Data & Resources: Evaluating the Ability of Spatial Indicators to Classify Fish Stock Status
Description
This repository contains all data and code to reproduce results, tables, and figures. The repository can also be cloned from https://github.com/peterjosephkidd/spatind_paper1. Abstract A positive relationship between fish abundance and spatial distribution suggests that indicators of spatial distribution could be used as proxies of abundance and stock status. This study evaluated the ability of 10 spatial indicators to classify the health status of 25 data-rich stocks, identifying the best performing indicators and the conditions that influence their performance using a combination of machine learning techniques. Spatial indicators were able to detect health status better than a random classifier for the majority of stocks studied. The occupancy and aggregation indicators showed the best overall performance, suggesting that years of healthy stock status were often associated with presence in more survey hauls or more evenly distributed fish density. However, indicator performance depended on the source of survey data used and varied stock-by-stock. Explanations of these findings are provided and future research is suggested to further explore the impact of survey design and environmental variables. Our findings highlight the potential for spatial indicators to improve the assessment of category 5 and 6 data-limited stocks managed by the International Council for the Exploration of the Seas (ICES). Data The data used to calculate spatial indicators is downloaded from ICES Database of Trawl Surveys (DATRAS). This repository contains the DATRAS data downloaded and used within this study. Redownloading data from DATRAS may result in changes in outputs due to retrospective corrections/additions to the data. It is therefore advised that users intending to replicate this papers findings use the DATRAS data provide in this repository. Software Reference to software used to complete this work is recorded in the PckgBibReport.html file
Files
Steps to reproduce
Data/: Contains initial data needed to replicate analysis in cluding stock information, survey data, ICES Divisions Functions/: Contains user defined functions for calculating spatial indicators, ROC curves, and communciating results Scripts/: R scripts to run analysis data_1_SelectStocks.R: Retrieve stock info and assessment outputs for data-rich stocks. (do not run) data_2_DownloadDATRAS.R: Download DATRAS survey data. (do not run) data_3_CleanExchange.R: Clean DATRAS survey data. data_3a_Mature.R: Subset survey data to matures based on L50. model_4_SpatIndCalc.R: Calculate spatial indicators using survey data. model_5_ROC.R: Assess classifcation skill of spatial indicators using ROC curves. model_6_regression_tree.R: Identify predictors of classifcation skill using RFE and regression trees. output_7_TabsAndPlots.R: Plot primary and supplementary tables and plots. Scripts 1 and 2 do not need to be run. They retrieve data rich stock information (e.g. reference points and SSB) and download DATRAS survey data which could change if ICES make retrospective corrections to data. Outputs of these scripts have been provided in Data/Initial/ so that results can be accurately replicated.