA_SC_002
SA_SC_001
SA_OK_005
0.040
SA_FL_002
SA_FL_001
SA_AR_002
SA_AR_001
SA_AL_001
Burn Probability (fraction)
0.045
SA_OK_003
SA_OK_002
SA_OK_001
SA_NC_003
SA_NC_002
Burn Probability (fraction)
RESULTS FOR SOUTHERN AREA
SA Figure 7-b. Burn Probability Distribution by FPU
Burn probability
0.050
Burn probabilities for this FPU
exceed the range of this plot.
0.035
0.030
0.025
0.020
0.015
0.010
0.005
0.000
31
SA_MS_002
SA_NC_001
SA_VA_001
SA_KY_003
SA_KY_001
SA_GA_001
SA_FL_004
SA_FL_003
SA_FL_002
SA_FL_001
SA_AR_002
SA_AR_001
SA_AL_001
SA_TX_006
0
SA_MS_001
2
SA_TX_005
4
SA_MD_001
6
SA_TX_004
8
SA_LA_003
10
SA_TX_003
12
SA_LA_001
Flame length
SA_TX_002
SA_TX_001
SA_TN_001
SA_SC_002
SA_SC_001
SA_OK_005
SA_OK_003
SA_OK_002
SA_OK_001
SA_NC_003
SA_NC_002
Mean Conditional Flame Length (ft)
Mean Conditional Flame Length (ft)
RESULTS FOR SOUTHERN AREA
SA Figure 7-c. Conditional Flame Length Distribution by FPU
Flame length
12
10
8
6
4
2
0
32
RESULTS FOR SOUTHERN AREA
SA Figure 7-d. Area of Highly Value Resources by FPU
33
RESULTS FOR SOUTHERN AREA
SA Figure 7-e (part 1). Wildfire Risk—TCE by HVR Value Category and FPU
TCE (Acres)
34
RESULTS FOR SOUTHERN AREA
SA Figure 7-e (part 2). Wildfire Risk—TCE by HVR Value Category and FPU
TCE (Acres)
35
RESULTS FOR SOUTHWEST AREA
SW Figure 8-a. Burnable Area by FPU
SW Figure 8-b. Burn Probability Distribution by FPU
Burn probability
0.050
0.040
0.035
0.030
0.025
0.020
0.015
0.010
0.005
SW_TX_004
SW_TX_002
SW_NM_007
SW_NM_006
SW_NM_005
SW_NM_004
SW_NM_003
SW_NM_002
SW_NM_001
SW_AZ_006
SW_AZ_005
SW_AZ_004
SW_AZ_003
SW_AZ_002
0.000
SW_AZ_001
Burn Probability (fraction)
0.045
36
SW_TX_004
SW_TX_002
SW_NM_007
SW_NM_006
SW_NM_005
SW_NM_004
SW_NM_003
SW_NM_002
SW_NM_001
SW_AZ_006
SW_AZ_005
SW_AZ_004
SW_AZ_003
SW_AZ_002
SW_AZ_001
Mean Conditional Flame Length (ft)
RESULTS FOR SOUTHWEST AREA
SW Figure 8-c. Conditional Flame Length Distribution by FPU
Flame length
12
10
8
6
4
2
0
SW Figure 8-d. Area of Highly Value Resources by FPU
37
RESULTS FOR SOUTHWEST AREA
SW Figure 8-e. Wildfire Risk—TCE by HVR Value Category and FPU
TCE (Acres)
38
Relative Risk Ranking for FPUs: Uncertainty-related Issues
Our approach has been to consider TCE as a relative measure of risk to inform management
priorities, and therefore we begin from the assumption that calculated risks should roughly align
with current understanding of risk and current management priorities. This is not to say that
results from this and subsequent analyses could not alter our understanding of risk or redirect
attention to other resources, but rather that current priorities are an appropriate baseline from
which to analyze simulation results. From this baseline we can therefore examine the relative
proportion of risk assigned to individual resources and recognize potential issues where certain
HVRs are responsible for a higher/lower proportion of risk than seems appropriate.
Five issues merit discussion before the results of this first approximation can be used to inform
prioritization discussions. In broad terms, all issues stem from uncertainty that needs to be
recognized, evaluated and addressed. This uncertainty manifests itself spatially, temporally, in
terms of unknown social preferences, and in terms of resource response function definition and
assignment. This uncertainty was managed to the extent possible in this first iteration. These
issues are not exclusive, but at present are the dominant challenges to using the results of this
first approximation to inform prioritization decisions.
1) Weights must be applied to the different HVR value categories in order to indicate the
importance of resources assigned to the Very High value category relative to the High
and Moderate categories. These weights are essentially prices that may have a
significant influence on final rankings, and therefore require careful consideration.
2) The initial value category assignments should reflect relative differences in national
priorities. The interaction of value category assignment and the weights assigned to
each value category may have a significant influence on rankings.
3) The HVR data layers were developed and compiled using different standards and for
different purposes. As a result the geospatial identification of some HVR layers is quite
refined whereas others are quite coarse.
4) Response functions effectively assume a constant time horizon (in that they incorporate
future resource value changes, including potential recovery or deterioration over time),
when in fact some resource changes occur at vastly different temporal scales.
5) Response functions were developed generally rather than for each resource specifically.
As a result resource response functions assigned to specific resources may not fully
encapsulate expected response across fire intensity levels.
39
Whereas issue #1 equally affects all HVR layers, issues #2-4 relate to specific HVR layers of
concern. Specifically, we identified the following HVR layers as problematic and in need of a
more refined approach:
Non-attainment areas and Class I airsheds
Temporal uncertainty: time horizon of response function
Sage-grouse key habitat
Spatial uncertainty: possible over-representation of habitat
◦
◦
◦
Preference uncertainty: incorrect value category assignment
Canada lynx critical habitat
Response function uncertainty: incorrect response function assignment
◦
We first address the issue of assigning weights to value categories. This section is important to
understand, as the approach we take is the foundation for our approaches to handle issues #2-4.
Weighting of Categorical Importance
The intent of our analysis was to aggregate risk summaries for each resource into a
common measure to facilitate monitoring and prioritization. In the ideal, the social value of all
resources could be monetized. Unfortunately, prices for many of the resources considered in
the analysis (e.g., critical habitat) are not easily available, challenging a monetization approach
to quantifying risk. To estimate prices for non-market resources would require substantial
financial investment, and results would not be available for some time. Therefore, some form of
multi-criteria analysis is required with relevant experts queried to develop a reasonable priority
ranking. Multi-criteria analysis involves a systematic approach to analyze relative worth and to
assign a weight to each HVR (or in this case, for simplicity, each HVR value category).
Identifying weights for each value category affects all HVR layers. If the weight distribution is
too skewed, resources in the Very High value category may disproportionately influence risk
calculations and therefore prioritization decisions. Conversely, if the weight distribution is too
narrow, resources in the Moderate value category may be given an inflated importance.
To move forward with aggregating these results into a single risk metric we simply assumed
that the ranking of the three individual categories maintained a simple proportional relationship.
We evaluated how changing the class weight differential affected the proportion of risk
assigned to individual categories and resource types. Four initial weight vectors were evaluated:
(1,1.5,2.25), (1,2,4), (1,3,9), and (1,4,16), reflecting value differential factors of 1.5, 2, 3, and 4,
respectively. Although we have little information related to the appropriate weights, we
selected the weight set (1,3,9) to move forward. In other words, this initial approach assumes
resources in the Very High value category are 3 times as important as resources in the High
value category, which in turn are 3 times as important as resources in the Moderate value
category. This multiplicative approach allows for the aggregation of TCE measures across
40
value categories into a single metric to facilitate prioritization efforts.
Adjustment Factor Approach
To account for the variable sources of uncertainty in the risk calculations, we adopted an
“adjustment factor” approach that is effectively a rough correction. Given the challenges of rerunning the nation-wide analysis to generate new TCE values, we decided this approach was the
most practical and transparent way to move forward at this point. These adjustment factors
allow expert panels to rapidly explore alternative adjustment schemes and determine how risk
profiles change in response. The end result is effectively an approximate re-run of the model,
without the burdensome computational effort required.
TCE calculations for HVRs are based on the integration of burn probability maps, geospatial
identification of resource presence, and expert-defined resource response functions. The
adjustment factor approach we adopt assumes that the burn probability mapping is correct, i.e.
the likelihood and intensity of fire is not biased towards certain geographic areas or specific fuel
types. Therefore we look to geospatial identification of resource presence, expert-defined
resource response functions, and initial value category assignment as possible sources of error.
The adjustment factor approach is based on the value category weighting approach described
above. As all TCE value categories are aggregated using a multiplicative approach, the
adjustment factors similarly employ a multiplicative factor to increase/decrease the calculated
TCE risk values for a given HVR layer. For instance, if we had reason to believe the spatial
extent for a given HVR layer was mistakenly input as twice its actual area, we would use an
adjustment factor of 0.5 to correct the TCE calculation.
Below we discuss in more detail the adjustment factor approaches we took for each HVR layer
identified as possibly erroneous. Specifically we identify our rationale for believing the
unadjusted TCE values to be incorrect, and describe the steps we took to arrive at adjustment
factors.
41
Non-attainment areas (Very High) and Class I airsheds (Moderate)
HVR-level risk as a share of nationwide total
Fire-adapted ecosystems
GACC-level share of nationwide risk
ski areas
2.4%
Energy Gen
3.1%
CA (1)
SA (2)
Recreation sites
11.3%
Communication towers
SW (5)
6.5%
National Trails
NR (6)
4.5%
Energy Gas
50.2%
NW (4)
6.4%
Energy Elec
GB (3)
15.5%
EA (8)
Class 1 airsheds
low density structures
municipal watersheds
mod and high-density WUI
fire-susceptible species
non-attainment areas
-10
0
10
20
30
40
50
Figure 9. Unadjusted nationwide risk for all HVR and unadjusted GACC-level share of
nationwide risk.
As can be seen in Figure 9, non-attainment areas accounted for nearly 50% of nation-wide risk.
Although we recognize that smoke issues are very important and can result in significant health
and economic impacts, particularly in highly populated areas, smoke impacts last only a short
duration (days to weeks) relative to the impacts to other resource types. (Class 1 airsheds also
have the same issue, but are assigned to the Moderate value category and therefore are not as
strong of a driver of national-scale risk.) Resource loss due to wildfire occurring within
moderate and high density populated areas are considered equivalent to loss associated with
smoke in these same areas. We felt this inflated the risk associated with wildfire in nonattainment areas. The issue here is the temporal nature of the response function.
To modify the TCE values we employed an adjustment factor of 1/52, based on the rationale
that the human health and safety issues associated with smoke only last for, on average, one
week per year.
42
Sage-grouse key habitat
2.5%
5.0%
0.9%
0.9%
0.8%
0.8%
Sage grouse (14.8%)
4.6%
Canada Lynx (4.3%)
Northern Spotted Owl (1.6%)
Mexican Spotted Owl (1.2%)
6.3%
17.2%
Cape Sable Seaside Sparrow (1.1%)
Marbled Murrelet (0.6%)
59.5%
Desert Tortoise (0.2%)
Coastal California Gnatcatcher (0.2%)
Peninsular Bighorn Sheep (0.2%)
All others combined (0.2%)
As can be seen in Figure 10,
unadjusted risk calculated for the
sage-grouse comprised 59.5% of
risk associated with fire-susceptible
species, and 14.8% of national risk
within the High value category.
We first looked to habitat maps as
possible sources of uncertainty and
subsequent error.
Table 1 displays the acreage of
all fire-susceptible species, and
suggests that in particular sagegrouse habitat is extensive. The
argument is that there is an
overrepresentation of habitat, in turn unduly
influencing TCE calculations. In absence of
more refined maps, we proceeded using the
adjustment factor approach to reduce the
influence of sage-grouse on TCE calculations.
Figure 10. Unadjusted contribution of fire-susceptible
species to nationwide risk (pie chart) and percent
contribution national risk (percentage shown with species
next to species label).
Table 1. Area of fire-susceptible species habitat
showing burnable and non-burnable pixels.
Fire-Susceptible Species
Acres
Sage grouse
50,406,639
Canada Lynx
27,720,188
Mexican Spotted Owl
9,869,519
Northern Spotted Owl
7,139,518
Desert Tortoise
6,452,177
Marbled Murrelet
3,296,559
Peninsular Bighorn Sheep
816,466
Bull Trout
717,425
California Red-legged Frog
449,503
Coastal California Gnatcatcher
373,232
Cape Sable Seaside Sparrow
197,487
Quino checkerspot butterfly
171,745
Alameda Whipsnake
154,632
Southwestern Willow Flycatcher
120,586
Arroyo Toad
104,463
Additional 26 species
288,314
Total
108,278,453
To arrive at an appropriate spatial adjustment
factor for key sage-grouse habitat we
proceeded as follows:
1) Through a literature review we identified
that wildfire occurrence within a 2-mile buffer
around lek locations would potentially
degrade habitat2.
2) Chet Van Dellen, GIS Coordinator for the
Nevada Department of Wildlife, created 2mile buffers around all leks occupied in 20032008, and derived the proportion of total
mapped area these buffers represented within
the key sage-grouse habitat map.
The
buffered lek area represented ~28.4% of the
total mapped key habitat.
2
http://sagemap.wr.usgs.gov/Docs/Greater_Sagegrouse_Conservation_Assessment_060404.pdf
43
3) Assuming a relatively
consistent ratio across mapped
habitat areas in other states, we
arrived at a spatial adjustment
factor of 0.284 to reduce TCE
values.
Recognizing that sage-grouse is not
a listed species under the ESA, we
took the further step of reevaluating the initial value category
assignment. That is, we addressed Figure 11. Adjusted contribution of fire-susceptible species
preference uncertainty relating to to nationwide risk (pie chart) and percent contribution
sage-grouse importance.
We national risk (percentage shown with species next to species
decided to re-assign sage-grouse to label).
the Moderate value category.
Using the (1,3,9) weighting scheme described above, we therefore further reduced the TCE
values for sage-grouse by a factor of 1/3.
After combining the spatial and preference uncertainty adjustment factors, the overall
adjustment factor applied to sage-grouse key habitat was 0.095. Figure 11 shows the
contribution of the adjusted fire-susceptible species.
Canada Lynx
To arrive at an appropriate response
function adjustment factor for Canada lynx
critical habitat we proceeded as follows:
Response function 13
Net Value Change (%)
Figure 12 displays the response function
assigned categorically to all firesusceptible species.
With respect to
Canada lynx, this response function was
deemed inappropriate because it did not
recognize the potentially beneficial (or at
least non-detrimental) effects of lowintensity fire. Adjustment factors were
therefore employed to handle response
function uncertainty.
100%
60%
20%
-20%
Low
Moderate
High
Very High
-60%
-100%
Fire intensity
Figure 12. Response function selected for firesusceptible species.
1. We identified two FPUs with substantial Canada lynx critical habitat.
2. Within these two FPUs, we calculated the overall percentage of risk attributable to low
flame lengths (which, according to the assigned response function, were deemed
44
detrimental). The analysis identified that 85% of TCE risk values for Canada lynx were
associated with low flame lengths. We removed these losses from TCE risk
calculations.
3. Assuming relatively consistent conditional flame length probability distributions for
other FPUs housing lynx critical habitat, we arrived at a response function adjustment
factor of 0.153.
Summary
To reiterate, re-running the national model would require a great deal of time and effort, and
was not feasible for generating these first approximation results. Hence we adopted an interim
solution using adjustment factors that roughly approximate a re-run of the model. These
adjustment factors address the various sources of uncertainty we believe could result in
erroneous results, which in turn could lead to inefficient or ineffective prioritization decisions.
The approaches taken to calculate adjustment factors were necessarily abbreviated due to time
constraints, and could be refined for further analysis. However, we feel they are useful for recalculating risk values to be more aligned with what we might reasonably expect, and do
address all sources of uncertainty in a reasonable manner.
GACC-level share of nationwide risk
3.6%
3.5%
6.1%
CA (2)
SA (1)
7.8%
32.3%
3.7%
SW (3)
NR (6)
10.1%
NW (4)
GB (5)
32.9%
EA (7)
Figure 13. Adjusted nationwide risk for all HVR and unadjusted GACC-level share of
nationwide risk.
45
Results of the risk adjustments can be seen in Figure 13. Non-attainment areas are now only
responsible for 2.4% of overall national-scale risk for the Very High value category. The
contribution of sage-grouse to risk across fire-susceptible species dropped to 17.9% (Figure 11),
and to 3.9% of national risk within the High value category. Canada lynx dropped from 17.2%
of fire-susceptible species and 4.3% of national risk to 8.4% and 1.8%, respectively (Figure 11).
For purposes of comparison (and due to time constraints) we retained sage-grouse in the High
value category. Future analyses will display results with sage-grouse assigned to the Moderate
value category from the onset.
All figures displaying TCE values in the sections preceding the discussion about uncertainty
issues and FPU rank are shown without an adjustment. All figures proceeding this section are
adjusted as described in the text above. In future analyses, there is a need for resource experts
and fire management experts to re-prioritize HVR layers within value categories, to refine
response function definitions and assignments, to improve upon data layers, and to consider the
variable impact of alternative weighting systems on national priorities. Given the significant
uncertainty and challenges of compiling enterprise-level datasets for myriad resources it is still
likely similar adjustments will need to occur, but these can be minimized by appropriately
mapping resources, appropriately assigning response functions, etc. The issue of how to
prioritize across resources will remain a challenge. Our coarse approach to define three value
categories (Moderate, High, Very High) is a useful first step, but could be refined in subsequent
analyses.
46
47
Figure 14. Relative Risk Ranking for California FPUs
48
Figure 15. Relative Risk Ranking for Eastern Area FPUs
49
Figure 16. Relative Risk Ranking for Great Basin Area FPUs
50
Figure 17. Relative Risk Ranking for Northern Rockies Area FPUs
51
Figure 18. Relative Risk Ranking for Northwest Area FPUs
52
Figure 19. Relative Risk Ranking for Rocky Mountain Area FPUs
53
Figure 20. Relative Risk Ranking for Southern Area FPUs
54
Figure 21. Relative Risk Ranking for Southwest Area FPUs