Machine Learning in Prediction of Parameter Uncertainty: MLUE Method

This chapter presents a novel approach to model and predict parametric uncertainty of rainfall-runoff model. The most widely used method to analyse such uncertainty (Monte Carlo simulation) involves multiple model runs. The computational load may prohibit the use of such method for real-time predictions of uncertainty valid for particular combinations of the input and state variables (hydrometeorological conditions). Machine learning techniques allow for the building of surrogate models for the results of such simulations; these models are fast and can be easily used in operation for real-time predictions of parameter uncertainty. The proposed method is referred to as MLUE (Machine Learning in parameter Uncertainty estimation).