Spatially Distributed Conceptual Hydrological Model Building: A Generic Top‐Down Approach Starting From Lumped Models

Tran, Quoc Quan; Niel, Jan De; Willems, Patrick

doi:10.1029/2018wr023566

Cited by 30 publications

(18 citation statements)

References 99 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…First, we did not pursue the highest spatial resolution. The graph neural network-based model GNRRM is developed at 4-km resolution in our case study, which is lower than other fully distributed rainfall-runoff modeling studies in 1km (Huang et al, 2019) or 250m (Tren andDe Niel, 2018) resolutions. However, the proposed method could be expanded to higher resolution theoretically with high-resolution rainfall data (Seo et al, 2019) and more computing resources or distributed computing (Agliamzanov et al, 2020).…”

Section: Discussionmentioning

confidence: 80%

Fully distributed rainfall-runoff modeling using spatial-temporal graph neural network

Xiang¹,

Demir²

2022

Preprint

View full text Add to dashboard Cite

Recent studies using latest deep learning algorithms such as LSTM (Long Short-Term Memory) have shown great promise in time-series modeling. There are many studies focusing on the watershed-scale rainfall-runoff modeling or streamflow forecasting, often considering a single watershed with limited generalization capabilities. To improve the model performance, several studies explored an integrated approach by decomposing a large watershed into multiple sub-watersheds with semi-distributed structure. In this study, we propose an innovative physics-informed fully-distributed rainfall-runoff model, NRM-Graph (Neural Runoff Model-Graph), using Graph Neural Networks (GNN) to make full use of spatial information including the flow direction and geographic data. Specifically, we applied a time-series model on each grid cell for its runoff production. The output of each grid cell is then aggregated by a GNN as the final runoff at the watershed outlet. The case study shows that our GNN based model successfully represents the spatial information in predictions. NRM-Graph network has shown less over-fitting and a significant improvement on the model performance compared to the baselines with spatial information. Our research further confirms the importance of spatially distributed hydrological information in rainfall-runoff modeling using deep learning, and we encourage researchers to incorporate more domain knowledge in modeling.

show abstract

Section: Discussionmentioning

confidence: 80%

Fully distributed rainfall-runoff modeling using spatial-temporal graph neural network

Xiang¹,

Demir²

2022

Preprint

View full text Add to dashboard Cite

show abstract

“…Same conclusion holds for the baseflow conditions at the internal stations, where all three coupled models improve the baseflow hydrographs compared to the simulation results of the distributed rainfall‐runoff model (Table ). Except for Meerhout station, where the measurement quality is limited (Tran et al ., ). Better performance at local gauging stations indicates the enhancement of the models over the distributed rainfall‐runoff model at simulating the spatially distributed groundwater discharge condition.…”

Section: Resultsmentioning

confidence: 97%

On the importance of river hydrodynamics in simulating groundwater levels and baseflows

Quan

Meert

Huysmans

et al. 2020

Hydrological Processes

Self Cite

View full text Add to dashboard Cite

Research to date affirmed the key role of stream-aquifer interactions in integrated water resources management. The importance of river hydrodynamics on the spatial and temporal behaviour of groundwater was, however, not yet fully investigated. In contrast to the common approach where topography-based estimates of riverbed elevation may lead to inappropriate discretization and constant river stages, this study couples a fully hydrodynamic and one-dimensional river model to a twodimensional catchment hydrological model. The surface and subsurface runoff, groundwater, and river components are integrated into a single modelling framework.The coupled model was applied to a medium sized catchment in Belgium with three model setups, in which the level of detail of representation of river hydrodynamics varies. Further model iterations were carried out for the most exhaustive setup to assess the importance of the bi-directional interactions between model components.Results show that higher details of river hydrodynamics help to improve the simulation of time-averaged groundwater levels. However, the impacts were not that clear for the time-varying groundwater levels. Moreover, visual and statistical model performance evaluation indicates a strong enhancement of the coupled models compared to the output from the hydrological model with respect to river discharge observations at catchment outlet and at internal stations. It also reveals the impact of river hydrodynamics on groundwater discharges when the most detailed setting delivered the highest performance among the three coupled models. K E Y W O R D Sgroundwater discharge, groundwater model, integrated modelling, river hydrodynamics

show abstract

“…For instance, Lobligeois et al (2014) showed that for basin exhibiting highly variable precipitation fields, the use of a distributed over a lumped model improved streamflow simulations. The benefits of distributed models is nevertheless wider than their ability to achieve better streamflow simulations (Tran et al, 2018). They provide spatially distributed information of key hydrological variables and processes.…”

Section: Discussionmentioning

confidence: 99%

Application of a High‐Resolution Distributed Hydrological Model on a U.S.‐Canada Transboundary Basin: Simulation of the Multiyear Mean AnnualHydrograph and 2011 Flood of theRichelieu River Basin

Lucas‐Picher

Arsenault

Poulin

et al. 2020

J Adv Model Earth Syst

View full text Add to dashboard Cite

During spring 2011, an extreme flood occurred along the Richelieu River located in southern Quebec, Canada. The Richelieu River is the last section of the complex Richelieu basin, which is composed of the large Lake Champlain located in a valley between two large mountains. Previous attempts in reproducing the Richelieu River flow relied on the use of simplified lumped models and showed mixed results. In order to prepare a tool to assess accurately the change of flood recurrences in the future, a state-of-the-art distributed hydrological model was applied over the Richelieu basin. The model setup comprises several novel methods and data sets such as a very high resolution river network, a modern calibration technique considering the net basin supply of Lake Champlain, a new optimization algorithm, and the use of an up-to-date meteorological data set to force the model. The results show that the hydrological model is able to satisfactorily reproduce the multiyear mean annual hydrograph and the 2011 flow time series when compared with the observed river flow and an estimation of the Lake Champlain net basin supply. Many factors, such as the quality of the meteorological forcing data, that are affected by the low density of the station network, the steep terrain, and the lake storage effect challenged the simulation of the river flow. Overall, the satisfactory validation of the hydrological model allows to move to the next step, which consists in assessing the impacts of climate change on the recurrence of Richelieu River floods.Plain Language Summary In order to study the 2011 Richelieu flood and prepare a tool capable of estimating the effects of climate change on the recurrence of floods, a hydrological model is applied over the Richelieu basin. The application of a distributed hydrological model is useful to simulate the flow of all the tributaries of the Richelieu basin. This new model setup stands out from past models due to its distribution in several hydrological units, its high-resolution river network, the calibration technique, and the high-resolution weather forcing data set used to drive the model. The model successfully reproduced the 2011 Richelieu River flood and the annual hydrograph. The simulation of the Richelieu flow was challenging due to the contrasted elevation of the Richelieu basin and the presence of the large Lake Champlain that acts as a reservoir and attenuates short-term fluctuations. Overall, the application was deemed satisfactory, and the tool is ready to assess the impacts of climate change on the recurrence of Richelieu River floods.

show abstract

Spatially Distributed Conceptual Hydrological Model Building: A Generic Top‐Down Approach Starting From Lumped Models

Cited by 30 publications

References 99 publications

Fully distributed rainfall-runoff modeling using spatial-temporal graph neural network

Fully distributed rainfall-runoff modeling using spatial-temporal graph neural network

On the importance of river hydrodynamics in simulating groundwater levels and baseflows

Application of a High‐Resolution Distributed Hydrological Model on a U.S.‐Canada Transboundary Basin: Simulation of the Multiyear Mean AnnualHydrograph and 2011 Flood of theRichelieu River Basin

Contact Info

Product

Resources

About