distribution_models

title	author	date	output
README	Renata L. Muylaert	3/28/2022	html_document df_print paged

knitr::opts_chunk$set(echo = TRUE)

Guidelines

This repository contains the code from Muylaert et al. (2022) and guidelines for replicating the workflow.

Please read the following guidelines:

• Read the paper.
• Make sure you download all data from Dryad.
• Unzip the data inmuylaert_et_al_data.zip.
• Clone the dynamic repository GitHub.
• Unzip the code repository and find the distribution_models folder.
• Move the extracted data files from muylaert_et_al_data to the distribution_models folder.
• Make sure you are working in the distribution_models R project.
• Run the 00_packages and 01_settings scripts first.
• Run the other scripts following the numerical 3-digit ascending order or as needed.

Folder structure

Please find the content description for each folder below:

• _env_27km_ss6 contains all environmental layers (*.tif) for the present, already scaled.
• _env_fut_27km_ss6 contains all environmental layers (*.tif) for the future, already scaled, for all global circulation models, periods and scenarios.
• dynamic_copies is an auxiliary data folder for when the user decides to update the species list used (*.xlsx and *.csv files).
• dynamic_master contains the text files for the master dataset with IUCN-intersected and non-intersected occurrences iwthin our working extent.
• hotspots contains data for species presence and future projections.
• IUCN_assessment_list_chiropteraIUCN assessments for the Order Chiroptera (*.xlsx).
• iucn_shapefile contains the IUCN ranges dataset for the Class Mammalia (terrestrial only).
• olson contains the shapefile files for all terrestrial ecosystems of the world (Olson et al. 2001).
• rasters_temp_forest contains the rasters used for hotspot calculation (*.tif).
• results_40o_ss6_maxent_15rep ENMTML data sctructure for non-intersected occurrences (text files and *.tif).
• results_iucni_40o_ss6_maxent_15repENMTML data sctructure for IUCN-intersected occurrences (text files and *.tif).

Workflow table

The reference for workflow table used is Zurell et al. 2020. A standard protocol for reporting species distribution models. Ecography 43, 1261–1277.

Section	Subsection	Element	Value
Overview	Authorship	Study title	Present and future distribution of bat hosts of sarbecoviruses: implications for conservation and public health.
Overview	Authorship	Author names	Renata L. Muylaert; Tigga Kingston; Jinhong Luo; Maurício Humberto Vancine; Nikolas Galli; Colin J. Carlson; Reju Sam John; Maria Cristina Rulli; David T. S. Hayman.
Overview	Authorship	Contact	R.deLaraMuylaert@massey.ac.nz
Overview	Authorship	Study link	https://doi.org/10.1098/rspb.rspb.2022.0397
Overview	Model objective	Model objective	Forecasting and transfer.
Overview	Model objective	Target output	Continuous occurrence probabilities and binary maps of potential presence.
Overview	Focal Taxon	Focal Taxon	Bats hosts of sarbecoviruses.
Overview	Location	Location	World.
Overview	Scale of Analysis	Spatial extent	-30, 160, -30, 70 (xmin, xmax, ymin, ymax)
Overview	Scale of Analysis	Spatial resolution	0.25 dd
Overview	Scale of Analysis	Temporal extent	Near-current and Future (2021-2100).
Overview	Scale of Analysis	Temporal resolution	Near-current, 2021-2040, 2041-2060, 2061-2080, 2081-2100.
Overview	Scale of Analysis	Boundary	Terrestrial areas of the world.
Overview	Biodiversity data	Observation type	Human observation of occurrences.
Overview	Biodiversity data	Response data type	Presence.
Overview	Predictors	Predictor types	Bioclimatic; karst; forest cover.
Overview	Hypotheses	Hypotheses	Implications for the conservation and public health through evaluation of species distribution change in response to climatic, karst, and forest cover.
Overview	Assumptions	Model assumptions	Bats occur within their bioregions where they were detected, and around their highest density of occurrence points (MSDMs). Bat distribution is driven bioclimatic covariates, karst and native forest cover. Accessibility bias partially drives observed occurrences. Sampling bias is minimized by filtering, spatial thinning and minimal occurrences for inclusion criteria (N=40).
Overview	Algorithms	Modelling techniques	Maxent through the ENMTML R package.
Overview	Algorithms	Model complexity	Six follwoing covariates were used bio 1, bio 4, bio 12, bio 15, karstm, primf tif files.
Overview	Algorithms	Model averaging	True skill statistics-weighted (TSS-weighted) averaging.
Overview	Workflow	Model workflow	ENMTML workflow.
Overview	Software	Software	R 4.
Overview	Software	Code availability	https://github.com/renatamuy/dynamic
Overview	Software	Data availability	Dryad.
Data	Biodiversity data	Taxon names	Aselliscus stoliczkanus Hipposideros armiger Hipposideros galeritus Hipposideros larvatus Hipposideros pomona (gentilis) Hipposideros pratti Hipposideros ruber Miniopterus schreibersii Chaerephon plicatus Tadarida teniotis Rhinolophus acuminatus Rhinolophus affinis Rhinolophus blasii Rhinolophus blythi Rhinolophus cornutus Rhinolophus creaghi Rhinolophus euryale Rhinolophus ferrumequinum Rhinolophus hipposideros Rhinolophus luctus Rhinolophus macrotis Rhinolophus malayanus Rhinolophus marshalli Rhinolophus mehelyi Rhinolophus monoceros Rhinolophus pearsonii Rhinolophus rex Rhinolophus shameli Rhinolophus siamensis Rhinolophus sinicus Rhinolophus stheno Rhinolophus thomasi Nyctalus leisleri Plecotus auritus
Data	Biodiversity data	Taxonomic reference system	Wilson D, Mittermeier R, editors. Handbook of the Mammals of the World. Barcelona: Springer; 2019.
Data	Biodiversity data	Ecological level	assemblage-level, species-level.
Data	Biodiversity data	Data sources	Darkcides v1, Global Biodiversity Information Facility (GBIF), Berkeley Ecoinformatics Engine (Ecoengine), Vertnet, Integrated Digitized Biocollections (IDigBio), iNaturalist, Obis, Vertnet, and data compiled for previous publications Darkcides v01, Rulli et al. (2020), Luo et al. (2013)
Data	Biodiversity data	Sampling design	ENMTML workflow.
Data	Biodiversity data	Clipping	Terrestrial areas of the world.
Data	Biodiversity data	Scaling	None.
Data	Biodiversity data	Cleaning	Data cleaning: Temporal range from 1970-2020. Cleaning process through CooordinateCleaner package including species with at least 40 occurrence points.
Data	Biodiversity data	Absence data	None.
Data	Biodiversity data	Background data	pres_abs_ratio = 1
Data	Biodiversity data	Errors and biases	Errors and biases: Sampling rates estimates through sampbias R package.
Data	Data partitioning	Training data	75:25 training:test.
Data	Data partitioning	Validation data	75:25 training:test.
Data	Data partitioning	Test data	Ratio of 75:25 training:test cross-validation splits with 10 repeats.
Data	Predictor variables	Predictor variables	Bioclimatic variables, Karst composite layer, Primary forest cover.
Data	Predictor variables	Data sources	Table S3.
Data	Predictor variables	Spatial extent	-30, 160, -30, 70 (xmin, xmax, ymin, ymax)'
Data	Predictor variables	Spatial resolution	0.25 dd.
Data	Predictor variables	Coordinate reference system	WGS84.
Data	Predictor variables	Temporal extent	Bioclimatic variables cover 1970-2000 for near-current conditions. Future projection periods: 2020-2040, 2040-2060, 2060-2080, 2080-2100.
Data	Predictor variables	Temporal resolution	Future projection periods: 2020-2040, 2040-2060, 2060-2080, 2080-2100.
Data	Predictor variables	Data processing	Covariates resampled to 0.25 dd.
Data	Predictor variables	Errors and biases	Assessed via sampbias R package.
Data	Predictor variables	Dimension reduction	None.
Data	Transfer data	Data sources
Data	Transfer data	Spatial extent	World.
Data	Transfer data	Spatial resolution	0.25 dd
Data	Transfer data	Temporal extent	1970-present
Data	Transfer data	Temporal resolution	Yearly
Data	Transfer data	Models and scenarios	Future bioclimatic data downloaded from Worldclim (CMIP6).
Data	Transfer data	Data processing	Future occurrence projections were made for each species and then ensembled per period per GCM and SSP.
Data	Transfer data	Quantification of Novelty	NA
Model	Variable pre-selection	Variable pre-selection	Relevance for our conceptual model of important native habitats for the selected species.
Model	Multicollinearity	Multicollinearity	All bioclimatic covariates, karst layer and forest layer were pre-selected and then filtered after correlation analysis (0.7 cutoff value).
Model	Model settings	Model settings (fitting)	MXS and MXD algorithms.
Model	Model settings	Model settings (extrapolation)	Extrapolations over near-current accessible areas assuming MSDM 'OBR' for the present.
Model	Model estimates	Coefficients	NA
Model	Model estimates	Parameter uncertainty	NA
Model	Model estimates	Variable importance	Correlative.
Model	Model selection - model averaging - ensembles	Model selection	NA
Model	Model selection - model averaging - ensembles	Model averaging	NA
Model	Model selection - model averaging - ensembles	Model ensembles	Weighted averaging of the algorithms through TSS.
Model	Analysis and Correction of non-independence	Spatial autocorrelation	NA
Model	Analysis and Correction of non-independence	Temporal autocorrelation	NA
Model	Analysis and Correction of non-independence	Nested data	NA
Model	Threshold selection	Threshold selection	We used the sensitivity‐specificity sum maximisation (max TSS) approach to select the optimal suitability threshold.
Assessment	Performance statistics	Performance on training data	NA
Assessment	Performance statistics	Performance on validation data	NA
Assessment	Performance statistics	Performance on test data	True skill statistics (TSS).
Assessment	Plausibility check	Response shapes	NA
Assessment	Plausibility check	Expert judgement	IUCN range polygons and the Handbook of the Mammals of the World.
Prediction	Prediction output	Prediction unit	Continuous suitability and estimated richness for hotspots inference (sum of final binary maps).
Prediction	Prediction output	Post-processing	Area calculation through raster R package.
Prediction	Uncertainty quantification	Algorithmic uncertainty	Ensemble over two algorithms and 10 repeats.
Prediction	Uncertainty quantification	Input data uncertainty	Sampling bias adjusted map in Figure 2. 2 SSPs and 2 GCMs for future scenarios.
Prediction	Uncertainty quantification	Parameter uncertainty	Table S2 for parameters used in sampbias.
Prediction	Uncertainty quantification	Scenario uncertainty	SSP-2.45 and SSP-5.85 scenario evaluation.
Prediction	Uncertainty quantification	Novel environments	NA