For society, the sediment and nutrient supply by large rivers maintains large natural and agricultural ecosystems, particularly in Southeast Asia. Riverine sand is also a major construction resource that is becoming increasingly scarce.
For scientists, accurate quantification of material exported by rivers is a reliable and efficient way to constrain key processes like erosion, chemical weathering, and the continental carbon cycle.
Largest rivers carry the most sediment but their fluxes are particularly
difficult to quantify because of hydrodynamic
sorting (i.e., most sand is carried at the bottom). Here we present another approach at tackling this problem.
River flow velocity data (measured via ADCP),
coupled with relatively few sediment depth samples
across the river channel, collected over varying
A new Rouse-based approach to synoptically
model suspended sediment concentration and composition in two
dimensions across a river section.
The Irrawaddy and the Salween are two of the last
un-dammed mega-rivers in Asia, although likely not
for long. They serve as major transportation
arteries, support agriculture feeding 100M+ people,
and provide a crucial supply of construction-grade
We used the model described here to quantify the
annual flux and the mean chemical composition of
the suspended sediment in the Irrawaddy and Salween Rivers.
However, the model is easily applicable to any
river where similar water velocity and sediment
data are available.
Due to hydrodynamic sorting, sediment grain size
and composition varies strongly with depth and
across the channel in large rivers, requiring
samples to be collected at various depths in order
to estimate the total sediment flux.
Using Acoustic Doppler Current Profilers (ADCP),
it is possible to measure flow velocity
distribution in two dimensions (laterally and with
depth) with sub-meter resolution in large river
Using a simple hydrodynamic (Rouse) model, ADCP
flow velocity can then be related to measured
suspended sediment concetrations (SSC) in alluvial
channels. This has been done previously for rivers
like the Amazon (Bouchez et al., 2011) and the Ganges
(Lupker et al., 2011).
More turbulent conditions in morphologically
complex channels result in more variable, "noisy"
SSC and grainsize data (see fig. on the left). Such
sites require a more explicit linkage of SSC and
local hydrodynamic conditions.
Our approach improves on previous efforts
by estimating SSC in both
dimensions (depth and width) across a river channel
while requiring fewer sediment samples for model
ADCP flow velocity cross-sections at different
water stages (4 per each river).
Suspended sediment depth profiles in different
parts of the channel (total n=30-40 per river).
ADCP flow velocity data while depth sampling,
providing local hydrodynamic conditions for each
Sediment grain-size distributions measured using
laser diffraction (see figure above).
In the Rouse model, turbulent lift that suspends
particles is taken proportional to shear stress (or
shear velocity), calculated here using
depth-averaged flow velocity.
Using measured grainsize distributions, each
suspended sediment sample was split into five
grainsize bins (0.2-4, 4-16,
16-63, 63-250, 250-2000 µm).
In contrast to previous approaches, we calibrated the measured SSC to both the normalized depth and the shear
velocity simultaneously, using a non-linear fitting algorithm in MATLAB. On the left, the symbols show measured sample values while the grid shows the fitted surface.
To account for differences in settling velocity, the Rouse model calibration was
done separately for each of the five grain-size
As expected, coarser SSC increased strongly with
depth and shear velocity (left, strongly slanted and bent fitted surface), whereas fine SSC
was relatively insensitive (above, relatively straight and upright fitted surface).
Using these fitted relationships and ADCP data,
SSC can be calculated synoptically in high
resolution across the river channel in both
dimensions (fig. on the right, top panel). The colored fields show calculated values across the river channel, while the square colors indicate the measured composition of suspended sediment samples.
This approach also estimates grainsize distribution across the channel (middle panel, shown here as D50 for visualization purposes).
Combining the fully resolved SSC estimates with the ADCP flow velocity cross-sections yields total sediment flux (bottom panel) values that can be up to 50% higher or lower (depending on the river stage), compared to values calculated using simple mean SSC.
The chemical composition of riverine sediment frequently strongly correlates with its grainsize (for example, Al/Si ratio, organic carbon concentration).
The estimated grainsize variation can thus be used to calculate the distribution of these chemical parameters across the river channel.
Similarly to total sediment flux, this approach yields significantly more accurate estimates of mean sediment chemical composition and the flux of components such as particulate organic carbon (POC).
Our results demonstrate that synoptic (i.e. spatially highly-resolved) sediment transport modeling is crucial for the accurate quantification of sediment composition and flux in large river channels, where wide sediment grain size distributions and variable
hydraulic conditions result in complex sediment transport patterns.
Irrawaddy and Salween in global context
Using the approach above, we established SSC-discharge rating curves and calculated monthly and annual sediment flux (and mean chemical composition) of the Irrawaddy and the Salween Rivers.
The combined sediment flux of the two rivers is 485 Mt/yr (68% confidence interval of 364-654 Mt/yr) and the particular carbon flux is 1.9 (1.0-3.3) Mt C/yr, accounting respectively for 2-3%
and 0.8-1.2% of the total global riverine export to the ocean (using the global values of Milliman & Farnsworth (2011) and Galy et al., (2015), respectively).
However, large dams are planned on both rivers, which, if built, will strongly reduce sediment delivery to the low-lying deltas, which serve as the main food supply for milions of people. Our results here will thus serve as a useful baseline to evaluate future changes in the sediment flux of these river systems.