Skip to main content
Mapping one million small reservoirs in Brazil highlights widespread environmental and policy implications

Mapping one million small reservoirs in Brazil highlights widespread environmental and policy implications

This is a Preprint and has not been peer reviewed. This is version 1 of this Preprint.

Add a Comment

You must log in to post a comment.


Comments

There are no comments or no comments have been made public for this article.

Downloads

Download Preprint

Authors

Kylen Solvik , Yaffa Truelove, Jennifer Balch, Michael Lathuillière, Thiago Fontenelle, Andrea Castanho, Michael Coe, Christina Shintani, Carlos Souza, Marcia Macedo

Abstract

Small dams and reservoirs disrupt hydrological connectivity and increase evaporative water loss across Brazil, yet most remain unmapped. Applying deep learning to Sentinel satellite data from 2021, we created a comprehensive map of 1.1 million small on-stream reservoirs (smaller than 50 ha). Their 7,597 km^2 cumulative surface area surpasses the combined area of Brazil’s three largest mega-dam impoundments. Especially concentrated in areas with high cattle herd densities and precipitation variability, 78% are found in headwater sub-basins, which are vital for downstream water quality and ecosystem health. The reservoirs lose an extra 4.2 km^3yr^-1 to surface evaporation—equivalent to 29% of total annual urban water use in Brazil. Our findings highlight the overlooked socio-environmental consequences of small reservoirs for land and water management.

DOI

https://doi.org/10.31223/X5Z20S

Subjects

Water Resource Management

Keywords

deep learning, remote sensing, water management

Dates

Published: 2026-08-21 18:26

Last Updated: 2026-08-21 18:26

License

CC BY Attribution 4.0 International

Additional Metadata

Conflict of interest statement:
None

Data Availability:
The complete reservoir dataset is available at: https://doi.org/10.5281/zenodo.14927227. The annotated training data, trained model weights, and manually validated reservoir detections are available at: https://doi.org/10.5281/zenodo.14927196.

Metrics

Views: 31

Downloads: 9