Artificial neural networks as emulators of process-based models to analyse bathing water quality in estuaries

García-Alba, Javier; Bárcena, Javier F.; Ugarteburu, Carlos; Garcı́a, Andrés

doi:10.1016/j.watres.2018.11.063

Cited by 103 publications

(46 citation statements)

References 53 publications

(38 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Future research should be directed to the validation of the model's biogeochemical component, and to the study of the biogeochemical processes in the TagusROFI domain. Moreover, it would be interesting to couple process-based-models (such as MOHID) with data-driven models [89] that employs machine learning techniques. This approach consists in feeding data-driven models with data from process-based-models that were already validated.…”

Section: Discussionmentioning

confidence: 99%

Validation of the 3D-MOHID Hydrodynamic Model for the Tagus Coastal Area

Pablo

Sobrinho

García

et al. 2019

Water

View full text Add to dashboard Cite

The hydrodynamics of the TagusROFI (Regions of Freshwater Influence) is affected by the coastal upwelling, the estuarine tidal flow, the thermohaline circulation that is modulated by the Tagus freshwater discharge, and by its complex bathymetry. The use of numerical models is the best way to explain the processes that characterize this region. These models are also crucial to answer important scientific and management questions. Nevertheless, the robustness of the products derived from models depend on their accuracy and therefore models must be validated to determine the uncertainty associated. Time and space variability of the driving forces and of bathymetry enhance flow complexity increasing validation difficulties, requiring continuous high-resolution data to describe flow and thermohaline horizontal and vertical variabilities. In the present work, to increase the precision and accuracy of the coastal processes simulations, the sub-systems coastal area and the Tagus estuary were integrated into a single domain, which considers higher resolution grids in both horizontal and vertical directions. The three-dimensiosal (3D)-MOHID Water model was validated for the TagusROFI by comparing statistically modelling results with in situ and satellite L4 data. Validation with a conductivity, temperature, and depth probe (CTD), an acoustic doppler current profiler (ADCP) and satellite data was performed for the first time. Validation against tidal gauges showed that the model is able to simulate tidal propagation inside the estuary with accuracy. A very good agreement between CTD data and surface sea water temperature (SST) and salinity simulations was observed. The validation of current direction and velocity from ADCP data also indicated a high model accuracy for these variables. Comparisons between model and satellite for SST also showed that the model produces realistic SSTs and upwelling events. Overall results showed that MOHID setup and parametrisations are well implemented for the TagusROFI domain. These results are even more important when a 3D model is used in simulations due to its complexity once it considers both horizontal and vertical discretization allowing a better representation of the heat and salinity fluxes in the water column. Moreover, the results achieved indicates that 3D-MOHID is robust enough to run in operational mode, including its forecast ability, fundamental to be used as a management tool.

show abstract

Section: Discussionmentioning

confidence: 99%

Validation of the 3D-MOHID Hydrodynamic Model for the Tagus Coastal Area

Pablo

Sobrinho

García

et al. 2019

Water

View full text Add to dashboard Cite

show abstract

“…Machine learning models have shown encouraging performances in a range of water resources applications, such as rainfallrunoff modelling (Minns and Hall, 1996;Khu et al, 2001;Babovic and Keijzer, 2002;Chiang et al, 2004), streamflow forecasting (Nourani et al, 2009;Meshgi et al, 2014Meshgi et al, , 2015Humphrey et al, 2016;Karimi et al, 2016), estimation of missing data (Elshorbagy et al, 2002), error correction (Sun et al, 2012), water quality modelling (Savic and Khu, 2005;Singh et al, 2011;García-Alba et al, 2019), sediment transport modelling (Babovic and Abbott, 1997;Afan et al, 2014;Safari and Mehr, 2018), reservoir management (Giuliani et al, 2015), prediction of climate variables (Dahamsheh and Aksoy, 2013;Ferreira et al, 2019), because of their potential to apprehend the noise complexity, non-linearity, non-stationarity and dynamism of data (Yaseen et al, 2015). Certainly, if we are only interested in better forecasting results then, the machine learning models might be the preferred choice over the conceptual or process-based models due to their better predictive capability.…”

Section: Machine Learning In Water Resourcesmentioning

confidence: 99%

“…CC BY 4.0 License. streamflow estimation (Nourani et al, 2009;Humphrey et al, 2016), water quality modelling (Singh et al, 2011;García-Alba et al, 2019), groundwater modelling (Nayak et al, 2006;Gholami et al, 2015), data assimilation Vojinovic et al, 2003), estimation of climate variables (Dahamsheh and Aksoy, 2013;Ferreira et al, 2019), flood and drought forecasting (Chang et al, 2014;Dehghani et al, 2014) and sediment transport modelling (Afan et al, 2014). Most of the above applications use supervised learning ANN models, such as Feed Forward Back Propagation (FFBP), Radial Basis Function Neural Network (RBFNN), and Generalized Regression Neural Network (GRNN).…”

Section: Artificial Neural Network (Ann)mentioning

confidence: 99%

Hydrologically Informed Machine Learning for Rainfall-Runoff Modelling: Towards Distributed Modelling

Herath¹,

Chadalawada²,

Babovic³

2020

Preprint

View full text Add to dashboard Cite

Abstract. Despite showing a great success of applications in many commercial fields, machine learning and data science models in general, show a limited use in scientific fields including hydrology. The approach is often criticized for lack of interpretability and physical consistency. This has led to the emergence of new paradigms, such as Theory Guided Data Science (TGDS) and physics informed machine learning. The motivation behind such approaches is to improve the physical meaningfulness of machine learning models by blending existing scientific knowledge with learning algorithms. Following the same principles, in our prior work (Chadalawada et al., 2020), a new model induction framework was founded on Genetic Programming (GP) namely Machine Learning Rainfall-Runoff Model Induction Toolkit (ML-RR-MI). ML-RR-MI is cable of developing fully-fledged lumped conceptual rainfall-runoff models for a watershed of interest using the building blocks of two flexible rainfall-runoff modelling frameworks (FUSE and SUPERFLEX). In this study, we extend ML-RR-MI towards inducing semi-distributed rainfall-runoff models. This effort is motivated by the desire to address the decreasing meaningfulness of lumped models which tend to particularly deteriorate within large catchments where the spatial heterogeneity of forcing variables and watershed properties are significant. Henceforth, our machine learning approach for rainfall-runoff modelling titled Machine Induction Knowledge-Augmented System Hydrologique Asiatique (MIKA-SHA) captures spatial variabilities and automatically induces rainfall-runoff models for the catchment of interest without any subjectivity in model selection. Currently, MIKA-SHA learns models utilizing the model building components of FUSE and SUPERFLEX. However, the proposed framework can be coupled with any internally coherent collection of building blocks. MIKA-SHA’s model induction capabilities have been tested on the Red Creek catchment near Vestry, Mississippi, United States. The resulted model architectures through MIKA-SHA are compatible with previously reported research findings and fieldwork insights of the watershed and are readily interpretable by hydrologists.

show abstract

“…CE was employed for two major reasons: (1) it is very commonly used in river and estuary model applications [52][53][54][55][56], and (2) [57] CE is also found to be the best objective function for reflecting the overall fit of a model output.…”

Section: Performance Metrics Of Numerical Modellingmentioning

confidence: 99%

Zonation of Positively Buoyant Jets Interacting with the Water-Free Surface Quantified by Physical and Numerical Modelling

García-Alba

Bárcena

Garcı́a

2020

Water

Self Cite

View full text Add to dashboard Cite

The evolution of positively buoyant jets was studied with non-intrusive techniques—Particle Image Velocimetry (PIV) and Laser Induce Fluorescence (LIF)—by analyzing four physical tests in their four characteristic zones: momentum dominant zone (jet-like), momentum to buoyancy transition zone (jet to plume), buoyancy dominant zone (plume-like), and lateral dispersion dominant zone. Four configurations were tested modifying the momentum and the buoyancy of the effluent through variations of flow discharge and the thermal gradient with the receiving water body, respectively. The physical model results were used to evaluate the performance of numerical models to describe such flows. Furthermore, a new method to delimitate the four characteristic zones of positively buoyant jets interacting with the water-free surface was proposed using the angle (α) shaped by the tangent of the centerline trajectory and the longitudinal axis. Physical model results showed that the dispersion of mass (concentrations) was always greater than the dispersion of energy (velocity) during the evolution of positively buoyant jets. The semiempirical models (CORJET and VISJET) underestimated the trajectory and overestimated the dilution of positively buoyant jets close to the impact zone with the water-free surface. The computational fluid dynamics (CFD) model (Open Field Operation And Manipulation model (OpenFOAM)) is able to reproduce the behavior of positively buoyant jets for all the proposed zones according to the physical results.

show abstract

Artificial neural networks as emulators of process-based models to analyse bathing water quality in estuaries

Cited by 103 publications

References 53 publications

Validation of the 3D-MOHID Hydrodynamic Model for the Tagus Coastal Area

Validation of the 3D-MOHID Hydrodynamic Model for the Tagus Coastal Area

Hydrologically Informed Machine Learning for Rainfall-Runoff Modelling: Towards Distributed Modelling

Zonation of Positively Buoyant Jets Interacting with the Water-Free Surface Quantified by Physical and Numerical Modelling

Contact Info

Product

Resources

About