Review of application of machine learning models for real time and short term prediction of water levels and discharge

Review and proposal of thorough review of application of machine learning models for real-time and short term prediction (forecasting) of water levels (stage) and discharge (streamflow, runoff, inflow). The emphasis is on methodology related to real time to short term prediction of high water events, but includes any machine learning (particulary supervised learning) modeling of levels and discharge. Some properties placed in the table may be estimated by reader, which is described in property "notes". For example, in some cases number of instances, historical dataset length, especially input shape and output shape should be estimated by reader. - prone to errors, they are estimated. Input shape also could be estimated in some contributions, Ratio of parts in dataset split is always in the form of percentage to sum up to 100 %, which often should be estimated by reader. Terms used in variety of papers for dataset split often differs (training, building, calibration, validation, verification, testing, etc.), but in the most of the cases it has two parts (which are here described as training and testing) or three parts (which are here described as called training, calibration and verification).