A review on how machine learning can be beneficial for sensor data quality control and imputation in water resources management
Quality control and data imputation are facing new challenges due to huge amounts of heterogeneous data that need to be qualified and, at the same time, new opportunities arising from innovations in machine learning (ML). A meta-level description of why a transformation from, or the integration of conventional methods to ML methods could be beneficial, which conditions need to be met, and how far the actual utilization possibilities have progressed is needed to leverage its potential. Our critical review closes this gap by qualitatively condensing the state of the art, pointing out open questions and possible solutions. The aim of this work is to contribute to a clearer outline of the current research state to better target future works. The results are that questions and challenges remain open in all three areas: benefits, conditions, and usage of machine learning in water resources management. Most importantly, ML models can currently only be poorly classified in terms of their performance and trustworthiness. Nevertheless, we hypothesize that ML will be capable of increasing the degree of automation in water resources management, even though there are still many steps to be taken for the efficient application of machine learning in water resources management.
Preview
Cite
Rights
Use and reproduction:
This work may be used under a
Creative Commons Attribution 4.0 License (CC BY 4.0)
.