This repository contains the workflows and methodologies used to detect Aquacultures and Saltcultures in Brazil within the MapBiomas Brazil project.

The mapping is based on Landsat Top-of-Atmosphere (TOA) imagery and combines machine learning and deep learning approaches to generate annual maps for multiple coastal environments.

To ensure consistency across collections, the repository is organized by MapBiomas Collection versions, with each collection implemented in its own directory.

The repository is subdivided into folders corresponding to MapBiomas collections, for example:

• MapBiomas09/
• MapBiomas10/

Each collection folder contains:

• The specific processing workflow adopted in that collection
• Scripts and notebooks for data preparation, training, classification, and post-processing
• Class-specific subdirectories when applicable

This structure allows methodologies to evolve across collections while preserving reproducibility and historical context.

Across collections, the aquaculture mapping follows a common high-level workflow:

1. Annual Landsat Mosaic Generation

2. Deep Learning Mapping (semantic segmentation)

3. Temporal Consistency and Integration Annual results are reviewed and integrated into the corresponding MapBiomas collection.

Implementation details may vary between collections and are documented within each collection folder.

A detailed description of the theoretical background and methodological decisions is available in the:

Coastal Zone Algorithm Theoretical Basis Document.

This README provides a general overview of the repository.
For collection-specific details, please refer to the README files inside each MapBiomasX/ directory.