home > data and maps > derived data

Data Derived from Foundational Data Layers

Fire Hazard and Crown Fire Behavior
Fire Risk
Post-Fire Erosion Potential
Post-Fire Sedimentation Potential
Post-Fire Flooding Potential
Relative Risk to Municipal Watersheds
Predicted Mexican Spotted Owl Nesting and Roosting Habitat
Mexican Spotted Owl Habitat based on Management Guidelines
Northern Goshawk Nesting Habitat
Passerine Avian Species Richness
Tassel-eared Squirrel Density and Recruitment
Pronghorn Potential Movement Corridors
Pronghorn Habitat Suitability
Wild Turkey Habitat Suitability
Individual Passerine Bird Species Habitat
Wildlife Corridors

Fire Hazard and Crown Fire Behavior

Click on maps for larger images

Fire hazard is a measure of the amount of fuel available to a burning fire. Fire modeling programs are often used to model fire hazard. The layer presented here is an index of fire hazard represented by a prediction of the amount of heat produced by a burning fire in a particular area. This output is based primarily on fuels available for the fire to burn, but is also modified by terrain and weather (e.g., under conditions of higher wind speed, fuels would be expected to burn more completely). The values in the fire hazard (heat) layer range from zero to nearly 8700 BTUs / ft2 (0 to nearly 100,000 KJ / m2) across the assessment area.

The crown fire behavior layer is a representation of the type of fire that would be burning at any given location on the landscape under a given set of weather conditions. There are three types of fires: surface fires, passive crown fires, and active crown fires. Surface fires occur wherever there is enough fuel on the ground to sustain a fire. These are the least dangerous type of fire and can even be considered beneficial and desirable for maintaining ecosystem function. Passive crown fire, or torching, occurs when flame lengths are long enough to reach the lower edge of the canopy, but there are not enough canopy fuels to carry a fire through the canopy. Thus, some individual trees, or small clumps of trees burn, but others do not. Active crown fire occurs when flames reach the canopy and spread through it. These are the most dangerous fires.

Layer creation

Both of these layers are outputs from the fire-modeling program FlamMap, developed by Mark Finney and the Fire Sciences Laboratory. FlamMap predicts various aspects of fire behavior across an entire landscape under a single set of weather conditions. We developed all of the necessary input layers for running FlamMap from either the Digital Elevation Model (slope, aspect, elevation), or our vegetation type (fuel models) and vegetation structure (canopy cover, crown bulk density, stand height, crown base height) layers. We used weather conditions approximating those of extreme drought and fire weather in this region. Fuel moistures were set very low (95th percentile dry) and sustained wind speeds were set to 30 mph. Winds were assumed to be coming from the southwest, which is the prevailing wind direction in the assessment area during the fire season.

Layer accuracy

There is no way to directly assess the accuracy of these layers because this would require documenting a fire burning under exactly the same weather conditions used as inputs to the fire modeling program. However, all of the layers used in the modeling are developed from either the Digital Elevation Model or our vegetation composition and structure layers. Thus, the accuracy of the input layers should approximate those of the layers they are created from. Also, many additional factors that affect how a fire will burn are not taken into account by the program (e.g., suppression efforts and changing weather conditions). Thus, FlamMap outputs should not be considered absolute representations of how a fire would behave if it were burning on the landscape, but can be used to effectively compare relative fire hazard and behavior across the landscape when all conditions are equal.

Crown fire behavior
Fire hazard

Fire Risk

Layer creation

To build this layer we obtained a 15-year database of fire ignitions (1986 - 2000) compiled by the Fire Sciences Lab and supplemented with additional data by Brett Dickson of Colorado State University. From this database we extracted all of the human and natural ignitions that had resulted in a fire larger than 50 acres (20 ha). We used a weights-of-evidence analysis to determine significant relationships between human or natural ignitions in relation to terrain, slope, aspect, dominant vegetation type, road densities, and human population densities in the 1km (100 ha or 240 acre) square surrounding the ignitions. Significant variables were then used to create a map of probability of an ignition resulting in a large fire. We combined the maps for human and natural ignitions to create a final map of the probability of a start of any large fire on the landscape. The final layer has a resolution of 1km (100 ha or 240 acres).

Layer accuracy

There is no way to assess the accuracy of this layer without additional ignition data. The values in the layer are based on actual relationships between large fires and predictive variables using historical data and can therefore be considered “real”. However, inaccuracies in the relationship could occur if ignitions are missing from the historical database or if ignition patterns have changed since the 1986 - 2000 period.

Citation: Dickson, B. G., J. W. Prather, Y. Xu, H. M. Hampton, E. N. Aumack, and T. D. Sisk. 2006. Mapping the probability of large fire occurrence in northern Arizona, USA. Landscape Ecology: 21(5):747-761.

Post-Fire Erosion Potential

This is a layer representing the potential for post-fire soil erosion in watersheds across the assessment area. Severe crown fires in forested areas often result in ecosystem degradation including immediate soil damage in burned areas and increased erosion often initiated during the first significant rainfall event after fire. This layer classifies watersheds across the study area into four erosion hazard ratings from one to four and is intended as a relative measure of erodibility across watersheds in the assessment area. Management attention in or around "fire-sensitive" watersheds, those with high erosion potential, could assist in strategies to mitigate ecosystem degradation.

Layer creation

The primary input data were developed by the U.S. Forest Service, U.S. Geological Survey and the ForestERA project. They are the soil's inherent susceptibility to erosion (K-factor), preliminary 6th order watersheds, slope, and the presence of course textured soils in areas expected to burn with high intensity in the event of a crown fire. This last data layer serves as an indicator of the potential for hydrophobic soil production, which can enhance erodibility.

Layer accuracy

We have not undertaken an accuracy assessment on this layer. Such an assessment would require ground data that are not currently available.

Post-Fire Sedimentation Potential

General description

This is a layer representing the potential for post-fire soil sedimentation in watersheds across the assessment area. Severe crown fires in forested areas often result in ecosystem degradation, including sediment deposition of eroded soils originating upslope. Sedimentation frequently occurs following the first significant rainfall event after fire. This layer classifies watersheds across the study area into four sedimentation hazard ratings from one to four and is intended as a relative measure of post-fire sedimentation across watersheds in the assessment area. The model it is based on includes an analysis of post-fire erosion potential within 100 meters of stream channels and the sediment transport characteristics of stream reaches. Management attention in or around watersheds with high sedimentation potential, could assist in strategies to reduce the risk of ecosystem degradation following catastrophic wildfire.

Layer creation

The primary input data were developed by the U.S. Forest Service, U.S. Geological Survey and the ForestERA project. They are stream channels, the soil's inherent susceptibility to erosion (K-factor), preliminary 6th order watersheds, slope, and the presence of course textured soils. In addition, digital orthophotoquadrangles (digital black and white aerial photographs) and topographic maps were used in a manual interpretation of stream morphology.

Layer accuracy

We have not undertaken an accuracy assessment on this layer. Such an assessment would require ground data that are not currently available.

Post-Fire Flooding Potential

General description

This is a layer representing the potential for post-fire flooding in watersheds across the assessment area. Severe crown fires in forested areas can result in flooding during and directly after rainfall events, especially during the period after fire and preceding the regrowth of sufficient understory vegetation to support infiltration. This layer classifies watersheds across the study area into four flood potential ratings from one to four and is intended as a relative measure of post-fire flooding across watersheds in the assessment area. Watersheds with higher road density, stream density, and ruggedness (a measure of steepness) are predicted to have a higher potential for post-fire flooding. Management attention in or around watersheds with high post-fire flooding potential, could assist in strategies to reduce the risk of ecosystem degradation following catastrophic wildfire.

Layer creation

The primary input data were developed by the U.S. Forest Service and U.S. Geological Survey. They are stream channels, roads, and elevation.

Layer accuracy

We have not undertaken an accuracy assessment on this layer. Such an assessment would require ground data that are not currently available.

Relative Risk to Municipal Watersheds

We developed this layer in consultation with USGS and Salt River Project (SRP) hydrologists who identified municipal water supplies inside and outside of the study area. Municipal water supplies were defined as reservoirs that provide municipal water to communities. This layer is not intended to capture other important water uses, such as irrigation, recreation, or fisheries. We evaluated the watersheds in the study area that drain into municipal water supplies for their risk to impact those water supplies. Watersheds containing reservoirs were given a value of one, whereas watersheds feeding more distant reservoirs were given values of one or less based on an assessment of the relative probability that sediment transport would reach those reservoirs following a wildfire. This analysis is based on preliminary 6th order watershed boundaries supplied by the US Forest Service.

Predicted Mexican Spotted Owl Nesting and Roosting Habitat

This layer is a prediction of the extent of nesting and roosting habitat for Mexican Spotted Owls (Strix occidentalis lucida) in the assessment area.

Layer creation

The extent of MSO nesting and roosting habitat was defined by using simple rules chosen through review of the scientific literature and an assessment of known owl nesting sites in the region. Habitat was defined as areas where the dominant overstory vegetation consisted of Pine-Oak, Mixed Conifer, or Ponderosa Pine on steep slopes (> 12 degrees), and where basal area exceeded 75 ft2 / acre (17 m2 / ha). These areas covered approximately 30% of the assessment area. In addition, predicted habitat was further assessed using the Mahalanobis distance statistic and vegetation (tree density, canopy cover, basal area) and terrain (slope, aspect) characteristics around known owl nest sites. This statistic is used to determine how divergent a given location on the landscape is as compared to the typical characteristics of the landscape at known nest sites. Thus, the likelihood that owls will use a particular area can be assessed. Nest site data and expertise on owls were provided by Joseph Ganey, Jeff Jenness, and Bill Block of the US Forest Service, Rocky Mountain Research Station.

Layer accuracy

Of the 132 known MSO nest sites in the assessment area, 113 (85%) fell within the boundaries of the extent of predicted habitat, indicating that the model provided a very high level of accuracy for identifying nesting habitat. In addition, several additional datasets of owl observation (audio, visual, radiotracking), were checked and 70 to 90% of the observations from these datasets fell within the extent of predicted habitat.

Contact us for data

Mexican Spotted Owl Management Guidelines Habitat Data

This is a layer representing management definitions of Mexican Spotted Owl habitat across the assessment area. The layer depicts MSO habitat as defined by the management guidelines for MSO produced by the Fish and Wildlife Service. Habitat is divided into protected and restricted habitat categories. These are management definitions that effect the locations and types of forest management that can take place within the boundaries of each habitat class.

Layer creation

This layer was created using our vegetation composition, slope, and special areas layers, a layer delineating MSO Protected Activity Centers (PACs), obtained from the Forest Service, and the MSO management guidelines. Protected habitat includes PACs, and areas of pine-oak and mixed conifer habitat on high slopes (> 40 degrees) or within wilderness areas. Restricted habitat includes other areas of pine-oak or mixed conifer habitat.

Layer accuracy

The layer is used to identify areas with owl activity or with specific habitat characteristics known to be important to owls for management purposes.

Contact us for data

Northern Goshawk Nesting Habitat Data

This layer is a prediction of the extent of nesting and post-fledging habitat for Northern Goshawks (Accipiter gentilis) in the assessment area.

Layer creation

The extent of goshawk nesting and roosting habitat was defined by using simple rules chosen through review of the literature and an assessment of known goshawk nesting sites in the region. Habitat was defined as areas where basal area exceeded 75 ft2 / acre (17 m2 / ha) and canopy cover exceeded 40%. These areas covered approximately 42% of the assessment area. In addition, predicted habitat was further assessed using the Mahalanobis distance statistic as described for Mexican Spotted Owl Habitat. This statistic is used to determine how divergent a given location on the landscape is as compared to the typical characteristics of the landscape at known nest sites. Thus, the likelihood that goshawks will use a particular area can be assessed. Nest site data and expertise on goshawks were provided by Paul Beier of Northern Arizona University and Micheal Ingraldi of the Arizona Game and Fish Department.

Layer accuracy

Of the 123 known goshawk nest sites in the assessment area, 102 (83%) fell within the boundaries of the extent of predicted habitat, indicating that the model provided a very high level of accuracy for identifying nesting habitat.

Passerine Avian Species Richness Data

This layer is a predictive model of passerine avian species richness across the assessment area. The predicted richness is the total number of species predicted to occur out of a group of 30 species typical of ponderosa pine and associated vegetation types in this region. Predicted species richness ranges from 6.8 to 14.3 within the assessment area.

Layer creation

The models were built using a classification tree methodology and point count data from 320 locations in the assessment area. Data were provided by Brett Dickson of Colorado State University, Kerry Griffis-Kyle of Syracuse University, Bill Block of the USDA Forest Service Rocky Mountain Research Station, Carol Chambers of Northern Arizona University, and Steve Rosenstock of the Arizona Game and Fish Department. Predictive variables in the models include basal area, tree density, canopy cover, slope, and aspect. Due to lack of data in other vegetation types, this model was only created in areas where the dominant vegetation is either ponderosa pine or pine-oak.

Layer accuracy

An internal accuracy assessment of this layer revealed that 44% of the variation in species richness in the training data was explained by the model (r² = 0.44). An external accuracy assessment was undertaken using linear regression and 56 points of data withheld from the model building procedure. For each node of the tree, the predicted value of the node was compared with the value of each of the points that fell within that node. This analysis revealed a highly significant relationship between the actual and predicted values (r² = 0.41, P < 0.0001).

Tassel-eared Squirrel Density and Recruitment Data

Click on maps for larger images

General description

These layers describe predicted Tassel-eared Squirrel (Sciurus aberti) population parameters (density and recruitment) in the region. Squirrel densities range from zero to 0.65 squirrels / ha (0 to 0.26 squirrels/acre) within the assessment area while squirrel recruitment ranges from zero to 0.26 juveniles / ha (0 to 0.11 juveniles/acre) within the assessment area. Tassel-eared Squirrels are heavily dependant on Ponderosa Pine and, thus, areas in vegetation types that do not include ponderosa pine are considered to have no squirrels. Also, since squirrel populations fluctuate dramatically in the region due to variation in weather and food resources, the values in these layers should be considered relative indicators of density and recruitment and not absolute numbers. We also have location data for known goshawk post-fledging areas (PFAs).

Layer creation

We created these layers using linear regression techniques and data collected by Norris Dodd of the Arizona Game and Fish Department. Dodd studied squirrels at 18 locations across the study area over a four-year period between 2000 and 2003. We assessed relationships between squirrel population parameters on these plots and forest structural variables for which we had data across the landscape. The strongest of these relationships were with basal area at each study site and between populations with canopy cover across an approximately 100 ha (240 acre) area surrounding each study site. These two variables were used as predictor variables in the regression models.

Layer accuracy

Linear regression analysis indicated that squirrel densities were strongly correlated with increasing basal area (P = 0.0003) and increasing canopy cover over an area of approximately 90 ha (P = 0.0029). The full model including both parameters was highly significant (P < 0.0001, r² = 0.87). Regression analysis indicated that squirrel recruitment was strongly correlated with increasing basal area (P = 0.0153) and increasing canopy cover over an area of approximately 120 ha (P = 0.0095). The full model including both parameters was highly significant (P = 0.0011, r² = 0.60).

Citation: Prather, J. W., N. L. Dodd, B. G. Dickson, H. M. Hampton, Y. Xu, E. N. Aumack, and T. D. Sisk. 2006. Landscape models to predict the influence of forest structure on Tassel-eared Squirrel populations. Journal of Wildlife Management: 70:723-731

Pronghorn Movement Corridors

In response to recommendations from the wildlife breakout group at the May '04 WMPALA Workshop we have created a layer that shows a number of locations where Pronghorn Antelope (Antelocapra americana) movement corridors might be placed to facilitate pronghorn movement between large patches of current habitat.

Creation of the layer was a three-step process. First, large patches (> 40ha) of existing habitat were identified. Based on the recommendations of the wildlife breakout group we identified all areas where in the pronghorn habitat suitability map with values of 0.7 or greater. We then filtered this layer to eliminate those patches that were smaller than 40 ha. As a second step we identified areas that would likely be unsuitable as corridors for pronghorn. These areas included Mexican Spotted Owl Protected Activity Centers, Northern Goshawk nesting and postfledging stands, areas with slopes greater than 20 degrees, private property, and areas within 1/4 mile of major highways. In the third step we digitized potential corridors between habitat patches that were less than 5km apart. We felt that it was unlikely that more isolated patches would be connected. We did not place corridors across any areas identified as unsuitable (see above) for pronghorn and placed corridors between patches along paths that help the highest quality habitat available between those patches. This was done because thinning areas that were already somewhat less forested would result in higher quality habitat within the corridor areas. All corridors have a width of 450m (~ 1/4 mile).

Pronghorn Habitat Suitability

We have created a layer representing predicted habitat suitability for Pronghorn Antelope (Antelocapra americana) across the assessment area. The layer was built using simple rules along with slope and canopy cover as predictor variables. The values for the area range from zero to 1 and are unitless. Note that this is not a layer representing habitat quality, which would require inclusion of some measure of survivability or reproductive success. Locational (radiotracking) data and expertise were provided by Rick Ockenfels of the Arizona Game and Fish Department.

Wild Turkey Habitat Suitability

We have created a layer representing habitat suitability for Merriam's Wild Turkey (Meleagris gallipavo merriami) across the assessment area. This layer is best viewed as representing roosting habitat suitability, but turkeys are almost always found near roost sites, so it is useable as a general habitat layer. The layer was built using simple rules along with slope, canopy cover, and basal area as predictor variables. The values for the area range from zero to 1 and are unitless. Note that this is not a layer representing habitat quality, which would require inclusion of some measure of survivability or reproductive success. Roost site data and expertise were provided by Brian Wakeling of the Arizona Game and Fish Department.

Individual Passerine Bird Species Habitat

We have created layers representing the probability of occurrence for nine individual species of passerine birds. The models were built using a classification tree methodology and point count data from 320 locations in the assessment area provided by the same collaborators listed above in the species richness section. Predictive variables in the models include basal area, tree density, canopy cover, slope, and aspect. Predictive variables in the models include basal area, tree density, canopy cover, slope, and aspect. Due to lack of data in other vegetation types, this model was only created in areas where the dominant vegetation is either ponderosa pine or pine-oak.

Wildlife Corridors

A full assessment of wildlife corridors is not available as there is little information on where animals are moving to and from, which is central to identification of corridors for movement. However, we have identified several locations in the assessment area that are known bottlenecks for the movements of Wild Turkeys (Meleagris gallipavo), Black Bear (Ursus americanus), and White-tailed Deer (Odocolius virginianus). In addition, we have identified the most likely routes for movements of these species to and from the bottleneck locations to locations 0.6 miles (1 km) away. This is a very general layer meant to identify areas where treatments may block animal movements due to existing constraints on their movements. The bottleneck locations were identified by Rick Miller of the Arizona Game and Fish Department and Paul Beier or Northern Arizona University.

Last updated January 10, 2007