Keywords

1 Introduction

River sediment connectivity is defined as the connected transfer of solid material mediated by water from erosion areas in a river catchment to deposition areas. Human presence and activities have deeply affected natural river sediment connectivity, whether directly, e.g., reservoirs and barriers construction, channels reshaping, sand and gravel mining, or indirectly, e.g., catchment land-use change, climate change, or water withdrawal for industrial, domestic or agricultural uses. Alterations in sediment delivery may have long-lasting and profound impacts on river systems, including the decline in fluvial biodiversity and floodplain fertility, delta land loss, and bank instability. Given the interconnected and distributed nature of sediment connectivity in rivers, multiple alterations may have cumulative effects, and these impacts may be displaced in space and time from their causes [11, 17].

Characterizing basin-scale sediment connectivity is paramount to improving our capacity to quantify possible future alterations [9, 10]. However, modeling such processes at the large temporal and spatial scales necessary for basin-scale decision-making is often challenging, due to their complexity and the lack of available datasets for initialization and validation. The recent increase in the availability of large-scale hydro-geomorphological datasets from remote sensing [3] has led to the development of new numerical and conceptual sediment connectivity models. These tools exchange part of the local-scale accuracy found in more traditional, physic-based morphodynamic models to reconstruct sediment delivery and transport patterns over the entire river network and for longer timeframes [4, 6, 13, 16, 19].

In this chapter, we introduce D-CASCADE (Dynamic CAtchment Sediment Connectivity And DElivery), a dynamic, network-based sediment connectivity model for quantifying spatiotemporal patterns of sediment supply and delivery across a basin [15]. The model uses a 1-D graph structure to represent the river system, and traces the position of individual sediment transport processes called cascades over time across this network.

By considering the movement of multiple cascades and the interactions between each other and the river channels, D-CASCADE can effectively characterize network sediment connectivity, its dynamic evolution over time, and its response to multiple heterogeneous drivers of change.

Add-on components can be included in the framework to model local processes that the 1D structure of D-CASCADE cannot capture. These processes may include 2D and 3D simulations of the morphological response of river channels to sediment delivery alterations, e.g., channel accretion or erosion or floodplain interactions, or the effects of human infrastructures like reservoirs on the fluvial system.

D-CASCADE is adaptable to different research purposes, output requirements, and data availability. In particular, the model is designed to be integrated with multi-objectives, optimization-based planning, and management frameworks, to derive sediment connectivity indicators and identify trade-offs and synergies between conflicting objectives.

The chapter is organized as follows. Section 2 describes the D-CASCADE model structure, core principles and components. Section 3 showcases two applications of D-CASCADE: on the Bega river system in Australia [15], and on the 3S river system in South East Asia. The two applications differ widely in terms of spatial and temporal scales, network hydromorphological features, data availability, and research objectives. Section 4 draws some concluding remarks and highlights future research opportunities.

2 Model Structure

In D-CASCADE, sediment transport is modeled by routing and tracing cascades as they move through the network over time. Cascades are identified by the reach where they were initially delivered or stored, and the total sediment volume transported [m\(^3\)], subdivided into different grain size classes.

The structure, core components, and procedures of the D-CASCADE model are shown in Fig. 1. The model setup is divided into two phases: initialization and main D-CASCADE loop.

Fig. 1
figure 1

Figure from Tangi et al. [15]

D-CASCADE framework modelling steps.

Full size image

2.1 Initialization

The initialization phase contains all the operations necessary to apply D-CASCADE on a fluvial system. The model’s routing is performed on a river network characterized as a direct, acyclic graph composed of nodes and reaches (Fig. 2a). The reach is the core modeling unit of D-CASCADE and represents a stretch of the river network with homogeneous or semi-homogeneous hydro-geomorphological properties. The level of detail in the network partitioning can be tailored according to data availability and computational effectiveness.

Each reach must be characterized by a unique set of parameters representing hydro-geomorphological features, e.g., channel active width, gradient, bed composition and roughness, and hydrological conditions such as discharge, water depth, and velocity.

Depending on the knowledge of the case study system and the data availability, a feature can be characterized as static, dynamic, or modeled. Static features do not change over time and are defined once for each reach. Dynamic features may change in each timestep, and must be specified as a time-varying data series. Modeled features are automatically updated in each timestep by specific add-on components, so they just require initialization for the first timestep.

Sediment volumes routed through the network may be derived from material initialized in the reaches as a sediment deposit or as external, user-defined sediment contributions supplied to individual reaches in specific timesteps and at specific rates, e.g., to reproduce sediment delivery from the hillslopes.

During initialization, the user defines the sediment classes to be considered in the analysis, according to the characteristic of the river system of the river and the level of detail desired.

Fig. 2
figure 2

Figure from Tangi et al. [15]

An example of how the mobilized sediment volume is defined in a reach in a single model timestep. a shows the modeled network as a graph composed of reaches and nodes, b to e show the passages to define the mobilized sediment volume for reach 4. The colors of the tiers indicate the reach of provenance in the network. In b the model extracts the incoming cascades and deposit layer in the timestep, in c the deposit is divided into active and substrate layers, in d the model calculates the transport capacity for the sediment in the active layer, according to the layer Grain Size Distribution (GSD). Finally, in e the mobilized volume and new deposit layer are defined.

Full size image

2.2 Main D-CASCADE Loop

D-CASCADE model structure consists of two nested loops. The inner loop contains operations to determine cascade mobilization, deposition, and routing for each network reach, as well as all other operations modelled by add-on components. The outer loop repeats such operations for the entire simulation timeframe.

As shown in Fig. 1, the reach D-CASCADE loop is comprised of a series of operations repeated for each reach: (i) sediment mobilization, erosion and deposition, (ii) geomorphic features changes modelization, (iii) and sediment transport and delivery.

Figure 2 summarizes the steps necessary to define the mobilized sediment in each reach for each timestep. D-CASCADE traces the evolution over time of each reach deposit layer, conceptualized as a series of tiers stacked on top of each other, with each tier comprised of a single cascade. Incoming cascades, instead, can be comprised of material delivered by upstream reaches or external sediment inputs defined by the user (Fig. 2b). The active layer thickness in Fig. 2c must be defined by the user and may depend on the type of material composing the uppermost layer of the deposit.

The volume mobilized in the reach in Fig. 2d depends on the sediment transport capacity, i.e., the amount of energy available in a reach for the transport of sediment of a particular grain class, determined by the hydromorphological characteristics of the reach and the composition of the sediment volume affected. The transport capacity is measured via empirical equations, selected according to the river’s hydromorphological features, the sediment classes considered, and the type of transport analyzed (total, suspended, or bedload transport).

Changes in geomorphic features can be modeled via selected add-on components to quantify changes in reach features between timesteps. These add-ons can include 2D or 3D morphodynamical models, simplified models calibrated on observed historic channel evolution, or conceptual models based on literature knowledge and field observations. Two-directional interactions between add-on components and D-CASCADE allow local changes to influence basin-scale sediment connectivity and vice-versa.

Finally, the destination of the mobilized volumes is determined according to the characteristic velocity of sediment transport, estimated via empirical equations [6]. According to the formula selected, sediments of different sizes may be transported at different velocities.

3 Case Studies

3.1 Bega River System

The D-CASCADE model was first applied to the Bega river network, in NSW, Australia (drainage area: 1930 km\(^2\)), to reconstruct the historical geomorphological response of the river network to anthropic pressures brought by European settlements (ES) in terms of both sediment connectivity disruption and the consequent river changes [15].

The geomorphic evolution of the Bega river catchment is well-documented in literature [5, 7], and data are available on sediment budgets, hydrological variability and magnitude, timing, and location of historical drivers of change over the last two centuries. These data are used to define the model boundary conditions and provide validation for the model outputs. Uncertainty in the reconstruction of the hydrological record are accounted for by simulating multiple scenarios.

D-CASCADE simulations are set up considering a temporal horizon from the first ES to the present days (1850-2020). The river network features pre-ES are defined from literature reconstructions. The most significant drivers of change: deforestation, swamp drainage, and re-forestation, are introduced in the simulation in their historical chronological sequence. Add-on components are added to reproduce two of the most prominent morphological processes observed historically: channel width expansion from bank erosion and overbank flooding. Several indicators are identified to compare model outputs to the observed river dynamics.

D-CASCADE simulates scenarios of morphological evolution in the Bega river network over two centuries which match the historical observations with suitable accuracy. Observed changes are reconstructed in the model with the correct timing and location, including lowland channel expansion, valley fill incision, and sediment slug mobilization and delivery.

Finally, multiple forecasting simulations have been conducted to test the model performances for the estimation of the future morphological evolution of the system. Sediment connectivity patterns are modeled with D-CASCADE under different future hydrological and land-use change scenarios. Simulations suggest that vegetation removal, e.g., from wildfires or human interventions, may trigger a new phase of sediment mobilization in the lowland, most likely accompanied by new channel adjustments.

3.2 3S River System

The river system composed of the Se Kong, Se San, and Sre Pok rivers, often referred to as 3S, is a large tributary network (drainage area: 82,500 km\(^2\)) of the lower Mekong river. In recent years, the 3S river basin has been interested in large-scale dam development projects aiming to exploit its hydropower potential, with approximately 40 dams are either planned or already built [12]. The construction of reservoirs, and the subsequent trapping of sediments in the impounded area, threatens to significantly reduce sediment delivery to the Mekong Delta, increasing the risk of delta subsidence and land loss [11].

D-CASCADE was applied to the 3S river system to assess the cumulative impact of multiple dams on the river sediment delivery and evaluate the benefits of coordinated reservoir sediment management techniques for sediment connectivity restoration.

As the 3S system is a data-scarce environment, we employed large-scale datasets, e.g., satellite imagery and digital elevation models to extract the river network and characterize the hydro-geomorphological features of its reaches. Different catchment sediment supply scenarios are tested using D-CASCADE and validated using literature estimates on outlet sediment delivery [8], to extract credible boundary conditions for future simulations with reservoirs.

Add-on components are developed and included in the simulations to integrate a dynamic representation of reservoir water and sediment management in D-CASCADE. These components simulate processes such as the alteration of morphological features in the flooded reaches inside the impoundment, the changes in water and sediment storage due to different management strategies and the effects of dam release on downstream hydrology.

Fig. 3
figure 3

Results for the application of D-CASCADE on the 3S system for the cumulative impact assessment of multiple dam development projects. a Dams location, impoundment area at full supply level for three dam development portfolios. b Average annual sediment yield to the Mekong river for the three dam development scenarios, partitioned by river provenance and grain size classes

Full size image

We evaluated the impact of three different dam development portfolios in the lower 3S river (Fig. 3a): two with large reservoirs, either existing or planned (LSS2Only and FullDam) [14], and one with an alternative configuration comprised of three smaller reservoirs (FullDam_Alt) [1].

The results show a significant reduction in sediment transport for all dam development portfolios. Portfolios with large reservoirs (LSS2Only and FullDam) would reduce the average annual sediment delivery to the Mekong by 50%. The alternative configurations FullDam_Alt, due to the lower trapping efficiency of its reservoir, would lead to a less significant but still noticeable reduction (26% of the total load) (Fig. 3b).

Sediment management strategies such as drawdown sediment flushing can be applied to reduce the impact of reservoirs on sediment connectivity [9]. During drawdown flushing, the reservoir is emptied via low-level gates, and river-like conditions are established in the impoundment for a fixed number of days, during which the inflow is used to scour deposited sediments and transport them downstream [2, 18] (Fig. 4a). These operations have the added benefit of removing sediment from the impoundment of the reservoir, thus increasing its water storage capacity and decreasing maintenance costs [2]. However, they are only feasible in smaller reservoirs like the ones in the FullDam_alt portfolio.

Fig. 4
figure 4

a Visualization of a successful drawdown flushing operation as modelled in D-CASCADE. The figure refers to a 25-days flushing operation in the LSS3 dam. b Average annual sediment yield for the FullDam_alt portfolio, with no flushing, or annual and biannual flushing operations

Full size image

To assess the potential benefits of these techniques for sediment connectivity restoration, we included drawdown sediment flushing operations in the management of the dams in the FullDam_alt portfolio via a specific add-on component. Sediment flushing is performed at the beginning of the wet season (July-August, as suggested from literature [1, 18, 19]), as reservoirs are typically at their lowest level after the dry season, and monsoon-induced floods increase both sediment scouring potential during flushing and refill velocity.

According to the results, drawdown sediment flushing operations repeated every one or two years may increase network sediment yield by 4%–10% respectively (Fig. 4b). Flushing implementation also removes a fraction of the reservoirs sediment deposits, although the benefits are contained (7–11% reduction in the total sediment storage for the upstream dams).

4 Conclusion

In this chapter, we contribute the large-scale, dynamic sediment connectivity model D-CASCADE. D-CASCADE conceptualizes sediment transport as composed of individual sediment processes routed across a graph-like network. This modeling framework returns disaggregated, spatiotemporal information on sediment transport in each section of the river, together with data on sediment provenance, composition, and properties.

The model is flexible and data-parsimonious, making it a versatile tool for many applications with different scales, data availability and objectives. Add-on components may be added to D-CASCADE to include the representation of local-scale morphological dynamics inside the large-scale connectivity modelling framework.

The application of D-CASCADE on the Bega river case study showcases the model’s potential for basin-scale geomorphological analysis to reconstruct historical sediment budget trajectories, increase our knowledge of the river sediment connectivity patterns, and forecast their future evolutions.

The study on the 3S system marks the first application of a dynamic, basin-scale sediment transport model to quantify the cumulative impact of a series of dams on river sediment delivery, and evaluate how sediment connectivity disruption could be mitigated by employing coordinated sediment management strategies.

Reservoir sediment management operations like drawdown sediment flushing come with noticeable costs, for example, due to losses in water storage and interruptions in hydropower production. Thus, these operations should be designed strategically and synchronized between multiple reservoirs to maximize benefits and reduce costs [19]. Future research will aim to integrate D-CASCADE in a multi-objective decision-making framework to extract optimal designs of water and reservoir sediment management strategies under competing objectives, e.g., hydropower generation, environmental conservation and sediment connectivity restoration.