Data and code from: Remote sensing reveals post-fire forest composition shifts in boreal Eastern Siberia

作者：O. Zheleznyy, Cornelius Senf, Julia Boike, Dirk Pflugmacher · 发表于：DRYAD · 年份：2026 · DOI：10.5061/dryad.k3j9kd5m7 · 研究领域：Remote sensing、Environmental science、Physical geography、Meteorology

This repository contains the code and data used to map and analyse post-fire forest composition shifts in boreal forests of Eastern Siberia. First, spectral indices were derived in Google Earth Engine from Landsat satellite imagery, helping us to semi-manually identify 1863 wildfires in 1990-1994. From this step, we provide vector burned area outlines, as well as points generated inside burned areas and used for further analysis. We then mapped the current fractional cover of five key boreal vegetation types using regression-based spectral unmixing of Sentinel-2 satellite imagery, combining R, Python and Google Earth Engine scripts. We provide manually collected training and validation data recording vegetation composition, as well as synthetically mixed endmembers and resulting fractional cover data, generalised for use in figures. Finally, we used R to quantify vegetation differences between burned and adjacent unburned areas, assessing the effects of burn severity, proximity to the fire edge, and average annual air temperatures on post-fire compositional changes. Here, we provide tabular data with fractional cover and predictor variables, extracted for burned and unburned points. Training and validation points, as well as fire polygons, can be used in further studies of fire ecology and forest composition of Eastern Siberia.