Abstract
Dissolved Oxygen (DO) in water enables marine life. Measuring the prevalence of DO in a body of water is an important part of sustainability efforts because low oxygen levels are a primary indicator of contamination and distress in bodies of water. Therefore, aquariums and aquaculture of all types are in need of near real-time dissolved oxygen monitoring and spend a lot of money on purchasing and maintaining DO meters that are either expensive, inefficient, or manually operated-in which case they also need to ensure that manual readings are taken frequently which is time consuming. Hence a cost-effective and sustainable automated Internet of Things (IoT) system for this task is necessary and long overdue. DOxy, is such an IoT system under research and development at Santa Clara University's Ethical, Pragmatic, and Intelligent Computing (EPIC) Laboratory which utilizes cost-effective, accessible, and sustainable Sensing Units (SUs) for measuring the dissolved oxygen levels present in bodies of water which send their readings to a web based cloud infrastructure for storage, analysis, and visualization. DOxy's SUs are equipped with a High-sensitivity Pulse Oximeter meant for measuring dissolved oxygen levels in human blood, not water. Hence a number of parallel readings of water samples were gathered by both the High-sensitivity Pulse Oximeter and a standard dissolved oxygen meter. Then, two approaches for relating the readings were investigated. In the first, various machine learning models were trained and tested to produce a dynamic mapping of sensor readings to actual DO values. In the second, curve-fitting models were used to produce a successful conversion formula usable in the DOxy SUs offline. Both proved successful in producing accurate results.