Machine-learning originated in computer science and is referred to as a type of Artificial Intelligence that allows computers to learn important relationships within data provided. Applied in environmental modelling, machine-learning is often referred to as a data-driven modelling approach, which extracts useful system information from existing data to predict future system behaviour without explicitly specifying the underling physical processes driving these behaviours. There are a lot of advances in environmental modelling using machine-learning techniques in the past three decades. Research focuses have expanded from the traditional machine-learning algorithms, such as Support Vector Machines (SVMs), Artificial Neural Networks (ANNs) and Bayesian networks (BNs), to ensemble learning techniques, such Random Forest, and the deep learning techniques recently becoming popular, including Deep Neural Networks (DNNs), Deep Belief Networks (DBNs), Convolution Neural Networks (CNNs) and Recurrent Neural Networks (RNNs). This session aims to bring together researchers from a wide spectrum of machine-learning in environmental modelling. Submissions on new developments of all learning techniques and their innovative applications to environmental systems are encouraged. We also welcome qualitative studies contributing to the discussion and communication of limitations, challenges and future research directions.
Key topics: Machine-learning, Neural networks, Environmental modelling