This project aims to help answer the surprisingly complex question of whether it is safe to swim at beaches that are subject to blue-green algae. It explores the efficacy of combining existing data sources with an automated analysis of user contributed photographs to improve real-time information.
There are a number of resources that available to the public, but none of them represent a timely and convenient way to determine the safety of a particular swimming area:
- Lake Champlain Basin Program: Testing and Beach Closures
- VT Dept of Health: Cyanobacteria Tracker Map
The UVM Rubinstein school is another potential source of reliable on water quality data. If the available data is combined, and augmented with properly filtered user generated content, it may be possible to help vermonters quickly evaluate the likely presence of dangerous cyanobacteria.
Note: this project does not intend to provide a definitive safety assessment – this will be made abundently clear in all final products.
Seeking Project Partners
Currently we do not have a organizational partner working with Code for BTV on this project. In order to ensure that our effort results in tools that serve a real need within the community, all our projects must have an external partner vetting and directing the work. Project partners can be any not-for-profit organization. Project partners provide the expert knowledge surrounding the problem, and verify the direction of the project. One of the project partners will ultimately end up hosting and maintaining any digital tools that are produced, with assistance & direction from Code for BTV.
If you or someone you know would be a suitable project partner, please contact email@example.com.
Protyping Image Analysis Datasource
The data that is currently available tends to be highly reliable, but frequently out of date. We are exploring the posibility of creating a new datasource that has the opposite properties: trading off some reliability for availability. The Image Analyzer prototype is a machine learning approach that takes images submitted via a simple mobile app and attempts to apply machine learning approaches to determine whether an image represents an algae bloom.
This project will begin simply, and then we will explore the extent to which actionable data can be sourced from images. Ultimately the hope is that this method can provide an additional, potentially real-time datasource to assist in determining whether an algea bloom exists in a given place and time.
This project already shows promise in it's ability to distinguish algae blooms from other false-positives. Development of the photo-recognition algorythm has been informed by the following papers to varying degrees: