Many of the world's deltas – home to major population centers – are rapidly
degrading due to reduced sediment supply, making these systems less resilient
to increasing rates of relative sea-level rise. The Mississippi Delta faces
some of the highest rates of wetland loss in the world. As a result,
multibillion dollar plans for coastal restoration by means of river
diversions are currently nearing implementation. River diversions aim to
bring sediment back to the presently sediment-starved delta plain. Within
this context, sediment retention efficiency (SRE) is a critical parameter
because it dictates the effectiveness of river diversions. Several recent
studies have focused on land building along the open coast, showing SREs
ranging from 5 to 30 %. Here we measure the SRE of a large relict
crevasse splay in an inland, vegetated setting that serves as an appropriate
model for river diversions. By comparing the mass fraction of sand in the
splay deposit to the estimated sand fraction that entered it during its life
cycle, we find that this mud-dominated sediment body has an SRE of
Most large rivers do not transport sufficient sediment to the coast to fill the accommodation that will be created on their delta plains due to rapid 21st century sea-level rise (Giosan et al., 2014). This shortfall ensures a global retreat of deltaic coasts and presents an existential threat to some of the densest human populations, most valuable economic infrastructure, and most vibrant ecologies on Earth (Ericson et al., 2006; Giosan et al., 2014). To mitigate land loss, sediments can be distributed to vulnerable or otherwise important locations with controlled diversions of sediment-laden river water (Day et al., 2007; Kim et al., 2009; Paola et al., 2011; Giosan et al., 2013; Smith et al., 2015; Auerbach et al., 2015; Coastal Protection and Restoration Authority of Louisiana, 2017). The most effective techniques for such diversions are subject to debate (Blum and Roberts, 2009; Kim et al., 2009; Nittrouer and Viparelli, 2014a, b; Blum and Roberts, 2014), but it is clear that maximizing sediment retention efficiency (SRE) is a critical concern (Blum and Roberts, 2009; Paola et al., 2011). Because fine sediments are highly mobile in suspension and delta plains are often regarded as inefficient traps for mud (Giosan et al., 2014), much of the literature on diversions has focused on extracting sandy material from the trunk channel (Nittrouer et al., 2012a; Nittrouer and Viparelli, 2014b; Meselhe et al., 2016). However, mud comprises 80 % or more of the incoming sediment load in most rivers (Giosan et al., 2014) and often dominates their delta-plain deposits.
Published estimates of SRE (Nittrouer et al., 1995; Allison et al., 1998; Goodbred and Kuehl, 1998; Törnqvist et al., 2007; Blum and Roberts, 2009; Day et al., 2016; Roberts et al., 2016) vary from 5 to 80 %, a range that is too wide to be useful for planning purposes but which suggests that the specific depositional setting is an important control. Here we propose that vegetated inland deltaic settings that are protected from wave and tide energy can be highly efficient in trapping sediment, especially mud, and thus offer desirable locations for diversions that target mud for coastal restoration. We test this hypothesis by measuring the SRE of a large crevasse splay at an inland setting in the Mississippi Delta (MD) and find that it exceeds 75 %. This is substantially higher than estimates of SRE in the Wax Lake delta (WLD), a well-studied prograding lobe on the open coast of the MD (Fig. 1a) that is often used as a diversion analog. The WLD is sand-dominated (Roberts et al., 2003; Kim et al., 2009) and has been estimated to have an SRE ranging from 5 to 30 % (Törnqvist et al., 2007; Kim et al., 2008; Roberts et al., 2016). Our results demonstrate the essential role that mud plays in vertically aggrading delta plains, and the contrast with the WLD highlights the importance of careful site selection for diversion projects.
We set this study in the Attakapas Crevasse Splay (ACS), a
Digital elevation models (DEMs) of the study area.
Normalized DEM of the LMR between Baton Rouge and the Bonnet Carre
Spillway and the adjacent portion of the delta plain for crevasse splay
identification. The DEM was obtained by subtracting a planar surface that
best fits the Mississippi River natural levee long profile from the DEM of
the studied reach. A
There are only a handful of studies that have attempted to tie the bulk sedimentary properties of a recent deposit to sediment-transport properties in the river that created it (Törnqvist et al., 2007; Kim et al., 2009; Giosan et al., 2013; Day et al., 2016), and we are unaware of any with a subsurface data set as rich as the one available for the ACS. While many workers have published on the “river side” issues concerning diversions, including the sediment available (Kesel, 1988; Blum and Roberts, 2009; Allison et al., 2012) and the physics of extracting sediment from the trunk channel (Allison and Meselhe, 2010; Meselhe et al., 2012; Nittrouer et al., 2012a; Allison et al., 2013), only recently have researchers begun to investigate “basin side” issues that impact SRE (Xu et al., 2016). The availability of detailed sediment-transport data from the modern Lower Mississippi River (LMR) (Allison et al., 2012) provides a unique opportunity to connect fluvial sediment budgets to the sediments preserved in the delta.
We estimate the SRE of the ACS with a 3-D model (Fig. 3) based on the 132 cores augmented by 53 grain-size analyses. These data are used to quantify
the sand fraction (> 62.5
Sediment texture data used to calculate
3-D geometry of the Attakapas Crevasse Splay. Vertical axis is in meters relative to NAVD 88. The relatively sandy channel deposits are shown in blue green; light green refers to the silt-dominated portion of the splay. Channel deposits make up 16 % of the total volume, and non-channel deposits compose the remaining 84 %. The transparent top bounding surface is the lidar-derived modern land surface, and the base is the clay to silt transition shown in Fig. 4. Wherever possible, the lateral boundary is chosen to be the intersection of the top and basal bounding surfaces. On the lateral edges, where the deposit does not pinch out and these two surfaces do not meet, the boundary was chosen manually.
Stratigraphic cross section along transect A to A
All 132 cores were described in the field at 10 cm increments following the
United States Department of Agriculture texture classification system (cf.
Shen et al., 2015). Texture classes encountered were very fine sand (vfS),
sandy loam (SL), silt loam (SiL), silty clay loam
(SiCL), silty clay (SiC),
and clay (C). Organic-rich clays are denoted as humic clay (HC). The sand
fraction for each texture class was determined by grain-size analysis of
53 samples taken from three separate cores (Table 1). For our analysis we
combined SL and vfS into a single class to which we applied the sand fraction
measured for vfS samples. This is consistent with our objective to estimate
an upper limit of sand content in the deposit. We combined SiC, HC, and C
into a single class as well. Samples were treated with hydrochloric acid to
remove carbonates and with hydrogen peroxide to remove organic matter and
then wet-sieved through a 106
All 3-D modeling was done by means of Schlumberger's Petrel geo-modeling software. The basal bounding surface of the ACS was generated with R, using the gstat package (Pebesma, 2004). The ACS consists of primarily muddy facies that overlie a wood peat bed that predates the occupation of the region by the precursor of the modern LMR, Bayou Lafourche (Törnqvist et al., 1996). The earliest Lafourche deposits feature a 1–3 m thick clay bed that transitions abruptly into a silty matrix with sandy ribbons embedded within it (Fig. 4). We interpret the clay to silt transition as the base of the ACS and the coarser sandy deposits as splay channel deposits.
The top bounding surface of the splay is the modern land surface as measured
by lidar. The picked subsurface elevation of the clay to silt transition was
linearly regressed on the elevation of the local land surface (
Channel bodies in the ACS are identifiable both as narrow alluvial ridges
and as coarser sediment bodies in the subsurface. All channel bodies were
modeled to extend through the full thickness of the splay. This choice makes
the channel bodies appear somewhat less sandy than they might be in reality but also makes them larger; the end result is a slightly sandier estimate of
We determined the average sand fraction for the channel and non-channel
portions of the ACS separately (Table 1). A volume-weighted average of these
sand fractions yields the estimate of overall
Data used to calculate
We estimated the yearly averaged sand fraction input into a 5 m deep crevasse channel emanating from a 30 m deep trunk channel. We obtained the depth of the trunk channel, Bayou Lafourche, from previous investigations in the region (Fisk, 1952). Shen et al. (2015) showed that the ACS deposit has a total thickness of up to 10 m, that it was active during two episodes of rapid aggradation, and that similar thicknesses accumulated near the inlet during each episode. In keeping with these data, we used 5 m as a representative estimate of the depth of the primary crevasse splay feeder channel during its lifetime. This value is similar to that of other well-studied crevasse systems in the MD (Farrell, 1987). The flow depth of the trunk channel was assumed not to vary significantly, consistent with measurements in the modern LMR, where at 100 km from the shoreline the flow depth varies less than 2 m throughout the year (Nittrouer et al., 2012b).
To estimate the input sand fraction (
To obtain the sand fraction in the uppermost 5 m of the water column we used
(1) a vertical profile of relative suspended sediment concentration for mud
and sand in each discharge bin, (2) the total suspended load for each
sediment class and discharge bin (estimated from gauge measurements at Belle
Chasse), and (3) log-law velocity profiles for each discharge bin. All
calculations can be seen in the “
We assumed that the sand load obtained from Belle Chasse was composed of 125 and 250
We adapted methods used in recent work in the modern LMR (Nittrouer et al., 2011) and the Rouse
equation to calculate profiles of relative concentrations of suspended sand.
The equation for the relative concentration of suspended sand is
Our Rouse profiles are not calculated using a near-bed reference
concentration, and thus they describe the shape of the sand concentration
profile but not its magnitude. We use these curves to define a set of
proportionality constants
Using the proportionality constants defined in Eq. (3) we now expand Eq. (4)
and obtain a value for
Note that we have calculated the width-integrated sand concentration in our
depth interval rather than the sediment load. We do this because crevasse
splays can distort local flow fields, but concentrations in a turbulent flow
can be inherited from upstream. To calculate the sand fraction entering the
ACS, we assume that the splay takes a constant fraction of the trunk
channel's discharge throughout the hydrograph, which we call
The sand fraction delivered to the ACS,
The 3-D model (Fig. 3) shows a half-lens-shaped deposit with a maximum
thickness of
Using modern suspended sediment data, we estimate that the yearly averaged
input sand fraction (
The ACS boundary is defined by the extent of silt-dominated deposits and therefore includes all sandy sediment bodies encased within it. Any reasonable splay boundary will exclude the most mobile sediments, so we proceed with the understanding that some fine sediments were lost to the surrounding environment. For example, the bank of Lake Verret near the downstream end of the ACS is convex where splay-derived sediments have partially filled it (Fig. 1b), suggesting a loss of fine sediments across our defined boundary.
Unlike
Spillways that are part of the modern flood-control system prevent the
largest floods, which had magnitudes substantially greater than what is
allowed today (Barry, 1998). The discharges in our largest bin
(> 30 000 m
Perhaps the largest uncertainty is the sand and mud fraction carried by the LMR prior to human modifications. It is well documented that the suspended sediment load of the modern LMR was dramatically reduced by dams in the mid-20th century (Kesel, 1988), but because it is likely that pre-dam sediment loads were elevated due to widespread agricultural activity in the drainage basin (Keown et al., 1986; Tweel and Turner, 2012), it is difficult to estimate the suspended sediment loads that prevailed from 1.2 to 0.6 ka. Recent modeling results suggest that the suspended sand load delivered to the MD is buffered from upstream change for timescales on the order of 1000 years (Nittrouer and Viparelli, 2014a). With that in mind, the events that may have substantially changed the suspended sand load are either so recent (dams in the tributaries, rapid deforestation associated with expanding agriculture; < 200 years ago) or so long past (glacial outwash floods; > 10 000 years ago; Rittenour et al., 2007) that it is reasonable to apply modern sand loads to the Lafourche channel.
The question of paleo-mud loads is more challenging, as there is no accepted
estimate for the magnitude of the sediment-load increase due to expanding
agriculture. Instead of choosing one particular mud load (and implied sand
fraction) we performed our
We performed independent sensitivity analyses on our estimates of
The ACS is overwhelmingly composed of mud, with sand-sized grains accounting
for only
The
Recent researchers have given considerable attention to the importance of
coarse-grained sediment as a restoration tool (Nittrouer et al., 2012a; Nittrouer and
Viparelli, 2014b), but proximal overbank deposits in the Mississippi Delta
are dominantly composed of mud (e.g., McFarlan, 1961; Frazier, 1967;
Törnqvist et al., 1996). Our results highlight the utility of the more
plentiful mud load. While significant land loss in the Mississippi Delta is
inevitable, accretion rates that persisted in the mud-dominated ACS for
centuries (1–4 cm yr
The LMR discharge was shared between Bayou Lafourche and the modern LMR when the ACS was active (cf. Törnqvist et al., 1996). Since the partitioning of the discharge between these two distributaries is unknown, we assume a roughly equal proportion of the sand and mud load. This is relevant because the ACS built a splay with an area comparable to the WLD even when the discharge was split between Bayou Lafourche and the LMR. The abundance of mud-dominated crevasse splays in the MD shows the importance of mud pathways to delta evolution. Their large numbers (Fig. 2) also make it conceivable that numerous crevasse splays were active at any given time, thus highlighting their potential as land builders compared to a single, terminal delta lobe such as the WLD.
Our results have implications for sediment management strategies in the MD.
A significant portion of the vegetated delta plain is within
The late Holocene stratigraphic record of the Mississippi Delta shows that
crevasse splay deposits consist of
The data, calculations, and sensitivity testing on important parameters can be seen in the Supplement.
CRE and ZS contributed equally to the paper and should be considered co-first authors. ZS and TET designed the project. ZS led all fieldwork. CRE was involved in fieldwork, constructed the 3-D model with geostatistical input from CW, compiled river surveying data, and undertook the suspended sediment modeling. JM contributed to fieldwork and conducted and interpreted the grain-size analyses. ZS and CRE composed the paper with major input from TET.
The authors declare that they have no conflict of interest.
Field assistance by Jennifer Kuykendall, Austin G. Nijhuis, Marc P. Hijma, Zhen Li, and Tulane University Spring 2008 and Fall 2012 Sedimentation and Stratigraphy classes is appreciated. The paper benefited from discussions with Mead A. Allison and Michael T. Ramirez. This research was supported by US National Science Foundation grant EAR-1148005 to Zhixiong Shen and Torbjörn E. Törnqvist and the Vokes Fellowship to Christopher R. Esposito, with additional support from the Long-term Estuary Assessment Group program through the Tulane/Xavier Center for Bioenvironmental Research to Zhixiong Shen, and from the Louisiana Sea Grant Undergraduate Research Opportunities Program to Jonathan Marshak. Schlumberger provided free access to the Petrel software suite. Kyle M. Straub provided lab equipment used in grain-size analyses. Edited by: Daniel Parsons Reviewed by: three anonymous referees