This dataset explores the behavioral responses of Black-tailed deer to a megafire in Mendocino County, CA, USA. Megafire dramatically changes landscapes, potentially placing pressure on native wildlife species. Recent work has found that Black-tailed deer alter their movement behaviors and home ranges following fire to compensate for the dramatic changes that follow these wildfire events. These analyses serve as a direct follow-up to that research and address how long these movement adjustments last on the landscape.

Data Collection

Collars were placed on deer from 2017-2020. We use recorded GPS locations to assess deer home range size (2017-2019), deer habitat selection (2018-2019), and movement behavior (2018-2019).

Analysis Methods

We use three primary analyses to assess how these changes in deer movement may change over time:

1. Comparison of Deer Home Range Sizes - Comparing deer home range size before and after the fire (as well as across seasons) using the 95% KUD method of home range size estimation.
2. Resource Selection Function (RSF) - How does habitat selection change over time across collared black-tailed deer following megafire?
3. Hidden Markov Movement Models (HMM) - How does movement behavior (behavioral states: resting vs. travel) change over time following a megafire?

Data

Note: NAs indicate the corresponding data is not available or applicable (e.g., The column 'Sex' in "Deer RSF Input Data.csv" is listed as either NA or False because all deer in the study are female/doe)

• "deer gps data.csv" - all collected GPS points (without error points) used to estimate LoCoH home ranges

ID: ID of the collared animal
Timestamp: Date and time of the GPS fix
Deployed: Date that the GPS collar was deployed onto the animal
Dropped: Date that the GPS collar was dropped from the animal
Period: Time period that the GPS data was collected within the study. The five study periods are: "fire/Recently Burned", "spring/First Spring", "fall/1 Year Post Fire", "prespring", and "prefire".

• "Deer RSF Input Data.csv" - input GPS data and environmental covariates extracted from each GPS point to run the resource selection function model

ID: ID of the collared animal
Sex: NA, all deer for analyses are Does
Timestamp: Date and time of the GPS fix
Used: binary (0/1) indicates whether the GPS point is a true recorded GPS collar fix (1), or a randomly generated non-use point (0).
TimePeriod: Time of day (categorical) that the GPS point was taken (Day, Night, or Crepuscular)
Ruggedness: The extracted value of ruggedness (a variation on demography) at the GPS location
Severity: Extracted value of fire severity at GPS location
Severity_scaled: Scaled Severity value at GPS point
Elevation: Extracted value of elevation at GPS location (meters)
Elevation_scaled: the scaled value of Elevation
predrisk: The extracted value of the probability of Mountain Lion presence at the GPS location
predrisk_scaled: the scaled value of predrisk
veg: Categorical value of vegetation type at GPS location (1 = chaparral, 2 = grassland, 3 = woodland)
distWater: Distance of GPS location from nearest stream/ravine (meters)
distWater_scaled: the scaled value of distWater
Severity_sq: Squared (non-linear) value of Severity at GPS location
Period: Time period that the GPS data was collected within the study. The five study periods are: "fire/Recently Burned", "spring/First Spring", and "fall/1 Year Post Fire".
Fire_Date: Date of 2018 Mendocino Complex Fire (July 27, 2018)
TimeSince_Burn: Number of Days since "Fire_Date" (days)
BurnLag: Scaled value of TimeSince_Burn

• "Deer RSF output.csv" - Tidy model output of the deer resource selection function

effect: Type of effect included in the model (fixed or random effect)
group: Grouping value of random effects
term: Covariate included in the RSF model
estimate: Estimated beta-coefficient for covariates within the model
std.error: Estimated standard error of estimate for each covariate
statistic: the value of T-statistic for each covariate in the RSF model
p.value: p-value associated with a statistic of the RSF model

• "deer_LoCoH_hrs.csv" - Estimated home range size for deer across all time periods

id: ID of the collared animal
Area: Estimated area of home range for individual animals (meters)
Period: Period: Time period that the GPS data was collected within the study. The five study periods are: "fire/Recently Burned", "spring/First Spring", "fall/1 Year Post Fire", "prespring", and "prefire".
area_km: Estimated area of home range for individual animals (kilometers)

• "evi severity comparison pts.csv" - randomly sampled points from the study area used to compare evi and severity values following the 2018 Mendocino complex fire

Number: Number of sampled points
Severity: Extracted fire severity value from the sampled point
NDVI: Extracted NDVI value from the sampled point
NDVI_scaled: Scaled NDVI value from the extracted point
EVI: extracted EVI value from the sampled point
EVI_scaled: Scaled EVI value from extracted point
veg: Categorical value of vegetation type at sampled point (1 = chaparral, 2 = grassland, 3 = woodland)
Vegetation Type: Categorical vegetation type at the sampled point
long: Longitude of recorded GPS point
lat: Latitide of recorded GPS point
Period: Time period that the GPS data was collected within the study. The five study periods are: "fire/Recently Burned", "spring/First Spring", and "fall/1 Year Post Fire".

• "HMM_input_data.csv" - Input GPS data and extracted covariate values for hidden Markov model (HMM)

ID: ID of the collared animal and time period of recorded GPS point
Timestamp: Date and time of the GPS fix
Deployed: Date that the GPS collar was deployed onto the animal
Dropped: Date that the GPS collar was dropped from the animal
Severity: Extracted value of fire severity at GPS location
Severity_scaled: Scaled Severity value at GPS point
predrisk: The extracted value of the probability of Mountain Lion presence at the GPS location
predrisk_scaled: scaled value of predrisk
veg: Categorical value of vegetation type at GPS location (1 = chaparral, 2 = grassland, 3 = woodland)
distWater: Distance of GPS location from nearest stream/ravine
distWater_scaled: the scaled value of distWater
Period: Time period that the GPS data was collected within the study. The five study periods are: "fire/Recently Burned", "spring/First Spring", and "fall/1 Year Post Fire".
Fire_Date: Date of 2018 Mendocino Complex Fire (July 27, 2018)
TimeSince_Burn: Number of Days since "Fire_Date"
BurnLag: Scaled value of TimeSince_Burn
chap: binary variable, yes (1) or no (0) if GPS location was within chaparral
grass: binary variable, yes (1) or no (0) if GPS location was within grassland
wood: binary variable, yes (1) or no (0) if GPS location was within woodland

• state_data.csv

Value: Number of estimated points (with error included) in each behavioral state
State: The estimated behavioral State (1 = Resting, 2 = Traveling)
Period: Time period that the GPS data was collected within the study. The five study periods are: "fire/Recently Burned", "spring/First Spring", and "fall/1 Year Post Fire".
perc: Percent of total GPS points in the associated behavioral state during the indicated period.

• "deer RSF model.RData" - Output model of deer resource selection function
• "HMM_model.RData" - Output model of deer hidden Markov model

R Scripts

1. "Home Range Analyis.R" - Script for estimating deer home ranges from recorded GPS locations. Uses paired t-tests to compare home range sizes across time periods
2. "Deer RSF Models.R" - Script to run resource selection function to estimate deer habitat selection by comparing true use points to available points.
3. "Deer HMM 2-State Model.R" - Script to run hidden Markov model to assign behavioral states ("resting" or "traveling") to deer GPS movement data. Also estimates the probability of a GPS point being in the behavioral state as a function of covariates.
4. Visualizations.R - Code to create figures and visuals from analyses.