Prediction Of Hydrological Models’ Uncertainty By A Committee Of Machine Learning-Models
N. Kayastha (IHE Delft Institute for Water Education)
Dmitri Solomatine (TU Delft - Water Resources, IHE Delft Institute for Water Education)
D. Lal Shrestha (CSIRO Land and Water)
More Info
expand_more
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.
Abstract
In the MLUE method (reported in Shrestha et al. [1, 2]) we run a hydrological model M for multiple realizations of parameters vectors (Monte Carlo simulations), and use this data to build a machine learning model V to predict uncertainty (quantiles) of the model M output. In this paper, for model V, we employ three machine learning techniques, namely, artificial neural networks, model tree, locally weighted regression which leads to several models results. We propose to use the simple averaging method (SA) and the weighted model averaging method (WMA) to form a committee of these models. These approaches are applied to estimate uncertainty of streamflows simulation in Bagmati catchment in Nepal. Tests on the different data sets show that WMA performs a bit better than SA.
help_outline