An interesting question, Pankaj.

This is an abbreviated and edited extract from a draft Review paper that our project team have written in draft form, and which will be published later this year once edits are complete.

Estimation of suspended sediment loads.

Introduction to the issues and problems

Suspended sediment loads (SSL) are derived from the product of suspended sediment concentration (SSC) and flow rate or discharge, Q . Although apparently simple, calculation of fluvial sediment fluxes is fraught with difficulty, and the process is often referred to as ‘estimation’. Suspended sediment flux estimates need to maximise accuracy (proximity to the actual value) and precision (high repeatability).

At the simplest level, it requires the determination of both river discharge, Q, and - this is the key - spatially and temporally representative suspended sediment concentrations. Thus, the instantaneous suspended sediment load, SSL (in kg s-1 for the units below), is given by:

SSL = Q x SSC

Where Q = water discharge (m3 s-1) and SSC is the instantaneous suspended sediment concentration in g l-1 (i.e. kg m-3).

Sediment fluxes over a given time period require integration of this simple equation over time.

However, problems include the following:

 1. The many non-linearities in the sediment transport system, especially in the relation between water discharge and suspended sediment concentration which are often power functions.

2. Temporal variability of river flows, and the variable SSC for a given flow.  Sediment rating curves show that SSC can vary by 2-3 orders of magnitude for a given flow (e.g. classic papers by Walling and Webb e.g. 1981 et seq. This is quite different from a stage-discharge rating equation where, in the absence of channel geometry changes at the gauging station, a given stage is always associated with the same discharge (notwithstanding rising and falling limb water surface slope differences as the flood wave passes.

3. Much suspended sediment transport occurs during a very limited time. For example, Walling et al. (1992) found, for the River Exe at Thorverton in SW England for 1978-1980, that 50% of the suspended sediment load occurred in just 1% of the time, and 90% of total suspended sediment load took place in 5% of the time. It is crucial, therefore, to sample or monitor in these key periods.

4. There is typically huge scatter in river discharge vs suspended sediment concentration relations which has to be addressed to produce the sediment rating curve), which is often best defined by log-linear relationships between discharge and suspended sediment concentration.

5. Hysteresis effects further complicate estimation.  Different suspended sediment concentrations are often obtained for a given flow (at storm event and seasonal timescales: section.

6. Cross-sectional SSC variations which may complicate single-point (e.g. channel-edge) sampling or monitoring etc.

7. Changing sediment particle size distribution effects, during and between storm events and seasons, which can degrade simple relationships between turbidity and SSC.

A key issue, therefore, is to ensure that sampling/monitoring is quasicontinuous. Walling and Webb’s (1981) pioneering study demonstrated this by using a resampling approach on data from the R. Creedy in Devon.  They first computed ‘true’ suspended sediment loads from a master high-resolution dataset collected at 60-minute intervals, and then compared these values with those obtained from restricting the analyses to successively coarser time resolutions (e.g. daily, weekly, 2-weekly), to mimic a low-resolution field investigation.  Such studies demonstrate the very large load errors that arise with coarse sampling (e.g. Walling, 1981, 1985 et seq). 

Several methods have been evaluated rigorously by Walling and Webb (1981; 1985), and further examinations of the statistical methodologies involved have been published by Ferguson (1987) and Clarke (1990). Phillips et al (1999) evaluated 22 suspended sediment load estimation equations for northern England catchments.

The problems continue to tax, and some examinations include: Webb et al. (1997) for British rivers; Philips et al. (1999) for the UK showing that errors are magnified in larger catchments; Horowitz (2003) on sediment rating curves for US and European rivers who recommends that sampling is done on a hydrological, rather than a calendar, basis to minimise errors; Crowder et al’s (2007) study of suspended sediment load–discharge relationships for US Midwest basins; and the work of Brown et al (2010) who used multiple load estimation approaches for basins in the UK, Germany and The Netherlands. As Moatar and Meybeck (2005, p. 430) argue, some of these issues also apply to nutrient load estimation: ‘concentration data are still commonly the limiting factor on the quality of river flux estimates’ .

High-resolution monitoring method: interpolation approaches

The most effective and accurate method to derive total suspended sediment loads is often field intensive.  It involves automated high-resolution, quasicontinuous sampling or monitoring, at a sampling interval which is short enough to ensure that SSC or flow does not change significantly i.e. where it is ‘safe’ to interpolate Q and SSC between sample values.  This interval is often set between 10 and 30 minutes (commonly 15 minutes) in small to medium scale catchments where flows and suspended sediment fluxes can change quickly (e.g. < 1000 km2 in drainage basin area).

Example high-resolution studies in the UK include D.E. Walling's pioneering work (many great papers!); Lawler et al. (1996; 2006) and Old et al. (2003; 2005), and examples for large glacial rivers in Iceland (Lawler, 2003). The monitoring period should also be sufficiently long to be representative of longer-term sediment transport conditions, and include a range of hydrological and meteorological events, especially storm sequences when most sediment transport occurs.

A few useful references are listed here (many of these are on Research Gate or Google Scholar) to get started, and many of these papers discuss field instrumentation options and details too, as part of the methodology section of the papers:

Horowitz, AJ, Kent A. Elrick and James J. Smith. 2001. ‘Estimating suspended sediment and trace element fluxes in large river basins: methodological considerations as applied to the NASQAN programme.’ Hydrological Processes, 15, pp.1107–1132.  DOI: 10.1002/hyp.206.

Lawler, DM, Petts, GE, Foster, IDL and Harper, S. 2006. Turbidity dynamics and hysteresis patterns during spring storm events in an urban headwater system: the Upper Tame, West Midlands, UK, Science of the Total Environment, 360, 109-126

Old, G.H., Lawler, D.M. and Snorrason, A. 2005. Flow and sediment dynamics of a glacial outburst flood from the Skaftá system in southern Iceland, Earth Surface Processes and Landforms, 30, 1441-1460.

WALLING, D. E., OWENS, P. N., WATERFALL, B. D., LEEKS, G. J. L. & WASS, P. D. 2000. The particle size characteristics of fluvial suspended sediment in the Humber and Tweed catchments, UK. Science of the Total Environment, 251-252, 205-222.

WALLING, D. E., RUSSELL, M. A., HODGKINSON, R. A. & ZHANG, Y. 2002. Establishing sediment budgets for two small lowland agricultural catchments in the UK. Catena, 47, 323-353.

WALLING, D. E., RUSSELL, M. A. & WEBB, B. W. 2001. Controls on the nutrient content of suspended sediment transported by British rivers. Science of the Total Environment, 266, 113-123.

WALLING, D. E. & WEBB, B. W. 1981a. The reliability of suspended sediment load data ( River Creedy, UK). Erosion and sediment transport measurement. Proc. Florence symposium, June 1981, (International Association of Hydrological Sciences, IAHS-AISH Publication 133), 177-194.

A search on Google Scholar will reveal a great many useful papers too.

I hope this helps, as a start!

An interesting question, Pankaj.

This is an abbreviated and edited extract from a draft Review paper that our project team have written in draft form, and which will be published later this year once edits are complete.

Estimation of suspended sediment loads.

Introduction to the issues and problems

Suspended sediment loads (SSL) are derived from the product of suspended sediment concentration (SSC) and flow rate or discharge, Q . Although apparently simple, calculation of fluvial sediment fluxes is fraught with difficulty, and the process is often referred to as ‘estimation’. Suspended sediment flux estimates need to maximise accuracy (proximity to the actual value) and precision (high repeatability).

At the simplest level, it requires the determination of both river discharge, Q, and - this is the key - spatially and temporally representative suspended sediment concentrations. Thus, the instantaneous suspended sediment load, SSL (in kg s-1 for the units below), is given by:

SSL = Q x SSC

Where Q = water discharge (m3 s-1) and SSC is the instantaneous suspended sediment concentration in g l-1 (i.e. kg m-3).

Sediment fluxes over a given time period require integration of this simple equation over time.

However, problems include the following:

 1. The many non-linearities in the sediment transport system, especially in the relation between water discharge and suspended sediment concentration which are often power functions.

2. Temporal variability of river flows, and the variable SSC for a given flow.  Sediment rating curves show that SSC can vary by 2-3 orders of magnitude for a given flow (e.g. classic papers by Walling and Webb e.g. 1981 et seq. This is quite different from a stage-discharge rating equation where, in the absence of channel geometry changes at the gauging station, a given stage is always associated with the same discharge (notwithstanding rising and falling limb water surface slope differences as the flood wave passes.

3. Much suspended sediment transport occurs during a very limited time. For example, Walling et al. (1992) found, for the River Exe at Thorverton in SW England for 1978-1980, that 50% of the suspended sediment load occurred in just 1% of the time, and 90% of total suspended sediment load took place in 5% of the time. It is crucial, therefore, to sample or monitor in these key periods.

4. There is typically huge scatter in river discharge vs suspended sediment concentration relations which has to be addressed to produce the sediment rating curve), which is often best defined by log-linear relationships between discharge and suspended sediment concentration.

5. Hysteresis effects further complicate estimation.  Different suspended sediment concentrations are often obtained for a given flow (at storm event and seasonal timescales: section.

6. Cross-sectional SSC variations which may complicate single-point (e.g. channel-edge) sampling or monitoring etc.

7. Changing sediment particle size distribution effects, during and between storm events and seasons, which can degrade simple relationships between turbidity and SSC.

A key issue, therefore, is to ensure that sampling/monitoring is quasicontinuous. Walling and Webb’s (1981) pioneering study demonstrated this by using a resampling approach on data from the R. Creedy in Devon.  They first computed ‘true’ suspended sediment loads from a master high-resolution dataset collected at 60-minute intervals, and then compared these values with those obtained from restricting the analyses to successively coarser time resolutions (e.g. daily, weekly, 2-weekly), to mimic a low-resolution field investigation.  Such studies demonstrate the very large load errors that arise with coarse sampling (e.g. Walling, 1981, 1985 et seq). 

Several methods have been evaluated rigorously by Walling and Webb (1981; 1985), and further examinations of the statistical methodologies involved have been published by Ferguson (1987) and Clarke (1990). Phillips et al (1999) evaluated 22 suspended sediment load estimation equations for northern England catchments.

The problems continue to tax, and some examinations include: Webb et al. (1997) for British rivers; Philips et al. (1999) for the UK showing that errors are magnified in larger catchments; Horowitz (2003) on sediment rating curves for US and European rivers who recommends that sampling is done on a hydrological, rather than a calendar, basis to minimise errors; Crowder et al’s (2007) study of suspended sediment load–discharge relationships for US Midwest basins; and the work of Brown et al (2010) who used multiple load estimation approaches for basins in the UK, Germany and The Netherlands. As Moatar and Meybeck (2005, p. 430) argue, some of these issues also apply to nutrient load estimation: ‘concentration data are still commonly the limiting factor on the quality of river flux estimates’ .

High-resolution monitoring method: interpolation approaches

The most effective and accurate method to derive total suspended sediment loads is often field intensive.  It involves automated high-resolution, quasicontinuous sampling or monitoring, at a sampling interval which is short enough to ensure that SSC or flow does not change significantly i.e. where it is ‘safe’ to interpolate Q and SSC between sample values.  This interval is often set between 10 and 30 minutes (commonly 15 minutes) in small to medium scale catchments where flows and suspended sediment fluxes can change quickly (e.g. < 1000 km2 in drainage basin area).

Example high-resolution studies in the UK include D.E. Walling's pioneering work (many great papers!); Lawler et al. (1996; 2006) and Old et al. (2003; 2005), and examples for large glacial rivers in Iceland (Lawler, 2003). The monitoring period should also be sufficiently long to be representative of longer-term sediment transport conditions, and include a range of hydrological and meteorological events, especially storm sequences when most sediment transport occurs.

A few useful references are listed here (many of these are on Research Gate or Google Scholar) to get started, and many of these papers discuss field instrumentation options and details too, as part of the methodology section of the papers:

Horowitz, AJ, Kent A. Elrick and James J. Smith. 2001. ‘Estimating suspended sediment and trace element fluxes in large river basins: methodological considerations as applied to the NASQAN programme.’ Hydrological Processes, 15, pp.1107–1132.  DOI: 10.1002/hyp.206.

Lawler, DM, Petts, GE, Foster, IDL and Harper, S. 2006. Turbidity dynamics and hysteresis patterns during spring storm events in an urban headwater system: the Upper Tame, West Midlands, UK, Science of the Total Environment, 360, 109-126

Old, G.H., Lawler, D.M. and Snorrason, A. 2005. Flow and sediment dynamics of a glacial outburst flood from the Skaftá system in southern Iceland, Earth Surface Processes and Landforms, 30, 1441-1460.

WALLING, D. E., OWENS, P. N., WATERFALL, B. D., LEEKS, G. J. L. & WASS, P. D. 2000. The particle size characteristics of fluvial suspended sediment in the Humber and Tweed catchments, UK. Science of the Total Environment, 251-252, 205-222.

WALLING, D. E., RUSSELL, M. A., HODGKINSON, R. A. & ZHANG, Y. 2002. Establishing sediment budgets for two small lowland agricultural catchments in the UK. Catena, 47, 323-353.

WALLING, D. E., RUSSELL, M. A. & WEBB, B. W. 2001. Controls on the nutrient content of suspended sediment transported by British rivers. Science of the Total Environment, 266, 113-123.

WALLING, D. E. & WEBB, B. W. 1981a. The reliability of suspended sediment load data ( River Creedy, UK). Erosion and sediment transport measurement. Proc. Florence symposium, June 1981, (International Association of Hydrological Sciences, IAHS-AISH Publication 133), 177-194.

A search on Google Scholar will reveal a great many useful papers too.

I hope this helps, as a start!

Check our www.wildlandhydrology.com.  Dave Rosgen approach too complex to explain here, some of it is free reference materials, some in books that they sell.. 

I completely agree with previous recommendations. The only one more advice. As I understand from your questions, you are interested in assessment the suspended sediemnt load for sveral points in your study catchment. So firstly, you should use similar sampling approach for each individual point and YOU SHOULD TAKE SAMPLES AND TO DETERMINE OF WATER DISCHARGE ACCORDING WATER PATHWAYS AND STREAM VELOCITY BETWEEN YOUR CROSS-SECTIONS. So you should start from upstream cross-sections and moving downstream. If you have few tributaries, so probably you need to meke measurement and take samples in few points almost in one time. In this case you need to have the other person(s), which help you or to use automatic equipment for sampling and water discharge measurement  

Sir, D. M. Lawler thank you very much for such a detail and wonderful explanation of the issues and challenges (not only) faced by (me, but by) the (other!) young researchers (also). I am happy to see some articles here that already I've gone through. I very often refer to your works whenever I have an academic gossip with my mentors. I have also cited some of your exemplary articles in my thesis. Its my humble pleasure to find direct suggestions from a scholarly person like you.

Finally, thanks once again for taking my queries such depth of seriousness!! Thank you!!!

With expectations for further suggestions in future___ Pankaj!

Thanks to William F. Hansen for your reference!!

Thanks to Valentin Nikolaevich Golosov for your valuable suggestions!

Yes, I have taken water discharge and velocity for the necessary sampling sites. Thanks!!

Suspended sediment concentrations: Manual field sampling methods

This is a follow-up to my earlier answer, and this focuses more on the specifics of field sampling and addresses several issues. It is based on a large Literature Review paper that I have lead-authored, in collaboration with colleagues from Ireland, and which will be published later in 2016.

Suspended sediment concentrations: Manual field sampling methods

Cross-sectional variations in suspended sediment concentration (SSC)

It has long been known that significant cross-sectional variation in suspended sediment concentration exists (e.g. early refs to this include Guy and Norman, 1970; Horowitz et al., 1989), with SSC increasing towards the bed, especially for coarser particles, and towards the channel centre.

Although such datasets are quite rare, channel centre SSCs can be twice that at the channel edge in some large rivers (as shown a long time ago by Nordin and Richardson, 1971). Acoustic Doppler Current Profiler data have recently helped to define spatial patterns of SSC in river cross-sections (e.g. see Lawler (2015) 'Turbidity, Turbidimetry and Nephelometry', Chapter in: 'Reference Module in Chemistry, Molecular Sciences and Chemical Engineering', Elsevier, Chapter: 11006)

For suspended sediment flux investigations based on sampling turbid waters, SSC values should be spatially representative of the whole stream cross-section. But this is often difficult to achieve, at least at a high temporal resolution.  Clearly, sampling should be temporally representative, too, and cover the full range of conditions, from low-flow SSC conditions right through to the all-important high-flow SSCs.

Often, just a single, fixed intake or monitoring point, is used at a representative point in the channel cross-section - or the channel edge if the channel is very large and difficult to instrument across the section.  Therefore, especially in large rivers of low turbulence, and/or with coarse suspended sediment where strong lateral and vertical suspended sediment concentration gradients develop, the following three additional procedures can be introduced, to help attain the most accurate and representative values of suspended sediment flux:

1. Attempt to achieve a spatially representative position for the sampler intake or turbidity monitoring point (e.g. not in a turbid ‘hot-spot’);

2. Compare suspended sediment concentrations at this single point with those from occasional depth-integrated manual sampling across the full channel width; and

3. Apply a suitable correction/calibration factor, thus determined, if the single point concentration value tends to significantly over- or under-estimate cross-sectional average SSCs.

Sampling to detect cross-sectional variation in suspended sediment.

For very turbulent flows in the channels, and/or for especially fine suspended sediment size distributions, lateral and vertical variability in stream channel SSC will normally be less notable. Sediment mixing is also likely to vary with river level (stage) and discharge for a given cross-section, though few datasets exist to demonstrate this. 

To check for the degree of mixing, spatial sampling should be done with specially designed point- or depth-integrated isokinetic samplers (see below), as pioneered by the US Geological Survey, and adopted by many other agencies.  An example is the USDH 48 depth-integrated sampler (or, if a bridge is available, the US D49 sampler, also of the US Geological Survey and FISC (US Federal Inter-Agency Sedimentation Committee). The USDH 48 sampler, now used internationally, consists of a US 1-pint (~0.5L) bottle fitted within a streamlined metal cowling which is fixed to a wading rod, or suspended by cables from river bridges.

Isokinetic samplers and sampling

N.B. Note that all field safety procedures of a researcher’s institution and/or government need to be observed by the whole party when field sampling.

The key point of this type of USGS sampler is that the sampling bottle is only allowed to fill with water through a special, carefully-machined nozzle of 6.3mm diameter (1/4 inch) which, when pointed upstream, allows water to flow into the sampler at the same velocity (+/- 10%) as water passing around the sampler. This is called ‘isokinetic sampling’ and is very important for representative suspended particulate sampling in flowing water (and air).

One advantage of isokinetic sampling is that it is less affected by uncertainties due to the selective collection of the larger, less easily deflected particulates. In principle, an isokinetic sampling device should have a collection efficiency of unity for all sizes of particulates in the sampled water.

At each sampling position (the ‘vertical’) in the stream, such samplers need to be lowered at an even rate to the bed and then raised again to the surface.  This allows sampling of most of the water column, except for the lowest few centimetres (e.g. 9cm for the USDH 48 sampler). Vertical Transits are repeated, if necessary, at the same vertical until the bottle is ~70 % full, thus avoiding overfilling which can bias a sample.  This is called the Equal Transit Rate method in USGS publications (see Guy and Norman, 1970; Edwards and Glysson, 1999; Gray and Gartner, 2010; Gray et al., 2005).

This vertical sampling procedure is then repeated at ‘no less than 3 and usually 5 to 10 verticals in a given (stream) cross section’ (Dunne and Leopold, 1978, p.715), to capture the lateral variations in SSC across the stream channel.  If you have multiple channels within a braided river system, then each channel should be measured, and total fluxes aggregated to give total sediment transport.  This is very laborious, so many workers choose to sample at a point where the channel is a single-thread stream.

In shallow rivers, depth-integrating samplers can be fitted to a wading rod (e.g. USDH 48) and safely hand-lowered to the bed by a fieldworker in waders or a dry-suit. In deep rivers, depth-integrated sampling may need to be carried out from boats, or at bridges by deploying the sampler on a long rod or cable, such as the US DH 59, US D-43 or US D-49.

Note also that point suspended sediment samplers exist to sample at a known depth (rather than integrating over the whole vertical).

Turbid water samples are carefully transferred from the sampler bottle to plastic sample bottles, which are then sealed, and labelled, and transported back to the laboratory for analysis of SSC. To do this, samples of known volume are simply filtered normally through 0.45 μm (micron) filter membranes.

SSC (mg/L) =  dry weight of suspended sediment (mg) / volume of water sample  (L)

Vacuum filtration can be used to speed up the filtration process if clogging of the filter paper/membrane occurs, for example in Icelandic glacial rivers:

e.g. see

Lawler et al. 2003 Hydrological Processes, 17, 3195–3223;  and

Old, Lawler and Snorrason 2005 Earth Surface Processes and Landforms, 30, 1441–1460.

I hope this note, and the references it contains, is helpful!  Good luck!

Wow, I finally took the opportunity to return back to see your progress.  I did not have time to review the first comment when I commented.  You have some excellent and complete advice.  I seldom had the time to implement all that is suggested, but understand so well that I could not imagine anyone giving better and such concise advice.  I did quite a bit of work with ISCO sediment samplers with stage actuators in Oregon, along with some DH-48 cross sections, but there was never any time to publish results.  Although the relationship between turbidity and suspended sediment changes by drainage, you can sometimes get some good correlations.  I know my reference to Dave Rosgen is not always well received, but since moving to SC and obtaining his training,  I have been impressed in his efforts to normalize the flow sediment concentration curve by normalizing by sediment and flow at bankfull, so you are plotting flow sediment information based on the comparison to the value at bankfull.  If you have or know of data sets that have tried this, it would be interesting to see the results in comparison to his.  Most researchers agree that the bankfull flow is the channel forming flow for stable channels and when streams reach their floodplain, the efficiency to move sediment reduces.  Rosgen's efforts in trying to estimate bank erosion have also been helpful tools or indicators.  In several gullied channels, we estimated that about 90% of the sediment was coming from bank erosion, something we had not imagined.  I am not too old to learn more though and will take some time to look up these references.  I am happy I came back and took another look, so much so, I copied into a word document for further study.  You know, I seldom give without getting something back.  I am so glad I found this researchgate, if nothing else to share and keep involved.

Dear Pankaj

in the end what did you do?

How did you sample the suspended sediments in your sections?

With which regularity?

Over which time period?

How did you compare point sampling with cross section sampling?

I am interested in what you could achieve in regard of your initial goal and of the field constraints.

Cheers

Gil