BlueDot Observatory – keeping an eye on our planet’s water resources

Managing water crises is one of the United Nations’ Sustainable Development Goals and the decline in the available quality and quantity of fresh water is ranked as one of the top ten most serious societal risks by the World Economic Forum’s 2018 Global Risks report. Using satellite imagery available through the AWS Open Data Program and the AWS Cloud, BlueDot Observatory is establishing a global monitoring system for all at-risk water bodies. This monitoring reveals a sad truth – the total loss of water bodies is in the not-too-distant future.

We invited Anze Zupanc, a data scientist who manages the BlueDot Observatory at Sinergise, to share how the AWS Open Data Program and the Amazon Sustainability Data Initiative support this work.

A guest post by Anze Zupanc, Data scientist, Sinergise

Leveraging AWS Open Data and the cloud to monitor at-risk bodies of water

Most of us who have the privilege to live in the developed world treat water as an unlimited resource. We don’t think about where water is coming from and how much is available. But even if we did, we might encounter challenges accessing detailed water-level information. This inspired us to build the BlueDot Observatory.

To prototype our solution using Sentinel data and save on computer resources, we decided not to process the entire land mass and instead focus only on water bodies and their surroundings. First, we needed to find a list of all water bodies around the world. Luckily, we were able to build our global database of water bodies – lakes, dams, and reservoirs – on top of existing databases, such as the Global Reservoir and Dam (GRanDv1.01) database, the WWF’s Global Lakes and Wetlands Database, and the OpenStreetMap.

To delineate the “current water extent,” we developed an open-source algorithm that checks for newly available Sentinel-2 data in the AWS Open Data Program and processes the data using Sentinel Hub services to filter out clouds and identify water pixels. These pixels are combined with an object and its area is compared with the “nominal water extent.” We regularly and automatically execute a water detection algorithm over water bodies from our database on a dedicated Amazon EC2 instance started by an AWS Lambda function and triggered with an Amazon CloudWatch event. We store all data in Amazon S3. The BlueDot Observatory retrieves data using web services and displays the satellite image, current and nominal water extent, and the chart displaying trends in surface water level.

The technology is not complex, as it uses AWS infrastructure and compute. By having the complete Sentinel-2 archive available on AWS and access to on-demand scalable compute resources, the workflow is simple to implement. This simplicity is one of the most important findings of our exercise. The cost to process one month of data for about 7,000 bodies of water currently in the system is 6 EUR. The overall cost of the system, including all backend resources and Sentinel Hub subscription, is around 100 EUR per month. It is possible to set up world-scale systems with a shoestring budget.

Next steps for the BlueDot Observatory

Our work is not finished yet. One of our immediate priorities is to add more bodies of water to the system, which will allow us to monitor more of those surfaces. The list of water bodies at risk, as monitored by the World Resources Institute’s Aqueduct project, continues to expand.

In addition, we are looking to use the Landsat-8 archive on AWS to extend the temporal coverage back to 2013 and, as soon as USGS stages the Landsat-5 data to AWS, we will look at the behavior of these water bodies since 1984. The use of SAR imagery from Sentinel-1 mission will help us improve the accuracy of water surface detection, especially in cloudy regions. We are looking into improving the depth of our analysis to go beyond detection of water level of a single body to provide information such as the high and low inter-season water level variability and identify lakes historic (low) levels. A simple map with this type of information will likely reveal many patterns. By correlating this information with other data sources available on AWS through the Amazon Sustainability Data Initiative (weather and climate data), we hope to create predictive capabilities for the water levels. Finally, we hope to streamline our workflows by filtering out irrelevant data, cleaning polygons, and improving efficiencies in our backend.

Our main objective when building this service is to raise awareness about the importance of how we treat water. We hope that by providing evidence of how vulnerable these water resources are, people will start accepting that the water crisis is no longer “somewhere far away” and that it affects us all.

The BlueDot Observatory project is supported with AWS Promotional Cloud Credits through the AWS Cloud Credits for Research Program, and with open data from the AWS Open Data Program and the Amazon Sustainability Data Initiative.