Suspended solids

One of the water quality parameters that has been most extensively studied in this way is the concentration of total suspended solids (TSS). Increases in the amount of suspended particles will, at wavelengths where the particles do not absorb strongly, increase the backscattering coefficient of the water more than the absorption coefficient, and so, in accordance with eqns 6.3 and 6.5, increase the emergent flux.

In the case of reservoirs in Mississippi, it was found that the irradiance reflectance measured above the water surface increased with the concentration of suspended (inorganic) sediment at all wavelengths from 450 to 900 nm (Fig. 7.6).1136 The best, linear, correlation between reflectance and suspended solids concentration was found between 700 and 800 nm. Correlations between suspended solids or nephelometric turbidity (proportional to suspended solids) and radiances in particular wavebands have been obtained with the Landsat satellites. Klemas et al. (1973,

Fig. 7.6 Intensity and spectral distribution of upwelling radiation measured 20 to 50 cm above the surface of freshwater bodies containing various levels of suspended solids (after Ritchie, Schiebe and McHenry, 1976). The concentration of suspended solids in mg is indicated next to each curve.

1974) found that in Delaware Bay the distribution of mineral suspended sediments (MSS) 600-700 nm radiance corresponded best with sediment load in the upper 1 m of the water (Fig. 7.7); the 500 to 600 nm radiance was more subject to interference from atmosphere haze while the 700 to 800 nm waveband did not penetrate sufficiently into the water column. Similarly, in Landsat imagery of Kenyan coastal waters, the 700 to 800 nm band revealed only high sediment concentrations near the surface, whereas the 600 to 700 nm and 500 to 600 nm bands were sensitive to lesser concentrations lower in the water column.156 Suspended solids values for the Bay of Fundy, Nova Scotia, Canada, correlated with particular functions of upward radiance in the 500 to 600, 600 to 700 and 700 to 800 nm bands, but the latter two bands gave the highest

Delaware City Cao CANAL


Concentrations form nonlinear regression. mg/liter

US 104 7.7

| unclassified


Rehoboth Beach



Ocean City,

Fig. 7.7 Distribution of suspended solids concentration in Delaware River estuary and Delaware Bay derived from Landsat 600 to 700 nm radiance measurements (by permission, from Klemas et al. (1974), Remote Sensing of Environment, 3, 153-74).



correlation coefficients.965 In the case of Lake Ontario, Canada, a good linear relation was observed between radiance reflectance in the 600 to 700 nm band and measured turbidity.181 For two turbid oxbow lakes of the Mississippi River (USA), Moon Lake (Mississippi) and L. Chicot (Arkansas), it was found that of the four Landsat MSS bands, the 700 to 800 nm band was the best for monitoring suspended sediment loads.1132,1135,1185

Munday and Alfoldi (1979) found that Landsat radiance values correlated somewhat better with the logarithm of TSS than with TSS alone, in the Bay of Fundy. Some algorithms for suspended sediments have been developed using combinations of Landsat radiances rather than the radiance values in a single waveband.683,1132,1133,1371 The quest for a universal con 1 1 ^^

algorithm for suspended sediments , is, however, never likely to succeed, no matter what combination of bands is used. The main reason is that the scattering efficiency of suspended particles per unit weight is very much a function of size (Fig. 4.3), and average particle size of suspended sediments is quite variable, both with time and place. There are additional reasons related to the absorption properties of the surrounding medium and of the particles themselves. Since reflectance varies inversely with a, at the same time as it increases with bb, a reproducible linear relation between emergent radiance and TSS is only to be expected if the total absorption coefficient of the aquatic medium remains constant while the concentration of suspended solids varies. This need not be the case, and variations in absorption coefficient from place to place or over a period of time, within the water body under study, would affect the reproducibility of any empirically derived relation between Lw and TSS, and so affect the accuracy with which the concentration of suspended solids can be determined from remotely sensed data. In the red and near-infrared wavebands, however, absorption is mainly due to water itself, and so fluctuations in absorption should not be large.

Confirming the advantage of the near-infrared for monitoring suspended particles, Doxaran et al. (2002) found, for the highly turbid waters of the Gironde estuary (France), that the (atmospherically corrected) reflectances in the visible-wavelength bands (green, 500-590 nm; red, 610-680 nm) of the HRV sensor on the SPOT satellite correlated poorly with suspended particulate matter (SPM), showing a logarithmic increase up to 500 mgl-1 and then saturating. Reflectance in the near-infrared (790-890 nm), and also the ratio of near-infrared reflectance to that in either visible band, by contrast, were highly correlated with SPM. On the basis of their in situ measurements of spectral reflectance and concentration of suspended particles they arrived at an algorithm, expressing the ratio of reflectance in the near-infrared (R790 890) to that in the green (R500 590) as a function of SPM in mgl-1.


Applied to a high-resolution (2.5 m) SPOT image of part of the Gironde, it revealed a detailed picture of sedimentary flow in the estuary.

Binding et al. (2005) studied the relationship between concentration of MSS and subsurface irradiance reflectance in six wavebands in the 400 to 700 nm region at a large number of stations in the moderately turbid Irish Sea. The most robust relationship between MSS and reflectance was found in the red (665 nm). These authors attributed the residual variability in the results to variation in the specific scattering coefficient (b*MSS) of the particles, which in fact extended over the range, 0.04 to 0.57 m2 g-1 for all stations. When their regression model was re-run using the local value for specific scattering coefficient applicable to each station, the agreement was much improved. Their results suggest that the errors in predicted MSS concentration can be reduced from 56% to as little as 12% with prior knowledge of the scattering properties of the sediments under study. Binding et al. suggest that in order to obtain quantitative estimates of MSS in moderately turbid waters from space it may be necessary to predetermine scattering efficiencies (b*MSS) within the area of interest.

The spectral distribution of light reflected by suspended inanimate particulate matter varies in accordance with its own absorption properties, which are determined by its chemical nature. For example, suspensions of white clay and red silt have quite different reflectance spectra.1000 Sydor (1980), on the basis of his measurements of the spectral distribution of light scattered upwards from suspensions of particular types (red clay, mine tailings) concluded that evidence on the identity as well as concentration of the suspended solids may be derived using the 500 to 600, 600 to 700 and 700 to 800 nm bands of Landsat.

The work that we have discussed so far, and indeed the majority of studies on TSS that have been carried out, have made use of Landsat MSS data, but there have been some studies with other sensors. Ritchie, Cooper and Schiebe (1990) found Thematic Mapper data to be as good as, but no better than, MSS data for estimating suspended sediments in Moon Lake (Mississippi), despite the narrower spectral band width of the former. They consider that the major justification for using Thematic Mapper data for sediment studies would be that its higher spatial resolution (30 m) would make it possible to monitor smaller lakes. The SPOT multispectral scanner has a ground-level resolution of 20 m. For the turbid waters of Green Bay, L. Michigan, Lathrop and Lillesand (1989) found that for extra-atmospheric reflectance in all three bands (green, red, near-infrared) the best correlation was with the logarithm of suspended sediment concentration: some improvement was obtained by regressing the logarithm of TSS against a combination of the three reflectances.

Stumpf and Pennock (1989) developed a method for determining suspended sediment concentrations in estuaries using the red and near-infrared channels of the AVHRR. They used a correction procedure similar to that developed for CZCS data to remove atmospheric effects, and worked in terms of the calculated values for reflectance. For the turbid waters along the Louisiana (USA) coast, Myint and Walker (2002) compared (atmospherically corrected) reflectances measured in two AVHRR wavebands with concentrations of total suspended solids (TSS) and suspended sediments (SS; SS = TSS after ignition) determined on water samples collected within a few hours of the satellite overpass. For Channel 1 (580-680 nm) TSS and SS as a function of reflectance were satisfactorily represented in the form of cubic equations. For the near-infrared Channel 2 (725-1100 nm), simple linear functions were found to work well.

Using atmospherically corrected reflectance values obtained with AVHRR band 1 (580-680 nm), Bowers et al. (1998) produced mean winter and summer maps of MSS concentration in the Irish Sea for the years 1982-88. For the Irish Sea as a whole, concentrations in winter were found to be higher than those in summer by a factor of 2.7. The highest sediment concentrations were found to occur in the more shallow areas and in the regions of strongest tidal currents.

Some, limited, use of SeaWiFS imagery has been made for remote sensing of the distribution of SS. Myint and Walker (2002) compared water-leaving radiance values, atmospherically corrected in accordance with Gordon and Wang (1994), with TSS and SS values from water samples collected off the Louisiana coast within a few hours of the satellite overpass. Correlation coefficients between Channel 5 (555 nm) radiances and TSS/SS were poor, only 0.46 to 0.47. In the case of Channel 6 (670 nm), however, the correlation coefficient was quite high, 0.85. Of various statistical models tried, expressing TSS or SS as a simple power function of L(670) gave the best results. Binding et al. (2003) measured subsurface irradiance reflectance in the SeaWiFS visible wavebands, together with the concentration of MSS at 124 stations in the Irish Sea. The ratio of reflectance in the red to that in the green, which might have been expected to be strongly correlated with MSS, was found to be a reliable indicator only in those waters where MSS dominated the optical properties: with increasing influence of phytoplankton or yellow substance the relationship broke down. Reflectance in the red region (665 nm) on its own, on the other hand, showed a satisfactory correlation with MSS throughout the region studied, which could be represented by the following algorithm.

The application of this algorithm to a SeaWiFS image of the Irish Sea accurately reproduced known regions of high turbidity, with realistic concentrations of MSS.

For turbid waters in the Bay of Biscay, Froidefond et al. (2002) found the (atmospherically corrected) SeaWiFS radiance at 555 nm to be strongly correlated with the concentration of suspended particulate matter (SPM) measured at sea level. From the data, the algorithm

where Y is SPM conc. in mgl-1, and X is the normalized water-leaving radiance at 555 nm, was derived.

Warrick et al. (2004) used SeaWiFS to map the distribution of SS in the turbid coastal waters of the Santa Barbara Channel (California). For atmospheric correction, the dark pixel method (see above) was used. The reflectance of the darkest single ocean pixel, patm, was assumed to be contributed entirely from the atmosphere. Reflectance due to water-leaving flux was then obtained from m -Pt(1) - Patm (1)

Where pt(1) is the total reflectance, and tv(1) is the diffuse atmospheric transmittance along the line of view of the sensor. Only reflectances in the 555, 670, 765 and 865 nm bands were used: the data from bands in the violet and blue (412, 443, 490 and 510 nm) were omitted because these are the bands most affected by phytoplankton and CDOM. To determine the sediment concentration, spectral mixture analysis was used, the end members for the analysis being provided by laboratory measurements of water-leaving spectral reflectance on sediment suspensions of known concentration. Images of sediment distribution up to 60mgl-1, in increments of 10mgl-1, were obtained.

The 1 km spatial resolution of SeaWiFS is entirely suitable for the open ocean, but is not optimal for small-scale coastal features such as estuaries and bays. The Landsat and Landsat TM satellites have much better spatial resolution (80msq, 30msq) but have a revisit time of —16 days, and are therefore not well suited for monitoring dynamic change in coastal ecosystems. Sensors such as SeaWiFS and MODIS, by contrast, achieve near-global coverage every one to two days. The MODIS sensor on the Terra spacecraft, in addition to the nine narrow (—10 nm) spectral bands in the visible and near-infrared used for ocean colour, which have a spatial resolution of —1km, has two broader wavebands, Band 1 (620670 nm) and Band 2 (842-876 nm), with 250 m sq resolution.924 Miller and McKee (2004) used MODIS Band 1 to map the concentration of total suspended matter in coastal and estuarine waters in the northern Gulf of Mexico. A linear relationship (r2 = 0.89) was found between Band 1 reflectances, atmospherically corrected by the dark pixel subtraction method, and the total suspended matter (TSM) concentration measured on water samples from 52 stations in this region. For determining TSM concentration from Band 1 reflectance, the following algorithm was derived

TSM (mg l-1) = 1140.25 (Band 1 reflectance, %) - 1.91

Was this article helpful?

0 0

Post a comment