This study aims to develop a probabilistic model to quantify the reliability of estimating riverbed elevations due to the uncertainties in the runoff and sediment-related factors (named PM_MBEE_1D); the above uncertainties are quantified by reproducing a considerable number of runoff-related and sediment-related factors via the multivariate Monte Carlo simulation approach. Using a sizeable number of simulated uncertainty factors, the proposed PM_MBEE_1D model is developed by coupling the rainfall–runoff model (SAC-SMA) and 1D sediment transport simulation model (CCHE1D) with the uncertainty/risk analysis advanced first-order second-moment (AFOSM) method as well as the logistic regression analysis. Validated by the historical data in the Jhuosdhuei River watershed, the proposed PM_MBEE_1D model could efficiently and successfully capture the spatial and temporal changes in the estimated riverbed elevations (i.e., scouring and siltation) due to the uncertainties in the river runoff and sediment with a high accuracy (nearly 0.983). Also, using the proposed PM_MBEE_1D model with given runoff and sediment factors under a desired reliability, the probabilistic-based riverbed elevations could accordingly be estimated as a reference to watershed treatment and management plan.

  • The proposed PM_MBEE_1D could quantify the reliability of the estimated riverbed elevations at various cross-sections due to the uncertainties in the river runoff and sediment discharges.

  • The proposed PM_MBEE_1D could provide the riverbed elevation estimates under a desired likelihood.

  • The resulting big data of the simulations of the rainfall-induced movable-bed elevations could be applied in AI model training.

Recently, sediment transport in the river frequently causes the severe aggradation and erosion of the riverbed elevation to result in flood-derived disaster (e.g., Neuhold et al. 2009; Kim & Julien 2018; Paul & Biswas 2019; Pfeiffer et al. 2019; Cheor & Tachikawa 2020; Chen et al. 2022; Liu et al. 2022, 2023); however, over-developing the alluvial river possibly destroys the bed stability, lowering the safety of the hydraulic structures and storage of the reservoirs (e.g., Chang et al. 1993; Teraguchi et al. 2011; Helal et al. 2013; Liu et al. 2016, 2023; Song et al. 2020); whereas, over-sedimentation in the riverbed might markedly raise the water level to trigger the overtopping risk of the embankment during rainfall-induced flooding. Accordingly, the change in the riverbed due to the sediment transport significantly results in the failure risk of flood-mitigation performance of the hydraulic structures with severe damage to people's lives and property. Therefore, the riverbed's stability analysis should be an essential issue in alluvial river management and improvement plans.

In the past, the changes in the riverbed were mainly quantified and analyzed based on the historical measurements at the various cross-sections along a river (e.g., Schappi et al. 2010; Jagallah et al. 2019; Maddahi & Rahimpour 2023); however, the above measurements are frequently recorded through a time-consuming and expensive process; it is probably challenging to reflect the effect of a heavy rainfall event on the riverbed, thereby inducing an underestimation in flood hazard management due to the lack of details on sediment composition (e.g., Kampf et al. 2015; Riahi-Madvar & Seifi 2018; Liu et al. 2021). Also, they are generally taken at the specific cross-sections in the main river channel, meaning that it is difficult to understand the spatial change in the whole watershed. Thereby, with the enhancement of computation power, the numerical models could simulate the changes in the movable riverbed elevations via the well-known sediment transport approaches based on mass conservation under different hydrological and topographical conditions The physical-based mobile-bed elevation simulation models are generally developed by collaborating the hydrodynamic numerical model with the empirical sediment transport equations between the sediment discharge and river runoff (e.g., Mustafa et al. 2008; Chao et al. 2018). For example, HEC-6 (1993) is a well-known numerical mode for the long-term aggradation and erosion of the river bed by integrating the backwater routing module with the bed-load computation approach for steady flow. Nonetheless, as a result of responding to the effect of the temporal varying trend of the river runoff and downstream boundary (i.e., tide depth), the unsteady flow modules are used in the erosion and deposition simulation of the riverbed, e.g., MIKE11 (1992), SRH-1D (Huang & Greimann 2007), and CCHE1D (Wu et al. 2004a, 2004b). Additionally, artificial intelligence (AI) techniques are widely applied in the relevant hydrological and hydraulic analysis; hence, a number of well-known AI methods are used to develop the data-derived sediment transport models. For example, Riahi-Madvar & Seifi (2018) proposed an intelligent model for predicting the bed load transport (i.e., sediment discharge) by coupling the artificial neural network (ANN) with adaptation neuro-fuzzy inference system (ANFIS) to proceed with the model training and validation, a significant number of the corresponding sediment discharges to the combination of 6 model inputs were achieved via the Monte Carlo simulation approach.

Due to the occurrences of extreme hydrological events and climate change, the variations in the hydrological and topographical data probably cause uncertainties in the calibrated parameters of the sediment transport numerical models and corresponding simulations (e.g., Neuhold et al. 2009; Lin et al. 2014; Ann & Steihschneider 2017; Beckers et al. 2018). For instance, Ann & Steihschneider (2017) proposed a time-varying relationship between the runoff and sediment discharge to estimate the suspend-sediment concentrations model by means of the dynamic regression analysis used in response to climate change. Moreover, a group of hydrodynamic numerical models were presented for the sediment transport simulation; however, their accuracy and reliability are likely to be impacted attributed to the limitation of the model concepts and relevant measurements, including the assumptions and empirical equations of interest (e.g., Chang et al. 1993; Ruark et al. 2011). To cite an instance, Chang et al. (1993) employed the first-order method with Latin hypercubic sampling technique to quantify the sensitivity in the parameters of the sediment transport simulation model (HEC2-SR) to quantify their uncertainties, indicating that the river-channel roughness coefficients make a more significant contribution to the estimation of the water surface and sediment discharge with positive and negative impact, respectively.

Despite numerous investigations proceeding with the uncertainty analysis regarding the movable riverbed elevation simulation in a river, they focused on either the variation in the hydrological and morphological factors (e.g., precipitation and runoff) (Lin et al. 2014; Ann & Steihschneider 2017; Riahi-Madvar & Seifi 2018; Pfeiffer et al. 2019) or the model parameters adopted in the process of simulating the change in the riverbed elevations triggered by the rainfall-induced runoff, including the rainfall–runoff modeling, river routing and sediment transport (Chang et al. 1993; Ruark et al. 2011; Sabatine et al. 2015; Beckers et al. 2018). Furthermore, it is noted that the time series of the river runoff and sediment discharge are used in the parameter calibration and application of the sediment transport model; the resulting riverbed elevation estimates at the various cross-sections during a rainstorm should be regarded as the spatial and temporal variables. Therefore, this study aims to develop a one-dimension (1D) probabilistic-based model for estimating the movable riverbed elevations at various cross-sections along a river under consideration of the uncertainties in the hydrological data and model parameters in time and space, named the PM_MBEE_1D model. The proposed PM_MBEE_1D model is anticipated to quantify the resulting reliability of the estimated riverbed's elevations at the specific cross-sections from the different conditions of the hydrological data, sediment features, and topographical factors. Additionally, the estimated riverbed elevations with a desired reliability by the proposed PM_MBEE_1D model could be considered as a reference to the flood-induced hazard assessment and the watershed treatment and management plan.

Since this study aims to develop a probabilistic-based riverbed elevation estimation model in response to the scouring and siltation effect, the Zhuoshui River watershed located in central Taiwan with magnificent supplied suspended load (54 Mt year−1) to the ocean (Figure 1) is selected as the study area. Zhuoshui River is the longest river in Taiwan, of which the drainage area is nearly 3,156.90 km2 with two main branches (Chenyoulan and Chingshui creeks). Moreover, the annual average rainfall is about 2,350 mm, and the annual sediment transport from the Zhuoshui River watershed approximates 63.87 Mt with 20% of the total amount of 322.76 Mt in the entire Taiwan. In addition, according to Figure 1, it is known that nine automatic raingauges, three water-level stations (Bashichi Bridge, Nei-mao-pu, and Longmen Bridge) and six sediment discharge stations are set up within the Zhoushui River watershed; the associated runoff-related and sediment-related historical data, including the gauged rainstorms, river runoff, river-channel roughness coefficients, tide depths, sediment discharge and bed material load, are adopted as the study data. The study data used in the development of the proposed PM_MBEE_1D model is briefly introduced below:
Figure 1

Map of the study area (Zhoushui River watershed) and locations of the rainfall gauges (RG), water-level gauges (WG), and sediment discharge gauges (SG).

Figure 1

Map of the study area (Zhoushui River watershed) and locations of the rainfall gauges (RG), water-level gauges (WG), and sediment discharge gauges (SG).

Close modal

Rainfall–runoff data

Among three water-level stations, the Baoshi Bridge gauge is located at the upstream in the Zhuoshui River watershed, meaning that its rainfall–runoff characteristics are poorly impacted by the lateral runoff coming from the branches. Thus, the observed rainfall and induced runoff data could be utilized in the parameter calibration of the rainfall–runoff model (SAC-SMA). Figure 2 presents the observed areal average rainfall and induced river runoff at the Baoshi Bridge water-level stations regarding typhoon events; it can be seen that the rainfall depths range from 107 to 760 mm for 120 h; the corresponding maximum rainfall intensity nearly varies between 21 and 43 mm/h. Thus, the induced peak discharges are located within 240 and 4,500 cm, indicating that ten typhoon events exhibit significant variations in the rainfall-induced floods. Therefore, the resulting calibrated model parameters could reasonably respond to changing the rainfall–runoff characteristics within the study area. Specifically, by observing Figure 2, the distribution of the rainfall in time exhibits a significant variation; for example, the maximum rainfall intensities regarding the EV1 (Typhoon Krosa) and EV10 (Typhoon Dujuan) occur in the middle time step during 120 h which can be treated as the Central-type storm pattern; whereas, as for the EV7 (Typhoon Trami) and EV2 (Typhoon Sinlaku), the maximum rainfall intensity could be nearly measured at the beginning and end of the events which could be defined as the advanced-type and delayed-type ones, respectively. With respect to EV3 (Typhoon Morakot), the rainfall uniformly distributes in a time period (i.e., rainfall duration) to be treated as the uniform-type one; a similar conclusion could be found in the remaining raingauges; hence, the historical typhoon events selected show a marked change in time and space; accordingly, they could be used in the simulation of rainstorms in respond to the impact to the variations of rainfall in time and space on the estimation of the riverbed elevation within the study area.
Figure 2

Historical hyetograph (rainfall) and corresponding runoff hydrographs (runoff_obs) of rainstorms at the upstream watershed for the calibration of the model parameters. (1) EV1: Typhoon Krosa (2007/10/4 00:00-2007/10/9 23:00), (2) EV2: Typhoon Sinlaku (2008/9/11 01:00-2008/9/17 00:00), (3) EV3: Typhoon Morakot (2017/8/7 01:00-2017/8/12 00:00), (4) EV4:Typhoon Fanapi (2010/9/18 12:00-2010/9/18 12:00), (5) EV5: Typhoon Saola (2012/7/31 07:00-2012/7/31 07:00), (6) EV6: Typhoon Soulik (2013/7/12 13:00-2013/7/14 22:00), (7) EV7: Typhoon Trami (2017/8/21 00:00-2017/8/25 23:00), (8) EV8: Typhoon Usagi (2017/9/21 00:00-2017/9/25 23:00), (9) EV9: Typhoon Matmo (2014/7/22 12:00-2014/7/25 11:00), (10) EV10: Typhoon Dujuan (2017/9/28 00:00-2017/10/2 23:00).

Figure 2

Historical hyetograph (rainfall) and corresponding runoff hydrographs (runoff_obs) of rainstorms at the upstream watershed for the calibration of the model parameters. (1) EV1: Typhoon Krosa (2007/10/4 00:00-2007/10/9 23:00), (2) EV2: Typhoon Sinlaku (2008/9/11 01:00-2008/9/17 00:00), (3) EV3: Typhoon Morakot (2017/8/7 01:00-2017/8/12 00:00), (4) EV4:Typhoon Fanapi (2010/9/18 12:00-2010/9/18 12:00), (5) EV5: Typhoon Saola (2012/7/31 07:00-2012/7/31 07:00), (6) EV6: Typhoon Soulik (2013/7/12 13:00-2013/7/14 22:00), (7) EV7: Typhoon Trami (2017/8/21 00:00-2017/8/25 23:00), (8) EV8: Typhoon Usagi (2017/9/21 00:00-2017/9/25 23:00), (9) EV9: Typhoon Matmo (2014/7/22 12:00-2014/7/25 11:00), (10) EV10: Typhoon Dujuan (2017/9/28 00:00-2017/10/2 23:00).

Close modal
Moreover, the baseflow is commonly required to estimate the rainfall-induced runoff via the rainfall–runoff model and apparently exhibits spatial variation. Figure 5 shows the baseflow at various branches estimated from the ten historical rainstorm events (see Figure 3), meaning the baseflow at the cross-sections of interest on average reaches six cm with a standard deviation of 1.5 cm, excluding those at the two branches (Chenyoulan and Chingshui creeks) and the upstream watershed with higher baseflow (on average 50 cm). Altogether, the spatial uncertainty in the baseflow should be considered regarding the model development.
Figure 3

Summary of baseflow estimated from 10 historical rainstorms in the branches and upstream within the study area.

Figure 3

Summary of baseflow estimated from 10 historical rainstorms in the branches and upstream within the study area.

Close modal

Sediment discharge data

This study's sediment-related data includes the bed material load, sediment transport, and the grand discharge. Figure 4 shows the temporal change in the sediment discharge (Qs) with the river runoff (Q) measured in 2007 and 2012 at the six sediment-measurement stations, which could be modeled a nonlinear relationship (i.e., Power function), called QQs rating curve (Walling 1997); the associated coefficients of the QQs relationship could be calibrated via the regression analysis as listed in Table 1. Given Table 1, the determination coefficient R2, accounting for the accuracy of the regression equation, varies from 0.35 to 0.82 (on average 0.7); this reveals that the above nonlinear QQs relationship can describe the change in sediment discharge Qs with river runoff Q under an acceptable accuracy (about 0.7); thus, the variation of the regression coefficients of QQs relationships needs to be considered in the model development.
Table 1

Regression coefficients and corresponding determination coefficients of QQs relations at six sediment discharge gauges

Sediment discharge gauge2007
2012
SG1 22.160 0.935 0.676 26.096 0.895 0.809 
SG2 57.024 0.786 0.768 52.503 0.745 0.768 
SG3 56.458 0.489 0.358 67.454 0.590 0.565 
SG4 29.238 1.084 0.476 18.240 1.226 0.764 
SG5 15.303 1.240 0.786 16.950 1.167 0.772 
SG6 25.974 1.146 0.816 52.902 0.773 0.662 
Sediment discharge gauge2007
2012
SG1 22.160 0.935 0.676 26.096 0.895 0.809 
SG2 57.024 0.786 0.768 52.503 0.745 0.768 
SG3 56.458 0.489 0.358 67.454 0.590 0.565 
SG4 29.238 1.084 0.476 18.240 1.226 0.764 
SG5 15.303 1.240 0.786 16.950 1.167 0.772 
SG6 25.974 1.146 0.816 52.902 0.773 0.662 
Figure 4

Historical sediment discharge with the river runoff at different gauges.

Figure 4

Historical sediment discharge with the river runoff at different gauges.

Close modal

Cross-section and roughness of river channel

In general, the bed elevation and roughness commonly play an essential role in the river routing as a result of mainly dominating the runoff moving along a river (Chang et al. 1993; Wu et al. 2011a, 2011b; Lau & Afshar 2013; Sanz-Ramos et al. 2021). Therefore, similar to the sediment discharge measurement, the elevation of the riverbed at 148 cross-sections along the study area recorded in 2007 and 2012 are adopted in the model development and validation as shown in Figure 5; it can be seen that the riverbed elevations at various cross-sections recorded in 2007 (110 m) are higher than those in 2012 (96 m). Especially for the distance of the cross-section to the estuary between 32,700 and 70,200 m, the difference in the riverbed elevation reaches 22 m, meaning the riverbed scouring and siltation significantly took place from 2007 to 2012. Thus, the riverbed elevation data measured in 2007 and 2012 selected as the study data are advantageous to the development of the proposed PM_MBEE_1D model, which can exhibit riverbed stability, scouring, or siltation due to the rainfall-induced runoff.
Figure 5

Historical riverbed elevation measured in 2007 and 2012 within the study area.

Figure 5

Historical riverbed elevation measured in 2007 and 2012 within the study area.

Close modal
Additionally, Figure 6 presents the river-channel roughness coefficients used in estimating the river stage, meaning the roughness coefficient has a significant increase with the distance to the river mouth from 0.27 to 0.042. Nevertheless, separated at 148 cross-sections into six elements, the roughness coefficient would be simulated regarding each cross-section along the study area.
Figure 6

Roughness coefficients at 148 cross-sections along the study area.

Figure 6

Roughness coefficients at 148 cross-sections along the study area.

Close modal

Tide depth

In addition to the lateral runoffs delivered from the branches and upstream required as the boundary conditions in the river routing, the tide depth near the river mouth is needed as the downstream boundary condition, which makes a significant contribution to the estimation of the river runoff and stage, especially for the neighboring estuary. In general, adopted in the river routing, the tide depths are derived from the observations during typhoon events. Therefore, similar to the parameter calibration of the rainfall–runoff (SAC-SMA) model, uncertainty analysis for the tide depth could be implemented with the historical hydrographs of 10 typhoons of 120 h, as shown in Figure 7; it can be observed that the tide depth commonly ranges from −2.5 to 2.5 m. However, a considerable difference in the temporal varying trend appears; for example, regarding EV2, the tide depth gradually decreases with time, whereas the corresponding tide depth to EV4 averagely rises with time, meaning that a sizeable temporal dispersion probably exists in the tide depth. Therefore, a considerable number of the tide depth hydrographs are then achieved to develop the proposed PM_MBEE_1D mode.
Figure 7

Historical tide depth hydrographs at the river mouth within the study area. (1) EV1: Typhoon Krosa (2007/10/4 00:00-2007/10/9 23:00), (2) EV2: Typhoon Sinlaku (2008/9/11 01:00-2008/9/17 00:00), (3) EV3: Typhoon Morakot (2017/8/7 01:00-2017/8/12 00:00), (4) EV4:Typhoon Fanapi (2010/9/18 12:00 2010/9/18 12:00), (5) EV5: Typhoon Saola (2012/7/31 07:00-2012/7/31 07:00), (6) EV6: Typhoon Soulik (2013/7/12 13:00-2013/7/14 22:00), (7) EV7: Typhoon Trami (2017/8/21 00:00-2017/8/25 23:00), (8) EV8: Typhoon Usagi (2017/9/21 00:00-2017/9/25 23:00), (9) EV9: Typhoon Matmo (2014/7/22 12:00-2014/7/25 11:00), and (10) EV10: Typhoon Dujuan (2017/9/28 00:00-2017/10/2 23:00).

Figure 7

Historical tide depth hydrographs at the river mouth within the study area. (1) EV1: Typhoon Krosa (2007/10/4 00:00-2007/10/9 23:00), (2) EV2: Typhoon Sinlaku (2008/9/11 01:00-2008/9/17 00:00), (3) EV3: Typhoon Morakot (2017/8/7 01:00-2017/8/12 00:00), (4) EV4:Typhoon Fanapi (2010/9/18 12:00 2010/9/18 12:00), (5) EV5: Typhoon Saola (2012/7/31 07:00-2012/7/31 07:00), (6) EV6: Typhoon Soulik (2013/7/12 13:00-2013/7/14 22:00), (7) EV7: Typhoon Trami (2017/8/21 00:00-2017/8/25 23:00), (8) EV8: Typhoon Usagi (2017/9/21 00:00-2017/9/25 23:00), (9) EV9: Typhoon Matmo (2014/7/22 12:00-2014/7/25 11:00), and (10) EV10: Typhoon Dujuan (2017/9/28 00:00-2017/10/2 23:00).

Close modal

Model concept

To quantify and evaluate the effect of the uncertainties in the hydrological data, sediment features, and topographical factors on the estimation of the riverbed elevation in a river, the proposed PM_MBEE_1D model is developed by coupling the uncertainty analysis with a considerable number of simulated riverbed elevations at various cross-sections; the resulting mobile-bed elevation simulations are from the 1D hydrodynamic numerical model with the sediment transport approach under the different conditions of the runoff and sediment caused by the rainfall events; this implies that hydrological and topographical variables (e.g., precipitation and roughness coefficient) and model parameters (e.g., rainfall–runoff model and sediment transport approach) related to the estimation of the river runoff and sediment could be defined as the runoff-related and sediment-related uncertainty factors, respectively.

Accordingly, while developing the proposed PM_MBEE_1D model, a significant number of the uncertainty factors are simulated using the Multivariate Monte Carlo simulation (MMCS) based on the statistical properties of their uncertainty. After that, a 1D sediment transport numerical model is configured under the study area's hydrological, hydraulic, topographic, and riverbed load conditions. Thus, the resulting simulations of the riverbed elevations at the cross-sections along the river could be achieved by the 1D sediment transpiration simulation numerical mode with a noticeable number of generated uncertainty factors; their corresponding reliabilities could be quantified via the uncertainty and risk analysis. Eventually, to efficiently assess the reliabilities of the estimated riverbed elevations under various uncertainty factors, the logistic regression analysis could derive a functional relationship of the reliabilities of the specific riverbed elevations with the uncertainty factors given.

In summary, the development of the proposed PM_MBEE_1D model could be grouped into six components: (1) identification of the uncertainty factors related to the estimation of the riverbed elevation; (2) configuration of the sediment transport numerical model; (2) generation of uncertainty factors; (4) simulation of the riverbed elevations at the various cross-sections (5) reliability assessment of the mobile-bed elevations at the various cross-sections; and (6) establishment of the exceedance-probability calculation equation. The detailed methods and concepts adopted in the development of the proposed PM_MBEE_1D model are addressed below:

Identification of uncertainty factors

When deriving the proposed PM_MBEE_1D model, the change in the riverbed elevations due to the uncertainty factors considered could be quantified by the 1D hydrodynamic numerical model under the various conditions of the river runoff and corresponding sediment discharge; also, the riverbed elevation could be regarded as the spatial and temporal variates; thus, the uncertainty factors considered in the proposed PM_BMEE_1D model could be identified within the hydrological and hydraulic analysis for estimating the river runoff and sediment in reaction to the change of the riverbed elevation in time and space, respectively.

Regarding the runoff estimation via the 1D river routing model, the variates adopted in the rainfall–runoff and hydraulic models should be treated as the uncertainty factors (Wu et al. 2011a, 2011b); that is, the event-based rainfall characteristics (i.e., event-based duration, rainfall depth, and storm pattern), baseflow, tide depth, external inflow, and roughness coefficients in the river channel (i.e., Manning coefficient) are supposed to be uncertainty factors. Specifically, the lateral and upstream runoff are frequently estimated via the rainfall–runoff modeling with the subbasin-based areal average rainfall; thus, the uncertainty in the lateral and upstream runoff could be attributed to the variations in the parameters of the rainfall–runoff routing and the areal average rainfall (Wu et al. 2011a, 2011b; Douinot et al. 2015).

Furthermore, the estimated sediment discharge in a river is generally obtained via a relationship with the river runoff Q, called the sediment rating curve (i.e., QQs relationship) (Isik 2013; Tfwala & Wang 2016; Sitorus & Susanto 2019); namely, the sediment discharge could be calculated in the case of the river discharge given. Nonetheless, the river sediment discharge possibly exhibits a considerable variation in time caused by the uncertainties in the river runoff and hydraulic characteristics (e.g., velocity, stress force, and energy slope) (Riahi-Madvar & Seifi 2018). By so doing, the coefficients of the Qs–Q relationship and parameters adopted in the sediment transport approaches are regarded as the uncertainty factors. Moreover, the initial condition of the riverbed elevation is commonly needed and obtained from historical digital elevation map in advance for the movable riverbed elevation simulation (Guo & Jin 2002); however, the initial riverbed elevation undergoes an apparent change with the variation in the hydrological (e.g., soil moisture) and topographical conditions (e.g., spatial resolution of DEM) in the watershed (Kim & Ivanov 2014; Hancock et al. 2016). Accordingly, the initial riverbed elevation should be considered as an uncertainty factor for the development of the proposed PM_MBEE_1D model.

Overall, the uncertainty factors corresponding to the riverbed elevation simulation in a river could be classified into runoff-related and sediment-related factors. The runoff-related uncertainty factors include the rainfall characteristics, baseflow, tide depth, river-channel roughness coefficient, and parameters of the rainfall–runoff modeling, whereas, the initial riverbed elevation, coefficients of the QQs relationship, and parameters of the sediment transport simulation model are listed in Table 2.

Table 2

Summary of the uncertainty factors for the mobile-bed elevation simulation

Type of uncertainty factorDefinition
Runoff-related factor Rainfall characteristics 
Baseflow (m3/s) 
Tide depth (m) 
Roughness coefficient 
Parameters of rainfall–runoff model 
Sediment-related factor Initial riverbed elevation (m) 
Parameters of sediment transport approach 
Type of uncertainty factorDefinition
Runoff-related factor Rainfall characteristics 
Baseflow (m3/s) 
Tide depth (m) 
Roughness coefficient 
Parameters of rainfall–runoff model 
Sediment-related factor Initial riverbed elevation (m) 
Parameters of sediment transport approach 

Configuration of 1D sediment-transport simulation model (CCHE1D)

Brief model concept

The computational hydroscience and engineering one-dimensional model (CCHE1D) aims to simulate the change in the riverbed elevations at the compound cross-sections under the condition of non-uniform and unsteady flow via the diffusion wave approximation equation within the ungauged watershed systems (Wu et al. 2004a, 2004b; Shen et al. 2016). Also, the CCHE1D model can quantify the effects of the hydraulic structures along a river-channel system, such as culverts, measuring flumes, drop structures, and bridge crossings, on the river runoff and sediment.

Within the CCHE1D model, the non-uniform sediment transport could be computed under the assumption of non-uniform total equilibrium by the following equation (Zhang & Langendoen 1998):
formula
(1)
where is the flow area; Qs and Q* stand for the actual and equilibrium sediment discharges, respectively; Cs serves as the depth-averaged total-load concentration; and Ls accounts for the adaptation length of non-equilibrium. Additionally, the above Qs is generally estimated using the river runoff through the following power function (Walling 1997):
formula
(2)
where α and β serve as the regression coefficients of the QQs relationship, respectively, which are commonly calibrated via the regression analysis with the historical measurement data.

Estimation of the river runoff

When using the CCHE1D model to simulate the riverbed elevation simulation at various cross-sections, the upstream and lateral runoff hydrographs are required as the boundary conditions. A group of the rainfall–runoff models is developed in estimating the runoff hydrograph; the Sacramento Soil Moisture Accounting Model (SAC-SMA) is a nonlinear semi-distributed hydrologic module under an assumption of the two soil layers (the upper and lower zone) in reaction to generate soil moisture states and resulting runoff components (Koren et al. 2000); Figure 8 shows the concept of rainfall–runoff routing in the SAC-SMA model with the 19 parameters, the definitions of which are introduced in Table 3 (Ajami et al. 2004; Wu et al. 2011a, 2011b). Therefore, in this study, the upstream and lateral runoff hydrographs estimated using the SAC-SMA model with the rainfall data are treated as the boundary conditions for carrying out the sediment transport simulation via the CCHE1D model.
Table 3

Description of parameters of the SAC-SMA model (Ajami et al. 2004; Wu et al. 2011a, 2011b)

ParametersDescription
UZTWM Upper zone tension water capacity (mm) 
UZFWM Upper zone free-water capacity (mm) 
UZK Upper zone recession coefficient 
PCTIM Percent of impervious area 
ADIMP Percent of additional impervious area 
SARVA Fraction of segment covered by streams, lakes, and riparian vegetation 
ZPERC Minimum percolation rate coefficient 
REXP Percolation equation exponent 
LZTWM Lower zone tension water capacity (mm) 
LZFSM Lower zone supplementary free-water capacity (mm) 
LZFPM Lower zone primary free-water capacity (mm) 
LZSK Lower zone supplementary recession coefficient (mm) 
LZPK Lower zone primary recession coefficient (mm) 
PFREE Percentage percolating directly to lower zone free water 
SIDE The ratio of deep recharge water going to channel baseflow 
RESERV Percentage of lower zone free water not transferable to lower zone tension water 
SSOUT Fixed-rate of discharge lost from the total channel flow (mm/Δt) 
Period of runoff distribution function 
DF_P Maximum ratio of runoff distribution function 
ParametersDescription
UZTWM Upper zone tension water capacity (mm) 
UZFWM Upper zone free-water capacity (mm) 
UZK Upper zone recession coefficient 
PCTIM Percent of impervious area 
ADIMP Percent of additional impervious area 
SARVA Fraction of segment covered by streams, lakes, and riparian vegetation 
ZPERC Minimum percolation rate coefficient 
REXP Percolation equation exponent 
LZTWM Lower zone tension water capacity (mm) 
LZFSM Lower zone supplementary free-water capacity (mm) 
LZFPM Lower zone primary free-water capacity (mm) 
LZSK Lower zone supplementary recession coefficient (mm) 
LZPK Lower zone primary recession coefficient (mm) 
PFREE Percentage percolating directly to lower zone free water 
SIDE The ratio of deep recharge water going to channel baseflow 
RESERV Percentage of lower zone free water not transferable to lower zone tension water 
SSOUT Fixed-rate of discharge lost from the total channel flow (mm/Δt) 
Period of runoff distribution function 
DF_P Maximum ratio of runoff distribution function 
Figure 8

Schematic concept of the rainfall–runoff process of the SAC-SMA model (Ajami et al. 2004).

Figure 8

Schematic concept of the rainfall–runoff process of the SAC-SMA model (Ajami et al. 2004).

Close modal
Because the physical-based SAC-SMA 19 parameters are simultaneously calibrated with considerable difficulty, Wu et al. (2011a, 2011b) carried out sensitivity analysis to recognize ten sensitive SAC-SMA parameters, UZTWM, UZFWM, UZK, PCTIM, ADIMP, ZPERC, REXP, LZTWM, DF_L and DF_P which noticeably influence the estimation of the runoff hydrograph; accordingly, using the historical rainfalls and induced runoff data, the above 10 SAC-SMA parameters could be calibrated with t utilizing a modified genetic algorithm (GA_SA) (Wu et al. 2011a, 2011b) with the following objective function (Madsen 2000):
formula
formula
(3)
where is the runoff duration; and are the observed and estimated discharge at the time step t; and is the mean of observed runoff volume. As a result, this study employs the SAC-SMA model with the ten sensitive parameters to estimate the upstream and lateral runoff adopted as the boundary conditions for the movable-bed elevation simulation coupled with the CCHE1D model.

Simulation of uncertainty factors

As shown in Table 2, a group of the runoff-related and sediment-related factors could be defined as the uncertainty factors; in this study, to quantify their effect in the estimation of the riverbed elevations, a considerable number of the runoff-rated and sediment-related uncertainty factors of interest are generated and used in the development of the proposed PM_MBEE_1D model. According to the results from the sensitivity analysis with the standard normal regression equation, the rainfall depth and initial riverbed elevation play a more critical role in estimating the riverbed elevations; thus, changing the rainfall depth and initial elevation account for the uncertainties in the rainfall and geographical factors.

The Monte Carlo simulation algorithm with the variate's statistical properties is widely applied in the uncertainty and induced risk quantification in the relevant hydrological analysis (Wu et al. 2017, 2022). Nevertheless, the hydrological, hydraulic, and geographical features generally exist in somewhat inherent correlation in time and space (Simpson & Schlunegger 2003; Wu et al. 2011a, 2011b, 2021a, 2021b). Therefore, this study employs the Monte Carlo simulation with the non-normal correlated multi-variates (Chang et al. 1997) to reproduce the runoff-related and sediment-related uncertainty factors; the relevant model concept and framework are briefly addressed below.

In general, the correlated non-normal MMCS algorithm is implemented in order to generate a sizeable number of the variables considered; however, the well-known normal distribution is constrained on the two correlated variables in response to their correlation structure. To improve the Monte Carlo simulation algorithm, which could be applied to a variety of correlated variables, Chang et al. (1997) proposed a modified Monte Carlo simulation method proceeding with the normalized-based algorithms, including the standardized, orthogonal, and inverse transformations, named the MMCS model. Within the normalized-based procedure, a relationship between the correlation coefficients of the multivariate in the real and normalized spaces should be derived in advance via the Nataf distribution (Nataf 1962):
formula
(4)
formula
where and are the correlated variables at points i and j, respectively, with the means and , the standard deviations and , and the correlation coefficient ; and serve as the corresponding bivariate standard normal variables to the variable and with the correlation coefficient and the joint standard normal density function . In detail, using the Nataf distribution, generated in the normalized space, the corresponding simulated standard normal variables should be transferred into one in the real space under consideration of inherent correlation structures within the normalized-based algorithm.

The runoff-related and sediment-related uncertainty factors (see Table 2) could be generated using the MMCS method with the associated statistical properties given in advance. In particular, the simulated rainfall characteristics should be combined as the hyetographs while developing the proposed PM_MBEE_1D model. In short, the simulation of the uncertainty factors subject to the riverbed elevation estimates would be executed through the rainfall characteristics and the remaining uncertainty factors.

Simulation of rainfall characteristics

To quantify the uncertainty of rainfall in time and space, a significant number of event-based rainstorms at all grids reproduced by the stochastic modeling of the gridded short-term rainstorms (called SM_GSTR model) (Wu et al. 2021a, 2021b) are used in the development of the proposed PM_MBEE_1D model. Within the SM_GSTR model, the event-based rainstorm is characterized in terms of three rainfall characteristics: the rainfall duration, gridded rainfall depths (regarded as the spatial variates), and gridded storm depths comprised of the dimensionless rainfalls at the various dimensionless times (treated as the Spatio-temporal correlated variates); specifically, the gridded storm pattern is supposed to be separated into two components, the areal average of the dimensionless rainfalls (i.e., the storm pattern) and the associated deviations at the various dimensionless times.

Characterized with the historical rainfall data, the gridded rainfall characteristics should be reproduced through the correlated MMCS method; note that the statistical analysis for the gridded rainfall characteristics should be carried out at first to quantify their uncertainties in time and space, including the first four statistical moments, correlation coefficients, and the appropriate probability functions. After that, the simulations of the gridded storm patterns are reproduced by combining the simulations of the areal averages of dimensionless cumulative rainfalls and the associated gridded bias; eventually, the gridded rainstorms are emulated by coupling the simulated storm patterns at all grids with the simulations of the gridded rainfall depths for the simulated event-based duration. Figure 9 shows the simulation process of gridded rainstorms via the SM_GSTR model. Eventually, the thorough introduction to the SM_GSTR model could be referred to in the investigation by Wu et al. (2021a, 2021b).
Figure 9

Schematic process of characterizing gridded rainstorm (Wu et al. 2021a, 2021b).

Figure 9

Schematic process of characterizing gridded rainstorm (Wu et al. 2021a, 2021b).

Close modal

Simulation of the remaining uncertainty factors

In addition to the generation of the rainfall characteristics, the MMCS approach would be implemented in the simulation of the remaining runoff-related and sediment-related uncertainty factors based on the corresponding statistical properties of the observation. Among these uncertainty factors subject to the riverbed elevation estimation, the model parameters, including the rainfall–runoff model, river routing, and sediment transport approach, should first be calibrated with the relevant observed rainfall–runoff-sediment discharge data. Also, concerning the remaining uncertainty factors (i.e., the upstream baseflow and tide depth), the statistical analysis is supposed to be carried out with the observation recorded in the historical events to quantify their uncertainties.

Identification of sensible uncertainty factors

Despite the model inputs and parameters commonly treated as uncertainty factors based on the model concept and configuration, sensitivity analysis should be implemented to find the uncertainty factors contributing more sensitivity to the model output in order to efficiently conduct hydrological reliability analysis (Chang et al. 1993; Wu et al. 2021a, 2021b). Thereby, while developing the proposed PM_MBEE_1D by coupling the Monte Carlo simulation with the uncertainty analysis algorithm, the sensitivities of the runoff-related and sediment-related uncertainty factors to the estimated riverbed elevation (see Table 2) are supposed to be quantified to identify the sensitive uncertainty factors exhibiting significant impact to the estimation of the riverbed elevation.

Two types of sensitivity analysis methods, standardized regression equation and analysis of the variance methods, could be widely applied in identifying the sensitive model inputs to the model outputs (Wu et al. 2011a, 2011b). As mentioned earlier, a group of the runoff-related and sediment-related factors are summarized for the estimation of the riverbed elevations in this study; thus, the standardized regression equation for the multivariate model inputs is adopted to specify the uncertainty factors with more sensitivity to the riverbed elevation estimates.

The standardized regression equation, linearly composed of the standard normal values of the model inputs and output, can be expressed as follows:
formula
(5)
where Y and X stand for the model output and inputs, respectively, with the mean () and standard deviation ( and ); and serves as the standard regression coefficients of the model inputs. The sensitivity of model inputs to the model output can be quantified subject to the standardized regression coefficient . In short, the greater absolute value of the regression coefficient has a more significant effect on the model output with a positive coefficient, meaning that the model input is directly proportional to the model output.

Therefore, within the proposed PM_MBEE_1D model, a sizeable number of the estimated riverbed elevations (i.e., the model outputs) under various combinations of the simulated runoff-related and sediment-related uncertainty factors as the model inputs are utilized in the derivation of the standard normal regression Equation (5). In the end, the identification of the sensitive uncertainty factors regarding the river-based elevation estimates could be achieved based on the results from the parameter calibration of the standard normal equation.

Reliability quantification of the simulated mobile-bed elevation

While developing the proposed PM_MBEE_1D model, the probabilities of the estimated riverbed elevation over the specific values (i.e., the exceedance probability) could be achieved using the well-known uncertainty analysis method, i.e., the advanced first-order and second-moment approach (AFOSM) (Tung & Mays 1981; Wu et al. 2017; Beckers et al. 2018; Tung 2018). The AFOSM approach is commonly utilized to quantify and evaluate the effect of the uncertainty factors in the hydrological-related analysis (Chang et al. 1993; Malkawi et al. 2000; Wu et al. 2011a, 2011b, 2017, 2021a, 2021b; Ganji & Jowkarshorijeh 2012). In detail, the AFOSM quantifies the risk by calculating the probability of the load variable () exceeding the resistance (), named exceedance probability, via the standard normal distribution with the two first orders of the statistical moments (i.e., mean and variance):
formula
(6)
In this study, the risk is defined as the probability of the estimated riverbed elevation exceeding a specific elevation (i.e., underestimated risk); the reliability is equal to one minus the underestimated risk. Thereby, performed by the AFSOM method, the resulting exceedance probability of the estimated riverbed elevation (i.e., underestimated risk) could be calculated through the following equation:
formula
(7)
formula
formula
where and stand for the estimated riverbed elevation and associated specific elevation, respectively; is the performance function; and serve as the mean and standard deviation of Z, respectively; denotes the standard normal distribution with a specific reliability index . Note that in Equation (7), and could then be achieved by the following equations:
formula
(8)
formula
(9)
where is the failure points of the ith uncertainty factor when the performance function z is equal to 0; and account for the mean and standard deviation of the ith uncertainty factor, respectively; serves as the estimated riverbed elevation regarding the uncertainty factors' failure points; and is the sensitivity coefficient of the ith uncertainty factor.
Note that when calculating the mean and variance of the model outputs via the AFOSM method, a functional relationship with model inputs is supposed to be derived in advance. Therefore, according to Table 2, the uncertainty factors for the riverbed elevation estimates can be grouped into runoff-related and sediment-related factors. Based on the results from the sensitivity analysis, the resulting sensitive uncertainty factors are used to establish the equation for estimating the riverbed elevation via the multivariate regression analysis:
formula
(10)
where account for the estimated riverbed elevation; stands for the runoff-related factors, and represents the sediment-related factors. In this study, to facilitate the reliability and accuracy of the regression equation for the riverbed elevation estimates with the uncertainty factors, the associated regression coefficients of the uncertainty factors in Equation (10) could be determined with a considerable number of the corresponding estimated riverbed elevations to the generated uncertainty factors via the aforementioned Monte Carlo simulation algorithm for the correlated non-normal variables.

Derivation of the exceedance-probability calculation equation

Despite coupling the multivariate Monte Carlo simulation with the AFOSM approach to employ the reliability analysis for the estimated riverbed elevation, a complicated process might be executed for calculating the exceedance probability (i.e., underestimated risk) through Equations (7)–(10). To effectively perform the reliability analysis for the riverbed elevation estimates, this study intends to establish an equation for calculating the corresponding exceedance probability of the estimated riverbed elevation to the specific uncertainty factors by means of the logistic regression analysis. The logistic analysis is mainly applied to establish a linear or nonlinear functional relationship between the numerous independent model inputs and the occurrence probability of a dependent model output:
formula
(11)
where Xi and P represent the ith independent model inputs and the corresponding occurrence probability to a specific model output, respectively. (i = 1,2, … n) are the regression coefficients, and is the intercept.

In this study, P in Equation (9) is regarded as the exceedance probability of the estimated riverbed elevation, i.e., , directly assigned or estimated by the numerical model; and Xi could be selected factors among the runoff-related and sediment-related factors (see Table 2). Although a group of the runoff-related and sediment-related uncertainty factors are of concern in the model development, the goal of the proposed PM_MBEE_1D is to effectively quantify the estimated riverbed elevation based on the observation data related to the hydrological and topographical features at the location of interest. Therefore, while deriving the exceedance-probability equation, the runoff-related and sediment-related uncertainty factors could be used in Equation (11) subject to the hydrological and topographical features. Therefore, using the resulting exceedance-probability equation, the reliability analysis for the estimated riverbed elevation under various river runoff and sediment conditions could be efficiently carried out without proceeding with the AFOSM approach with a complicated analysis process.

Model framework

According to the concepts and methods addressed in the previous sections, the proposed 1-D probabilistic-based mobile-bed elevation estimation model (PM_MBEE_1D) comprises the five main components: the identification of uncertainty factors, generation of uncertainty factors, estimation of the riverbed elevation, calculation of the exceedance probability for the riverbed elevation estimate, and derivation of the exceedance-probability calculation equation; the aforementioned exceedance probability accounts for the risk of underestimating the riverbed elevation due to the uncertainties in the runoff-related and sediment-related factors (see Table 2) required in the well-known rainfall–runoff routing (SAC_SMA) and sediment-transportation simulation numerical modeling (CCHE1D). In short, the proposed PB_MBEE_1D model mainly provides the stochastic estimation of the riverbed elevations along the river under consideration of the uncertainties related to the runoff and sediment characteristics; thus, while developing the proposed PM_MBEE_1D model, two equations could be established and utilized for estimating the riverbed elevations and quantifying the corresponding exceedance probability (i.e., underestimated risk). The thorough procedure of model development and application could be expressed as follows:

Model development

  • Step [1]: Collect the hydrological data (rainfall, river runoff, and stage, baseflow, and tide depth), hydraulic features (river-channel cross-section and roughness coefficient), and local soil properties information (riverbed grain size and sediment discharge) in the study area to configure the rainfall–runoff model (SAC-SMA) and the numerical sediment transport simulation for the movable riverbed (i.e., CCHE1D).

  • Step [2]: Calibrate the parameters of the runoff-runoff model (SAC-SMA) and numerical mobile-bed sediment transport model (CCHE1D).

  • Step [3]: Extract the corresponding gridded rainfall characteristics from the historical radar-rainfall data and calculate their statistical properties in time and space; reproduce a significant number of gridded rainstorms.

  • Step [4]: Generate the runoff-related and sediment-related factors, excluding the gridded rainfall characteristics.

  • Step [5]: Simulate the runoff hydrographs at the branches with the generated rainfall hyetographs and SAC-SMA parameters as the boundary conditions for the CCHE1D model.

  • Step [6]: Obtain the simulations of the temporally-based riverbed elevations at the specific locations (i.e., cross-section) along the river under consideration of generated lateral and upstream runoff hydrographs as well as the sediment-related factors used in the CCHE1D model.

  • Step [7]: Identify the sensitive uncertainty factors via the standard regression analysis with the simulations of riverbed elevations and corresponding generated uncertainty factors.

  • Step [8]: Derive the nonlinear functional relationship of the corresponding simulated riverbed elevations to the sensitive uncertainty factors determined at the previous step.

  • Step [9]: Calculate the exceedance probabilities of specific magnitudes of the riverbed elevations via the advanced first order and second-moment (AFOSM) approach with the mean and variance of the sensitive runoff-related and sediment-related factors given in advance.

  • Step [10]: Establish the exceedance-probability calculation equations for the specific riverbed elevations under consideration of the sensitive uncertainty factor concerned via the logistic regression analysis.

Model application

When the proposed PM_MBEE_1D is developed, the reliability analysis could be carried out via the exceedance-probability calculation equations derived within the proposed model under the runoff-related and sediment-related factors given; also, the riverbed elevation could be estimated with a desired likelihood in case of the uncertainty factors assigned. Accordingly, the procedure mentioned above of estimating the riverbed elevation with an exceedance probability can be addressed:

  • Step [1]: Ensure the sensitive uncertainty factors related to the hydrological and topographical characteristics subject to the change in the riverbed elevation.

  • Step [2]: Calculate the exceedance probability of the specific riverbed elevations at the different cross-sections under various conditions of the hydrological and topographical factors given. Accordingly, the reliabilities of the estimated riverbed elevations could be achieved with varying the uncertainty factors concerned.

  • Step [3]: Estimate the riverbed elevations at various locations corresponding to the desired reliability to determine the riverbed scour and siltation.

Configuration of CCHE1D for the study area

As mentioned in the model development framework, the sediment-transport simulation numerical model (CCHE1D) is used in the estimation of the riverbed elevations with the generated runoff-related and sediment-related uncertainty factors; thus, the CCHE1D model should be configured in advance based on the hydrological and topological characteristics as well as hydraulic facilities within the study area. According to the introduction on the study area with two main branches, the schematic model structure of the CCHE1D could then be set up; note that the riverbed elevation simulation carried out via the CCHE1D model mainly focuses on the Zhuoshui River and resulting lateral runoff hydrographs from the two branches, Chenyoulan and Chingshui Creek (i.e., boundary conditions). Altogether, while proceeding with the sediment transport simulation via the CCHE1D model to simulate the change in the riverbed elevations with time, the lateral and upstream runoff hydrographs are reproduced via the SAC-SMA model with the subbasin-based average rainfall.

Determination of model parameters adopted

According to the aforementioned model concept and framework, the parameters of the rainfall–runoff model (SAC_SMA) and sediment transport simulation numerical model (CCHE1D) should be calibrated with the relevant observations as addressed in Section 3. The relevant results are expressed as:

SAC-SMA model

According to the introduction on the 19-parameter SAC-SMA model in Section 2.3.2, 11 calibrated parameters could be obtained with the observed rainfall-induced runoff of 10 typhoon events (Figure 2); their statistical properties could then be calculated as listed in Table 4; it can be seen that there is a significant difference in the statistical properties, especially for the high skewness coefficients (from −0.5 to 2.1) and correlation coefficients (from −0.6 to 0.65). Specifically, the above negative coefficient indicates that the two SAC-SMA parameters make a completely different contribution to the runoff estimation; for instance, the parameter UZTWN and LZTWN are related to the surface runoff and infiltration, respectively; accordingly, they should have a negative correlation coefficient. As a result, the SAC-SMA parameters could be treated as non-normal correlated multivariate.

Table 4

Summary of the parameters of the CCHE1D model used in the model development

Wash-load adaptation coefficient αMix-layer height (m)Sediment transport equationBed-load adaptation length Ls (m)
0.3 SEDTRA 1,000 
1,333 
2,000 
2,500 
3,000 
3,500 
4,000 
5,200 
6,400 
7,600 
8,800 
9,120 
10,000 
11,000 
13,333 
Wash-load adaptation coefficient αMix-layer height (m)Sediment transport equationBed-load adaptation length Ls (m)
0.3 SEDTRA 1,000 
1,333 
2,000 
2,500 
3,000 
3,500 
4,000 
5,200 
6,400 
7,600 
8,800 
9,120 
10,000 
11,000 
13,333 
Table 5

Statistical properties of the SAC-SMA parameters calibrated with observed rainfall–runoff data of 10 historical typhoons

Statistical propertiesUZTWMUZFWMUZKPCTIMADIMPZPERCLZTWMLZFSMLZSKDF_LDF_P
Mean 171.538 173.757 0.321 0.118 0.099 57.617 571.284 165.938 0.148 151,230.385 2,058.076 
Standard deviation 55.576 130.071 0.128 0.028 0.016 18.735 81.376 73.353 0.087 35,732.371 682.201 
Skewness − 0.467 2.126 1.183 0.285 0.159 0.693 0.065 0.718 1.309 0.386 0.846 
Kurtosis 2.164 6.702 2.980 1.633 1.746 2.341 2.167 2.262 3.940 3.538 3.388 
Correlation coefficient UZTWM 1.000           
UZFWM 0.074 1.000          
UZK 0.019 − 0.327 1.000         
PCTIM 0.606 − 0.253 0.144 1.000        
ADIMP − 0.229 − 0.176 − 0.460 − 0.175 1.000       
ZPERC 0.208 − 0.030 0.071 0.204 0.041 1.000      
LZTWM − 0.223 0.268 − 0.187 − 0.217 − 0.035 0.100 1.000     
LZFSM 0.240 0.294 0.317 − 0.104 0.020 − 0.230 − 0.073 1.000    
LZSK 0.110 0.398 − 0.041 − 0.106 0.260 0.303 − 0.058 0.357 1.000   
DF_L − 0.045 − 0.101 0.236 − 0.088 − 0.072 − 0.454 0.029 0.392 0.035 1.000  
DF_P 0.357 − 0.083 − 0.256 − 0.136 0.103 − 0.144 0.054 0.062 − 0.226 0.391 1.000 
Defaults of remaining parameters SARVA REXP LZFAM LZPK PFREE SIDE RESERV UZTWC UZFWC LZTWC LZFSC LZFPC 
200 0.04 0.2 0.3 100  
Statistical propertiesUZTWMUZFWMUZKPCTIMADIMPZPERCLZTWMLZFSMLZSKDF_LDF_P
Mean 171.538 173.757 0.321 0.118 0.099 57.617 571.284 165.938 0.148 151,230.385 2,058.076 
Standard deviation 55.576 130.071 0.128 0.028 0.016 18.735 81.376 73.353 0.087 35,732.371 682.201 
Skewness − 0.467 2.126 1.183 0.285 0.159 0.693 0.065 0.718 1.309 0.386 0.846 
Kurtosis 2.164 6.702 2.980 1.633 1.746 2.341 2.167 2.262 3.940 3.538 3.388 
Correlation coefficient UZTWM 1.000           
UZFWM 0.074 1.000          
UZK 0.019 − 0.327 1.000         
PCTIM 0.606 − 0.253 0.144 1.000        
ADIMP − 0.229 − 0.176 − 0.460 − 0.175 1.000       
ZPERC 0.208 − 0.030 0.071 0.204 0.041 1.000      
LZTWM − 0.223 0.268 − 0.187 − 0.217 − 0.035 0.100 1.000     
LZFSM 0.240 0.294 0.317 − 0.104 0.020 − 0.230 − 0.073 1.000    
LZSK 0.110 0.398 − 0.041 − 0.106 0.260 0.303 − 0.058 0.357 1.000   
DF_L − 0.045 − 0.101 0.236 − 0.088 − 0.072 − 0.454 0.029 0.392 0.035 1.000  
DF_P 0.357 − 0.083 − 0.256 − 0.136 0.103 − 0.144 0.054 0.062 − 0.226 0.391 1.000 
Defaults of remaining parameters SARVA REXP LZFAM LZPK PFREE SIDE RESERV UZTWC UZFWC LZTWC LZFSC LZFPC 
200 0.04 0.2 0.3 100  

CCHE1D parameters

As introduced in Section 2.3.1, the sediment transport equations should be determined in advance when configuring the CCHE1D model to simulate the change in the riverbed elevation at the various cross-sections along a river. In general, the four commonly used sediment transport equations, SEDTRA formula (Garbrecht et al. 1995), Wu-Wang-Jia formula (Wu et al. 2000), modified Ackers-White and modified Engelund-Hansen formula (Proffitt & Sutherland 1983). In this study, subject to the investigations (e.g., Wu et al. 2004a, 2004b; Ding & Langendoen 2016; Norouzi et al. 2022), the SEDTRA model is selected to describe the runoff-induced change in the riverbed elevation, i.e., scoring and silting, within the study area. In addition, the adaptation length (Ls) markedly contributes to estimating the riverbed elevation. Although adopting a shorter adaptation length causes a significant variation in riverbed scour and siltation with a large simulation error in comparison to observations, it could result in a higher correlation between the simulated and measured riverbed elevations. Bed-load adaptation lengths could be given from 4,000 to 20,000 m. Thus, the mean and standard deviation could be assigned as 5,920 and 3,870 m, respectively, in the simulations of the uncertainty factors for the model development. Note that the remaining CCHE1D parameters (e.g., wash-load adaptation coefficient and mix-layer height) mainly contribute to the runoff-induced riverbed elevation estimation (see Table 5).

Generation of uncertainty factors

Before deriving the proposed PM_MBEE_1D model, a number of uncertainty factors related to estimating the riverbed elevation should be generated via the Monte Carlo simulation with the correlated non-normal multivariate. Of the runoff-related and sediment-related uncertainty factors, the parameters of the SAC-SMA and CCHE1D models would be generated with the results from the model calibrations in the previous section. Furthermore, the simulations of the hyetographs of rainstorms should be completed by combining the gridded rainfall characteristics, including the event-based rainfall durations, gridded rainfall depths, and gridded storm patterns. Their simulations could be achieved with the measurements (i.e., Manning roughness coefficient, tide depth, and baseflow) regarding the remaining uncertainty factors. Therefore, the simulations of uncertainty factors could be separated into four parts: SAC-SMA parameters, CCHE1D parameters, and rainfall characteristics and remaining factors. Figure 10 illustrates the 1,000 simulations of the above uncertainty factors used in the model development.
Figure 10

Illustration of 1,000 simulations of uncertainty factors. (1) SAC-SMA parameters, (2) bed-load adaptation length, (3) regression coefficient of QQs relationships, (4) gridded rainstorm events, (5) baseflow, (6) river-channel roughness coefficient, and (7) tide depth.

Figure 10

Illustration of 1,000 simulations of uncertainty factors. (1) SAC-SMA parameters, (2) bed-load adaptation length, (3) regression coefficient of QQs relationships, (4) gridded rainstorm events, (5) baseflow, (6) river-channel roughness coefficient, and (7) tide depth.

Close modal

Simulation of the riverbed elevation

Given the development procedure of the proposed PM_MBEE_1D model, while carrying out the sediment transport analysis to quantify the change in the riverbed elevation caused by the river runoff, the rainfall-derived runoff should be estimated in advance; hence, the simulations of the runoff from the upstream basin (upstream runoff) and two branches (i.e., lateral runoff) could be obtained by configuring the SAC-SMA model with the generated gridded rainstorms, baseflow and tide depths; thus, the resulting lateral and upstream runoff should be used as the boundary conditions for carrying out the sediment-transportation simulation via the CCHE1D model. Moreover, to take into account the effect of the uncertainty in the initial riverbed elevation, the simulated elevations at the last time step for the specific cross-sections regarding the ith simulation case are treated as the simulated initial elevations for the (i + 1)th case, except for the first simulation case with the measured cross-section elevation in 2012. As mentioned earlier, the estimated riverbed elevation should be defined as the spatiotemporally-based variates; however, this study focuses on generally evaluating the effect of temporally- and spatially-varied uncertainty factors on estimating the riverbed elevations. Therefore, the riverbed elevations at the last time step (named final riverbed elevation) are mainly adopted in the model development and demonstration.

Figure 11 shows the simulations of the final riverbed elevations and associated statistical properties (i.e., mean and standard deviation) at the specific cross-sections under specific simulation cases within the study area; it can be observed that the riverbed elevation has a considerable decrease with the cross-section from the upstream to the downstream. For example, at the 140th cross-section, the maximum and minimum of the final simulated riverbed elevation are approximately 324.4 and 299.2 m, respectively, with a standard deviation of 1.74 m; in contrast with the previous cross-section, at the downstream 118th and 86th cross-section, the final simulated riverbed elevation is significantly associated with the less standard deviations of 1.57 and 0.16 m. As a result, the final simulated riverbed elevations at the upstream cross-sections noticeably exhibit a larger variation than those at the downstream due to the uncertainties in the runoff-related and sediment-related factors. In addition, regarding the 118th cross-section where the Jiji Diversion Weir is located, its riverbed elevation nearly reaches a constant (about 199 m); namely, its standard deviation approximates zero.
Figure 11

Simulations of the riverbed elevations at the cross-sections and associated statistical properties (mean and standard deviation) within the study.

Figure 11

Simulations of the riverbed elevations at the cross-sections and associated statistical properties (mean and standard deviation) within the study.

Close modal

In total, the spatial and temporal uncertainties in the runoff-related and sediment-related factors could significantly impact the changes in the riverbed elevations at the specific cross-sections within the study area. Hence, it is necessary to quantify and evaluate their effect on estimating the riverbed elevations at the particular cross-sections. As a result, this study aims to develop a probabilistic-based model that can provide the section-based elevations of the movable riverbed with a corresponding likelihood under consideration of the spatial and temporal uncertainties.

Sensitivity quantification of uncertainty factors to riverbed elevation

Despite a group of runoff-related and sediment-related uncertainty factors triggering the dramatic variation in the riverbed elevation, it probably causes a complicated process of quantifying the reliability of the riverbed elevation estimate using the AFOSM approach for developing the proposed PM_MBEE_1D model. This is because a functional relationship of the riverbed elevation with the uncertainty factors (i.e., Equation (10)) should be derived in advance while conducting the reliability analysis via the AFOSM approach with the runoff-related and sediment-related uncertainty factors. Therefore, in this study, using the standard normal equation addressed in Section 2.5, the sensitive uncertainty factors could be identified among the runoff-related and sediment-related uncertainty factors based on their corresponding regression coefficients of the relationship calibrated regarding the riverbed elevation (see Equation (9)) at the specific cross-sections selected from the downstream to the upstream. Also, the sensitivity assessment would be carried out based on four groups classified under their types, i.e., the rainfall characteristics (5 factors), SAC-SMA parameters (11 factors), CCHE1D parameters (4 factors), and remaining factors (3 factors); of which the uncertainty factors considered, the rainfall characteristics are mainly generated to produce the hyetograph in which the rainfall duration (Dur_rain), depth (Depth_rain) and the maximum rainfall intensity (RI_max) which could account for the change in the rainfall in time (Wu et al. 2011a, 2011b, 2015). In the end, the sensitive uncertainty factors would be found out individually from four groups at the six cross-sections selected from the downstream to the upstream, including the first (CS_1st), 72nd (CS_72th), 100th(CS_100th), 124th (CS_120th), 136th (CS_136th) and 146th (CS_146th). Therefore, the rainfall duration, depth, and maximum rainfall intensity are the rainfall factors used in the reliability/uncertainty quantification via the AFOSM method. Similarly, the tide depth hydrograph's maximum value is also treated as the uncertainty factor for the model development.

Figure 12 represents the absolute averages of the regression coefficients regarding the uncertainty factors calibrated with 1,000 simulations of the final riverbed elevations and the corresponding generated factors at the above specific cross-sections. From Figure 12, as for the CCHE1D parameters, it can be found that the initial riverbed elevation (Z_ini) has a markedly larger regression coefficient (about 0.85) than other sediment-related uncertainty factors; also, the regression coefficient of adaptation length (Ls) averagely reaches 0.16. Therefore, the initial riverbed elevation and adaptation length should both be treated as the sensitivity factors. Although the coefficients of QQs relationships (α and β) merely approximate 0.02, they play an important role in the sediment concentration in the river, mainly dominating the riverbed elevation; thus, similar to the initial riverbed elevation and adaptative length, the regression coefficients α and β should be regarded as the sensitive sediment-related uncertainty factors.
Figure 12

Summary of the average coefficients of the uncertainty factors used in the standard normal regression equations regarding the riverbed elevation.

Figure 12

Summary of the average coefficients of the uncertainty factors used in the standard normal regression equations regarding the riverbed elevation.

Close modal

In the case of the lateral and upstream runoff hydrograph, the average regression coefficients of the SAC-SMA parameters approximate 0.012, but they primarily contribute to the estimation of the river runoff triggered by the rainfall. Accordingly, the SAC-SMA parameters are necessarily used in developing the proposed PM_MBEE_1D model. Moreover, the baseflow is frequently required to estimate the river runoff via the rainfall–runoff routing, especially in drought periods. Since the baseflow in the study area mainly comes from the upstream catchment and two branches (Chenyoulan and Chingshui Creeks) by observing Figure 12, the regression coefficients of the baseflows at the upstream catchment (i.e., upstream baseflow) (BF1) (about 0.021) noticeably exceed those at the two branches (BF2 and BF3) (around 0.008); thus, the baseflow from the upstream catchment is adopted as the uncertainty factor.

In addition to the SAC-SMA and CCHE1D parameters, the tide depth is needed as the downstream conditions in the river routing. According to Figure 9, although the average regression coefficient is nearly 0.007, the associated river runoff is commonly impacted by the tide depth at the cross-sections close to the estuary (e.g., the first and 70th sections). Likewise, the average regression coefficients of the river-channel roughness coefficients are merely about 0.0014; nonetheless, the roughness coefficients should be given to estimate the runoff and induced water level for calculating the corresponding sediment discharge. As a result, the tide depth and river-channel roughness coefficient should both be considered in the model development.

To sum up the results from the sensitivity analysis, the uncertainty factors, including the rainfall depth, maximum rainfall intensity, tide depth, baseflow from the upstream, river-channel roughness coefficient, the SAC-SAM and CCHE1D parameters (i.e., initial riverbed elevation and adaptation length), defined as the sensitive uncertainty factors. By so doing, the above-derived functional relationships for estimating the riverbed elevations at various sections are adopted in the AFOSM method. The resulting reliability of the simulated riverbed elevation (i.e., exceedance probability) could be quantified and utilized in developing the proposed PM_MBEE_1D model.

Development of the proposed PM_BMEE_1D model

Based on the model framework (see Section 2.8), while developing the proposed PM_MBEE_1D model via the AFOSM algorithm, 1,000 simulations of the riverbed elevations at various cross-sections are used in advance to establish the corresponding relationship with the sensitive runoff-related and sediment-related uncertainty factors (see Equation (10)); the derived relationship of the riverbed elevation would then be utilized for calculating the probability of the riverbed elevation exceeding a given magnitude named the exceedance probability. Eventually, the resulting exceedance probabilities of the riverbed elevations and corresponding uncertainty factors selected would be applied in the establishment of the exceedance-probability calculation equation; thus, the reliability analysis for the rivebed elevation given at the particular cross-section could be efficiently carried out under the known magnitudes of the uncertainty factors. Overall, the detailed model development process could be addressed in three parts: establishment of the riverbed elevation estimation equation, the reliability quantification of the estimated riverbed elevation, and derivation of the exceedance-probability calculation.

Establishment of riverbed elevation estimation equation

While developing the proposed PM_MBEE_1D model, the underestimated risk and reliability of the simulated riverbed elevation at various cross-sections should be quantified by calculating the corresponding exceedance probabilities via the AFOSM approach due to the uncertainties in the runoff-related and sediment-related uncertainty factors. According to the introduction to the AFOSM in section 2.6, a functional relationship between the riverbed elevation and associated sensitive uncertainty factors should be derived in advance; thus, a linear regression equation of the final estimated riverbed elevation with the sensitive uncertainty factors identified in the previous section can be defined as:
formula
(12)
where and serve as the estimated riverbed elevation; and the number of sensitive uncertainty factors, respectively, accounts for those above-mentioned runoff-related and sediment-related uncertainty factors (see Table 6); and are the associated regression coefficients. Table 7 illustrates the regression coefficients of sensitive uncertainty factors used in the riverbed elevation estimation equation at the specific cross-sections, 1st (CS_1st), 72nd (CS_72nd), 100th (CS_100th), 118th (CS_118th), 136th (CS_136th) and 146th (CS_146th). It can be seen that despite Equation (12) being a linear equation, the resulting determination coefficients R2 from 0.917 to 0.988, on average, approaching 0.95, revealing that Equation (12) can reasonably describe the varying tendency of the riverbed elevation estimate with the change in the sensitive runoff-related and sediment-related certainty factors with high likelihood. By coupling Equation (12) with the AFOSM approach, reliability analysis for the estimated riverbed elevations at various cross-sections within the study area could be carried out under consideration of the variations in the runoff-related and sediment-relation factors.
Table 6

Summary of the sensitive uncertainty factors regarding the estimation of the riverbed elevation

Type of uncertainty factorUncertainty factorSymbol of factors
Runoff-related factor Rainfall factor Rainfall depth θ1 
Maximum rainfall intensity θ2 
Upstream baseflow θ3 
Maximum tide depth θ4 
Roughness coefficient θ5 
SAC-SMA parameters UZTWM θ6θ16 
UZFWM 
UZK 
PCTIM 
ADIMP 
ZPERC 
LZTWM 
LZFSM 
LZSK 
DF_L 
DF_H 
Sediment-related factor Initial riverbed elevation Z_ini θ17 
Adaptative length Ls θ18 
QQs relationship Regression coefficient α θ19 
Regression coefficient β θ20 
Type of uncertainty factorUncertainty factorSymbol of factors
Runoff-related factor Rainfall factor Rainfall depth θ1 
Maximum rainfall intensity θ2 
Upstream baseflow θ3 
Maximum tide depth θ4 
Roughness coefficient θ5 
SAC-SMA parameters UZTWM θ6θ16 
UZFWM 
UZK 
PCTIM 
ADIMP 
ZPERC 
LZTWM 
LZFSM 
LZSK 
DF_L 
DF_H 
Sediment-related factor Initial riverbed elevation Z_ini θ17 
Adaptative length Ls θ18 
QQs relationship Regression coefficient α θ19 
Regression coefficient β θ20 
Table 7

Regression coefficients of sensitive uncertainty factors regarding the riverbed elevation estimation equations at the specific cross-sections

Cross-sectionβ0β1β2β3β4β5β6β7β8β9β10β11β12β13β14β15β16β17β18β19β20R2
CS_1st −0.031 −0.0001 0.0084 −0.003 −0.001 −6.700E-06 2.68E-05 −1.70E-05 0.002 −0.017 0.124 8.95E-05 4.21E-05 3.40E-05 −0.016 −2.00E-08 −1.60E-06 −0.052 1.595 1.036 3.29E-06 0.938 
CS_72th −4.854 0.0001 0.0052 1.022 −0.001 2.720E-05 −1.20E-04 1.63E − 05 −0.009 0.069 0.059 −6.20E-05 −1.30E-04 −9.30E-05 −0.005 6.00E-08 3.13E-06 0.152 −0.886 1.109 −1.10E-05 0.988 
CS_100th 3.307 0.0000 0.0008 −0.023 −0.010 6.500E-04 6.30E-04 2.16E-04 −0.040 −0.586 −1.943 2.99E-04 3.86E-04 4.82E-05 0.172 6.20E-07 −2.70E-05 −0.156 22.897 0.962 −1.60E-06 0.949 
CS_124th −0.416 0.0003 0.0003 0.008 0.004 7.130E-05 −7.70E-04 −1.10E-05 0.091 0.030 0.077 1.23E-04 −4.200E-04 −2.900E-04 −0.060 3.40E-07 5.91E-06 0.670 4.418 0.998 −7.30E-06 0.965 
CS_136th 14.023 0.0011 0.0003 0.005 0.009 1.590E-04 −2.11E-03 −4.90E-05 0.195 0.583 −0.165 −1.00E-04 −1.35E-03 −9.90E-04 −0.018 4.90E-07 4.92E-05 −0.002 −38.442 0.951 −2.10E-05 0.917 
CS_146th −8.532 0.0010 −0.0090 0.670 0.074 2.360E-04 −1.11E-03 3.81E-05 −0.253 1.465 −0.732 −1.06E-03 −1.06E-03 −3.40E-04 0.409 6.30E-07 3.53E-05 3.630 −15.825 1.012 −2.00E-05 0.938 
Cross-sectionβ0β1β2β3β4β5β6β7β8β9β10β11β12β13β14β15β16β17β18β19β20R2
CS_1st −0.031 −0.0001 0.0084 −0.003 −0.001 −6.700E-06 2.68E-05 −1.70E-05 0.002 −0.017 0.124 8.95E-05 4.21E-05 3.40E-05 −0.016 −2.00E-08 −1.60E-06 −0.052 1.595 1.036 3.29E-06 0.938 
CS_72th −4.854 0.0001 0.0052 1.022 −0.001 2.720E-05 −1.20E-04 1.63E − 05 −0.009 0.069 0.059 −6.20E-05 −1.30E-04 −9.30E-05 −0.005 6.00E-08 3.13E-06 0.152 −0.886 1.109 −1.10E-05 0.988 
CS_100th 3.307 0.0000 0.0008 −0.023 −0.010 6.500E-04 6.30E-04 2.16E-04 −0.040 −0.586 −1.943 2.99E-04 3.86E-04 4.82E-05 0.172 6.20E-07 −2.70E-05 −0.156 22.897 0.962 −1.60E-06 0.949 
CS_124th −0.416 0.0003 0.0003 0.008 0.004 7.130E-05 −7.70E-04 −1.10E-05 0.091 0.030 0.077 1.23E-04 −4.200E-04 −2.900E-04 −0.060 3.40E-07 5.91E-06 0.670 4.418 0.998 −7.30E-06 0.965 
CS_136th 14.023 0.0011 0.0003 0.005 0.009 1.590E-04 −2.11E-03 −4.90E-05 0.195 0.583 −0.165 −1.00E-04 −1.35E-03 −9.90E-04 −0.018 4.90E-07 4.92E-05 −0.002 −38.442 0.951 −2.10E-05 0.917 
CS_146th −8.532 0.0010 −0.0090 0.670 0.074 2.360E-04 −1.11E-03 3.81E-05 −0.253 1.465 −0.732 −1.06E-03 −1.06E-03 −3.40E-04 0.409 6.30E-07 3.53E-05 3.630 −15.825 1.012 −2.00E-05 0.938 

Calculation of exceedance probabilities of mobile riverbed elevations̀

As the regression coefficients of riverbed elevation estimation equations at various cross-sections are calibrated as listed in Table 7, the corresponding exceedance probabilities to the estimated riverbed elevations could then be quantified via the AFOSM approach with the statistical properties, mean, and variance of the uncertainty factors given in advance. Since the estimated riverbed elevation exhibits a larger dispersion in space, the specific magnitudes are assigned by adding an increment of 0.25 m to the initial elevation of each cross-section, 0 m (CS_1ST), 44 m (CS_72nd), 119 m (CS_72th), 219 m (CS_124th), 272 m (CS_136th), 334 m (CS_146th). Figure 13 shows the exceedance probabilities of the specific magnitudes of the riverbed elevations at six cross-sections in the study area, indicating that the exceedance probability has a noticeable increase with the cross-section located from the downstream to the upstream. For illustration, at the 146th cross-section, when the riverbed elevation rises from 336 to 338.25 m, the resulting exceedance probability dramatically declines from 0.97 to 0.54 with a large difference of 2.25 m; however, as for the first cross-section, the exceedance probability sharply drops from 1 to 0 within a short difference in the elevation (about 0.2 m); this is because the riverbed at the upstream with high slope might be scoured due to low water level and high runoff velocity; on the contrary, the stilting riverbed would be commonly found at the downstream due to low river velocity (Wang 1999). Altogether, the corresponding exceedance probability to a specific river elevation exhibits a noticeable difference in space.
Figure 13

Exceedance probabilities of the riverbed elevations at the specific cross-sections.

Figure 13

Exceedance probabilities of the riverbed elevations at the specific cross-sections.

Close modal

In conclusion, the exceedance probability of the riverbed elevation exhibits a noticeable variation in space; namely, the results from reliability analysis for the estimated riverbed elevations should change with space. Therefore, it is proven that the estimation of the riverbed elevation is significantly impacted due to the uncertainties in the runoff-related and sediment-related factors; that is, the inherent spatial variations of the riverbed elevations should be considered in the sediment-transport simulation for the mobile riverbed.

Derivation of the exceedance-probability calculation equation

Despite the reliability of the riverbed elevations at various cross-sections being quantified via the AFOSM approach, it might take expensive computation time (Wu et al. 2017). In this study, to efficiently achieve the goal of implementing the reliability assessment for the estimated riverbed elevation, an exceedance probability calculation equation could be derived via the logistic regression analysis based on Equation (11). After establishing the above logistic regression equation, the model inputs corresponding to the exceedance probability of the specific magnitudes of the model outputs are supposed to be identified among the potential impact factors (Wang et al. 2023). According to the results from the sensitivity quantification to estimate the riverbed elevation, 23 runoff-related and sediment-related uncertainty factors should be considered. However, among 23 uncertainty factors, except for the observed data-derived factors (i.e., the rainfall factors, baseflow, maximum tide depth, initial riverbed elevation, and roughness coefficient), the remaining uncertainty factor should be calibrated in advance via the numerical models with the relevant observations; it might hardly effectively respond to the impact of the change in the hydrological and topographical features in the watershed on the variation in the riverbed. Thereby, the above-observed data factors are classified into two types: hydrological factors (rainfall depth and upstream baseflow) and topographical factors (initial riverbed elevation and river-channel roughness coefficient), and they are adopted in establishing the logistic regression equations for calculating the exceedance probability of the estimated riverbed elevation as:
formula
(13)
where accounts for the riverbed elevation at the cross-section; serves as the baseflow at the upstream catchment; and stand for the initial and estimated riverbed elevation, respectively; denotes the river-channel roughness coefficient; and are the regression coefficients. Note that after calculating the exceedance probability of the estimated riverbed elevation, the corresponding reliability, 1-, could be achieved through Equation (13). As well as calculating the exceedance probability of the estimated riverbed elevation due to the hydrological and topographical factors given, the corresponding estimated riverbed elevations at various cross-sections subject to the desired reliability (i.e., one minus exceedance probability) could be obtained by modifying Equation (13) as:
formula
(14)
where is named the probabilistic-based final estimated riverbed elevation.

In this study, the AFSOM approach is employed to reproduce a significant number of the exceedance probabilities of the riverbed elevations at the various cross-sections under different conditions of various rainfall and topographical factors given with the statistical properties of the remaining runoff-related and sediment-related uncertainty factors; then, their coefficients in Equation (13) could be calibrated via the multivariate regression analysis. Table 7 lists the regression coefficients of uncertainty factors of interest and deterministic coefficient (R2) of the exceedance-probability calculation equations at the 1st, 72nd, 100th, 124th, 136th, and 146th cross-sections located from the downstream and upstream within the study area. In referring to Table 7, the corresponding R2 value to the various cross-sections, on average, reaches 0.95, meaning that the resulting exceedance-probability calculation equations could reasonably reflect the change in the reliability of the estimated riverbed elevation with the rainfall depth, baseflow, initial riverbed elevation, and river-channel roughness coefficient. In particular, according to the absolute values of the regression coefficients of hydrological and topographical factors, the absolute coefficient of the initial elevation is significantly superior to those of the remaining factors, especially for the first cross-section (about 193), implying that the reliabilities of the estimated riverbed elevations at the downstream cross-sections are mainly affected due to the corresponding initial elevations.

In total, despite the identical uncertainty factors being utilized in the reliability quantification of the estimated riverbed elevation, their variation due to the varying hydrological factor (rainfall depth and baseflow) and topographical features (roughness coefficient and initial elevation) could be accordingly achieved; this is advantageous to assess the corresponding reliability of the estimated riverbed elevation to the change in the hydrological and topographical factors via the derived exceedance-probability equations without using the AFOSM approach to save expensive computation time.

Model validation and application

In this study, to validate the applicability of the proposed PM_MBEE_1D model in the reliability quantification of the estimated section-based riverbed elevation via the model application procedure (see Section 3.8.2), the estimated riverbed elevations of three validation events at the specific sections via the CCHE1D model with the corresponding runoff-related and sediment-based factors given (see Table 8). The model validation could then be made by comparing the initial riverbed elevation with the final estimated elevation based on the associated exceedance probability, which is calculated via Equation (12) under consideration of the hydrological and geographical factors known. In addition, by varying rainfall and geographical factors used in Equation (12), the logistic regression equation regarding the estimated riverbed elevation could be applied in the evaluation of the effect of their spatial variations on the reliability of the specific riverbed elevations at various cross-sections. As well as the reliability assessment for the estimated riverbed elevation of interest via Equation (12), the estimation of the riverbed elevation at a specific cross-section with a likelihood could be achieved via Equation (13). Furthermore, through the proposed PM_MBEE_1D model with the runoff-related factor given, the riverbed scour and siltation could be identified by comparing the resulting final estimated riverbed elevation with the initial values; in detail, in case of the final estimated riverbed elevation at a cross-section being more significant than the initial elevation, the riverbed would be stilted; on the contrary, the scouring riverbed is possibly detected based on the lower final estimated elevation than initial one. The detailed results and discussion could be referred to as follows:

Table 8

Conditions of rainfall and physiographic factors at the cross-sections used in the model validation

Cross-sectionValidation eventUncertainty factor
Final estimated riverbed elevation (m)
Rainfall depth (mm)Baseflow (m3/s)Initial elevation (m)Roughness coefficient
CS_1st EV1 326.3 340.1 0.544 0.027 0.518 
EV2 452.4 294.5 0.521 0.027 0.480 
EV3 217.8 334.3 0.559 0.023 0.549 
CS_72th EV1 326.3 340.1 46.168 0.034 46.395 
EV2 452.4 294.5 46.048 0.035 46.236 
EV3 217.7 334.3 45.838 0.035 45.992 
CS_100th EV1 326.3 340.1 121.281 0.037 120.947 
EV2 452.4 294.5 121.600 0.039 120.803 
EV3 217.7 334.3 122.230 0.039 121.456 
CS_124th EV1 326.3 340.1 219.233 0.040 219.387 
EV2 452.4 294.5 219.181 0.041 219.331 
EV3 217.7 334.3 219.700 0.041 219.770 
CS_136th EV1 326.3 340.1 270.346 0.041 270.478 
EV2 452.4 294.5 270.319 0.041 270.493 
EV3 217.7 334.3 271.463 0.042 271.581 
CS_146th EV1 326.3 340.13 328.826 0.042 328.849 
EV2 452.4 294.5 328.998 0.042 329.187 
EV3 217.7 334.3 329.369 0.043 329.387 
Cross-sectionValidation eventUncertainty factor
Final estimated riverbed elevation (m)
Rainfall depth (mm)Baseflow (m3/s)Initial elevation (m)Roughness coefficient
CS_1st EV1 326.3 340.1 0.544 0.027 0.518 
EV2 452.4 294.5 0.521 0.027 0.480 
EV3 217.8 334.3 0.559 0.023 0.549 
CS_72th EV1 326.3 340.1 46.168 0.034 46.395 
EV2 452.4 294.5 46.048 0.035 46.236 
EV3 217.7 334.3 45.838 0.035 45.992 
CS_100th EV1 326.3 340.1 121.281 0.037 120.947 
EV2 452.4 294.5 121.600 0.039 120.803 
EV3 217.7 334.3 122.230 0.039 121.456 
CS_124th EV1 326.3 340.1 219.233 0.040 219.387 
EV2 452.4 294.5 219.181 0.041 219.331 
EV3 217.7 334.3 219.700 0.041 219.770 
CS_136th EV1 326.3 340.1 270.346 0.041 270.478 
EV2 452.4 294.5 270.319 0.041 270.493 
EV3 217.7 334.3 271.463 0.042 271.581 
CS_146th EV1 326.3 340.13 328.826 0.042 328.849 
EV2 452.4 294.5 328.998 0.042 329.187 
EV3 217.7 334.3 329.369 0.043 329.387 

Model validation

To demonstrate the applicability of the exceedance probability calculation equations within the proposed PM_MBEE_1D model on estimating the stochastically-based final riverbed elevations, three validation events with various simulations of the final estimated elevations at different cross-sections are simulated in advance via the CCHE1D model under consideration of the uncertainty factors of interest. In detail, the exceedance probability of the initial riverbed elevation could be achieved via Equation (12) based on the rainfall and geographical factors given (see Table 8).

Figure 14 shows that the exceedance probabilities of the initial riverbed elevations at the downstream cross-sections CS_1st, on average, merely approximate 0.017, meaning that the final estimated riverbed elevations have high reliability (nearly 0.983) to be less than the initial elevations, namely, the riverbed might be silted at the downstream; this can be proven that the final estimated riverbed elevation employing the CCHE1D model is undoubtedly more significant than the initial riverbed. Also, siltation would occur at section CS_100th according to the final estimated riverbed elevation lower than the initial elevation with an exceedance probability nearly equal to zero.
Figure 14

Exceedance probabilities of the initial riverbed elevations (Zini) calculated via the proposed PM_MBEE_1D model in comparison to final estimated ones (Zfin_CCHE1D) via the CCHE1D model. (1) CS_1st, (2) CS_72nd, (3) CS_100th, (4) CS_124th, (5) CS_136th, and (6) CS_146th.

Figure 14

Exceedance probabilities of the initial riverbed elevations (Zini) calculated via the proposed PM_MBEE_1D model in comparison to final estimated ones (Zfin_CCHE1D) via the CCHE1D model. (1) CS_1st, (2) CS_72nd, (3) CS_100th, (4) CS_124th, (5) CS_136th, and (6) CS_146th.

Close modal

However, in contrast with the cross-sections CS_1st and CS_100th, the exceedance probabilities of the initial riverbed elevations at the remaining cross-sections (i.e., CS_72nd, CS_124th, and CS_136th) reach from 0.75 to 0.999, indicating that the estimated riverbed elevations are less than the initial ones; thus, it can be verified that the scouring riverbeds could be found at the particular cross-sections. However, with respect to the upstream cross-section CS_100th, the associated exceedance probabilities of the initial riverbed elevations vary from 0.46 to 0.96 for different validation events; this reveals that the riverbed at this cross-section might be both scoured and silted based on the known rainfall and geographical conditions.

In total, the quantified reliabilities of the estimated riverbed elevations by the proposed PM_MBEE_1D model could successfully account for the result from the comparison between the initial riverbed elevations and final estimated ones; namely, when a low exceedance probability is obtained, the final estimated riverbed elevation should be less than the initial one with high likelihood. That is to say, the results from the proposed PM_MBEE_1D model have a good agreement with the variation of the riverbed attributed to the spatial and temporal variation of the rainfall-induced runoff quantified via the CCHE1D model. As a result, the resulting exceedance probability of the final estimated riverbed from the proposed PM_MBEE_1D model could reasonably reflect the change in the riverbed due to uncertainties in the runoff-related and sediment-related factors on the elevation of the movable riverbed.

Model application

Evaluation of variation in the rainfall depth
In this study, as well as demonstrating the proposed PM_MBEE_1D model, the proposed PM_MBEE_1D model could be applied to evaluate the effect of the variations in the rainfall and geographical factors on the estimated riverbed elevations. Also, according to the results of the model validation, the cross-sections CS_100th and CS_136th could be treated as the scouring and silting sections, respectively; thus, the above two sections are applied in the model application on the evaluation of the variation in the rainfall depth on change of the riverbed elevations. Table 9 lists the conditions of the rainfall depth and initial riverbed elevation used in the application of the PM_MBEE_1D model on quantifying the varying trend of the exceedance probabilities on the estimated riverbed elevations caused by the uncertainties in the rainfall and geographical factor (see Figure 15(1)). Given Figure 15, at the 136th cross-section (i.e., CS_136th), the exceedance probability of the estimated riverbed elevation markedly rises with the rainfall depth; in detail, when the rainfall depth changes from 200 to 1,000 mm, the corresponding exceedance probability has a significantly rapid increase from 0.001 to 0.997; this implies that at the cross-section CS_136th, the rainfall depth could boost the riverbed siltation during a heavy rainstorm. In contrast, at the cross-section CS_100th, the probability of the estimated riverbed elevation (around from 0.343 to 0.184) varies adversely with the rainfall depth, rising from 100 to 1,000 mm, revealing that at the 100th cross-section, the riverbed scour is more likely to be triggered due to the heavy rainfall. As a result, the spatial change in the riverbed elevation (i.e., scouring and siltation) is markedly influenced by the rainfall depth. It could be carried by the proposed PM_MBEE_1D model with varying rainfall depth.
Table 9

Conditions of rainfall and physiographic factors at the cross-sections used in the model application

Application caseCross-sectionUncertainty factor
Final estimated riverbed elevation (m)
Rainfall depth (mm)Baseflow (m3/s)Initial elevation (m)Roughness coefficient
CS_100th 100
300
500
700
900
1,000 
150 273 0.04 274 
CS_136th 122 121.2 
II CS_136th 500 150 274 0.04 273.0
273.5
274.0
274.5
275.0
275.5
276.0 
Application caseCross-sectionUncertainty factor
Final estimated riverbed elevation (m)
Rainfall depth (mm)Baseflow (m3/s)Initial elevation (m)Roughness coefficient
CS_100th 100
300
500
700
900
1,000 
150 273 0.04 274 
CS_136th 122 121.2 
II CS_136th 500 150 274 0.04 273.0
273.5
274.0
274.5
275.0
275.5
276.0 
Figure 15

Exceedance probabilities of riverbed elevations at different cross-sections calculated via the proposed PM_MBEE_1D model for the model application. (1) Application case I and (2) Application case II.

Figure 15

Exceedance probabilities of riverbed elevations at different cross-sections calculated via the proposed PM_MBEE_1D model for the model application. (1) Application case I and (2) Application case II.

Close modal
Evaluation of variation in the initial riverbed elevation

In addition to the rainfall depth, the geographical factor (i.e., initial riverbed elevation) is verified to quantify the effect of its uncertainty on the estimation of the riverbed elevation, as shown in Figure 15(2), in which the final riverbed elevation is estimated by adding an increment of 0.5 m into the elevation (273 m) under the condition of the rainfall depth of 500 mm. Given Figure 15, at the cross-section CS_136th, the exceedance probability of the final riverbed elevation of 274 m has a significant decrease with the initial riverbed elevation. In detail, the corresponding exceedance probability to the initial riverbed elevation over 274 m (i.e., silting probability) significantly reaches 0.999 under an initial elevation of 274 m given; in contrast, in the case of the initial elevation being less than 274 m, the exceedance probability (i.e., scouring probability) sharply decline to 0.001. This implies that at the cross-section CS_136th, siltation takes place with a high likelihood attributed to its mild slope at the downstream.

Estimation of the probabilistic-based riverbed elevation
Apart from the evaluation of rainfall and geographical uncertainty factors, the proposed PM_MBEE_1D model could be utilized to estimate the final riverbed elevation with design reliability (called potential riverbed elevation) under consideration of rainfall and geographical factors known using Equation (13). Figure 16 presents that the final estimated riverbed elevations at various cross-sections for the desired exceedance probabilities from 0.1 to 0.99 in the case of the hydrological and geographical factors regarding the first validation event (see Table 9); it can be seen that the resulting estimated riverbed elevations adversely vary with the exceedance probability, meaning that less riverbed elevation might lead to the higher exceedance probability. In detail, the corresponding final estimated riverbed elevations to the downstream cross-sections CS_1st, CS_72nd, and CS_100th obviously depart from the initial elevations for various desired exceedance probabilities; on the contrary, the estimated riverbed elevations at the remaining cross-sections might be lower than the initial ones for the high exceedance probability.
Figure 16

Summary of final estimated riverbed elevations at specific cross-sections via the proposed PM_MBEE_1D model with a desired reliability of 0.8. (1) CS_1st, (2) CS_72nd, (3) CS_100th, (4) CS_124th, (5) CS_136th, and (6) CS_146th.

Figure 16

Summary of final estimated riverbed elevations at specific cross-sections via the proposed PM_MBEE_1D model with a desired reliability of 0.8. (1) CS_1st, (2) CS_72nd, (3) CS_100th, (4) CS_124th, (5) CS_136th, and (6) CS_146th.

Close modal

In summary, the riverbed elevations at the downstream cross-sections appear to have a stable varying trend in space, scouring (CS_1st and CS_100th), and silting (CS_72nd), whereas about the cross-sections at the upstream (CS_124th, CS_136th and CS_146th), scouring riverbed is frequently detected, but the riverbed might be scoured with a low chance. As a result, the probabilistic-based riverbed elevations given within the study area are significantly consistent with the results from the uncertainty analysis via the AFOSM approach as a reference to riverbed improvement.

This study aims to develop a probabilistic movable riverbed elevation estimation (PM_MBEE_1D) model to quantify the effect of uncertainties in the runoff-related and sediment-related factors on the estimation of the riverbed elevations at various cross-sections in terms of the corresponding exceedance probability. The 1,000 simulations of the riverbed elevations and corresponding uncertainty factors are achieved via the MMCS with the historical topographical, hydrological, and sediment data in the Jhuosdhuei River watershed for the model development and validation. As a result, the exceedance probability calculated via the proposed PM_MBEE_1D model could reasonably respond to the change in the riverbed elevation (scouring and siltation) in comparison to the given initial elevation under consideration of the hydrological and geographical factors given with high accuracy. Also, under consideration of the uncertainties in the riverbed elevations at the cross-sections with a desired reliability (named probabilistically-based riverbed elevation) could be effectively achieved by means of the proposed PM_MBEE_1D model with given runoff-related and sediment-related factors.

This published paper was funded by the National United University Projects (grant 112-NUUPRJ-04).

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

The authors declare there is no conflict.

Ajami
N. K.
,
Gupta
H.
,
Wagener
T.
&
Sorooshian
S.
2004
Calibration of a semi-distributed hydrologic model for streamflow estimation along a river system
.
Journal of Hydrology
298
,
112
135
.
Ann
K. H.
&
Steihschneider
S.
2017
Time-varying suspended sediment-discharge rating curves to estimate climate impacts on fluvial sediment transport
.
Hydrological Processes
32
(
1
),
102
117
.
Beckers
F.
,
Noack
M.
&
Wieprecht
S.
2018
Uncertainty analysis of a 2D sediment transport model: an example of the Lower River Salzach
.
Journal of Soils Sediments
18
,
3133
3144
.
Chang
C. H.
,
Yang
J. C.
&
Tung
Y. K.
1993
Sensitivity and uncertainty analyses of a sediment transport model: a global approach
.
Journal of Stochastic Hydrology and Hydraulic
7
(
4
),
299
314
.
Chang
C. H.
,
Yang
J. C.
&
Tung
Y. K.
1997
Incorporate marginal distributions in point estimate methods for uncertainty analysis
.
Journal of Hydraulic Engineering
123
(
3
),
244
251
.
Chao
Y. C.
,
Chen
C. W.
,
Li
H. C.
&
Chen
Y. M.
2018
Riverbed migration in Western Taiwan under climate change
.
Water
10
.
doi:10.3390/w10111631
.
Chen
T. Y.
,
Hung
C. Y.
,
Chiang
Y. C.
,
Hsieh
M. L.
&
Capart
H.
2022
A stochastic model of geomorphic risk due to episodic river aggradation and degradation
.
Engineering Geology
309
.
https://doi.org/10.1016/j.enggeo.2022.106845
.
DHI Software
1992
User Guides and Reference Manuals for MIKE11 Model
.
Danish Hydraulic Institute
,
Denmark
.
Ding
Y.
&
Langendoen
E.
2016
Simulation and control of sediment transport due to dam removal
.
Journal of Applied Water Engineering and Research.
6
(
2
),
95
108
.
Ganji
A
&
Jowkarshorijeh
L.
2012
Advance first order second moment (AFOSM) method for single reservoir operation reliability analysis: a case study
.
Stochastic Environmental Research and Risk Assessment
26
(
1
),
33
42
.
doi:10.1007/s00477-011-0517-1
.
Garbrecht
J.
,
Kuhnle
R.
&
Alonso
C.
1995
A sediment transport capacity formulation for application to large channel networks
.
Journal of Soil and Water Conservation
50
(
5
),
527
529
.
Guo
Q. C.
&
Jin
Y. C.
2002
Modeling nonuniform suspended sediment transport in Alluvial Rivers
.
Journal of Hydraulic Engineering
128
(
9
),
839
847
.
Hancock
G. R.
,
Coulthard
T. J.
&
Lowry
J. B. C.
2016
Predicting uncertainty in sediment transport and landscape evolution-the influence of initial surface conditions
.
Computers and Geosciences
90
,
117
130
.
Helal
E. Y.
,
Nassralla
T. H.
&
Abdelaziz
A. A.
2013
Minimizing of scour downstream hydraulic structures using sills
.
International Journal of Civil Structural Engineering
3
(
3
),
591
602
.
Huang
J. V.
&
Greimann
B.
2007
User's Manual for SRH-1D V2.1
.
Bureau of Reclamation, U.S. Department of the Interior
,
Denver
.
Jagallah
M.
,
Camenen
B.
,
Paquier
A.
&
Jodeau
M.
2019
An optimized use of limited ground-based topographic data for river applications
.
International Journal of Sediment Research
34
,
216
225
.
Kampf
L.
,
Mueller
P.
,
Hollerer
H.
,
Plessen
B.
,
Naumann
R.
,
Thoss
H.
,
Guntner
A.
,
Mera
B.
&
Brauer
A.
2015
Hydrological and sedimentological processes of flood layer formation in Lake Mondsee
.
The Depositional Record
1
(
1
),
18
37
.
Kim
H. Y.
&
Julien
P. Y.
2018
Hydraulic thresholds to mitigate sedimentation Problems at Sangju Weir, South Korea
.
Journal of Hydraulic Engineering
144
(
6
),
5018005
.
Koren
V. I.
,
Smith
M.
,
Wang
D.
&
Zhang
Z.
2000
Use of soil property data in deriving conceptual rainfall–runoff model parameters
. In
15th Conference on Hydrology
,
Long Beach, California
, pp.
103
106
.
Lau
T. W.
&
Afshar
N. R.
2013
Effect of roughness on discharge
.
Journal of Civil Engineering Science and Technology
4
(
3
),
29
33
.
Lin
C. P.
,
Wang
Y. M.
,
Tfwala
S. S.
&
Chen
C. N.
2014
The variation of riverbed material due to tropical storms in Shi-Wen River, Taiwan
.
The Scientific World Journal
2014
.
https://doi.org/10.1155/2014/580936.
Liu
H.
,
Yi
Y.
&
Jin
Z.
2021
Sensitivity analysis of flash flood hazard on sediment load characteristics
.
Frontiers in Earth Science
9
.
doi:10.3389/feart.2021.683453
.
Liu
H.
,
Du
J.
&
Yi
Y.
2022
Reconceptualising flood risk assessment by incorporating sediment supply
.
Catena
217
,
106503
.
https://doi.org/10.1016/j.catena.2022.106503
.
Liu
M.
,
Lv
S.
,
Qiao
Q.
&
Song
L.
2023
Design and numerical simulation of the headworks in the Shizuishan Section of the Yellow River
.
Sustainability
15
,
4564
.
https://doi.org/10.3390/su15054564
.
Malkawi
A. I. H.
,
Hassan
W. F.
&
Abdulla
F. A.
2000
Uncertainty and reliability analysis applied to slope stability
.
Structural Safety
22
,
161
187
.
Mustafa
H. Q.
,
Altinakar
M. S.
,
Vieira
D. A. N.
&
Alidee
B.
2008
Application of Tabu search algorithm with a coupled ANNAGNPS-CCHE1D model to optimize agriculture land use
.
Journal of the American Water Resources Assocation
44
(
4
),
866
878
.
Nataf
A.
1962
Determination des distributions don't les marges sontdonnees
.
Comptes Rendus L'académie Sci.
225
,
42
43
.
Neuhold
C.
,
Stanzel
P.
&
Nachtnebel
H. P.
2009
Incorporating river morphological changes to flood risk assessment: uncertainties, methodology and application
.
Natural Hazards and Earth System Sciences
9
,
789
799
.
Norouzi
S.
,
Kolahdoozan
M.
&
Zarrati
A. R.
2022
Development of a 2D depth-averaged model for calculating scouring and deposition in alluvial streams
.
Scientia Iranica
29
(
3
),
941
950
.
Pfeiffer
A. M.
,
Collins
B. D.
,
Anderson
S. W.
,
Montgomery
D. R.
&
Istanbulluoglu
E.
2019
River bed elevation variability reflects sediment supply, rather than peak flows, in the uplands of Washington state
.
Water Resources Research
55
.
https://doi.org/10.1029/2019WR025394
.
Proffitt
G. T.
&
Sutherland
A. J.
1980
Transport of nonuniform sediments
.
Journal of Hydraulic Research
21
(
1
),
33
43
.
Riahi-Madvar
H.
&
Seifi
A.
2018
Uncertainty analysis in bed load transport prediction of gravel bed rivers by ANN and ANFIS
.
Arabian Journal of Geosciences
11
,
688
707
.
Ruark
M. D.
,
Niemann
J. D.
,
Greimann
B. P.
&
Arabi
M.
2011
Method for assessing impacts of parameter uncertainty in sediment transport modeling applications
.
Journal of Hydraulic Engineering
137
(
7
),
623
636
.
Sabatine
S. M.
,
Niemann
J. D.
&
Greimann
B. P.
2015
Evaluation of parameter and model uncertainty in simple applications of a 1D sediment transport model
.
Journal of Hydraulic Engineering
141
(
5
),
04015002
.
doi:10.1061/(ASCE)HY.1943-7900.0000992
.
Sanz-Ramos
M.
,
Blade
E.
,
Gonzalez-Escalona
F.
,
Oilvaras
G.
&
Aragon-Hernandez
J. L.
2021
Interpreting the manning roughness coefficient in overland flow simulations with coupled hydrological-hydraulic distributed models
.
Water
12
(
23
),
3433
.
https://doi.org/10.3390/w13233433
.
Shen
D.
,
Jia
Y.
,
Altinakar
M.
&
Bingner
R. L.
2016
GIS-based channel flow and sediment transport simulation using CCHE1D coupled with AnnAGNPS
.
Journal of Hydraulic Research
.
doi:10.1080/00221686.2016.1168883
.
Simpson
G.
&
Schlunegger
F.
2003
Topographic evolution and morphology of surfaces evolving in response to coupled fluvial and hillslope sediment transport
.
Journal of Geophysical Research
108
(
b6
),
2300
.
doi:10.1029/2002JB002162
.
Song
S. T.
,
Hu
T. F.
&
Chiou
D. J.
2020
Influence of riverbed scour on the performance of bridges subjected to lateral seismic loads
.
Journal of Earthquake Engineering
26
(
5
),
2251
2282
.
Teraguchi
H.
,
Nakagawa
H.
&
Kawaike
K.
2011
Effects of hydraulic structures on river morphological processes
.
International Journal of Sediment Research
26
(
3
),
283
303
.
Tung
Y. K.
&
Mays
L. W.
1981
Risk models for flood levee design
.
Water Resources Research
17
(
4
),
833
841
.
U.S. Army Corps of Engineering
1993
HEC-6: Scour and Deposition in Rivers and Reservoirs Use's Manual
.
Wang
Z. Y.
1999
Experimental study on scour rate and river bed inertia
.
Journal of Hydraulic Research
37
(
1
),
17
37
.
Wu
W.
,
Wang
S. S. Y.
&
Jia
Y.
2000
Nonuniform sediment transport in alluvial rivers
.
Journal of Hydraulic Research
38
(
6
),
427
434
.
Wu
W.
,
Vieira
D. A.
&
Wang
S. S. Y.
2004a
A 1-D numerical model for nonuniform sediment transport under unsteady flows in channel networks
.
Journal of Hydraulic Engineering, ASCE
130
(
9
),
914
923
.
Wu
W.
,
Vieira
D. A.
&
Wang
S. S. Y.
2004b
Ono-dimensional numerical model for nonuniform sediment transport under unsteady flows in channel networks
.
Journal of Hydraulic Engineering
130
(
9
),
914
923
.
Wu
S. J.
,
Hsu
C. T.
&
Chang
C. H.
2021b
Stochastic modeling of gridded short-term rainstorms
.
Hydrology Research
52
,
876
904
.
Wu
S. J.
,
Mai
J. S.
,
Lin
Y. H.
&
Yeh
K. C.
2022
Modeling probabilistic-based reliability analysis for irrigation water supply due to uncertainties in hydrological and irrigation factors
.
Sustainability
14
(
9
),
12747
.
https://doi.org/10.3390/su141912747.
Zhang
Y.
&
Langendoen
E. J.
1998
An Introduction to BEAMS (Bed and Bank Erosion Analysis Model for Streams) (Technical Report No. CCHE-TR-98-2)
.
Center for Computational Hydroscience and Engineering, The University of Mississippi. Oxford, MS
.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY-NC-ND 4.0), which permits copying and redistribution for non-commercial purposes with no derivatives, provided the original work is properly cited (http://creativecommons.org/licenses/by-nc-nd/4.0/).