Advancing integrated river basin management and flood forecasting in the Cagne catchment: a combined approach using deterministic distributed models

Integrated river basin management (IRBM) is a holistic approach to address complex water resource challenges that arise from the interdependence of hydrological, ecological, and socio-economic systems within a river basin. IRBM aims to promote the coordinated development and management of water resources in a sustainable and equitable manner, considering the diverse needs of various stakeholders and the natural environment. One of the key aspects of IRBM is flood forecasting, which plays a vital role in mitigating the adverse impacts of floods on human lives, property, and the environment.

The Cagne catchment, with an area of 96 km², is in the French Mediterranean region and exhibits the typical Mediterranean hydrological conditions that are observed in most of coastal catchments along the North Mediterranean coast. The Cagne River originates at Coursegoules, at an altitude of 1,535 m, and flows through the city of Cagnes-sur-Mer before entering the sea. The catchment is steep in its upstream part and then can be affected by intense runoff processes. The catchment's hydrology is primarily influenced by autumn precipitations, leading to increased discharge from October to January (Folton 2020) as in most of Mediterranean regions. Over past decades, under the fast urban development, the catchment vulnerability has strongly increased within the downstream area, along the coast. To better anticipate extreme situations as those observed in October 2015 (Charpentier-Noyer et al. 2020), extensive topographic, climatic, and hydrological data have been collected within the catchment in the past decade, enabling the development of hydrological and hydrodynamic models for flood prediction. The initial approach based on multiple tools has quickly evolved towards a Decision Support System (DSS) that can provide an efficient support for daily management of water resources and also helps to forecast floods during extreme events. The main challenge is then to identify and implement the simulation models best suited to the catchment's size and hydrological conditions, and subsequently integrate them within a DSS environment (Gourbesville et al. 2022).

Managing the water resources within this catchment requires an in-depth understanding of the hydrogeological characteristics and the interactions between surface water and groundwater systems. MIKE SHE (Abbott et al. 1986), MIKE 21 FM (DHI 2017), and FEFLOW (Diersch 1999) are advanced deterministic distributed hydrogeological and hydraulic modelling tools that have been widely used for simulating various aspects of water resources management, including flood forecasting. MIKE SHE is an integrated hydrological model that simulates surface water and groundwater interactions, while MIKE 21 FM is a flexible mesh model for simulating two-dimensional shallow water flows in rivers, estuaries, and coastal areas. FEFLOW, on the other hand, is a finite-element subsurface flow system used for groundwater modelling (DHI 2017). Over the past decades, numerous studies have been conducted utilizing these three numerical modelling approaches. Zhang et al. (2008) evaluated the ability of MIKE SHE, to simulate basin runoff in a water-limited region, demonstrating its usefulness in understanding rainfall-runoff mechanisms while highlighting the need for higher-resolution data for regional applications. Sahoo et al. (2006) apply the MIKE SHE to the Manoa–Palolo stream system in Hawaii to investigate watershed response to storm events and show that a well-calibrated model with a single-valued hydraulic conductivity can produce consistent results, with rainfall distribution, Manning's roughness coefficient, and hydraulic conductivities playing significant roles in determining flood peak characteristics. Nigussie & Altunkaynak (2019) use MIKE 21 FM model to assess the impact of different land-use policy scenarios on flood risk in Ayamama Watershed, Turkey, concluding that unrestricted urbanization significantly increases flood risk and recommending limited development and improved drainage systems to mitigate this risk. Using MIKE 11 and FEFLOW software, a coupled surface water-groundwater interface was developed by Monninkhoff & Li (2009) to study interactions in river branches, polders, and wetlands, enabling the estimation of groundwater storage influence and the development of an optimized flooding strategy for the river Elbe. Zavattero et al. (2016) present a methodology for building a MIKE 21 FM model to understand sediment movement and analyse impacts on river-aquifer exchange areas, ultimately aiding in simulating the behaviour of physical processes within the riverbed. Using FEFLOW software, a groundwater model for the Zayandeh Rud Basin was developed by Sklorz et al. (2017), revealing a drawdown of over 60 m in 15 years and highlighting the need for reduced groundwater extraction to prevent further decline in groundwater levels and increased surface water leakage. Du et al. (2019) present a validated integrated hydrodynamic model for a Mediterranean alluvial aquifer in the lower Var valley, which effectively estimates undocumented pumping volumes and is ready to be used as a decision support tool for local water supply managers in assessing pollution scenarios and hazard management. These studies have demonstrated the viability of individual models. However, integrating these models could yield even better results and performance. The integration of these models enables a comprehensive understanding of the hydrological processes in the Cagne catchment and supports the development of effective flood forecasting and IRBM strategies.

The issue related to flood forecasting in the Cagne catchment as in many coastal Mediterranean catchments is the need to predict and mitigate the adverse impacts of flood events on human lives, property, and the environment during intense events mainly taking place at the beginning of the autumn when the sea temperature exceed 24 °C. Due to its steep topography, impermeable soils, and the occurrence of intense rainfall events, the Cagne catchment is particularly susceptible to flooding. Additionally, the complex interactions between surface water and groundwater systems further complicate the hydrological processes in the catchment. Therefore, there is a pressing need for an integrated modelling approach that can accurately simulate these complex processes and provide reliable flood forecasts for effective flood risk management.

An integrated modelling approach that combines the strengths of advanced deterministic hydrogeological and hydraulic models, such as MIKE SHE, MIKE 21 FM, and FEFLOW, offers a comprehensive understanding of the hydrological processes in the Cagne catchment. Within a context of limited available hydrological data, the assumption is that this approach will allow the development of accurate and timely flood forecasts, which are crucial for implementing effective preparedness and response measures to minimize the vulnerability of communities and ecosystems to flood hazards. Moreover, the integration of these models will also support the broader objectives of IRBM, such as sustainable water resources management, ecosystem protection, and climate change adaptation.

Water information technology, especially modelling tools (hydrological, hydraulic, and geohydrological modelling), have been widely used in the integrated water management field to analyse the historical hydrological events and to make a forecast of potential flood events. To provide efficient support to the field decision-makers, these models should be accurate and then will play an important role in the DSS. In this study, the distributed hydrological model MIKE SHE, the hydraulic model MIKE 21 FM and geohydrological model FEFLOW (all developed by DHI) were selected for the local DSS operated by the operating teams.

Hydrological models can be primarily classified into two types – lumped hydrological models and distributed deterministic hydrological models – which can be integrated within a DSS. Lumped models simplify the distributed characteristics of a catchment homogeneously in simulations. Although parameters in these models may possess physical meaning, they are often determined through calibration rather than being derived from catchment physical attributes. Consequently, lumped models lack the ability to accurately represent hydrological processes over space and can only analyse catchment responses at the outflow without considering specific sub-basin responses. On the other hand, distributed models represent catchments by dividing the entire research area into several smaller areas. These models maintain physical information with a certain level of accuracy and consider the spatial nature of hydrological variables, including soil, land use, and slopes. The primary challenge of distributed models is the scarcity of high-quality input data. However, advancements in data collection technology have facilitated systematic data collection for topography (including high-resolution data obtained through LIDAR solutions), climate, and hydrological variables. This has allowed for a comprehensive understanding of the main hydrological processes within the Cagne catchment and has enabled the implementation of a distributed approach of MIKE SHE for hydrological analysis.

The application of diffusive wave approximation allows the flow depth to have significant variation between neighbouring calculation grids and backwater conditions to be simulated.

The hydrologic model generates flood hydrographs, while the flooding process in the downstream area is simulated using a high-resolution 2D deterministic hydraulic model. This hydraulic model can represent discharges within the riverbed and simulate the extent of inundation processes. To accurately depict the complex hydraulic processes, the MIKE 21 FM hydrodynamic two-dimensional flexible mesh model was chosen. This model can represent a broad range of physical and chemical processes, including hydrodynamic behaviour, pollutant transport, and morphological changes (Diersch 1999).

MIKE 21 FM utilizes a flexible mesh approach, where the nonorthogonal triangular construction of the mesh provides resolution flexibility across the model area, as compared to a rigid raster grid. This flexibility allows to focus on specific sections of the model. For instance, particular attention may be needed to accurately represent a small region with complex elevation conditions or to provide a realistic portrayal of buildings crucial to the hydraulic situation. Conversely, reduced resolution can be implemented in less important regions, resulting in fewer mesh components and faster processing times. In essence, the flexible mesh enables to allocate computing power from homogeneous low-importance areas to heterogeneous high-importance areas (Durand et al. 1993).

The entire water cycle encompasses not only surface water but also groundwater; therefore, it is crucial to consider the role of groundwater in integrated water management. To address this aspect, the hydrogeological model FEFLOW is employed to simulate the groundwater cycle in the study area. FEFLOW is a powerful numerical modelling tool that accounts for various subsurface flow and transport processes, enabling comprehensive groundwater analysis and management.

S_s is the specific storage (m⁻¹); B is the thickness of the unconfined aquifer (m); S_y is the specific yield also called effective porosity(dimensionless); is the hydraulic head of groundwater (m): for unconfined aquifer, its value is equal to the piezometric level; q is the groundwater Darcy's velocity or specific discharge (m/s); K is the tensor of hydraulic conductivity of the porous medium (m/s); ε is the total porosity (dimensionless); Q is the specific mass supply per unit time per unit depth (s⁻¹); q_ex and q_r are, respectively, the increased mass per unit area due to the river-aquifer exchange and groundwater recharge (m/s); and q_w is the groundwater abstraction per unit area (m/s).

The model domain is defined by the boundaries of permeable layers and geological faults present in the study area (Figure 4). This delineation ensures that the model accurately represents the geohydrological characteristics of the region, allowing for precise groundwater flow simulations. The downstream boundary is represented by the coastline at the Mediterranean Sea, where the influence of seawater on the groundwater system is considered.

Precipitations play a vital role in recharging the unconfined aquifer in the valley. The hydrogeological model accounts for this recharge process, simulating the infiltration of water through the soil and its subsequent movement through the subsurface. This information is essential for understanding the interactions between surface water and groundwater and for effectively managing water resources in the catchment.

The model was set up with a resolution of 5 m in the river area and 100 m in the other areas to provide a detailed representation of the hydrological processes involved. The vertical extent of the model consists of five distinct layers: alluvium, Pliocene conglomerate, Pliocene marls, Jurassic limestone, and karst. Each layer possesses unique hydrogeological properties that influence the groundwater flow patterns and aquifer dynamics in the region. By incorporating these layers into the FEFLOW model, a more accurate representation of the subsurface geology and groundwater behaviour can be achieved, allowing for better predictions and management of the region's groundwater resources.

In conclusion, the integration of the hydrogeological model into the study of the Cagne catchment's water cycle allows for a more comprehensive understanding of groundwater dynamics and their interactions with surface water. This information is crucial for the development of effective integrated water management strategies and the sustainable use of water resources in the catchment area. Future research may benefit from the collection of additional measurement data to refine model parameters and improve the accuracy of the groundwater simulations. Meanwhile, since the models have been successfully validated in the Cagne catchment, this integrated modelling approach can be replicated in other catchments exhibiting similar hydrological characteristics, thereby expanding its applicability and impact.

The modelling procedure commences with a simulation of the catchment hydrological cycle using MIKE SHE. The application of MIKE SHE in the Cagne catchment can provide accurate representations of major hydrological and geohydrological characteristics at any location within the catchment. This detailed information supports the use of MIKE 21 FM for more in-depth simulations of the Cagnes-sur-Mer urban area. By leveraging high-resolution simulations on daily and hourly time intervals, a comprehensive understanding of the hydrodynamic processes in the urban area can be obtained, allowing for more effective water management and flood mitigation strategies.

The analysis reveals that the model can capture the trend with a topographic resolution of 30 m; however, there is an abrupt drop following each peak, and certain areas do not align with the observed data, even considering the inherent uncertainties. The simulated data with a resolution of 20 m exhibit greater accuracy. The root mean square error (RMSE) for the 30 m resolution result is 0.107, while the RMSE for the 20 m resolution result is 0.119, representing a 10% increase in accuracy.

In this study, an integrated modelling approach using hydrological (MIKE SHE), hydraulic (MIKE 21 FM), and hydrogeological (FEFLOW) models was employed for flood forecasting in the Cagne catchment, France. The main objective of the project was to demonstrate the added value of the deterministic models' combination for an effective and sustainable river basin management. The results showed that the integrated modelling framework can effectively simulated the global hydrological cycle, the groundwater levels, and the flood events as well within the Cagne catchment. The findings highlight the importance of using a comprehensive approach to address complex hydrological processes with deterministic tools that can be mobilized for daily river basin management and extreme events forecasting. The obtained results can strongly support the decision-making process within the Cagne catchment. By mobilizing an integrated modelling approach, authorities and stakeholders can develop effective flood management strategies, identify potential areas of concern, and adapt to future changes in climate conditions and land uses. Future efforts could focus on refining model calibration, incorporating additional data sources, and extending the modelling domain to consider regional-scale processes by linking with neighbour decision support system (DSSs) such as AquaVar. These improvements would enhance the understanding of hydrological processes and flood risks in the Cagne catchment and contribute to more effective river basin management. In conclusion, the integrated modelling approach using MIKE SHE, MIKE 21 FM and FEFLOW has demonstrated its potential for flood forecasting and IRBM in the Cagne catchment. The study's findings contribute to the development of more effective and sustainable management strategies, ultimately benefiting local communities, ecosystems, and economies.

The work benefited from the data provided by the Métropole Nice Côte d'Azur, Conseil Départemental 06 and Météo France. This work is funded by China Scholarship Council.

Data cannot be made publicly available; readers should contact the corresponding author for details.

The authors declare there is no conflict.