Effects of cropland abandonment on biodiversity

Cropland abandonment benefits most birds and mammals but rarely compensates for habitat loss

Code archive: | Data archive:

This repository houses analysis scripts and documentation for:

Crawford CL*, Wiebe RA, Yin H, Radeloff VC, and Wilcove DS. 2024. Effects of cropland abandonment on biodiversity. Nature Sustainability.

*@chriscra, ccrawford@alumni.princeton.edu, School of Public & International Affairs, Robertson Hall, Princeton University, Princeton, NJ

This project builds on Crawford et al. 2022 (https://github.com/chriscra/abandonment_trajectories) and Yin et al. 2020 to assess the impact of cropland abandonment on the area of habitat for over 1,322 species of birds and mammals at 11 sites across 4 continents. We combined annual maps of cropland abandonment with IUCN habitat and elevation preferences for individual species of birds and mammals and a map of IUCN habitat types Jung et al. 2020 to estimate the area of habitat (AOH; Brooks et al. 2019) for each species in each year from 1987-2017.

We then use simple linear models to assess habitat trends for each species, and further investigate how these trends are associated with various species traits. Our primary analysis isolates the effect of abandonment on habitat by calculating AOH provided by abandoned croplands during cultivation and subsequent abandonment (and recultivation, where applicable) [Calculation 1]. In order to place the habitat effects of abandonment into the broader context of land cover changes at each site, we also calculate 2) the net habitat change in croplands that were abandoned at some point from 1987 to 2017 (therefore factoring in habitat that was cleared for croplands prior to abandonment), and 3) the net habitat change across the full extent of each site from 1987-2017 (factoring in habitat loss that occurred elsewhere at the site). Following Crawford et al. 2022, we also calculate the potential AOH under a scenario in which no recultivation took place, in order to quantify the benefit of reducing recultivation rates.

The annual land cover maps (1987-2017, 30-meter resolution) that underlie our analysis were developed on Google Earth Engine using publicly available Landsat satellite imagery (Yin et al. 2020, Remote Sensing of Environment). These data are archived and publicly available at Zenodo (DOI: 10.5281/zenodo.5348287), along with other data produced for Crawford et al. 2022.

The derived and supporting data for this project are archived on Zenodo at DOI: 10.5281/zenodo.13766321.

This is a living repository. See the public release and Zenodo archive for the code used at time of publication.

Components of this repository

This repository contains one primary directory housing the scripts used in our analysis. The "scripts" directory contains the primary working scripts for the project and two additional directories: "cluster" and "util."

The working R scripts and Rmarkdown files are listed below based on the general order in which parts of the analysis were conducted. Note that the .Rmd scripts are intended to be run interactively chunk by chunk (sometimes line by line), not knit together all at once. They also include code for data visualization and exploratory data analysis; not all code is strictly required. Some chunks produce and save data files, which other chunks often reload and use for further analysis or visualization (see "util/_util_files.R").

These files include:

• 0_start.R serves as the starting place for all scripts, setting the project directory and loading required packages and custom functions.
• habitats.Rmd contains code to prepare the underlying spatial habitat data required to calculate species' Area of Habitat [AOH] (interpolating the annual Yin et al. 2020 land cover types to match the IUCN Level 2 habitat types mapped by Jung et al. 2020) and analyze landscape configuration and fragmentation metrics via {landscapemetrics}.
• IUCN.Rmd handles spatial (i.e., range map) and tabular (i.e., Red List Assessment) data from IUCN and BirdLife International for study species, including: a) loading, prepping, and filtering range maps (after processing on the cluster, see below), b) loading Red List Assessment data, including habitat and elevation preferences used for the area of habitat (AOH) calculation, and c) filtering out species we consider "mature forest obligate species," which we determined to require habitats >30 years in age (i.e., longer than our study period).
• AOH.Rmd models Area of Habitat (AOH) as a function of time while accounting for temporal autocorrelation, and extracts AOH trends and effect sizes. This script relies on results produced on Princeton's High Performance Computing Clusters (see "cluster/aoh.R").
• traits.Rmd contains the primary R scripts for our traits analysis. The key code chunk is labeled "*many-models," which is where our three regression models are run. All data files required for these models are included in the accompanying Zenodo data repository.
• figures.Rmd contains scripts to produce figures for the manuscript and Supplemental Information, along with many figures exclusively produced for exploratory data analysis.
• biodiv_abn_MS.Rmd contains the text of the main text of the manuscript as well as the supporting information. Data is loaded and manipulated directly and programmatically in this document, and the document requires access to specific data files and figures in order to knit correctly.
• preamble.tex contains common LaTeX code for both the manuscript and the supporting information file.

cluster scripts

The "cluster" folder contains scripts performing much of the heavy-lifting of the analysis on Princeton's High Performance Computing (HPC) cluster system, primarily calculations of Area of Habitat for each species at each site over time. Most of these R scripts are accompanied by one or more "slurm" scripts (e.g., "calc_range_area.R" accompanied by "calc_range_area.slurm"), which set the parameters for processing each script on Princeton's HPC clusters through the Slurm scheduling system:

• Area of Habitat (AOH) calculations.

  • aoh.R calculates Area of Habitat for a set of species at a given site for a range of different calculations and scenarios ("aoh1.slurm" through "aoh11.slurm" correspond to 11 different AOH calculations, as described in the script; the analysis in Crawford et al. 2024 focuses on indices 9 [Calc. 1a], 10 [Calc. 1b], 6 [Calc. 2a], 8 [Calc. 2b], 4 [Calc. 3a], and 11 [Calc. 3b]). This relies on maps of land cover (Yin et al. 2020) interpolated to IUCN Level 2 Habitat Types based on the June et al. (2020) map for the year 2015. This script is designed to run in parallel on an array of nodes for high performance.
  • aoh_terra_jung.R calculates the Area of Habitat for each species at each site using the IUCN Level 2 habitat map for the year 2015 from (Jung et al. 2020).

• Input Data Preparation.

  • noncrop_precrop_mask.R produces the final input maps for Calculations 1a and 1b ("crop_abn_iucn" observed and potential), by masking out noncrop land covers / habitat types that occurred prior to cultivation in pixels that experienced abandonment at some point during the time series.
  • potential_full_iucn.R produces rasters with IUCN Habitat types at each site for a scenario in which no recultivation took place (i.e., all abandoned croplands remained abandoned through the end of the time series). These data feed into one of the calculations in aoh.R.
  • resample_iucn.R resamples the IUCN Habitat map (Jung et al. 2020) for each coarse land cover type (Yin et al. 2020), using a focal moving window filter to directly assign of IUCN habitat types to land cover classes based on the modal habitat type.

• Validate, crop, and save bird and mammal ranges from BirdLife International and IUCN respectively.

  • crop_sp_ranges_array.R loads, filters, validates, crops, and saves species range maps. Because some bird range maps have huge numbers of vertices (many shorelines), they required more specialized code to handle (below), but if you have the computing power and time, this script can be run in place of the follow scripts.
  • crop_mammal_ranges.R [optional, development] does the same as "crop_sp_ranges_array.R," but only for mammal species (if birds are done separately, so too can mammals).
  • save_bird_ranges_as_blocks.R calculates the number of vertices in each range, then saves small blocks of bird ranges based on the number of vertices in each range.
  • validate_bird_ranges_array.R sequentially loads and validates the bird range chunks produced in the previous script.
  • merge_crop_bird_ranges.R loads, merges, and crops all individual bird ranges to site locations.
  • calc_range_area.R with validated range maps for both brids and mammals, this script loads the validated ranges, calculates the area of each range polygon, and crops to sites again.

• Clean and save data.tables and rasters (SpatRasters).

  • remove_0s.R removes 0s from rasters {raster} and rewrite to file.
  • save_spatraster_as_dt.R saves {terra} SpatRasters as data.tables {data.table}.
  • save_parquet.R saves data.tables (read in from .csv files on disk using fread() {data.table}) as .parquet files (using {arrow}), which take up much less space on disk. save_spatraster_as_dt.R and save_parquet.R are used to produce the tabular parquet files files for {data.table} to use as inputs to the AOH calculations in aoh.R.
  • troubleshooting.R is exactly what it seems: a totally optional script for troubleshooting on the computing cluster.

• Calculate fragmentation statistics using the {landscapemetrics} package.

  • lsm_2024.R calculates a range of landscape metrics for the observed landscape throughout the entire time series. Metrics include: clumpiness index per class (an aggregation metric), mean patch area per class (an area and edge metric), and mean perimeter-area ratio (a shape metric).
  • lsm_two_2024.R extracts the signal of abandonment for the year 2017 by comparing the observed landscape in 2017 with a hypothetical landscape in which all abandoned croplands remained in cultivation through 2017.

util scripts

The "util" folder contains scripts that include custom analysis functions and pathnames for managing the project:

• _util_main.R includes user-specific switches (e.g., "run_label") and pathnames for managing the project, which are used throughout the rest of the analysis scripts. This script also loads "_util_functions.R" and "_util_misc.R." Note, however, that cluster scripts make use of alternative pathnames (specific to the Princeton high-performance computing clusters used for the analysis), as specified in each script.
• _util_files.R is a convenience script that loads many of the relevant input and derived data that are used throughout the project (some of which are required for the manuscript .Rmd file). The primary required data files for reproducing this work are archived in Zenodo (https://doi.org/10.5281/zenodo.13766321). This script also includes information about where certain files can be accessed (if external) or created throughout the various .R and .Rmd files in this repository.
• _util_functions.R contains custom functions written by Christopher L. Crawford (@chriscra), which underlie the bulk of these analyses and are used throughout the scripts above. These functions accomplish a variety of tasks, including: cleaning, filtering, and organizing data; calculating Area of Habitat (AOH), using both {terra} and {data.table}; saving plots and tables; and helping with miscellaneous tasks.
• _util_misc.R contains miscellaneous things like labels and color palettes for figures.