Wildlife Roadkill Mortality Estimates on a Highway in the Brazilian Semiarid Region: Effects of Pavement and Sampling Biases
Author
Marina de Souza
GenEst workflow for estimation of wildlife road mortality
This script provides the workflow and analytical steps used in “Wildlife road mortality on paved and unpaved roads in the Cerrado–Caatinga ecotone: accounting for biases in carcass persistence and searcher efficiency”. All analyses were conducted using the GenEst software. This study aimed to estimate wildlife road mortality from vehicle collisions in paved and unpaved roads in the Cerrado–Caatinga ecotone, accounting carcass persistence ans searcher efficiency biases using the GenEst framework, and to compare mortality patterns between road types (paved x unpaved). To account for potential differences in road characteristics, analyses were conducted separately for paved and unpaved roads.
GenEst
GenEst is an R-based software package designed to estimate wildlife mortality while accounting for imperfect carcass detection and carcass persistence. It integrates information from carcass persistence trials, searcher efficiency trials, search schedules, and carcass observations to correct for biases associated with carcass removal and missed detections. GenEst provides statistically robust estimates of mortality and associated uncertainty and has been widely used in studies assessing wildlife mortality related to anthropogenic structures and activities, including roads, wind energy facilities, and power lines.
Additional information on GenEst, including the user guide, software documentation, and example datasets, is available at the USGS website: GenEst – A Generalized Estimator of Mortality
Load required packages
library(GenEst)
Warning: pacote 'GenEst' foi compilado no R versão 4.4.3
library(dplyr)
Anexando pacote: 'dplyr'
O seguinte objeto é mascarado por 'package:GenEst':
desc
Os seguintes objetos são mascarados por 'package:stats':
filter, lag
Os seguintes objetos são mascarados por 'package:base':
intersect, setdiff, setequal, union
── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
✖ dplyr::desc() masks GenEst::desc()
✖ dplyr::filter() masks stats::filter()
✖ dplyr::lag() masks stats::lag()
ℹ Use the conflicted package (<http://conflicted.r-lib.org/>) to force all conflicts to become errors
library(MASS)
Anexando pacote: 'MASS'
O seguinte objeto é mascarado por 'package:dplyr':
select
library(patchwork)
Anexando pacote: 'patchwork'
O seguinte objeto é mascarado por 'package:MASS':
area
Loading input data - paved
1) Search Efficiency
SE_p <-read.csv("V1_01_SE_p.txt")
2) Carcass persistence
CP_p <-read.csv("V1_02_CP_p.txt")
3) Search Schedule
SS_p <-read.csv("V1_03_SS_p.txt")
4) Density-Weighted Proportion
DWP_p <-read.csv("V1_04_DWP_p.txt")
5) Carcass Observation
CO_p <-read.csv("V1_05_CO_p.txt")
SS, DWP and CO verification
Ensure that dates are correctly interpreted by R
all(CO_p$DateFound %in% SS_p$DateSearched)
[1] TRUE
Ensure that dates are in ascending order
SS_p <- SS_p[order(SS_p$DateSearched), ]
CO_p <- CO_p[order(CO_p$DateFound), ]
CO_p$DateFound <-as.Date(CO_p$DateFound, format ="%Y-%m-%d")
Alignment between the DWP categories and the CO dataset
Paved - Bias-corrected estimates accounting for observer detection efficiency and carcass persistence time
The total sampling effort on the paved road section was calculated based on two survey schemes. First, 335 km were surveyed during 43 field expeditions, with each expedition covering the entire transect twice, resulting in a total effort of 335 × 2 × 43 = 28,810 km. Second, an additional 17 field expeditions covered the same 335 km transect four times per expedition, resulting in an additional effort of 335 × 4 × 17 = 22,780 km. Combined, the total distance surveyed on paved roads was 51,590 km.
We will calculate the wildlife road mortality rate for each field expedition. After organizing the sampling months according to the survey scheme used, the number of carcasses recorded during each expedition will be divided by the corresponding distance surveyed. For the first survey scheme, the entire 335-km transect was surveyed twice per expedition, resulting in a sampling effort of 670 km per expedition (335 × 2). For the second survey scheme, the transect was surveyed four times per expedition, resulting in a sampling effort of 1,340 km per expedition (335 × 4).
Unpaved - Bias-corrected estimates accounting for observer detection efficiency and carcass persistence time
The total sampling effort on the unpaved road section was calculated based on two survey schemes. First, 45 km were surveyed during 43 field expeditions, with each expedition covering the entire transect twice, resulting in a total effort of 45 × 2 × 43 = 3,870 km. Second, an additional 17 field expeditions covered the same 45 km transect four times per expedition, resulting in an additional effort of 45 × 4 × 17 = 3,060 km. Combined, the total distance surveyed on unpaved roads was 6,930 km.
We will calculate the carcass rate for each field expedition. After organizing the sampling months according to the survey scheme used, the number of carcasses recorded during each expedition will be divided by the corresponding distance surveyed. For the first survey scheme, the entire 45-km unpaved transect was surveyed twice per expedition, resulting in a sampling effort of 90 km per expedition (45 × 2). For the second survey scheme, the transect was surveyed four times per expedition, resulting in a sampling effort of 180 km per expedition (45 × 4).
