anaya (2025) Obas

Identification

• Journal: Mendeley Data
• Year: 2025
• Date: 2025-12-05
• Authors: anaya, Jesús
• DOI: 10.17632/5d8kxmmwtz

Research Groups

• Universidad de Medellin

Short Summary

This document describes the methodology for creating a pre-processed Sentinel-2 multispectral data cube, designed as input for a Convolutional Neural Network to predict burned areas.

Objective

• To detail the process of generating a clean, analysis-ready Sentinel-2 multispectral raster data cube, suitable for training and inference with a Convolutional Neural Network for burned area prediction.

Study Configuration

• Spatial Scale: Sentinel-2 data with resolutions of 10 m, 20 m, and 60 m. Data processing involves a global surface water mask and an official land cover map of Colombia.
• Temporal Scale: Not specified for the data acquisition period.

Methodology and Data

• Models used: Convolutional Neural Network (CNN) (intended recipient of the data cube).
• Data sources: Sentinel-2 (S2) multispectral data from the Copernicus Program (Sentinel-2A and 2B), official land cover map of Colombia, global surface water mask (Pekel et al., 2016), S2 Scene Classification Layer (SCL) shadow (class 3) and cloud (classes 8-9) masks (ESA, 2025). Data processing performed in Google Earth Engine.

Main Results

This document describes the methodology for preparing a Sentinel-2 multispectral data cube for burned area prediction using a CNN model. It does not present specific results from a burned area prediction study.

Contributions

The original value lies in the detailed methodology for generating a pre-processed Sentinel-2 multispectral data cube, optimized for input into a Convolutional Neural Network for burned area prediction, incorporating robust masking procedures for water, clouds, and shadows.

Funding

• Saxon State Ministry for Science and the Arts Germany (Grant ID: 3-7304/35/6-2021/48880)

Citation

@article{anaya2025Obas,
  author = {anaya, Jesús},
  title = {Obas},
  journal = {Mendeley Data},
  year = {2025},
  doi = {10.17632/5d8kxmmwtz},
  url = {https://doi.org/10.17632/5d8kxmmwtz}
}