Dataset DOI: https://doi.org/10.5061/dryad.34tmpg4zw
Associated article DOI: https://doi.org/10.1002/ece3.74131

Description of the data and file structure

The occurrence file provides a spatially generalized public version of 368 original field-survey records collected by our research group in the Qinling–Daba Mountains, China, during 2018–2025. Collection period and provenance are supplied by the authors; observation dates are not stored in this three-column table. The occurrence file is distinct from the cleaned, deduplicated, multi-source occurrence inputs used for modelling in the associated article.

Public files

• nanorana_quadranus_occurrence_data.csv: genuine comma-delimited UTF-8 text with a byte-order mark (BOM), one header row and 368 data rows. Coordinates are generalized to 0.01°. No rows have been removed, reordered, or added during conversion.
• Code_S1.txt: associated R analysis code. The local source copy is named Code S1.txt; this README uses the uploaded filename reported in the submission. The code has not been modified or executed during this data conversion.

Variables

No missing values were found in these three fields. Coordinate decimal precision is a generalization setting, not a measurement of the original positional accuracy.

Data provenance and third-party sources

All 368 records in the deposited occurrence table were collected through field surveys by our research group. The table contains no occurrence records downloaded from GBIF or extracted from published literature. GBIF and literature references in the associated study refer to additional third-party sources used to prepare the integrated modelling inputs, not to the provenance of the 368 deposited field-survey records.

Third-party occurrence sources used in the associated study include the GBIF download cited at https://doi.org/10.15468/dl.788ct4 and published literature identified in the associated article. These are sources used by the study, not alternative public locations for the deposited survey table. Third-party records are not redistributed in this occurrence file or represented as author-owned data under a new licence; their original provider terms remain applicable.

Spatial generalization and conservation precautions

The source coordinates were original observation coordinates, not grid-cell or polygon centroids. For this public version, both longitude and latitude were rounded to the nearest 0.01° using decimal rounding with half-way values rounded away from zero (ROUND_HALF_UP). Only the rounded values are stored in the public CSV. This is coordinate rounding, not a transformation to centroids of environmental raster cells. No random displacement was applied.

Exact observation coordinates are not included in this public file. The generalized positions should not be used for precise site relocation or fine-scale habitat inference. Angular resolution does not correspond to a constant distance in both directions across latitudes. Spatial generalization may change environmental values extracted at the released coordinates and does not guarantee exact reproduction of analyses based on original coordinates.

Record duplication and quality limitations

The source table contains 368 records representing 160 distinct coordinate pairs before generalization. After rounding to 0.01°, the public file contains 105 distinct coordinate pairs. Shared coordinates may reflect surveys in different years, observations of different individuals, or possible duplicate entries. The available fields do not allow these cases to be distinguished reliably. All 368 records are retained, and no coordinate-based deduplication was performed. Additional coincident coordinates resulting from generalization are not evidence of duplicate observations.

Record counts must not be interpreted as abundance, numbers of individuals, or numbers of independent survey sites. Users should assess spatial sampling bias and apply analysis-appropriate quality control; the table does not contain dates or individual identifiers for event-level deduplication.

Code, inputs and reproducibility limitations

The code covers selected modelling and visualization steps, including five-algorithm biomod2 modelling and evaluation (GLM, GAM, GBM, RF and MAXNET), ENMeval tuning, correlation visualization, habitat-area plotting and centroid-distance plotting. The ENMeval section supplied uses random 10-fold cross-validation (randomkfold), not spatially blocked cross-validation. R version 4.4.1 was reported in the original submission; package versions are not fully documented in the supplied script.

The files Nanorana quadranus.csv and lgw.csv referenced by the script are processed modelling inputs prepared through cleaning, deduplication and integration of field-survey and third-party occurrence data. They are not included here and are not equivalent to the 368-record survey file. Do not directly substitute the deposited survey table for those files. The script also refers to external environmental layers, plotting input tables and intermediate objects whose preparation is not fully included. The submitted materials are not a self-contained, end-to-end reproduction package.

External environmental inputs

Environmental raster files are not included. The original submission identifies WorldClim v2.1 (https://www.worldclim.org/), the Harmonized World Soil Database (https://www.fao.org/soils-portal/), and Global Terrestrial Human Footprint data (https://doi.org/10.5061/dryad.052q5) as external sources. Consult the associated article and original providers for methods, versions and access conditions. Listing these sources does not relicense or redistribute their data.

Rights and sensitive information

The authors submit the author-collected public occurrence table under Dryad's CC0 dedication. This statement does not apply a new licence to third-party data cited above. The table contains no human-subject information. Spatial generalization is applied to reduce conservation risks; no claim is made that rounding eliminates all location-disclosure risk.