Long-term benefits of burns for large mammal habitat undermined by large, severe fires in the American West
Data files
Nov 12, 2025 version files 1.82 GB
-
Data.zip
1.82 GB
-
fire-RSF-models.R
28.10 KB
-
fire-SSF-models.R
30.93 KB
-
README.md
8.76 KB
-
rsf-predictions.R
28.56 KB
-
source_setup.R
2.36 KB
Abstract
Escalating wildfire frequency and severity are altering wildland habitats worldwide. Yet investigations into fire impacts on wildlife habitat rarely extend to the macroecological scales relevant to species conservation and global change processes. We evaluate the effects of wildfire on habitat quality and selection by large mammals spanning three trophic levels in the Western United States. We analyze 12 years of GPS telemetry data for 2,966 mule deer (Odocoileus hemionus), 52 black bears (Ursus americanus), and 74 cougars (Puma concolor) across Utah and Nevada, USA. Over 800 areas burned between 1990-2022 overlapped with the home ranges of 1,892 animals, resulting in almost 23,000 km2 of burned habitat and representing 12.8% of the total home range area for animals in our sample. Habitat suitability models for 664 mule deer, 14 black bears, and 11 cougars indicated that burns improved summer home range quality for mule deer and black bears by 7% and 14%, respectively, highlighting the benefits of fires for nutrient cycling, understory herbaceous growth, and resultant caloric value for animal nutrition. When making fine-scale movement decisions, however, mule deer avoided burned habitats, and all three species generally avoided high-severity burns for up to 30 years post-fire. Thus, the effects of burns on wildlife habitat selection appear to be dependent on spatial scale. Given projected increases in large, severe fires, our results suggest potential reductions in beneficial habitat for wildlife in the long term. However, our results also suggest that prescribed burns, because of their smaller spatial footprints and lower severity relative to wildfires, can benefit wildlife habitat quality through improvements in forage, cover, and other vegetation characteristics. Therefore, managing for low-severity burns and limiting large, severe wildfires, e.g., via prescribed burns or fire control policies, could positively impact the habitat quality of these three common species and, therefore, the economic and ecosystem services they provide.
Dataset DOI: 10.5061/dryad.fqz612k4d
Description of the data and file structure
This data and code supports a study on wildlife responses to burned habitats at broad spatiotemporal scales. Data are available for cougar, black bear, and mule deer habitat selection models and other analyses.
Files and variables
File: fire-SSF-models.R
Description: Code for building individual step-selection functions, including model evaluation & coefficient bootstrapping
File: rsf-predictions.R
Description: Includes code (commented) to demonstrate the extraction process for RSF predictions in burned and simulated unburned habitats to calculate MQD (median quality difference). The output MQD data are available for the statistical analyses below the extraction procedures. Also includes comparison of wildfires versus prescribed burns.
File: source_setup.R
Description: Source file containing functions and package libraries. All R files have source command at beginning which should be run using this file.
File: fire-RSF-models.R
Description: Code for building individual resource selection functions, including model evaluation & coefficient bootstrapping
File: Data.zip
Description: 4 data files: (1) data for SSF models, (2) data for RSF models, (3) data for MQD statistical analyses, and (4) data on burn characteristics to compare wildfires and prescribed burns. All latitude and longitude values in GPS datasets have been scaled and centered.
Data File: final-RSF-model-data.rds
Description: Animal GPS data used for RSF models. Each row represents one animal ID-season, with a nested list-column "data" containing each animal's model dataset (R data frame with list-column). Data columns: "SAID" = animal ID; "species" = species of animal; "studyarea" = study area; "season" = season, summer or winter; "data" = list-column containing animal's GPS dataset; "pct_fire" = percent of animal's used GPS points within burn perimeters; "n_row" = number of rows in 'data'.
Columns within "data": "species" = species of animal; "studyarea" = study area; "animalID" = animal ID; "UID" = animal ID; "sex" = animal sex; "age" = animal age; "lat" = GPS location latitude; "long" = GPS location longitude; "ts_utc" = GPS location timestamp; "season" = season, summer or winter; "y_var" = real (1) or random location (0) for response variable in RSF model; "dem" = elevation; "tri" = terrain ruggedness index; "NDVI" = Normalized Difference Vegetation Index; "NDSI" = Normalized Difference Snow Index; "biomass" = satellite derived biomass; "ptid" = unique record ID; "FireDate" = date of most recent fire at location; "fire" = fire status, burned/unburned; "fire_dist" = distance to nearest fire perimeter; "fire_cbi" = composite burn index or most recent fire at location; "fire_yrs" = years since most recent fire at location; "fire_yrs_cat" = categorical years since fire; "fire_logdist"= log-distance to fire perimeter; "log_dist_road" = log-distance to road; "dem_s" = scaled elevation; "tri_s" = scaled TRI; "NDVI_s" = scaled NDVI; "NDSI_s" = scaled NDSI; "biomass_s" = scaled biomass; "log_dist_road_s" = scaled log-distance to road; "Hab_type" = NLCD habitat type; "Hab_Shrub" = binary shrub habitat, yes (1) or no (0)
Data File: final-SSF-model-data.rds
Description: Animal GPS 'steps' used for RSF models. Each row represents one animal ID-season, with a nested list-column "data" containing each animal's model dataset (R data frame with list-column). Data columns: "species" = species of animal; "studyarea" = study area; "UID" = animal ID; "season" = season, summer or winter; UID_s = animal ID + season; "n_row" = number of rows in 'data'; "data" = list-column containing animal's GPS steps dataset.
Columns within 'data': "UID" = animal ID; "species" = species of animal; "studyarea" = study area; "animalID" = animal ID; "sex" = animal sex; "age" = animal age; "burst_" = burst ID; "t1_" = timestamp at beginning of step; "t2_" = timestamp at end of step; "dt_" = time difference during step; "case_" = real step end point (TRUE) or randomly generated available endpoint (FALSE) used as response variable in SSF models; "step_id_" = Id of step for grouping available endpoints in model (strata); "sl_" = step length (m); "ta_" = turn angle; "log_sl" = log-step length; "cos_ta"= cosign of turning angle; "random_id" = unique point id; "season" = season; "lat" = latitude at step endpoint; "lon" = longitude at step endpo*;* "ts" = timestamp at end of step; "fire" = fire status, burned/unburned; "fire_dist" = distance to nearest fire perimeter; "fire_cbi" = composite burn index or most recent fire at location; "fire_yrs" = years since most recent fire at location; "fire_yrs_cat" = categorical years since fire; "fire_logdist"= log-distance to fire perimeter; "tri" = terrain ruggedness index; "distance_to_road" = distance to road (m); "NDVI" = Normalized Difference Vegetation Index; "NDSI" = Normalized Difference Snow Index; "biomass" = satellite derived biomass; "DEM" = elevation; "nlcd" = NLCD habitat type; "Hab_Shrub" = binary shrub habitat, yes (1) or no (0); "log_dist_road" = log-distance to road; "biomass_s" = scaled biomass; "NDVI_s" = scaled NDVI; "NDSI_s" = scaled NDSI; "log_dist_road_s" = scaled log-distance to road; "dem_s" = scaled elevation; "tri_s" = scaled TRI;
Data File: rsf-prediction-results-df.rds
Description: Records of fire-animal 95% home range overlap areas generated from RSF model predictions of MQD and MTBS fire perimeters (R data frame). Data columns: "Event_ID" = MTBS Fire ID; "Incid_Name" = MTBS fire name; "Incid_Type" = MTBS fire type; "Ig_Date" = fire date; "BurnBndLat" = fire latitude; "BurnBndLon" = fire longitude; "yrs_since_fire" = years between fire date and animal GPS data median date; "fire_yrs_cat" = cateogrical representation of years since fire; "UID" = animal ID; "season" = season, winter or summer; "id" = animal ID (minus study site); "hr_area" = animal 95% home range size; "yrs_min" = earliest GPS data year; "yrs_max" = latest GPS data year; "yrs_med"= median GPS data year; "date_min" = earliest GPS data date; "date_max" = latest GPS data date; "species" = animal species; "UID_s" = animal ID x season; "date_med" = median GPS data date; "nlcd_majority" = most common NLCD category in animal home range; "fire_area" = size of fire in sq-km; "mean_CBI" = mean Composite Burn Index value within fire perimeter; "pct_hr_fire" = percent of animal 95% home range overlapping with fire perimeter; "mean_rsf_diff" = average value of RSF difference between fire x HR overlap area; "med_rsf_diff" = median value of RSF difference between fire x HR overlap area; "fire_yrs_cat0-9" = RSF model coefficient for effect of recent fires on habitat selection; "fireburned:fire_logdist" = RSF model coefficient for effect of distance to fire perimeter, outside of fire perimeters, on habitat selection; "fireunburned:fire_logdist" = RSF model coefficient for effect of distance to fire perimeter, inside of fire perimeters, on habitat selection; "fire_yrs_cat0-9:fire_cbi" = RSF model coefficient for effect of CBI of recent fires on habitat selection; "fire_yrs_cat10-30" = RSF model coefficient for effect of older fires on habitat selection; "fire_yrs_cat10-30:fire_cbi" = RSF model coefficient for effect of CBI of older fires on habitat selection
Data File: rx-fire-plot-data.rds
Description: Records of all fires from MTBS overlapping with 95% animal home ranges included in our analyses (R data frame). Data columns: [1] "Event_ID" = MTBS Fire ID; "Incid_Name" = MTBS fire name; "Incid_Type" = MTBS fire type; "Ig_Date" = fire date; "BurnBndLat" = fire latitude; "BurnBndLon" = fire longitude; "fire_area" = size of fire in sq-km; "median_cbi" = median Composite Burn Index value within fire perimeter; "mean_cbi" = mean Composite Burn Index value within fire perimeter; "nlcd" = most common NLCD category in fire perimeter; "log_area" = log-transformed area of fire
Access information
Data was derived from the following sources:
- MTBS fire perimeter dataset: https://www.mtbs.gov/
- NLCD habitat type data: https://www.usgs.gov/centers/eros/science/national-land-cover-database
