Data for "Mapping the ghost: Estimating probabilistic snow leopard distribution across Mongolia"

Data and code used for a country-wide occupancy survey of snow leopards in Mongolia, accompanying the paper "Mapping the ghost: Estimating probabilistic snow leopard distribution across Mongolia".

This data contains the results of a survey of 1017 20x20km sampling units, out of a total of 1200 sampling units identified as potential snow leopard habitat (183 could not be sampled for various reasons), a near complete survey of potential snow leopard habitat in Mongolia, nearly 500,000 square kilometers, and an enormous effort by many researchers. If you make use of the data, please cite the following sources:


Data for "Mapping the ghost: Estimating probabilistic snow leopard distribution across Mongolia". (2021). Gantulga Bayandonoi, Koustubh Sharma, Justine Shanti Alexander, Purevjav Lkhagvajav, Ian Durbach, Darryl MacKenzie, Chimeddorj Buyanaa, Bariushaa Munkhtsog, Munkhtogtokh Ochirjav, Sergelen Erdenebaatar, Bilguun Batkhuyag, Nyamzav Battulga, Choidogjamts Byambasuren, Bayartsaikhan Uudus, Shar Setev, Lkhagvasuren Davaa, Khurel-Erdene Agchbayar, Naranbaatar Galsandorj, David Borchers. doi: https://doi.org/10.5281/zenodo.5257572
Mapping the ghost: Estimating probabilistic snow leopard distribution across Mongolia. (2021). Gantulga Bayandonoi, Koustubh Sharma, Justine Shanti Alexander, Purevjav Lkhagvajav, Ian Durbach, Darryl MacKenzie, Chimeddorj Buyanaa, Bariushaa Munkhtsog, Munkhtogtokh Ochirjav, Sergelen Erdenebaatar, Bilguun Batkhuyag, Nyamzav Battulga, Choidogjamts Byambasuren, Bayartsaikhan Uudus, Shar Setev, Lkhagvasuren Davaa, Khurel-Erdene Agchbayar, Naranbaatar Galsandorj, David Borchers. To appear in Diversity and Distributions


Contents of zip file

Data

The main dataset is contained in `data\Mongolia_occupancy_inputs.Rdata` . Please see the paper for more detail on data collection. The following objects are contained in the file:

- Pres: presence/absence occupancy survey results, used for model fitting
- Site_Cov: unit-specific covariates, used for model fitting
- SurvCov: survey-specific covariates, used for model fitting
- Mongolia_studyarea: covariates for whole survey area, used for prediction
- Mongolia_fullrange: covariates across whole expected snow leopard range, used for prediction

Code

Code is cloned from the GitHub repository https://github.com/iandurbach/mongolia-occupancy, which may contain updates. The version here reproduces the analyses in the paper above. The run these analyses:

- run *occupancy-analysis.R* to fit the main occupancy models (these are also saved in the `\output` folder), do model selection, and plot covariate effects
- run *occupancy-goodness-of-fit.R* to calculate the c-hat statistic giving an indication of model fit for the best model
- run *comparing-maps.R* to compare the occupancy results with similar metrics generated using a presence-only analysis (using MaxEnt) or an expert map generated through qualitative discussion (reproduces Figure 3 in the paper).

Code in *occupancy-data-preproc.R* is not needed but included for completeness. It converts the csv files in `data\csv`, which contain various input datasets used by the occupancy model, into a single .Rdata file (`data\Mongolia_occupancy_inputs.Rdata`), which is then used by the scripts above. Some minimal pre-processing (excluding ununsed variables, renaming for consistency, etc) is performed.