Data and code from: The continuity-contiguity problem in fragmentation-biodiversity research
Data files
Dec 31, 2024 version files 22.63 GB
-
README.md
2.41 KB
-
rf_means.csv
15.61 MB
-
rfClasses.tif
102.85 KB
-
woodBetaCCR1.rds
22.60 GB
-
woodland_var.csv
5.47 MB
Abstract
A primary question for researchers in the field of conservation science concerns the fragmentation of biodiversity-supporting habitat. Key debates revolve around the relevance of habitat composition and configuration for conservation outcomes. Central to this debate is the measurement and delineation of the habitat patch, as the analytical unit from which fragmentation-related metrics are assessed and evaluated. Despite the need to quantify and model habitat fragmentation, the habitat patch concept itself has been subject to criticism concerning its ability to adequately reflect the heterogeneity of resource distributions in complex landscapes. In this paper, we present the continuity-contiguity problem, which is one of the fundamental challenges related to the delineation of habitat in space, and discuss its implications for fragmentation-biodiversity research. We outline the potential contribution of recent developments in spatial-ecological methods and the leveraging of uncertainty in the modelling process to address four common issues related to the habitat patch concept. We conclude with several recommendations for studies on fragmentation-biodiversity outcomes where the continuity-contiguity problem may influence the research process.
README: Data and code from: The continuity-contiguity problem in fragmentation-biodiversity research
Data DOI: https://doi.org/10.5061/dryad.6hdr7sr9m
Description of the data and file structure
This repository contains the code and necessary data to reproduce the analysis in the Dennis and Huck (2025): The continuity-contiguity problem in fragmentation-biodiversity research
File list:
"cc_For_review_Rev2.R"
"rfClasses.tif"
"rf_means.csv"
"woodland_var.csv"
"woodBetaCCR1.rds"
Data description:
"rfClasses.tif" is a GeoTiff format raster of a classified colour infrared image (Planet Labs PBC, 2024), see below. Class 1 = Woodland, Class 2 = Grassland, Class 3 = Shrub, Class 4 =Water, Class 5 = Urban
"rf_means.csv" contains the membership values for each cell in the study landscape to each class resulting from the random forest classifier.
The file has 6 headings: "Cell" denoting the cell number starting from top left in the raster layer and a column for each land-cover class given in "rfClasses.tif"
"woodland_var.csv" contains the variance around the membership values of each cell to the "Woodland" class for the study landscape. The file has two headings "Cell" denoting the cell number starting from top left in the raster layer and "variance" denoting the variance for each cell computed as the variance of predictions of all decision trees for the "Woodland" class for each cell.
"woodBetaCCR1.rds" consists of a matrix wherein rows represent locations (cells) in the landscape and each column gives 10,000 random variables drawn from a beta distribution based on the mean and variance of cell membership to the "Woodland" class taken from "rf_means.csv" and "woodland_var.csv", respectively.
Code/software
R code for use with this dataset is in the file cc_For_review_Rev2.R included in this upload. This code can be used to reproduce the figures in the article and a distribution of habitat values for the study area landscape.
Access information
The data can also be accessed from the corresponding author upon request made to matthew.dennis@manchester.ac.uk
The data were derived from the following sources:
- Planet Labs PBC. 2024. Planet Application Program Interface: In Space for Life on Earth. URL: https://api.planet.com
Methods
These data were computed from a Random Forest classification of the study area landscape represented in Dennis and Huck (2025): DOI: 10.1111/jbi.15077.
They can be used to generate a Boolean classification of the landscape into 5 classes (Woodland, Grassland, Shrubs, Wetland and Urban) and a distribution of possible habitat values for the "Woodland" class derived from a Beta distribution for each location (cell) in the landscape. The classification is based on the use of Planet Scope Colour Infra-red imagery (Planet Labs PBC. 2024. URL: https://api.planet.com) and 4,500 training points.
Methodological steps can be found in Dennis and Huck (2025): DOI: 10.1111/jbi.15077