This code examines the influence of drought on human-wildlife conflict across the state of California, USA. Drought and other climate change-driven disturbances are expected to amplify human-wildlife conflict in the future. However, limited empirical studies have explicitly tested whether this theorized relationship is consistent across a diverse group of species and land-use contexts. In this study, we leverage an extensive community science-derived dataset from the California Department of Fish and Wildlife (CDFW) of human-wildlife interaction incident reports to measure the influence of precipitation, human population density, seasonality, and tree cover on the frequency of reported incidents. Human-wildlife conflict remains one of the greatest challenges to wildlife conservation globally, and understanding the potential influences of climate change and climatic disturbances on conflict will be key to guiding future successful conservation strategies.

Data Collection

We used the CDFW's Wildlife Incident Reporting System (WIR) to examine the influence of changes in precipitation on the frequency of reported human-wildlife conflicts. Individual reports are added to the database through either self-reporting via the online Wildlife Incident Reporting (WIR) interface (https://apps.wildlife.ca.gov/wir) or direct input by regional CDFW biologists and specialists on behalf of individuals between the years of 2017 and 2023.

Analysis Methods

We use two primary analysis methods to examine how changes in precipitation influence human-wildlife conflict reporting:

1. All Conflict Reports Negative-Binomial - Negative Binomial regression that models the number of aggregated incidents (Depredation and General Nuisance reports) for each annual month-grid cell as a function of precipitation, human population density, seasonality, and tree cover. In this model, all incidents were grouped together (i.e., not separated by species) in order to examine the influence of drought on the overall phenomenon of human-wildlife conflict.
2. Diet-Group Multispecies Negative-Binomial Model - Bayesian hierarchical model that analyses the influence of drought and other environmental covariates on human wildlife conflict for diet-groups and individual species. Modeled species are split across 3 diet groups according to their life histories: herbivores, omnivores, and carnivores

We use an additional 4 analyses to compare whether precipitation impacts the type of report (Depredation, Sighting, Potential Human Conflict, and General Nuisance) differently.

1. Depredation Reports Negative-Binomial - Negative Binomial regression that models the number of aggregated depredation reports for each annual-month grid cell.
2. Nuisance Reports Negative-Binomial - Negative Binomial regression that models the number of aggregated general nuisance reports for each annual-month grid cell.
3. Potential Human Conflict Reports Negative-Binomial - Negative Binomial regression that models the number of aggregated potential human conflict reports for each annual-month grid cell.
4. Sighting Reports Negative-Binomial - Negative Binomial regression that models the number of aggregated sighting reports for each annual-month grid cell.

Data

CA-Fire-HWC.zip

Data Files

1. "all_monthly_data.csv" - all aggregated human-wildlife conflict reports by year-month-cells

order: numerical order of rows
month.year: unique ID representing each month-year (e.g., January 2017 = 1)
gridID: unique ID of the grid cell (50x50km), conflicts were aggregated into
year: year that the aggregated conflicts were reported in
month: numerical month (1-12) in which the aggregated conflicts were reported in
n.incidents: number of aggregated incidents within each year-month-cell

1. "data-rnD08.csv" - Historic precipitation data from California used for visualizations. Obtained from: https://oehha.ca.gov/climate-change/epic-2022/changes-climate/precipitation

Year: year of recorded precipitation data
Precipitation (inches): Annual statewide precipitation in inches

1. "depredation_monthly_data.csv" - all aggregated depredation conflict reports by year-month-cells

order: numerical order of rows
month.year: unique ID representing each month-year (e.g., January 2017 = 1)
gridID: unique ID of the grid cell (50x50km), conflicts were aggregated into
year: year that the aggregated conflicts were reported in
month: numerical month (1-12) in which the aggregated conflicts were reported in
n.incidents: number of aggregated incidents within each year-month-cell

1. "diet_group model_data.Rda" - Model output data of diet-group conflict multi-species analysis
2. "diet_multispecies_monthly_data.csv" - aggregated human-wildlife conflict reports aggregated by year-month-cells for the 15 species modeled by diet groups.

order: numerical order of rows
month.year: unique ID representing each month-year (e.g., January 2017 = 1)
gridID: unique ID of the grid cell (50x50km), conflicts were aggregated into
year: year that the aggregated conflicts were reported in
month: numerical month (1-12) in which the aggregated conflicts were reported
n.incidents: number of aggregated incidents within each year-month-cell
species: species that the aggregated conflicts are associated with

1. "metadata_depredation.csv" - metadata for associated depredation conflict type analysis

order: numerical order of rows
month.year: unique ID representing each month-year (e.g., January 2017 = 1)
gridID: unique ID of the grid cell (50x50km), conflicts were aggregated into
year: year that the aggregated conflicts were reported in
month: numerical month (1-12) in which the aggregated conflicts were reported in
census: mean human population density on the log scale for each grid cell
treecover: mean percent tree cover for each grid cell
income: mean value of median household income for each grid cell
ppt.lag.3: summed precipitation at a grid cell from the previous 3 months
ppt.lag.6: summed precipitation at a grid cell from the previous 6 months
ppt.lag.12: summed precipitation at a grid cell from the previous 12 months
ppt.lag.3.scale: scaled and centered values of summed precipitation at each grid cell from the last 3 months
ppt.lag.6.scale: scaled and centered values of summed precipitation at each grid cell from the last 6 months
ppt.lag.12.scale: scaled and centered values of summed precipitation at each grid cell from the last 12 months
census.scale: scaled and centered values of human population density at each grid cell
treecover.scale: scaled and centered values of percent tree cover at each grid cell
income.scale: scaled and centered values of median household income at each grid cell
month_sin: sine term of circular transformation for month covariate
month_cos: cosine term of circular transformation for the month covariate
month.scale: scaled and centered values of numerical month data
month.sin.scale: scaled and centered values of sine transformed month data
month.cos.scale: scaled and centered values of cosine-transformed month data

1. "metadata_nuisance.csv" - metadata for associated general nuisance conflict type analysis

order: numerical order of rows
month.year: unique ID representing each month-year (e.g., January 2017 = 1)
gridID: unique ID of the grid cell (50x50km), conflicts were aggregated into
year: year that the aggregated conflicts were reported in
month: numerical month (1-12) in which the aggregated conflicts were reported in
census: mean human population density on the log scale for each grid cell
treecover: mean percent tree cover for each grid cell
income: mean value of median household income for each grid cell
ppt.lag.3: summed precipitation at a grid cell from the previous 3 months
ppt.lag.6: summed precipitation at a grid cell from the previous 6 months
ppt.lag.12: summed precipitation at a grid cell from the previous 12 months
ppt.lag.3.scale: scaled and centered values of summed precipitation at each grid cell from the last 3 months
ppt.lag.6.scale: scaled and centered values of summed precipitation at each grid cell from the last 6 months
ppt.lag.12.scale: scaled and centered values of summed precipitation at each grid cell from the last 12 months
census.scale: scaled and centered values of human population density at each grid cell
treecover.scale: scaled and centered values of percent tree cover at each grid cell
income.scale: scaled and centered values of median household income at each grid cell
month_sin: sine term of circular transformation for month covariate
month_cos: cosine term of circular transformation for the month covariate
month.scale: scaled and centered values of numerical month data
month.sin.scale: scaled and centered values of sine transformed month data
month.cos.scale: scaled and centered values of cosine-transformed month data

1. "metadata_potential.csv" - metadata for associated potential human conflict analysis

order: numerical order of rows
month.year: unique ID representing each month-year (e.g., January 2017 = 1)
gridID: unique ID of the grid cell (50x50km), conflicts were aggregated into
year: year that the aggregated conflicts were reported in
month: numerical month (1-12) in which the aggregated conflicts were reported in
census: mean human population density on the log scale for each grid cell
treecover: mean percent tree cover for each grid cell
income: mean value of median household income for each grid cell
ppt.lag.3: summed precipitation at a grid cell from the previous 3 months
ppt.lag.6: summed precipitation at a grid cell from the previous 6 months
ppt.lag.12: summed precipitation at a grid cell from the previous 12 months
ppt.lag.3.scale: scaled and centered values of summed precipitation at each grid cell from the last 3 months
ppt.lag.6.scale: scaled and centered values of summed precipitation at each grid cell from the last 6 months
ppt.lag.12.scale: scaled and centered values of summed precipitation at each grid cell from the last 12 months
census.scale: scaled and centered values of human population density at each grid cell
treecover.scale: scaled and centered values of percent tree cover at each grid cell
income.scale: scaled and centered values of median household income at each grid cell
month_sin: sine term of circular transformation for month covariate
month_cos: cosine term of circular transformation for the month covariate
month.scale: scaled and centered values of numerical month data
month.sin.scale: scaled and centered values of sine transformed month data
month.cos.scale: scaled and centered values of cosine-transformed month data

1. "metadata_sightings.csv" - metadata for associated sightings conflict type analysis

order: numerical order of rows
month.year: unique ID representing each month-year (e.g., January 2017 = 1)
gridID: unique ID of the grid cell (50x50km), conflicts were aggregated into
year: year that the aggregated conflicts were reported in
month: numerical month (1-12) in which the aggregated conflicts were reported
census: mean human population density on the log scale for each grid cell
treecover: mean percent tree cover for each grid cell
income: mean value of median household income for each grid cell
ppt.lag.3: summed precipitation at a grid cell from the previous 3 months
ppt.lag.6: summed precipitation at a grid cell from the previous 6 months
ppt.lag.12: summed precipitation at a grid cell from the previous 12 months
ppt.lag.3.scale: scaled and centered values of summed precipitation at each grid cell from the last 3 months
ppt.lag.6.scale: scaled and centered values of summed precipitation at each grid cell from the last 6 months
ppt.lag.12.scale: scaled and centered values of summed precipitation at each grid cell from the last 12 months
census.scale: scaled and centered values of human population density at each grid cell
treecover.scale: scaled and centered values of percent tree cover at each grid cell
income.scale: scaled and centered values of median household income at each grid cell
month_sin: sine term of circular transformation for month covariate
month_cos: cosine term of circular transformation for the month covariate
month.scale: scaled and centered values of numerical month data
month.sin.scale: scaled and centered values of sine transformed month data
month.cos.scale: scaled and centered values of cosine-transformed month data

1. "metadata.csv"

order: numerical order of rows
month.year: unique ID representing each month-year (e.g., January 2017 = 1)
gridID: unique ID of the grid cell (50x50km), conflicts were aggregated into
year: year that the aggregated conflicts were reported in
month: numerical month (1-12) in which the aggregated conflicts were reported in
census: mean human population density on the log scale for each grid cell
treecover: mean percent tree cover for each grid cell
income: mean value of median household income for each grid cell
ppt.lag.3: summed precipitation at a grid cell from the previous 3 months
ppt.lag.6: summed precipitation at a grid cell from the previous 6 months
ppt.lag.12: summed precipitation at a grid cell from the previous 12 months
ppt.lag.3.scale: scaled and centered values of summed precipitation at each grid cell from the last 3 months
ppt.lag.6.scale: scaled and centered values of summed precipitation at each grid cell from the last 6 months
ppt.lag.12.scale: scaled and centered values of summed precipitation at each grid cell from the last 12 months
census.scale: scaled and centered values of human population density at each grid cell
treecover.scale: scaled and centered values of percent tree cover at each grid cell
income.scale: scaled and centered values of median household income at each grid cell
month_sin: sine term of circular transformation for month covariate
month_cos: cosine term of circular transformation for the month covariate
month.scale: scaled and centered values of numerical month data
month.sin.scale: scaled and centered values of sine transformed month data
month.cos.scale: scaled and centered values of cosine-transformed month data

1. "monthly_precip_df.csv" - Average statewide precipitation from each month drawn from PRISM 30-year average data (https://prism.oregonstate.edu/normals/)

Month: month of average precipitation
Precipitation: 30-year average amount of precipitation across the state within each month
month_num: numerical number of month (1-12)

1. "nuisance_monthly_data.csv" - all aggregated nuisance conflict reports by year-month-cells

order: numerical order of rows
month.year: unique ID representing each month-year (e.g., January 2017 = 1)
gridID: unique ID of the grid cell (50x50km), conflicts were aggregated into
year: year that the aggregated conflicts were reported in
month: numerical month (1-12) in which the aggregated conflicts were reported in
n.incidents: number of aggregated incidents within each year-month-cell

1. "potential_monthly_data.csv" - all aggregated potential human conflict reports by year-month-cells

order: numerical order of rows
month.year: unique ID representing each month-year (e.g., January 2017 = 1)
gridID: unique ID of the grid cell (50x50km), conflicts were aggregated into
year: year that the aggregated conflicts were reported in
month: numerical month (1-12) in which the aggregated conflicts were reported in
n.incidents: number of aggregated incidents within each year-month-cell

1. "sightings_monthly_data.csv" - all aggregated sighting reports by year-month-cells

order: numerical order of rows
month.year: unique ID representing each month-year (e.g., January 2017 = 1)
gridID: unique ID of the grid cell (50x50km), conflicts were aggregated into
year: year that the aggregated conflicts were reported in
month: numerical month (1-12) in which the aggregated conflicts were reported in
n.incidents: number of aggregated incidents within each year-month-cell

Analysis

R Scripts

1. "All_Species_Model_Setup.R" - Script to set up the input model data, constants, inits, and supporting functions for the All Species Negative Binomial Regression
2. "All_Species_Model.R" - Script to run the Bayesian negative binomial regression for the All Species Model
3. "Diet_Multispecies_Model_Setup.R" - Script to set up the input model data, constants, inits, and supporting functions for the Diet-Group Multispecies Negative Binomial Regression Model
4. "Diet_Multispecies_Model.R" - Script to run the Bayesian negative binomial regression for the Diet-Group Multispecies Model.
5. "Depredation_Model_Setup.R" - Script to set up the input model data, constants, inits, and supporting functions for the depredation reports Negative Binomial Regression
6. "Depredation_Conflict_Model.R" - Script to run the Bayesian negative binomial regression for the all depredations model
7. "Nuisance_Model_Setup.R" - Script to set up the input model data, constants, inits, and supporting functions for the nuisance reports Negative Binomial Regression
8. "Nuisance_Conflict_Model.R" - Script to run the Bayesian negative binomial regression for the all general nuisances model
9. "Potential_Model_Setup.R" - Script to set up the input model data, constants, inits, and supporting functions for the potential human conflict reports Negative Binomial Regression
10. "Potential_Conflict_Model.R" - Script to run the Bayesian negative binomial regression for the all potential human conflict model
11. "Sightings_Model_Setup.R" - Script to set up the input model data, constants, inits, and supporting functions for the sighting reports Negative Binomial Regression
12. "Sightings_Conflict_Model.R" - Script to run the Bayesian negative binomial regression for the all sightings model
13. "all_conflicts_samples_coda.Rda" - Posterior samples of the All Species Conflict Model

Results

Data Files

1. "all_conflicts_3monthlag_summary.csv" - Summarized model coefficients of the All Conflict Model using the summed 3 months prior to precipitation

mean: Average value of parameter
sd: Standard deviation of parameter
90%_HPDL: Lower bound value of 90% Confidence Interval of the parameter
90%_HPDU: Upper bound value of 90% Confidence Interval of the parameter
Rhat: Rhat values for parameter calculated across 3 chains
n.eff: Number of effective samplers for each parameter
nonzero: Denotes whether the 90% Confidence Interval overlaps zero or not
param: Code syntax name of parameter
species: Animal species or species group for the parameter estimate
layer: Denotes whether the parameter is a covariate or other model parameter
parname: Name of each parameter

1. "all_conflicts_6monthlag_summary.csv" - Summarized model coefficients of the All Conflict Model using the summed 6 months prior to precipitation

mean: Average value of parameter
sd: Standard deviation of parameter
90%_HPDL: Lower bound value of 90% Confidence Interval of parameter
90%_HPDU: Upper bound value of 90% Confidence Interval of parameter
Rhat: Rhat values for parameter calculated across 3 chains
n.eff: Number of effective samplers for each parameter
nonzero: Denotes whether the 90% Confidence Interval overlaps zero or not
param: Code syntax name of parameter
species: Animal species or species group for the parameter estimate
layer: Denotes whether the parameter is a covariate or other model parameter
parname: Name of each parameter

1. "all_conflicts_samples_coda.Rda" - Posterior samples of the all-species conflict model
2. "all_conflicts_summary.csv" - Summarized model coefficients of the All Species Conflict Model

mean: Average value of parameter
sd: Standard deviation of parameter
90%_HPDL: Lower bound value of 90% Confidence Interval of parameter
90%_HPDU: Upper bound value of 90% Confidence Interval of parameter
Rhat: Rhat values for parameter calculated across 3 chains
n.eff: Number of effective samplers for each parameter
nonzero: Denotes whether the 90% Confidence Interval overlaps zero or not
param: Code syntax name of parameter
species: Animal species or species group for the parameter estimate
layer: Denotes whether the parameter is a covariate or other model parameter
parname: Name of each parameter

1. "all_depredations_summary.csv" - Summarized model coefficients of the Depredations Conflict Model

mean: Average value of parameter
sd: Standard deviation of parameter
90%_HPDL: Lower bound value of 90% Confidence Interval of parameter
90%_HPDU: Upper bound value of 90% Confidence Interval of parameter
Rhat: Rhat values for parameter calculated across 3 chains
n.eff: Number of effective samplers for each parameter
nonzero: Denotes whether the 90% Confidence Interval overlaps zero or not
param: Code syntax name of parameter
species: Animal species or species group for the parameter estimate
layer: Denotes whether the parameter is a covariate or other model parameter
parname: Name of each parameter

1. "all_nuisance_summary.csv" - Summarized model coefficients of the Nuisance Conflict Model

mean: Average value of parameter
sd: Standard deviation of parameter
90%_HPDL: Lower bound value of 90% Confidence Interval of parameter
90%_HPDU: Upper bound value of 90% Confidence Interval of parameter
Rhat: Rhat values for parameter calculated across 3 chains
n.eff: Number of effective samplers for each parameter
nonzero: Denotes whether the 90% Confidence Interval overlaps zero or not
param: Code syntax name of parameter
species: Animal species or species group for the parameter estimate
layer: Denotes whether the parameter is a covariate or other model parameter
parname: Name of each parameter

1. "all_potential_summary.csv" - Summarized model coefficients of the Potential Human Conflict Model

mean: Average value of parameter
sd: Standard deviation of parameter
90%_HPDL: Lower bound value of 90% Confidence Interval of parameter
90%_HPDU: Upper bound value of 90% Confidence Interval of parameter
Rhat: Rhat values for parameter calculated across 3 chains
n.eff: Number of effective samplers for each parameter
nonzero: Denotes whether the 90% Confidence Interval overlaps zero or not
param: Code syntax name of parameter
species: Animal species or species group for the parameter estimate
layer: Denotes whether the parameter is a covariate or other model parameter
parname: Name of each parameter

1. "all_sightings_summary.csv" - Summarized model coefficients of the Sightings Conflict Model

mean: Average value of parameter
sd: Standard deviation of parameter
90%_HPDL: Lower bound value of 90% Confidence Interval of parameter
90%_HPDU: Upper bound value of 90% Confidence Interval of parameter
Rhat: Rhat values for parameter calculated across 3 chains
n.eff: Number of effective samplers for each parameter
nonzero: Denotes whether the 90% Confidence Interval overlaps zero or not
param: Code syntax name of parameter
species: Animal species or species group for the parameter estimate
layer: Denotes whether the parameter is a covariate or other model parameter
parname: Name of each parameter

1. "derived_month_effect_plot_data.csv" - Predicted coefficients of the effect of seasonality (month) on reported conflict across species.

Month: Month for which the predicted coefficient is estimated
Pred.Low: Lower bound value of 90% Confidence Interval of predicted estimate
Pred.High: Upper bound value of 90% Confidence Interval of predicted estimate
Pred.Mean: Average value of predicted coefficient estimate
Species: Species for which the predicted estimate is made

1. "derived_ppt_efect_plot_data.csv" - Predicted coefficients of the effect of precipitation on reported conflict across species.

Precipitation: Precipitation for which the predicted coefficient is estimated
Pred.Low: Lower bound value of 90% Confidence Interval of predicted estimate
Pred.High: Upper bound value of 90% Confidence Interval of predicted estimate
Pred.Mean: Average value of predicted coefficient estimate
Species: Species for which the predicted estimate is made

1. "deviance_gof.csv" - Observed deviances of fit from posterior predictive check used for plotting goodness-of-fit
   type - categorical distinction if value is from the distribution of deviances from fit ("PPC") or the mean value of observations ("Observed")
   value - numerical value of deviation for model fit
2. "diet_multispecies_conflict_3monthlag_summary.csv" - Summarized model coefficients of the Diet Multispecies Model using the summed 3 months prior to precipitation

mean: Average value of parameter
sd: Standard deviation of parameter
90%_HPDL: Lower bound value of 90% Confidence Interval of parameter
90%_HPDU: Upper bound value of 90% Confidence Interval of parameter
Rhat: Rhat values for parameter calculated across 3 chains
n.eff: Number of effective samplers for each parameter
nonzero: Denotes whether the 90% Confidence Interval overlaps zero or not
param: Code syntax name of parameter
species: Animal species or species group for the parameter estimate
layer: Denotes whether the parameter is a covariate or other model parameter

1. "diet_multispecies_conflict_6monthlag_summary.csv" - Summarized model coefficients of the Diet Multispecies Model using the summed 6 months prior to precipitation

mean: Average value of parameter
sd: Standard deviation of parameter
90%_HPDL: Lower bound value of 90% Confidence Interval of parameter
90%_HPDU: Upper bound value of 90% Confidence Interval of parameter
Rhat: Rhat values for parameter calculated across 3 chains
n.eff: Number of effective samplers for each parameter
nonzero: Denotes whether the 90% Confidence Interval overlaps zero or not
param: Code syntax name of parameter
species: Animal species or species group for the parameter estimate
layer: Denotes whether the parameter is a covariate or other model parameter
parname: Name of each parameter

1. "diet_multispecies_conflict_samples_coda.Rda" - Posterior samples of the multi-species diet conflict model
2. "diet_multispecies_conflict_summary.csv" - Summarized model coefficients of the Diet Multispecies Conflict Model

mean: Average value of parameter
sd: Standard deviation of parameter
90%_HPDL: Lower bound value of 90% Confidence Interval of parameter
90%_HPDU: Upper bound value of 90% Confidence Interval of parameter
Rhat: Rhat values for parameter calculated across 3 chains
n.eff: Number of effective samplers for each parameter
nonzero: Denotes whether the 90% Confidence Interval overlaps zero or not
param: Code syntax name of parameter
species: Animal species or species group for the parameter estimate
layer: Denotes whether the parameter is a covariate or other model parameter
parname: Name of each parameter

Figures

R Scripts

1. "Visualizations.R" - Script to visualize results from analyses