A hydraulic modelling approach to study flood sediment deposition in floodplain lakes

Abandoned river channels on alluvial floodplains represent areas where sediments, organic matter, and pollutants preferentially accumulate during overbank flooding. Theoretical models describing sedimentation in floodplain lakes recognize the different stages in their evolution, where the threshold for hydrological connectivity increases in older lakes as a plug‐bar develops. Sedimentary archives collected from floodplain lakes are widely used to reconstruct ecological and hydrological dynamics in riverine settings, but how floodplain lake evolution influences flow velocities and sedimentation patterns on an event scale remains poorly understood. Here we combine sediment samples collected in and around a floodplain lake with hydraulic modelling simulations to examine inundation, flow velocity, and sedimentation patterns in a floodplain lake along the Trinity River at Liberty, Texas. We focus our analyses on an extreme flood event associated with the landfall of Hurricane Harvey in August 2017 and develop a series of alternative lake bathymetries to examine the influence of floodplain lake evolution on flow velocity patterns during the flood. We find that sediments deposited in the lake after the Hurricane Harvey flood become thinner and finer with distance from the tie‐channel in accordance with simulated flow velocities that drop with distance from the tie‐channel. Flow velocity simulations from model runs with alternative plug‐bar geometries and lake depths imply that sedimentation patterns will shift as the lake evolves and infills. The integration of sediment sampling and hydraulic model simulations provides a method to understand the processes that govern sedimentation in floodplain lakes during flood events that will improve interpretations of individual events in sedimentary archives from these contexts.

ent stages in their evolution, where the threshold for hydrological connectivity increases in older lakes as a plug-bar develops. Sedimentary archives collected from floodplain lakes are widely used to reconstruct ecological and hydrological dynamics in riverine settings, but how floodplain lake evolution influences flow velocities and sedimentation patterns on an event scale remains poorly understood. Here we combine sediment samples collected in and around a floodplain lake with hydraulic modelling simulations to examine inundation, flow velocity, and sedimentation patterns in a floodplain lake along the Trinity River at Liberty, Texas. We focus our analyses on an extreme flood event associated with the landfall of Hurricane Harvey in August 2017 and develop a series of alternative lake bathymetries to examine the influence of floodplain lake evolution on flow velocity patterns during the flood. We find that sediments deposited in the lake after the Hurricane Harvey flood become thinner and finer with distance from the tie-channel in accordance with simulated flow velocities that drop with distance from the tie-channel. Flow velocity simulations from model runs with alternative plug-bar geometries and lake depths imply that sedimentation patterns will shift as the lake evolves and infills. The integration of sediment sampling and hydraulic model simulations provides a method to understand the processes that govern sedimentation in floodplain lakes during flood events that will improve interpretations of individual events in sedimentary archives from these contexts.

K E Y W O R D S
flooding, floodplain lakes, hurricane Harvey, hydraulic modelling, sediment transport 1 | INTRODUCTION Alluvial floodplains constitute a major store of material transported by rivers including sediments, organic material, and pollutants (Asselman & Middelkoop, 1995;Ciszewski & Grygar, 2016;Phillips et al., 2004;Thoms, 2003;Walling & Bradley, 1989). Abandoned channels-floodplain depressions that form as a result of avulsion, meander cutoffs, or lateral channel migration-are particularly efficient sediment traps due to their geometry and proximity to the active river channel (Gilli et al., 2012;Munoz et al., 2015;Rowland et al., 2005;Toonen et al., 2012). Theoretical models describing the long-term evolution and infilling of abandoned channels generally recognize two major phases: a transitional phase and the abandoned phase (Gagliano & Howard, 1984;Toonen et al., 2012). The transitional phase follows lake formation and is marked by regular inundation of the lake (Gagliano & Howard, 1984;Toonen et al., 2012).
Sediments deposited in the lake during the transitional phase are typically coarse-grained sands deposited in thick sequences, and result in the formation of a plug-bar at the lake entrance, and narrowing and shallowing of the lake arms (Douglas Shields & Abt, 1989;Toonen et al., 2012). As a mature plug-bar develops, the lake floods less frequently and moves towards the abandoned phase (Amoros & Bornette, 2002;Hudson et al., 2012). Sedimentation during the abandoned phase is recognized as laminated fill, in which coarse-grained bedload accumulates in the lake during infrequent overbank flooding while finer-grained autochthonous material is deposited during nonflood conditions (Toonen et al., 2012. The process of infilling continues to decrease water depth in the lake until the abandoned channel is completely filled with sediment (Gagliano & Howard, 1984).
Controls on the overall rate, distribution, and characteristics of floodplain lake infilling are contingent on the sediment load of the main channel, the trapping efficiency of the lake, and the hydrological connectivity of the lake (Citterio & Piégay, 2009;Cooper & McHenry, 1989). Sediment concentrations in the river are mediated by watershed scale processes, such as the geology, land use, and basin climatology. Individual lake hydrological connectivity and trapping efficiency (i.e., the ability of the lake to retain sediments) depend on local geomorphic conditions and change over time as the lake evolves and infills (Citterio & Piégay, 2009;Douglas Shields & Abt, 1989).
Trapping efficiency is controlled by lake geometry (Citterio & Piégay, 2009;Constantine et al., 2010)-for example, the shape, channel slope, depth, and diversion angle from the river (Douglas Shields & Abt, 1989;Fisk, 1948). Hydrological connectivity represents the synchronicity between stage levels in the river and lake (Hudson et al., 2012), and depends on the magnitude and frequency of flooding on the river, and topographic features in and around the lake (Amoros & Bornette, 2002). Plug-bar and tie-channel morphology are particularly important in the transfer of water and sediment between the river and lake (Hudson et al., 2012;Rowland et al., 2005), partly because the increased elevation of the plug-bar over time increases the threshold for hydrological connectivity (Hudson et al., 2012). As such, younger floodplain lakes frequently receive water and sediment through tie-channels whereas older lakes become increasingly isolated from the main channel (Citterio & Piégay, 2009;Hudson et al., 2012).
Theoretical models and observations provide valuable parameters to understand long-term sediment accumulation in floodplain lakes, but they are less well suited to explain the event-to-event variation in flood deposits within sedimentary archives. Sedimentary archives from floodplain lakes are widely used to reconstruct ecological (Bartlein et al., 2011;Bhattacharya et al., 2016;Brugam & Munoz, 2018;Liu et al., 2012;McGlue et al., 2016;Munoz et al., 2014;2019) and hydrological (Leigh, 2018;Munoz et al., 2015Munoz et al., , 2018Toonen et al., 2015Toonen et al., , 2020 dynamics in riverine environments. In these latter applications, laminated sequences are interpreted in terms of sedimentation during floods, where coarse-grained event beds are used as proxies for flood occurrence (Munoz et al., 2015;Toonen et al., 2016) and magnitude (Toonen, 2015), and finer-grained material deposited during non-flood conditions can, for example, preserve geochemical and biological proxies (Bhattacharya et al., 2016).
The interpretation of these sedimentary archives relies on geomorphic stability of the floodplain and river-a condition that is challenging to meet given the dynamic nature of alluvial systems (Dee et al., 2018;Lecce et al., 2004).
The long-term process of floodplain lake evolution is likely to progressively affect patterns of sediment accumulation through its influence on hydrological connectivity and lake bathymetry, making individual event layers in sediment archives highly time and context dependent. Hydraulic models could offer an opportunity to better understand the processes that mediate sediment accumulation in floodplain lakes during floods without having to rely on extensive data collection. Here, we combine field-based sampling and observations of sediments in and around a floodplain lake with hydraulic modelling to examine patterns of sedimentation, flow velocity, and inundation extent in the context of an extreme flood. We focus our analysis on a floodplain lake of the Trinity River (Texas) and the recent extreme flooding associated with the landfall of Hurricane Harvey in August of 2017 (Blake & Zelinsky, 2018). We construct a hydraulic model using HEC-RAS (Hydrologic Engineering Center-River Analysis System), a publicly available modelling platform developed by the US Army Corps of Engineers that is commonly used in flood inundation mapping (Brunner, 2021a). We use grain-size analysis of sediment samples from point-bars, a tie-channel, and the lake-bottom to document how sediments are mobilized and deposited during an extreme flood, while hydraulic simulations of the same event provide insights into the flow velocities that give rise to these sedimentation patterns. We then use our hydraulic model to evaluate how floodplain lake evolution influences patterns of flow velocity by developing a series of alternative lake bathymetries where the plug-bar and lake depth are varied. These analyses demonstrate that the properties of the sediment layer deposited by a certain flood magnitude change as floodplain lake geometry and morphology evolve over time and that this signal is nonlinear. Our work also shows that hydraulic model can be used to study the characteristics of floods and associated sedimentation through time to improve interpretations of sedimentary archives from floodplain lakes.

| STUDY AREA: TRINITY RIVER AND PORT OF LIBERTY LAKE, TEXAS
The Trinity River is a major drainage of the Texas Gulf Plain, with a drainage area of $46,100 km 2 and mean annual discharge of 224 m 3 /s (TWDB, 2022), which empties into the Gulf of Mexico via Trinity Bay (Phillips et al., 2004). Below the Livingston dam (completed in 1969), the Trinity River occupies a low-relief valley bottom bounded by Quaternary-age terraces (Blum & Aslan, 2006;Blum et al., 2002), with a riverbed below modern sea level (Phillips et al., 2005;Smith & Mohrig, 2017). The floodplain contains lakes formed during the Holocene and enlarged meander scrolls from channels occupying the floodplain during the Pleistocene (Phillips & Slattery, 2008;Phillips et al., 2005). Flooding associated with precipitation from Hurricane Harvey in August 2017 resulted in the largest annual peak flow event on the Trinity River at Liberty since 1995 (USGS, 2022). Harvey-induced flooding inundated large areas of the Trinity River floodplain, and mobilized and deposited large amounts of sediment (Blake & Zelinsky, 2018).
Our study focuses on a floodplain lake located on the southwest border of Liberty, Texas, which is referred to herein as Port of Liberty Lake ( Figure 1). Port of Liberty Lake has an area of $0.4 km 2 and is located $85 km downstream of the Livingstone dam and $3 km downstream of a USGS stream gauge (ID 08067000) (USGS, 2022).
Based on historical USGS maps, the floodplain lake formed through a neck cut-off event between and 1986(USGS HTMC, 1984, 1986, which may have been influenced by the construction of an embankment and road (Port Drive) that cut off the upstream arm of the lake from the river. The lake behaves as what Citterio and Piégay (2009) refer to as a backwater lake, where water from the main river channel primarily enters the floodplain lake through the downstream arm because the embankment prevents water from entering through the upstream arm, as opposed to a lotic lake, where water enters via the upstream arm. Backwater lakes typically display a sediment gradient with thick deposits in the downstream lake arm, and little sediment accumulation in the upstream lake arm (Citterio & Piégay, 2009).

The sandy point bars at the downstream limb of the Port of Liberty
Lake support the inference that it is backwater lake (Supporting Information Figure S1). The Port of Liberty Lake represents a typical oxbow lake in the early stages of the abandoned phase, where an active tie-channel directs flow into the lake during high water events.

| Hydraulic modelling
To simulate inundation and flow velocities in the floodplain lake during flooding associated with precipitation from Hurricane Harvey (26-30 August), we employed the HEC-RAS 6.1 hydraulic modelling platform developed by the US Army Corps of Engineers (Brunner, 2021a). We chose HEC-RAS for its ability to simulate processes on different floodplain geomorphologies that can be manipulated through the internal GIS interface RASMapper (Yalcin, 2020).
We ran our simulation using the HEC-RAS 2D model to simulate overbank flow through a floodplain lake without predetermining the overflow locations (Quirogaa et al., 2016). The 2D HEC-RAS model solves the Saint-Venant equations through an implicit finite volume (IFV) algorithm in an unstructured computational mesh (Brunner, 2021a;Kumar et al., 2020). In RASMapper we generated an initial structured mesh with 25 m grid cells in the flow area, which we refined with $15 m grid cells for the river, lake, and tie-channel (Supporting Information Figure S2). Our unstructured mesh was further refined by adding breaklines along the banks of the river and lake, as well as other areas of rapid elevation change. Breaklines ensure a proper representation of topographic features and prevent water from 'leaking' between cells unrealistically in the simulation.
We ran an unsteady flow simulation based on the stage hydrograph from USGS gauge 08067000 at Liberty for the period 18 August (DEM)-derived normal depth boundary condition was used for both the downstream river (friction slope = 0.0075) and floodplain (friction slope = 0.005) boundaries to ensure that water could move out of the system (Supporting Information Figure S2). To classify land use types, we used the National Land Cover Database 2016 products (Dewitz, 2019) and the associated Manning's roughness values suggested by the HEC-RAS manual (Brunner, 2021b). To achieve model stability, the timestep for each model iteration was determined through the Courant-Friedrichs-Lewy condition (Equation 1), where T is the computational time step (s), x is the average cell size (m), V is the flood wave velocity (m/s), and C is the Courant number. If we assume x to be 25 m and the maximum V to be 5 m/ s, we find a T of approximately 5 s (Brunner, 2021b;Shustikova et al., 2019): With this time step, the Courant value remains <0.5 in the lake and floodplain and <1.5 in the river, which we determine to be stable (Brunner, 2021a). We calibrated the model by varying the Manning's roughness values of the river around Manning's values suggested by the HEC-RAS manual, and the slope value at the downstream boundary to match the discharge values measured by the streamflow gauge at Liberty. From a sensitivity analysis we found that, of the two variables just for calibrating, simulated discharge is primarily affected by the roughness values. Missing discharge measurements (n = 9) from before and after the flood peak were reconstructed using a rating curve based on Trinity River discharges levels below 1700 m 3 /s during the period 1996-2021 (Supporting Information Figure S3).

| Bathymetric data and scenarios
The where Modification elevation is the height added to each grid cell of the lake, Maximum modification elevation is the largest desired elevation change, C is a bathymetry depth-specific constant (0, 0.5, 1, respectively, for +1, +2, and +4), and Lake depth is the depth at the grid cell in the original bathymetry. Additionally, we created three alternative scenarios in which we modified the elevation of the plugbar: (1) a partially infilled tie-channel (PlugBar+5); (2) a completely filled-in tie-channel (PlugBar+6); (3) original downstream tie-channel but without the embankment at the upstream lake limb (PlugBar-Up).
Each of these alternative bathymetric scenarios were applied to the unsteady flow analyses described above, resulting in a total of seven scenarios including the original bathymetry.

| Sediment sampling and coring
To examine sediment mobilization and deposition after Hurricane We performed a grain-size analysis at 0.01 m resolution on the top 0.80 m of each core to identify the texture and thickness of coarse-grained deposits associated with recent flooding due to Hurricane Harvey. Each sediment subsample was pre-treated with 10 mL of 30% H 2 O 2 to remove organic matter (van Hengstum et al., 2007) and 10 mL of dispersant (0.5% sodium hexametaphosphate) and mixed for 5-15 min on an orbital plate shaker before grain-size analysis using a Malvern Mastersizer 2000-LV laser diffraction particle size analyser (Vasskog et al., 2016). We unmixed the grain-size distributions in each core into different populations using end-member modelling (EMM) in the EMMAgeo package in RStudio (Dietze & Dietze, 2019). The coarsest end-members and the 90th percentile of the grain-size distribution (D 90 ) were used to identify coarse-grained event-beds that are deposited during extreme floods Toonen et al., 2015).

| Hydraulic modelling
The hydraulic model successfully simulates the peak discharge magni-  Figure S4). as water continues to flow in from the river (Figure 4b) or out of the lake into the river during the flood crest.

| Sediment sampling and analysis
We Grain-size distributions of the sediment cores show that the uppermost deposits are coarser than underlying material, and that these uppermost deposits decrease in thickness and grain size and with increased distance from the tie-channel (Figure 7). End-member (EM) modelling for each core results in three distinct end-member populations, where EM3 represents an extreme flood grain-size distribution with a left skew and narrow peak around 100 μm , and dominates the uppermost deposits in all three cores.
Adjacent to the tie-channel, core LIB1 has a mean D90 value of 220 μm from 0 to 0.48 m, which then drops to 78.5 μm deeper in the core. Further from the tie-channel, core LIB2 has a mean D90 of 113 μm from 0 to 0.12 m, while core LIB3 has a mean D90 from 0 to 0.04 m of 47 μm. The thinning and finning with distance from the tiechannel of the coarse-grained Hurricane Harvey deposits in the uppermost section of the core is consistent with the flow velocity profiles generated by our hydraulic model for this same event. Although the EM3 curves all represent the coarsest sediment population within their core, we do observe notable differences between the three EM3 populations. As we move away from the tie-channel, the grain-size distribution of EM3 shifts towards smaller grain-size values (peaks at 153 μm, 71 μm, 29 μm) and becomes narrower. These differences imply that the core closest to the tie-channel consists of sediments that can also be found further-and are captured in, for example, EM1 and EM2-out in the lake, but that the coarsest sediments are primarily deposited at the lake entrance.
A comparison between grain-size distributions of lake-bottom sediments and sediments collected from the tie-channel and active river channel shows that sediments deposited in the lake during an extreme flood reflect sorted sediments in the tie-channel (Figure 8). Finer-grained sediments are carried further into the lake, causing a larger share of fine grain-sizes in cores further from the tie-channel (LIB2 and LIB3) relative to the core closest to tie-channel (LIB1).

| Alternative floodplain bathymetries
To examine the influence of floodplain lake evolution on flow velocity patterns, we developed a series of alternative bathymetries that vary plug-bar geometry and lake depth. First, we examined two scenarios with a matured plug-bar: (i) PlugBar+5, where the tie-channel has been filled up to 5 m; and (ii) PlugBar+6, where the entire tie-channel is filled up to 6 m to match the elevation of the surrounding floodplain ( Figure 2f,g). These modifications serve to reduce hydrological connectivity of the lake, and imply that as the plug-bar matures higher river stages are required to transfer floodwaters and sediments from the river to the lake (Figure 9). In our simulations, the river stage at which the lake floods in the original tie-channel bathymetry is 4 m, while for scenario PlugBar+5 it is 5.3 m and for PlugBar+6 it is 6.6 m.
In all scenarios, lake level rises to 7.2 m to match those of the river.
Lake levels drop as river stage decreases following the flood crest, stabilizing at 4.5 m for the original bathymetry, 5.2 m for PlugBar+5, and 6.0 m for PlugBar+6, reflecting the importance of plug-bar elevation in mediating the transfer of water and sediments between the river and lake during flood and non-flood conditions. These simulations also show that plug-bar elevation influences flow velocity through the lake such that a completely infilled plug-bar (PlugBar+6 scenario) attenuates the initial burst of floodwaters and only allows floodwaters into the lake via overbank flow across the floodplain (Figure 9c). landforms around the lake such as plug-bars and lake geometry play an essential role in lake infilling and sediment deposition.

| DISCUSSION
In this study, we combine field-based sediment sampling with hydrau-

| Theoretical models of floodplain lake sedimentation
Our findings support and build on key parameters of theoretical models describing long-term floodplain lake evolution (Gagliano & Howard, 1984;Toonen et al., 2012), by providing insights into the mechanisms that generate observed sedimentation patterns. The hydraulic simulations reproduce the expected behaviour of a young floodplain lake in its transitional phase with a developing plug-bar and lake shallowing as described by Toonen et al. (2012) during an extreme flood event. Plug-bar growth can be observed as flow velocities peak at the end of the downstream lake arm, but do not propagate into the lake, implying that deposition of coarse-grained sediments occurs around the lake entrance during this stage. Moreover, flow velocities reduce towards the banks of the lake, resulting in preferential deposition that narrows the lake (Douglas Shields & Abt, 1989). The sediment cores also reflect transitional floodplain lake fills (Toonen et al., 2012) and support the simulation results. The uppermost deposits in core LIB1 consist of thick ($0.4 m) coarse-grained (>200 μm) sediment, typical of a floodplain lake with a low plug-bar that allows water to flow in easily, while the uppermost coarse beds captured in the other cores winnow as distance from the plug-bar increases. We show that bed material samples from the tie-channel match these coarse-grained lakebed sediments, demonstrating that the tie-channel acts as the conveyor belt that brings water and coarse sediment to the lake entrance. The lack of thinly laminated layers below the uppermost coarse-grained deposits further supports that the lake has not yet reached full abandonment from the river (Gagliano & Howard, 1984;Toonen et al., 2012). In short, the HEC-RAS software can be used to describe both the general patterns of floodplain lake evolution and the individual sediment signal after an extreme flood event.
The alternative bathymetry simulations further support observations and theoretical models that describe the transformation from the transitional lake phase to an abandoned floodplain lake with thinly laminated coarse sediment layers (Citterio & Piégay, 2009;Gagliano & Howard, 1984;Toonen et al., 2012). The hydraulic simulations show that, as the lake becomes shallower and narrower in the transitional phase, flow velocities extend further into the lake, which would result in more extensive transport of coarse-grained material into the lake. At the same time, increased plug-bar height increases the threshold for hydrological connectivity, creating a barrier that increasingly impedes sedimentation in the lake, except during the largest flood events. In these ways, lake and plug-bar infilling generate the laminated sedimentary fills that are characteristic of floodplain lakes in the abandoned stage (Bábek et al., 2008;Citterio & Piégay, 2009;Munoz et al., 2018). When floodwaters enter the lake via multiple entrances through overland flow across the floodplain instead of being only funnelled through the tie-channel, sedimentation will be more uniform across the lakebed (Hudson et al., 2012). In our simulations, however, elevated areas of the floodplain between the lake and river obstruct floodwaters ( Figure 3a), implying that in-lake sedimentation patterns will also be influenced by floodplain morphology beyond the plug-bar and lake bathymetry, as examined in this study. Together, our simulations with alternative lake bathymetries show how the sediment signal from a given flood is contingent on progressive changes in the lake and floodplain morphology.
Finally, the PlugBar-Up simulation documents the differences in flow velocity patterns and inundation for floodplain lakes that receive floodwaters through a downstream entrance (backwater lakes) and those that receive floodwaters primarily through an upstream entrance (lotic lakes), and support observations of sedimentation patterns in these lakes ( Figure 11) (Citterio & Piégay, 2009). In the hydraulic model, opening the upstream entrances increased the hydrological connectivity, overall flow velocities, and reversed the direction of flow in the lake ( Figure 11). These simulations imply that coarse sediments in a lotic lake will either be distributed more homogeneously or flow out of the lake entirely, as lotic lakes exhibit reduced trapping efficiency relative to backwater lakes (Constantine et al., 2010). Observed sedimentation patterns support these inferences, where sedimentation in backwater lakes occurs primarily in the downstream lake arm, whereas in lotic lakes sedimentation is minimal and distributed more evenly across the lake because the lake acts as an anabranch of the active river channel (Citterio & Piégay, 2009). The trapping efficiency of lotic and backwater lakes will also be dependent on lake geometry, where straighter lakes with lower diversion angles allow flow velocities to remain high and limit sedimentation in the lake (Constantine et al., 2010). This is reflected by the reduced flow velocity in the lake bends, which suppresses sedimentation potential. Here, the hydraulic model allows us to decipher site-specific sedimentation patterns based on lake geometry and within the context of long-term floodplain lake evolution of lake models. This study outlines an advance in the use of hydraulic modelling on floodplains by relating flow velocities to sedimentation patterns in floodplain lakes, and the use of alternative bathymetries to evaluate the sensitivity of these patterns to lake evolution and infilling. Sediment samples from the lake bottom document how coarse-grained event beds are sorted within the lake during extreme flood events, demonstrating that these deposits thin and fine with distance from the tie-channel (Figure 7). Our findings show how flow velocity patterns change as the lake infills ( Figure 10), implying that depositional patterns will change over time to generate foreset bedding via progradation across the lake arm ( Figure 12). These findings are particularly relevant in the use of grain-size distributions to infer flood magnitudes (Jones et al., 2010;Wilhelm et al., 2019), because they imply that at a single core location the relationship between river discharge and grain size will be nonlinear and non-stationary. This relationship is nonlinear because flow velocities in the tie-channel and lake do not peak when river discharge crests. Instead, energy to erode material from the tie-channel is maximized during a pulse when floodwaters from the river initially flow into the lake. As a result, peak flow velocities in the tie-channel will be a function of tie-channel geometry, lake level immediately preceding the flood, and the rate of stage increase in the river at the onset of flooding. In addition to being nonlinear, the relationship between river discharge and grain size is non-stationary because, as the lakebed infills, high flow velocities extend further into the lake such that coarse-grained event beds will become thicker and coarser for a flood with a similar discharge. Additional changes to the tie-channel and plug-bar via infilling and vegetation succession will further alter the relationship between flood magnitudes and event bed properties. These findings do not negate prior work inferring flood magnitudes from grain-size distributions Toonen et al., 2015, but instead add nuance to these interpretations and point towards opportunities to reduce their uncertainties.

| Applications and implications
The application of alternative bathymetries in hydraulic modelling utilized in this study provides an avenue to improve on prior work estimating the hydrological characteristics of a flood from sediments.
The use of a linear model to estimate peak river discharge from grainsize distributions generates 1σ uncertainties of the order of 5-20% (Fuller et al., 2019;Munoz et al., 2018;Toonen et al., 2015, and these large uncertainties can be detrimental to the accuracy of flood frequency analysis (Reinders & Muñoz, 2021). To improve on these discharge estimates, an inverse-modelling approach could be used that simulates flow velocities and sedimentation patterns under different river discharge and bathymetric scenarios. Our work demonstrates how bathymetric scenarios can be developed using simple assumptions of infilling patterns, and these patterns could be constrained using sedimentation rates derived from sediment cores. For example, the sediment cores recovered in this study could be used to reconstruct lakebed morphology and prior to the Hurricane Harvey floods, or at earlier points in time, which could then be examined within a hydraulic model under a range of flood scenarios. The resulting flow velocity, sediment transport thresholds, and sediment distribution patterns could assist in the interpretation and analysis of the lake's observed sedimentary characteristics. Hydraulic models could also serve the analysis of sediment transport on floodplains more generally, for example by studying avulsions. The integration of sediment transport into hydraulic models, including in HEC-RAS 6, could further facilitate these efforts by simulating deposition resulting from a flood of a given magnitude. The integration of hydraulic models into palaeoflood hydrology has focused primarily on slackwater deposits in bedrock canyons, but our work implies that a similar approach could be applied in dynamic alluvial settings.

| CONCLUSIONS
Floodplain lakes constitute important stores of sediment (Asselman & Middelkoop, 1995;Phillips et al., 2004;Walling & Bradley, 1989), nutrients (Thoms, 2003), and pollution (Ciszewski & Grygar, 2016), yet the sedimentation patterns in these lakes are non-stationary and nonlinear due to changes in flow velocity patterns as lake geometry and floodplain topography evolve over time (Gagliano & Howard, 1984;Toonen et al., 2012). Therefore, existing models of floodplain lake evolution are not well suited to guide the interpretation of individual flood events in sedimentary records. In this study, we examine sedimentation and flow velocity patterns in a floodplain lake during a flood event by combining field-based sampling of flood deposits resulting from the extreme flooding in the wake of Hurricane Harvey with hydraulic model simulations of this event implemented in HEC-RAS. Our simulations demonstrate the mechanisms that result in the deposition of coarse-grained sediments in the floodplain lake via the tie-channel, and how these deposits thin and fine with distance from the tie-channel. We also use a series of hydraulic model simulations using alternative lake and plug-bar bathymetries to examine the sensitivity of flow velocity patterns to floodplain lake evolution and infilling. These modelling exercises imply that sedimentation patterns in floodplain lakes are highly sensitive to local floodplain geomorphology and lake evolution. Subsequently, the sediment signal of a particular flood will change over time due to the progressive infilling of the lake. Our findings build on theoretical models describing sedimentation in floodplain lakes (Gagliano & Howard, 1984;Toonen et al., 2012) by showing that numerical models such as HEC-RAS can provide a mechanistic understanding of observed sedimentation patterns and how these patterns change over time.
We demonstrate how hydraulic modelling can aid in interpretations of individual events in sedimentary records in alluvial settings to improve reconstructions of ecological and hydrological dynamics in riverine environments.
grants from the US National Science Foundation (EAR1804107 and EAR-1833200).