×J2. Data science for water resources: the application of deep learning and machine learning

The availability of data with high spatio-temporal resolution has increased significantly in recent times with extended measurement networks and improved remote sensing techniques. High-resolution environmental modelling data that are now available include Digital Elevation Models; radar, gauge and satellite-derived precipitation products; gridded temperature, solar radiation, humidity, pressure, wind and potential evapotranspiration products, and Numerical Weather Prediction forecasts. High-performance computing also makes it easier to take advantage of these datasets. As a result, machine learning (ML) algorithms are now widely used in Earth and Environmental modelling studies. ML algorithms allow researchers to identify relationships within datasets and extract useful system information that can help predict future behaviour without explicitly specifying the underling physical processes. The research community has been developing new data-driven modelling methods and applications to support informed decision-making: traditional ML has been extended to Artificial Neural Networks and Bayesian networks, including Deep Neural Networks and Convolution Neural Networks. The objective of this session is to bring together a diverse group of researchers who use ML to solve different problems. We welcome data scientists, including those focussed on data analysis, statistics, data visualization, computer science and/or mathematics, hydrologists, hydro-meteorologist to present their innovative work with environmental systems. We also encourage submissions that discuss new developments, applications, limitations and future directions related to ML in environmental modelling.

Key topics: Machine learning, Deep learning, Streamflow forecasts and NWP forecasts, Water resources