Machine learning method makes accurate water temperature prediction

Water temperature plays an important role in the assessment of water’s purity. It affects the level of dissolved oxygen in water. It also affects the rate of biological and chemical reactions in the water. To ensure the water is safe for drinking purposes it is important to check its temperature level.

 

Water temperatures tend to fluctuate in streams and rivers. As there is no fixed criteria to map the water levels it becomes difficult to assess its purity. To fix this problem of water temperature assessment a team of researchers from University of Minnesota and University of Pittsburgh have developed a machine learning method where the algorithm is trained with the rules of the physical world to make predictions about water temperatures. 

 

The model can predict the water temperature of streams and rivers with little data.  There is no data available for water temperatures of streams and rivers. It can perform a lot with no data. The method uses process guided knowledge based machine learning. It was used for the water temperature assessment in the Delaware river basin. As the method uses process guided knowledge it has overcome the common errors associated with machine learning prediction models. The model can be used to check water temperatures at different periods of time. 

 

(Must Check: 4 Applications of Big Data in Waste Management)

 

The method informs the machine learning algorithm by the consideration of factors like correlation through time, connection between streams, and energy budget equations. The Delaware River Basin in the US, is subject to extreme fluctuations in water temperature therefore it was chosen for testing machine learning algorithm. 

 

Knowledge guided machine learning techniques are better compared to traditional machine learning models. They can be used for accurate temperature prediction in waters and streams; it will also help in ensuring purity in the drinking water level. 

 

According to Tech Xplore Xiaowei Jia, assistant professor in the University of Pittsburgh's Department of Computer Science  said, "Accurate  prediction of water temperature and streamflow also aids in decision making for resource managers, for example helping them to determine when and how much water to release from reservoirs to downstream rivers.

 

Researchers plan to extend the usability of this machine learning algorithm. The correct predictions will also help in clean water supply.