These R scripts and dataset support the article “Afforestation reduces fire occurrence in Colombia’s tropical savannas,” which investigates the effects of commercial forest plantations on fire dynamics in the tropical savannas of Vichada, Colombia. The study area is located in the Orinoquia region and spans various land cover types, including native grasslands, mixed-use agro-pastoral mosaics, and commercial plantations of Acacia mangium, Eucalyptus pellita, and Pinus caribaea. The dataset contains: 1) coordinates of 96 spatially independent sampling points, stratified by land‑cover type (44 plantation sites, 52 grassland sites); 2) csv tables with buffer‑based summaries of land cover, burned area, climate, and distances to infrastructure; 3) R scripts for spatial sampling, data preprocessing, buffer‑based extraction, model fitting, diagnostics, and plotting; and 4) point GIS layers (GeoPackage format) for plantation and grassland sampling sites, compatible with standard GIS software. No human subjects are included in these data.

File inventory and descriptions

Spatial Data

plantaciones.gpkg
Contents: 44 point features representing plantation sampling sites.

Geometry: Points. Each point corresponds to a commercial forest plantation site included in the analyses.

Use: Input for buffer creation, spatial extraction of land‑cover, climate, and burned‑area variables, and mapping of plantation sites

pastos_dep.gpkg
Contents: 52 point features representing grassland (pasture) sampling sites.

Geometry: Points. Each point corresponds to a native grassland/pasture site included in the analyses.

Use: Input for buffer creation, spatial extraction of land‑cover, climate, and burned‑area variables, and mapping of grassland sites.

CSV data

CoversAreaClimaterevDA.csv

Format: Comma‑separated values (CSV) with header row.

Content: Buffer‑level summaries of land cover, burned area, climate, and distances to infrastructure and roads, for multiple radii around each sampling point and year.

Structure: Each row corresponds to one point–year–radius combination (e.g. a 1 km, 2 km, … 6 km buffer around a given point in a given year).

Column name	Units / categories	Description
Unique ID	String	Unique identifier for each point–radius–year record; encodes land‑cover type, point, and year
point_id	1-96	Numeric ID of the sampling point. All radii around the same point share this `point_id`
Type	Grasslands/ Plantations	Land‑cover stratum of the focal point: grassland sites vs plantation sites.
Year	Calendar year	Year for which land‑cover, burned‑area, and climate summaries were calculated
radius_km	Kilometres (1, 2, 3, 4, 5, 6)	Radius of the circular buffer used to summarize land cover and fire around each site
Mosaic_area_ha	Hectares	Area within the buffer classified as agricultural and pasture mosaics (mixed crops and pastures).
Plantations_area_ha	Hectares	Area of the forest plantations class (MapBiomas Colombia) within the buffer.
Grasslands_area_ha	Hectares	Area within the buffer classified as grasslands/savannas (MapBiomas Colombia).
Grasslands_Mosaics	Hectares	Combined area of grasslands plus agricultural/pasture mosaics (the “Mixed” class used in the analyses).
Burnt_area_ha_CCI	Hectares	Annual burned area within the buffer from ESA FireCCI51 burned‑area product (MODIS Fire_cci v5.1, 250 m).
Burnt_area_ha_MCD	Hectares	Annual burned area within the buffer from MODIS MCD64A1 burned‑area product (500 m)
Temperature_C	Degrees Celsius	Mean annual air temperature from NOAA CFSv2 reanalysis, averaged over the buffer for that year.
Precipitation_mm	Millimetres	Total annual precipitation from the GPM dataset, aggregated over the buffer for that year.
dist_to_infraestructure_km	Kilometres	Euclidean distance from the sampling site to the nearest urban infrastructure (towns and built‑up areas).
dist_to_roads_km	Kilometres	Euclidean distance from the sampling site to the nearest road (MapBiomas roads layer).
Area_radio_ha	Hectares	Total area of the circular buffer defined by radius_km; geometric buffer area in hectares.

Notes:

Land cover comes from MapBiomas Colombia Collection 1.0, annual land‑cover maps (1985–2022, 30 m resolution); analyses focus on forest plantations, grasslands, and agricultural/pasture mosaics.

Burned‑area variables (Burnt_area_ha_CCI, Burnt_area_ha_MCD) are derived from monthly products aggregated to annual burned area per buffer; the main analyses use FireCCI51 (250 m), with MCD64A1 (500 m) used for robustness.

All covariates are harmonized at annual resolution to match fire and land‑cover data; the statistical analyses use records from 2010 onwards, when plantations become widespread in the region

R Scripts

ScriptCoversFireSampling.R, Spatial sampling and construction of buffer‑level covariates used in the analysis.
burnt_area_gamma_vAug_cci.R, Fit hierarchical gamma models of annual burned area
burnt_area_gamma_vAug.R, Replicate the main burned‑area models using MCD64A1 as an alternative burned‑area dataset.
burnt_gamma_check_vAug_cci.R, Diagnostic checks for FireCCI51 gamma models
burnt_gamma_check_vAug.R, Diagnostic checks for MCD64A1 gamma models
burnt_replot_vSep_cci.r, Reproduce figures related to FireCCI51 burned‑area models
burnt_replot_vSep.r, Reproduce figures based on MCD64A1 models.
mix_area_gamma_vSep.R, Model relationships between mixed land cover (grasslands + agricultural/pasture mosaics) and fire
mix_gamma_check_vSep.R, Diagnostic checks for mixed land‑cover gamma models.
mix_replot_vSep.r, Recreate figures summarizing mixed land‑cover composition and fire outcomes across buffer radii

Data and R code from: Afforestation reduces fire occurrence in Colombia’s tropical savannas

Data files

Abstract

File inventory and descriptions

Spatial Data

CSV data

Data and R code from: Afforestation reduces fire occurrence in Colombia’s tropical savannas

Data files

Abstract

README

File inventory and descriptions

Spatial Data

CSV data