---
title: "JARA: Just Another Redlist Assessment"
author: "Henning Winker, Nathan Pacoureau & Richard Sherley
date: "Cape Town, 2020"
output: html_document
---
JARA (Just Another Redlist Assessment) is a Bayesian state-space trend analysis tool that is designed to objectively incorporate uncertainty into the IUCN Red Listing evaluation process. To ensure to a high degree of transparency and reproducibility, JARA has now implemented as an R package that hosted online on the global open-source platform GitHub (https://github.com/henning-winker/JARA).
Installing JARA requires the librabry(devtools), which can be install by 'install.packages('devtools')' and a R version >= 3.5. Then simply install JARA from github with the command:
install_github("henning-winker/JARA")
library(JARA)
The name ‘Just Another Red List Assessment’ is a reference to JAGS (Just Another Gibbs Sampler, Plummer, 2003), which is the software called from R to run the Bayesian state-space model application. The name reference, together with user-friendly R interface and modulated coding structure of JARA follows the example of the new open source fisheries stock assessment R Package ‘Just Another Bayesian Biomass Assessment‘ (JABBA; Winker et al. 2018)
The fully commented R JARA functions enable to easy adjustments of the input data and parameters by conservation practitioners to apply JARA to their count or relative abundance data. We provide a detailed Technical Documentation of JARA in the form of a Preprint Draft (Winker and Sherley, 2019) available on bioRxiv. We provide an easy "test-drive" of the JARA options JARApkg_TestDrive.R. Worked examples of the following seven datasets compiled in jaradata can be reproduced by running JARA_7examples.R
- Cape gannet (Breeding pair counts from multiple colonies)
- African Penguin (Breeding pair counts from multiple colonies)
- Mountain Zebra (Census data with Carrying Capacity option)
- Yellowspotted skate Leucoraja wallacei (South African demersal survey abudance indices)
- Smoothhound shark Mustelus mustelus (Combining South African demersal survey indices and recreational angling survey indices from the De Hoop MPA)
- Indian Ocean Striped Marlin (Fitting multiple fisheries Catch-Per-Unit Effort (CPUE) from the 2018 IOTC Striped Marlin Assessment)
- 2018 ICCAT Atlantic Blue Marlin stock assessment (Comparing estimated Biomass trends and fitted CPUE indices)
The JARA framework provides the option to simultaniously fit multiple relative abundance indices and estimate the underlying mean trend (Figure 1).
Figure 1. (a) Relative abudance abundance indices with 95% CIs for smoothhound shark Mustelus mustelus, depecting for abudance indices from demersal trawl surveys along the South African South Coast and one from research angling surveys conducted in the De Hoop MPA and (b) Standardized Catch-Per-Unit-Effort indices for Indian Striped marlin.
.
To evaluate model fit, JARA provides the user with three plots. The first shows the observed and predicted abundance values for each time series together with the 95% posterior predictive credibility intervals (Figure 2).
Figure 2. Time-series of observed (color coded dots) relative indices and a joint predicted (balck solid line) trend for (a) smoothhound shark and (b) Indian Ocean striped marlin. Shaded grey area indicates 95% credibility intervals.
.
The second shows individual fits, as well as the 95% credible intervals (CI) derived from the observation variance (Figure 3).
Figure 3. JARA fits to six standardized Catch-Per-Unit-Effort indices for Indian Ocean striped marlin. The solid black line is the model predicted CPUE and the circles are observed CPUE values. Grey-shaded areas denote the 95% Credibility Intervals.
.
The third plot illustrates the fits on log scale.
Figure 4. JARA fits (on log-scale) to six standardized Catch-Per-Unit-Effort indices for Indian Ocean striped marlin. The solid blue line is the model predicted CPUEand the circles are observed CPUE values. Error bars denote the assumed observation variance (in 95% CIs) for the observed CPUE values.
If absolute abundance estimates are available (e.g. census data from different breeding colonies) JARA can also produce a summed population trend for the total population, where each subpopulations' trend is modelled seperately within JARA given a common process variance (Figure 5).
Figure 5. (a) Trends in estimated numbers (coloured points) and JARA fits
(coloured lines) with 95% credible intervals (grey polygons) of African penguin breeding pairs at its eleven colonies and (b) total estimated numbers for the 'global' population obtained by summing the 11 posteriors from each colony. The colored dashed lines denote 1 x generation length (GL), 2 x GL and 3 X GL, going backwards from the last year.
If the time series is longer than than three generation lengths (GL) the %decline is automatically estimated from the last assessment year minus 3 x GL. If the time series is shorter than 3 x GT, JARA projects forward to by passing the number of desired future years without observations to the state-space model to achive a time horizon of 3 x GL + 2 (Figure 6).
To facilitate Red List assessment decision-making, JARA automatically produces two key plots for each input dataset showing: (1) the median and posterior probabilities for the percentage annual population change calculated from all the observed data, and from each of the most recent 1 GL, 2 GL, and 3 GL (depending on the length of the observed time-series), shown relative to a stable population (%C = 0) (Figure 7a); and (2) how the posterior distribution for the percentage change in abundance (%C) over 3 GL aligns against the thresholds for the Red List categories (Least Concern LC, Near Threatened NT, Vulnerable VU, Endangered EN or Critically Endangered CR) under criteria A2–A4 (Figure 7b) or under A1.
Figure 7. (a) Posterior probability densities for the annual rate of change of the combined population African Penguin over three generation lengths (3 x GL), 2 x GL, the most recent 1 x GL and for all available years (All.yrs) and (b) posterior probability density of the percentage change (C%) over 3 x GL. Note that the probabilities have become more negative in more recent 1 x GL and 2 x GL and the median rate of change over 3 x GL is estiamted −54.4%.
To assist with presenting the JARA outputs in scientific reports, we also provide a plotting R script JARAplotting.R that entails examples of the here presented plots . We also include R code for reproducing the JARA output format used for production of the IUCN Redlist Supplemetary Information for recent and on-going global shark assessments by the Shark Specialist Group of the IUCN (see e.g. Figure 8).
