Optimising fire and predator management for conservation
Data files
Dec 04, 2025 version files 9.08 GB
-
data_JAPPL-2024-00590.zip
9.08 GB
-
README.md
13.53 KB
Abstract
Interactions between threatening processes compound and accelerate biodiversity decline. Conservation managers need to understand how co-occurring threats interact and account for such interactions when prioritising when, where, and how to manage landscapes to recover declining species. Using the Upper Warren region in south-western Australia as a case study, we develop a framework for identifying optimal fire age classes while also considering predation by introduced red foxes (Vulpes vulpes) – two co-occurring processes affecting the recovery of a threatened faunal community (woylie Bettongia penicillata, chuditch Dasyurus geoffroii, quenda Isoodon fusciventor, and numbat Myrmecobius fasciatus). We fitted a multi-species relative abundance model to a dataset from 548 camera trap sites and tested for associations between each species’ relative abundance and an interaction between fox baiting intensity and time since fire. We then used linear programming optimization to identify the optimal distribution of time since fire values across the study region that maximizes the abundance of four focal species under alternative fox baiting intensity and fire severity scenarios. Fire and baiting both influenced the relative abundance of the four species in our study, with baiting intensity having a positive association with woylie relative abundance. The optimal distribution of time since fire values across the study region varied with the intensity of fox baiting. The importance of older fire ages increased in some locations when fox baiting intensity was high, but these results were highly uncertain and varied spatially. High fox baiting intensities combined with optimal fire age distributions also led to a higher relative abundance of three focal species overall, namely woylie, quenda, and numbat. Our study demonstrates an end-to-end framework for using field data to derive optimal fire regimes for biodiversity in a way that explicitly acknowledges uncertainty and remains useful for conservation decisions. Approaches such as these are essential for managing ecosystems with compounding threats to biodiversity.
https://doi.org/10.5061/dryad.b8gtht7s3
Scripts and data for analysis of the manuscript:
Optimising fire and predator management for conservation
William L. Geary, Ayesha I.T. Tulloch, Tim S. Doherty, Dale G. Nimmo, Euan G. Ritchie, Jeffrey O. Hanson, Marika A. Maxwell & Adrian F. Wayne
Corresponding author: William L. Geary (billy.geary@unimelb.edu.au)
Description of the data and file structure
Files and variables
Folder: data_JAPPL-2024-00590.zip
All files necessary to replicate this analysis are contained within this folder. The ReadMe below outlines each file in detail.
R Project: UW_MultiSppFire.Rproj
This analysis can be replicated by opening the R project and running each script sequentially, except for:
- Script 1 which creates the species detection histories used for modelling. The output detection histories of Script 1 can be found in the Data Processing folder.
- Script 2 which does background calculations to generate the environmental covariates used in the modelling. The outputs of this script can be found in the Data Processing folder (Rasters; Detection Histories with site covariates added).
To replicate the remaining analysis, start from Script 3. Scripts 6 and 7 generate the figures included in the publication.
Folder: Scripts
1. 1_Create detection histories_Final.R - Creates detection histories from the camera data. One file created is presence-absence and the other is counts.
2. 2_Create covariate dataframe and rasters_Final.R - Creates spatial covariate layers and extracts them for the camera sites
3. 3_Run the MultiSpp N-mixture model_Final.R - Fits a Multi-species N-mixture model in Nimble
4. 4_Abundance predictions for alternative management scenarios_Final.R - Uses the model to make predictions of abundance under a range of scenarios
5. 5_Find optimal management actions with abundance samples_Final.R - Uses prioritizr to run optimisations that find the fire age class distribution under alternate scenarios
6. 6_Plot optimisation results for publication_Final.R - Plots the optimisation results for publication
7. 7_Additional plots for publication.R - Makes some additional plots for inclusion in the publication
Helper functions:
* covariate_helper_functions.R - Functions for processing and managing covariate data
* model_functions.R - Core functions for model fitting and analysis
* model_helper_functions.R - Auxiliary functions supporting model operations
Folder: data_clean
This folder contains final datasets created at key stages in the analysis process.
Model outputs:
These files are Nimble output objects for each of the five abundance models fit with the data.
* nmix_nimblemodel_final_linear.RDS - Final N-mixture model with linear relationships - This model was used in subsequent predictions.
* nmix_nimblemodel_final_linear_noint.RDS - Final N-mixture model with linear relationships (no interaction terms)
* nmix_nimblemodel_final_spline.RDS - Final N-mixture model with spline relationships
* nmix_nimblemodel_final_poly.RDS - Final N-mixture model with polynomial relationships
* nmix_nimblemodel_final_sqrt.RDS - Final N-mixture model with square root transformations
Spatial covariates:
* covariates_stack.RData - Stack of environmental covariates used in n-mixture modeling
Optimization results - Abundances:
Files containing predicted species abundances under different baiting and fire scenarios at the transect scale. File naming convention: baitfireoptimisationresults_transectscale_abundances_[baiting]_[fire].csv
Baiting scenarios: s_b0 (baseline level), s_bA (aerial-only), s_bAG (aerial and ground baiting), s_bL (low intensity of baiting)
Fire severity scenarios: sH (high), sM (medium), sL (low)
* baitfireoptimisationresults_transectscale_abundances_s_b0_sH.csv
* baitfireoptimisationresults_transectscale_abundances_s_b0_sL.csv
* baitfireoptimisationresults_transectscale_abundances_s_b0_sM.csv
* baitfireoptimisationresults_transectscale_abundances_s_bA_sH.csv
* baitfireoptimisationresults_transectscale_abundances_s_bA_sL.csv
* baitfireoptimisationresults_transectscale_abundances_s_bA_sM.csv
* baitfireoptimisationresults_transectscale_abundances_s_bAG_sH.csv
* baitfireoptimisationresults_transectscale_abundances_s_bAG_sL.csv
* baitfireoptimisationresults_transectscale_abundances_s_bAG_sM.csv
* baitfireoptimisationresults_transectscale_abundances_s_bL_sH.csv
* baitfireoptimisationresults_transectscale_abundances_s_bL_sL.csv
* baitfireoptimisationresults_transectscale_abundances_s_bL_sM.csv
Column descriptions in abundance files
* .id: Iteration number (1000 iterations based on draws from the N-mixture model)
* summary: Time since fire character value
* feature: Feature ID for the species
* total_amount: total abundance of the species
* absolute held: absolute abundance of the specific species in parts of the landscape with this year since last fire
* relative held: relative abundance of the specific species in parts of the landscape with this year since last fire
* tsf: years since last fire related to abundance information in each row
* Species: Species related to abundance information in each row
* Scenario: Optimisation scenario name (uses same notation as filename)
Optimization results - Outcomes:
Files containing optimization outcomes under different baiting and fire scenarios at the transect scale.
File naming convention: baitfireoptimisationresults_transectscale_outcomes_s_[baiting]_[fireseverity].csv
* baitfireoptimisationresults_transectscale_outcomes_s_b0_sH.csv
* baitfireoptimisationresults_transectscale_outcomes_s_b0_sL.csv
* baitfireoptimisationresults_transectscale_outcomes_s_b0_sM.csv
* baitfireoptimisationresults_transectscale_outcomes_s_bA_sH.csv
* baitfireoptimisationresults_transectscale_outcomes_s_bA_sL.csv
* baitfireoptimisationresults_transectscale_outcomes_s_bA_sM.csv
* baitfireoptimisationresults_transectscale_outcomes_s_bAG_sH.csv
* baitfireoptimisationresults_transectscale_outcomes_s_bAG_sL.csv
* baitfireoptimisationresults_transectscale_outcomes_s_bAG_sM.csv
* baitfireoptimisationresults_transectscale_outcomes_s_bL_sH.csv
* baitfireoptimisationresults_transectscale_outcomes_s_bL_sL.csv
* baitfireoptimisationresults_transectscale_outcomes_s_bL_sM.csv
Column descriptions in outcomes files
* .id: Iteration number (1000 iterations based on draws from the N-mixture model)
* Site: The transect that this row of data is related to
* tsf: years since last fire related to prop_landscape information in this row
* prop_landscape: The optimal proportion of the site that should be this time since fire age for this iteration
* TSF_CAT: Factor describing the broad time since fire category this row of data falls into
* Scenario: Optimisation scenario name (uses same notation as filename)
Optimization results - Summaries:
Files containing summary statistics of optimization results under different baiting and fire scenarios at the transect scale.
File naming convention: baitfireoptimisationresults_transectscale_summaries_s_[baiting]_[fireseverity].csv
* baitfireoptimisationresults_transectscale_summaries_s_b0_sH.csv
* baitfireoptimisationresults_transectscale_summaries_s_b0_sL.csv
* baitfireoptimisationresults_transectscale_summaries_s_b0_sM.csv
* baitfireoptimisationresults_transectscale_summaries_s_bA_sH.csv
* baitfireoptimisationresults_transectscale_summaries_s_bA_sL.csv
* baitfireoptimisationresults_transectscale_summaries_s_bA_sM.csv
* baitfireoptimisationresults_transectscale_summaries_s_bAG_sH.csv
* baitfireoptimisationresults_transectscale_summaries_s_bAG_sL.csv
* baitfireoptimisationresults_transectscale_summaries_s_bAG_sM.csv
* baitfireoptimisationresults_transectscale_summaries_s_bL_sH.csv
* baitfireoptimisationresults_transectscale_summaries_s_bL_sL.csv
* baitfireoptimisationresults_transectscale_summaries_s_bL_sM.csv
Column descriptions in outcomes files
* .id: Iteration number (1000 iterations based on draws from the N-mixture model)
* summary: Time since fire character value
* n: Proportion of sites at this transect selected for this time since fire
* tsf: years since last fire related to n information in this row
* TSFCAT: Factor describing the broad time since fire category this row of data falls into
* Scenario: Optimisation scenario name (uses same notation as filename)
Folder: data_processing
This folder contains either key input data or intermediary data products created in the analysis process. RData files can be opened with function readRDS().
Detection histories:
* camtrapR.counthist.UpperWarren.multispp09122023.RData - Count-based detection histories from camera trap data
* camtrapR.dethist.UpperWarren.multispp09122023.RData - Presence-absence detection histories from camera trap data
These files are lists containing detection information for each species at each site.
* Slot 1 is a list of the species in the dataset and their daily detections at each site in the dataset
* Slot 2 is detailed covariate information about the sites in the dataset. Column headings relate to the covariates outlined below.
* Slot 3 is the number of nights surveyed in the dataset at each site (integer)
Abundance scenario predictions:
Files containing predicted abundances under different management scenarios. These abundances feed directly into the PrioritizR process.
File naming convention: UpperWarrenAbundance_[baiting]_[fireseverity]_abundance_scenarios.RData
Each file is a list with 1000 slots. Each slot is a draw (1000 total) from the abundance model containing abundance predictions for each species, at each site (within a transect), for each time since fire value.
* UpperWarrenAbundance_s_b0_sH_abundance_scenarios.RData
* UpperWarrenAbundance_s_b0_sL_abundance_scenarios.RData
* UpperWarrenAbundance_s_b0_sM_abundance_scenarios.RData
* UpperWarrenAbundance_s_bA_sH_abundance_scenarios.RData
* UpperWarrenAbundance_s_bA_sL_abundance_scenarios.RData
* UpperWarrenAbundance_s_bA_sM_abundance_scenarios.RData
* UpperWarrenAbundance_s_bAG_sH_abundance_scenarios.RData
* UpperWarrenAbundance_s_bAG_sL_abundance_scenarios.RData
* UpperWarrenAbundance_s_bAG_sM_abundance_scenarios.RData
* UpperWarrenAbundance_s_bL_sH_abundance_scenarios.RData
* UpperWarrenAbundance_s_bL_sL_abundance_scenarios.RData
* UpperWarrenAbundance_s_bL_sM_abundance_scenarios.RData
Column descriptions in each slot in each list
* tsf: the time since fire the row of data relates to (scaled)
* Sample: Iteration number (1000 iterations based on draws from the N-mixture model)
* Scenario: Optimisation scenario name (uses same notation as filename), with tsf value appended
* Site: The transect that this row of data is related to
* LocationName: The site name within each transect the row of data is related to
* tsf_actual: the time since fire the row of data relates to (unscaled)
* n: Proportion of sites at this transect selected for this time since fire
* Chuditch: The predicted abundance of Chuditch at this Location/Transect at this tsf for this sample
* Quenda: The predicted abundance of Quenda at this Location/Transect at this tsf for this sample
* Woylie: The predicted abundance of Woylie at this Location/Transect at this tsf for this sample
* Numbat: The predicted abundance of Chuditch at this Location/Transect at this tsf for this sample
* Vulpes: The predicted abundance of Red Fox at this Location/Transect at this tsf for this sample
Folder: data_processing/rasters
Spatial environmental covariate layers in GeoTIFF format.
* bait_intensity_2019.04.01_38month_lag_13052022.tif - Baiting intensity with 38-month lag
* dist_to_ag_13052022.tif - Distance to agricultural land
* dist_to_allhydro_13052022.tif - Distance to all hydrological features
* dist_to_majorhydro_13052022.tif - Distance to major hydrological features
* dist_to_minorhydro_13052022.tif - Distance to minor hydrological features
* prop_ag_3km_13052022.tif - Proportion of agricultural land within 3km
* prop_filtered_roads_3km_13052022.tif - Proportion of roads within 3km
* prop_roads_3km_13052022.tif - Proportion of roads within 3km
* prop_nv_1km_13052022.tif - Proportion of native vegetation within 1km
* prop_nv_3km_13052022.tif - Proportion of native vegetation within 3km
* topographic_wetness_index_13052022.tif - Topographic wetness index
All data products key for undertaking the analysis for this study are referenced to in Scripts 3-7 and are included in this repository.
Code and software
All analyses were done in R v4.4.0 or later.
Access information
The code for this project is also stored at: https://github.com/billygeary/upperwarren_fire_predator_optimisation
