1/2

29 files

Machine learning concepts and methods for hydrological post-processing and forecasting (invited)

presentation

posted on 2021-10-27, 13:04 authored by Georgia PapacharalampousGeorgia Papacharalampous, Hristos TyralisHristos Tyralis

During the past decades, hydrological forecasting was traditionally performed using process-based (i.e., physically-based or conceptual) and/or statistical (e.g., time series or distributional) models. Among other advantages, such models offer a certain degree of interpretability, which is considered important in the hydrological literature. In this presentation, we extensively discuss machine learning concepts and methods for hydrological post-processing and forecasting. We also discuss some new benefits that these concepts and methods could offer to our field from a technical point of view, and their integration with the existing hydrological experience and knowledge. For supporting our discussions, we go through published large-scale benchmark tests.

History

Usage metrics

Categories

Keywords

ARFIMA ARIMA artificial intelligence artificial neural networks autocorrelation automatic forecasting benchmarking big data big data hydrology catchment hydrology climatic regime combining forecasts combining probabilistic forecasts complex exponential smoothing computer science conceptual model generalization generalized random forests geoscience global-scale hydrology gradient boosting machine ensemble learning entropy environmental informatics environmental science equal-weight combiner explainable machine learning exponential smoothing extremely randomized trees feature extraction feature-based time series analysis feature-based time series clustering feature-based time series forecasting forecast combinations forecastability forecasting hydroclimatic change hydroclimatic feature hydroclimatic regime hydroclimatic signature hydroclimatic time series hydroclimatic variability hydroinformatics hydrological forecasting hydrological modelling hydrological post-processing hydrological prediction hydrological processes hydrological science hydrological time series forecasting hydrological uncertainty hydrology interval score lag-1 sample autocorrelation large datasets large-sample hydrology large-scale benchmarking large-scale hydrology lasso linear regression loess long-range dependence machine learning MARS metalearning model-based boosting neural networks no free lunch nonlinearity physically-based model polyMARS precipitation predictability prediction probabilistic forecasting probabilistic modelling process-based model Prophet quantile regression quantile regression forests quantile regression neural networks quantile score random forests river flow rivers sample autocorrelation seasonal strength seasonality simple averaging simple combinations simple exponential smoothing spectral entropy stacked generalizations stacking statistical characterization statistical hydrology statistical learning stochastic hydrology stochastic model streamflow streams super learning support vector machines temperature temporal dependence time series time series analysis time series characterization time series clustering time series forecasting time series feature trend strength trends water water informatics water science wisdom of the crowd XGBoost Environmental Science Hydrology Applied Computer Science Conceptual Modelling Applied Statistics Statistics Stochastic Analysis and Modelling Water Resources Engineering

Licence

Exports

RefWorks

BibTeX

Ref. manager

Endnote

DataCite

NLM

DC