Ensemble machine learning paradigms in hydrology: A review

Zounemat‐Kermani, Mohammad; Batelaan, Okke; Fadaee, Marzieh; Hinkelmann, Reinhard

doi:10.1016/j.jhydrol.2021.126266

Cited by 357 publications

(149 citation statements)

References 218 publications

Supporting

Mentioning

148

Contrasting

Unclassified

Order By: Relevance

“…typically combine the forecasts from multiple models and are designed to outweigh any contributing ensemble member. Applications of different types of ensemble AI techniques in hydrology have been recently reviewed by Zounemat-Kermani et al (2021). This study introduces the ensemble GP-SARIMA model in which the prediction process is composed of three main stages (Fig.…”

Section: The Proposed Evolutionary Gp-sarima Modelmentioning

confidence: 99%

A new evolutionary time series model for streamflow forecasting in boreal lake-river systems

Mehr

Ghadimi

Marttila

et al. 2022

Theor Appl Climatol

View full text Add to dashboard Cite

Genetic programming (GP) is an evolutionary regression method that has received considerable interest to model hydro-environmental phenomena recently. Considering the sparseness of hydro-meteorological stations on northern areas, this study investigates the benefits and downfalls of univariate streamflow modeling at high latitudes using GP and seasonal autoregressive integrated moving average (SARIMA). Furthermore, a new evolutionary time series model, called GP-SARIMA, is introduced to enhance streamflow forecasting accuracy at long-term horizons in a lake-river system. The paper includes testing the new model for one-step-ahead forecasts of daily mean, weekly mean, and monthly mean streamflow in the headwaters of the Oulujoki River, Finland. The results showed that a combination of correlogram and average mutual information (AMI) analysis might yield in the selection of the optimum lags that are needed to be used as the predictors of streamflow models. With Nash-Sutcliffe efficiency values of more than 99%, both GP and SARIMA models exhibited good performance for daily streamflow prediction. However, they were not able to precisely model the intramonthly snow water equivalent in the long-term forecast. The proposed ensemble model, which integrates the best GP and SARIMA models with the most efficient predictor, may eliminate one-fourth of root mean squared errors of standalone models. The GP-SARIMA also showed up to three times improvement in the accuracy of the standalone models based on the Nash-Sutcliff efficiency measure.

show abstract

Section: The Proposed Evolutionary Gp-sarima Modelmentioning

confidence: 99%

A new evolutionary time series model for streamflow forecasting in boreal lake-river systems

Mehr

Ghadimi

Marttila

et al. 2022

Theor Appl Climatol

View full text Add to dashboard Cite

show abstract

“…One key aspect of many subsurface energy applications is a coupled process that involves hydrogeology and geomechanics. Although ML-based data-driven modeling has been increasingly studied for reservoir modeling, most are still limited to uncoupled processes (e.g., Lange and Sippel (2020); Miah (2020)) or relatively homogeneous fields (e.g., Zounemat-Kermani et al (2021); Zhao (2021)). The CcGAN approach proposed in this study demonstrates its capability to handle coupled hydro-mechanical processes with relative RMSE less than 2% of the transient pressure and displacement responses in the worst case.…”

Section: Prediction Accuracy and Geophysical Applicationsmentioning

confidence: 99%

Continuous conditional generative adversarial networks for data-driven solutions of poroelasticity with heterogeneous material properties

Kadeethum,

O'Malley,

Choi

et al. 2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Data-driven models have the advantage of often being naturally suitable to effectively perform uncertainty quantification (e.g., via resampling methods, Friedman et al 2001); in some cases, they are also characterized by a lower number of input parameters to be calibrated (hereafter called hyperparameters). Zounemat-Kermani et al (2021) presented a review on the use of ensemble machine learning in a variety of hydrological applications-including the estimation of suspended sediment transport-reporting that a superiority of ensemble machine learning compared to ordinary learning had been claimed in many literature studies. Bhattacharya et al (2007) used artificial neural networks (ANNs) and model trees (MTs) to predict bed load and total sediment fluxes; they found that the machine-learning approach performed better than several commonly used empirical equations (with ref-erence to a prior compilation of laboratory and field data)-with similar performance for ANNs and MTs.…”

Section: Introductionmentioning

confidence: 99%

A Comparison Between Machine Learning and Functional Geostatistics Approaches for Data-Driven Analyses of Sediment Transport in a Pre-Alpine Stream

et al. 2022

View full text Add to dashboard Cite

The problem of providing data-driven models for sediment transport in a pre-Alpine stream in Italy is addressed. This study is based on a large set of measurements collected from real pebbles, traced along the stream through radio-frequency identification tags after precipitation events. Two classes of data-driven models based on machine learning and functional geostatistics approaches are proposed and evaluated to predict the probability of movement of single pebbles within the stream. The first class built upon gradient-boosting decision trees allows one to estimate the probability of movement of a pebble based on the pebbles’ geometrical features, river flow rate, location, and subdomain types. The second class is built upon functional kriging, a recent geostatistical technique that allows one to predict a functional profile—that is, the movement probability of a pebble, as a function of the pebbles’ geometrical features or the stream’s flow rate—at unsampled locations in the study area. Although grounded in different perspectives, both models aim to account for two main sources of uncertainty, namely, (1) the complexity of a river’s morphological structure and (2) the highly nonlinear dependence between probability of movement, pebble size and shape, and the stream’s flow rate. The performance of the two methods is extensively compared in terms of classification accuracy. The analyses show that despite the different perspectives, the overall performance is adequate and consistent, which suggests that both approaches can provide modeling frameworks for sediment transport. These data-driven approaches are also compared with physics-based ones that are classically used in the hydrological literature. Finally, the use of the developed models in a bottom-up strategy, which starts with the prediction/classification of a single pebble and then integrates the results into a forecast of the grain-size distribution of mobilized sediments, is discussed.

show abstract

Ensemble machine learning paradigms in hydrology: A review

Cited by 357 publications

References 218 publications

A new evolutionary time series model for streamflow forecasting in boreal lake-river systems

A new evolutionary time series model for streamflow forecasting in boreal lake-river systems

Continuous conditional generative adversarial networks for data-driven solutions of poroelasticity with heterogeneous material properties

A Comparison Between Machine Learning and Functional Geostatistics Approaches for Data-Driven Analyses of Sediment Transport in a Pre-Alpine Stream

Contact Info

Product

Resources

About