Brazil accounts for 13 percent of the fresh water in the world, but approximately 35 million Brazilians don’t have access to clean water. The contamination of many of Brazil’s rivers leads to high rates of waterborne illnesses among the human population and poses a threat to the area’s biodiversity—especially that of the Atlantic Forest. Although it stretches from the coast of Brazil inland to Argentina and Paraguay and encompasses major cities such as Rio de Janeiro and São Paulo, the Atlantic Forest has suffered greatly from deforestation; today it covers just 12.4 percent of the land it once did. SOS Mata Atlântica, a nonprofit organization dedicated to preserving the Atlantic Forest and its species and ecosystems, has organized thousands of volunteers across the country to collect water samples and create sets of longitudinal data (collected monthly) on the quality of Brazil’s water. Without sophisticated analysis, however, the data couldn't be used to its full potential.


With Azure Machine Learning and other Azure services, SOS Mata Atlântica has moved from collating and summarizing data to extracting insights and making predictions about water quality across Brazil. As a result, the organization can get closer to answering key questions about the impact of forestation and sanitation on water quality. With a clearer understanding of the implications of the data collected over the past two decades, SOS Mata Atlântica can use data more effectively than ever to advocate for public action and policy change.

How SOS Mata Atlântica uses Azure

  • Functions for consumption of external bases via API and scraping.
  • Data Factory to orchestrate the consumption of bases and execution of databricks notebooks.
  • Blob Storage to store data collected directly from sources, without any treatment.
  • Databricks to perform analysis and treatment of data collected on water quality, deforestation, public health, and sanitation.
  • SQL Server to store analytical data already processed by Databricks.
  • App Services to integrate the donation service and the dashboard and donation pages.
  • Azure Machine Learning to test predictive models and select the model with the best accuracy. In this case, the selected model was the Elastic Net.
A woman testing water outside in the forest.

Working to defend the Atlantic Forest and river ecosystems.

A river in the forest.