Full text of DOI:10.1080/01431160050144938

This article was downloaded by: [UQ Library]
On: 02 November 2014, At: 10:42
Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number:
1072954 Registered office: Mortimer House, 37-41 Mortimer Street,
London W1T 3JH, UK
International Journal of
Remote Sensing
Publication details, including instructions for
authors and subscription information:
http://www.tandfonline.com/loi/tres20
The influence of surface
and interstitial moisture on
the spectral characteristics
of intertidal sediments:
Implications for airborne
image acquisition and
processing
M. P. Rainey a , A. N. Tyler a , R. G. Bryant a , D. J.
Gilvear a & P. McDonald b
a Department of Environmental Science ,
University of Stirling , Stirling, Scotland, FK9 4LA,
UK
b Westlakes Scientific Consulting Ltd , Westlakes
Science and Technology Park , Moor Row,
Cumbria, CA24 3LN, England, UK
Published online: 25 Nov 2010.
To cite this article: M. P. Rainey , A. N. Tyler , R. G. Bryant , D. J. Gilvear & P.
McDonald (2000) The influence of surface and interstitial moisture on the spectral
characteristics of intertidal sediments: Implications for airborne image acquisition
and processing, International Journal of Remote Sensing, 21:16, 3025-3038, DOI:
10.1080/01431160050144938
To link to this article: http://dx.doi.org/10.1080/01431160050144938
PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all
the information (the “Content”) contained in the publications on our
platform. However, Taylor & Francis, our agents, and our licensors
make no representations or warranties whatsoever as to the accuracy,
completeness, or suitability for any purpose of the Content. Any opinions
and views expressed in this publication are the opinions and views of
the authors, and are not the views of or endorsed by Taylor & Francis.
The accuracy of the Content should not be relied upon and should be
independently verified with primary sources of information. Taylor and
Francis shall not be liable for any losses, actions, claims, proceedings,
demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in
relation to or arising out of the use of the Content.
This article may be used for research, teaching, and private study
purposes. Any substantial or systematic reproduction, redistribution,
reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access
and use can be found at http://www.tandfonline.com/page/terms-and-
conditions

int. j. remote sensing, 2000, vol. 21, no. 16, 3025–3038
The in uence of surface and interstitial moisture on the spectral
characteristics of intertidal sediments: implications for airborne image
acquisition and processing
M. P. RAINEY†‡, A. N. TYLER, R. G. BRYANT‡, D. J. GILVEAR
Department of Environmental Science, University of Stirling, Stirling FK9 4LA,
Scotland, UK
and P. McDONALD
Westlakes ScientiŽ c Consulting Ltd, Westlakes Science and Technology Park,
Moor Row, Cumbria CA24 3LN, England, UK
(Received 6 October 1998; in Ž nal form 29 November 1999)
Abstract. Intertidal sediment distribution is an important factor in the under-
standing of estuarine ecological, morphological and pollution processes. Airborne
remote sensing is identiŽ ed as a cost-eŒective tool that may be used to provide
accurate synoptic maps of estuarine sediment distributions. However, prior to the
collection and analysis of airborne imagery, spectral characterization of intertidal
sediments is fundamental. Through a series of in situ and laboratory re ectance
(0.35–2.5 mm) experiments the spectral complexity of the Ribble Estuary, UK,
intertidal sediments has been investigated. Positive relationships between
Daedalus Airborne Thematic Mapper band 9 and band 10 simulated re ectance
and percentage sand of intertidal sediments have been identiŽ ed in the Ž eld
2
2
(r 5 0.51 and r 5 0.61).
These associations are sensitive to the interstitial and surface moisture condi-
tions of the intertidal environment and, consequently, temporally unstable.
Laboratory experiments demonstrated that interstitial moisture in uences the
spectral contrast between sediments of contrasting grain size, consequently deter-
mining the ability to map sediment domains through remote sensing. As a result,
it is recommended that airborne imagery be collected of intertidal areas after a
prolonged period of exposure to summer drying conditions. This ensures that the
spectral in uence of moisture is minimized and the spectral contrast between
sediments of diŒerent grain size is maximized.
1. Introduction
Conventional mapping of intertidal sediment domains using point samples can
be spatially unrepresentative, due to errors introduced through sediment sampling
and subsequent interpolations (Tyler et al. 1996). To address this, airborne remote
sensing is being used to derive synoptic information and quantitative data on
sediment grain-size distributions within estuaries. Such data are fundamental in the
†Corresponding author; email: m.p.rainey@she eld.ac.uk.
‡Present address: Department of Geography, University of She eld, Winter St., She eld
S10 2TN, England, UK.
International Journal of Remote Sensing
ISSN 0143-1161 print/ISSN 1366-5901 online © 2000 Taylor & Francis Ltd
http://www.tandf.co.uk/journals

3026
M. P. Rainey et al.
monitoring of estuarine processes, for example to map sources and sinks of particle
reactive pollutants, e.g. anthropogenic radionuclides (e.g. Tyler et al. 1995). Equally,
an understanding of the distribution and stability of estuarine sediments can provide
an important insight into littoral estuarine ecology (e.g. Goss-Custard and Yates
1992, Black and Paterson 1997).
A number of satellite remote sensing studies have previously attempted to map
the distribution and nature of intertidal sediments (Folving 1984, Bartholdy and
Folving 1986, DoerŒer and Murphy 1989, Yates et al. 1993, Populas et al. 1995).
However, the poor spatial, spectral and temporal resolution of satellite datatypes,
makes them unsuitable for detailed morphological studies of tidal  ats. This is
primarily due to the complex spatial variability of the sediment distributions and
the need for cloud-free imagery coincident with particular tidal conditions. In con-
trast, airborne data oŒers superior spectral and spatial resolutions in comparison to
the majority of satellite sensors. More importantly, airborne data is not constrained
to a regular orbit and, consequently, imagery can be collected of an estuary during
a particular tidal state whenever cloud-free conditions permit. These advantages
make airborne imagery ideal for detailed quantitative monitoring of surface intertidal
sediment grain sizes.
Successful airborne mapping of intertidal sediment facies necessitates accurate
deŽ nition of the spectral properties of sediments of contrasting grain sizes. The
re ectance (0.4–2.5 mm) of intertidal sediment is controlled by a range of properties
such as grain size, moisture content, organic content, cyanobacteria, iron oxide
content and mineralogy (Asrar 1989). To spectrally determine grain size, it is therefore
necessary to recognize how variations in this sediment characteristic in uence the
re ectance spectrum and how this is aŒected by other sediment properties, primarily
surface and interstitial moisture.
This study focuses on the Ribble Estuary, Lancashire, England (Ž gure 1), and
forms part of a wider study of the estuary using airborne remote sensing. The results
of in situ spectroradiometry work at Warton Bank in 1994 with the GER-SIRIS
spectroradiometer demonstrated that a strong negative relationship existed between
the clay content of the sediments and re ectance within the 0.4–2.5 mm range (Tyler
et al. 1995, Bryant et al. 1996). Outlined in this paper are the results of further in
situ and laboratory experiments, carried out in 1995, 1996 and 1997, to quantify the
in uence of interstitial and surface moisture conditions on the relationship between
sediment grain size and re ectance. The implication of these results on the timing
of future airborne remote sensing campaigns is also discussed.
2. Methodology
2.1. Study area
The Ribble Estuary joins the Irish Sea at Lytham St Anne’s on the north-west
coast of England (Ž gure 1). Tides in the estuary have an amplitude between 5 and
9 m and are asymmetrical in nature with a strong  ood, which gives rise to a
turbulent, well-mixed system (Burton et al. 1995). The estuary receives direct inputs
of liquid and particulate e‚ uents containing mainly Th and U isotopes and decay
products (BNFL SpringŽ elds, close to the head of the estuary) and anthropogenic
radionuclides from the Irish Sea attached to marine sediment (BNFL SellaŽ eld).
Two sites of contrasting location, scale, sediment type and topography were
selected for spectral investigation (Ž gure 1). Warton Bank is representative of the
narrow sloped intertidal area, typical of the middle and inner reaches of the estuary

Spectral characteristics of intertidal sediments
3027
Figure 1. Location of the Lytham St Anne’s and Warton Bank sample sites, with the 1995,
1996 and 1997 Warton Bank sampling transects and sample numbers outlined.
and Lytham St Anne’s is an example of the relatively  at large intertidal stretches
in the outer reaches of the estuary. Warton Bank contains a range of sediment types
including muds, silts and sands, whilst Lytham St Anne’s is dominated by Ž ne
sediment. The intertidal regions of both Lytham and Warton have been shown to
be rapidly accreting sediments (Brown 1997).
2.2. Spectroradiometers
The Geophysical and Environmental Research Corporation (GER) IRIS Mk IV
spectroradiometer (Natural Environment Research Council Equipment Pool for
Field Spectrometry, NERC EPFS) was used to collect bi-directional re ectance
spectra (350–2500 nm) of the intertidal sediments in 1995. This instrument is a dual-
beam Ž eld spectroradiometer that can collect a simultaneous measurement of target
and reference panel, oŒering the best correction for changes in incident illumination.
During 1996 and 1997, the Analytical Spectral Devices (ASD) high-performance
single-beam Ž eld spectroradiometer (NERC EPFS), which was not available for the
earlier Ž eldwork, was selected in preference to the GER IRIS Mk IV spectroradi-
ometer. The ASD, which measures between the visible and short-wave infrared
wavelength range, was used in a series of in situ and laboratory re ectance experi-
ments. Its superior spectral collection and storage capabilities, combined with its
portability and real-time spectra viewing capabilities, enabled a more detailed spectral
investigation of the intertidal environment. The demand for the NERC EPFS

3028
M. P. Rainey et al.
hardware was so great in 1996 and 1997 that it was not possible to use both
spectroradiometers at the same time.
2.3. Sediment sampling and analysis
To determine the relationship between in situ re ectance and sediment character-
istics, sediment samples were collected where re ectance measurements were made.
The sediment surface scrapings ( 0.5 cm) were collected, doubled bagged and labelled
<
immediately after spectra collection. A qualitative visual assessment of surface mois-
ture conditions, cyanobacteria cover, surface mud draping and polygonation was
also undertaken at each site. The grain size of the samples was determined using a
Malvern Longbed particle-sizer. The sediment moisture content was estimated by
drying in a fan-assisted oven at 100ß C.
2.4. In situ experimental technique
Field sampling was undertaken over a range of years and conditions to maximize
the variation with the sample set. Sampling was repeated each year along transects
at Warton Bank to highlight any seasonal in uence on the re ectance of the intertidal
sediments. In August 1995, the GER IRIS spectroradiometer was used to collect
spectra along two transects from the saltmarsh edge to the main channel at Warton
Bank (Ž gure 1). The intertidal area was relatively dry after being exposed to summer
drying conditions for a few hours before sampling. At each sample site, Ž ve repeat
spectra and a sediment sample were collected simultaneously.
During the 1996 study the ASD spectroradiometer was used to collect 10 repeat
sediment spectra and an associated reference panel spectra (NERC: Spectralon Panel
1011) at each site along one perpendicular and three parallel transects to the main
channel (Ž gure 1). The perpendicular transect was Ž rst sampled between 09:30 and
10:30 before tidal inundation. At 15:30, immediately after sediment re-exposure,
sampling was repeated along this transect and the three parallel transects were
subsequently sampled. Sampling along the parallel transect closest to the saltmarsh
was partially repeated following a short period of exposure. A preliminary investi-
gation was also undertaken at Lytham St Anne’s, where re ectance spectra with
associated sediment samples were collected along a short transect perpendicular to
the channel.
As a continuation of the 1996 study, Ž eldwork was carried out in February 1997
to characterize the direct eŒect of moisture and exposure on the re ectance values
of intertidal sediments. At Warton Bank sediment and reference spectra (NERC:
Spectralon Panel 1011) were collected, with associated samples, using the ASD
spectroradiometer along three transects perpendicular to the main channel (Ž gure 1).
At Lytham St Anne’s, similar sampling was also carried out along two transects
perpendicular to the main channel. In both study areas, sampling was repeated along
one of the transects following a period of exposure. However, the intertidal areas
appeared relatively saturated throughout this Ž eld study, due to evaporation rates
being typically low during the month of February.
2.5. L aboratory experimental technique
To quantitatively investigate the spectral in uence of moisture, a series of experi-
ments were carried out under controlled laboratory conditions. Six samples from
the Ribble Estuary were prepared for analysis. These include three bulk sediment
samples, of known grain size (table 1), and three wet sieved samples ( 63 mm,
<

Spectral characteristics of intertidal sediments
3029
Table 1. Characteristics of the 1996 laboratory sediment samples.
Sample A
Sample B
Sample C
Mean grain size (mm)
% Sand
% Silt
186.63
97.72
1.55
135.73
86.26
12.13
1.61
44.97
28.11
65.18
6.71
% Clay
0.73
63–125 mm and 125–215 mm) which were placed in shallow black containers to a
depth of 2 cm. Each sample was saturated, weighed and then 10 repeat re ectance
spectra of the sediments and a reference spectrum (NERC: Spectralon panel 1011)
were collected using the ASD spectroradiometer, with a halogen lamp as the illumina-
tion source. The samples were then placed in a fan-assisted drying oven (100ß C),
and at regular intervals were re-weighed and a set of spectra collected. This procedure
continued until each sample weight stabilized, i.e. the sample was dry. These experi-
ments were carried out in a darkroom, with photographic tape and black card used
to prevent light scatter from the instrumental and background surfaces.
2.6. Spectral analysis
Using the associated reference spectrum and the reference panel calibration Ž le,
each re ectance spectra collected was initially converted to absolute re ectance. The
repeat absolute re ectance Ž les were then averaged for each site/sample and the
atmospherically aŒected windows between 1353 and 1408 nm, 1803 and 1969 nm and
2483 and 2500 nm were removed. To allow comparison with Airborne Thematic
Mapper (ATM) imagery, the resultant spectra were convolved into ATM bandwidths,
using Ž lter function Ž les provided by NERC EPFS.
2.7. Statistical analyses
A Pearson’s correlation coe cient analysis was undertaken on the data to identify
any signiŽ cant relationships between the simulated ATM re ectance and the
intertidal sediment characteristics. The conventional signiŽ cance level of 0.05 and
below was used to identify appropriate correlations. Where appropriate, regression
analysis was also undertaken.
Principal component analysis (PCA) produces an optimal presentation of multi-
variate data by projecting the raw data into the space deŽ ned by the eigenvectors
of the data variance–covariance matrix (Smith et al. 1985). This technique was
applied to all the simulated ATM bands of the in situ spectra collected during 3
years of study and to the simulated bands of the laboratory spectra. From each set
of PCA results, the Ž rst and second principal components were then plotted against
each other in a scatterplot, which was subsequently used to identify the spectral
endmembers.
3. Results
3.1. In situ characterization
Analysis of the in situ spectral datasets, identiŽ ed a signiŽ cant positive relationship
between percentage sand ( 63 mm) content of intertidal sediment, which was exposed
>
for a long period of time ( 3 h) to strong summer drying conditions, and almost all
>
the simulated ATM bandwidths (table 2). A similar, though negative, relationship

3030
M. P. Rainey et al.
Table 2. Pearson correlation and level of signiŽ cance for ATM bands against percentage
sand content of intertidal sediment.
Pearson correlation with %
sand (P)
Pearson correlation with %
sand (P)
ATM bandwidth
Õ
Õ
Õ
Õ
Õ
Õ
Õ
Õ
Õ
Õ
ATM 1
ATM 2
ATM 3
ATM 4
ATM 5
ATM 6
ATM 7
ATM 8
ATM 9
ATM 10
0.53 (NS)
0.41 (NS)
0.45 (NS)
<
0.56 ( 0.05)
<
0.56 ( 0.05)
0.66 ( 0.02)
<
<
0.66 ( 0.01)
0.65 ( 0.02)
<
0.54 ( 0.05)
<
<
0.56 ( 0.05)
0.61 (0.02)
0.45 (NS)
0.49 (NS)
0.50 (NS)
0.34 (NS)
0.38 (NS)
<
0.68 ( 0.01)
<
0.61 ( 0.02)
<
0.78 ( 0.01)
<
0.72 ( 0.01)
existed between the simulated ATM bandwidths and percentage clay ( 2 mm) con-
<
tent. In both cases the strongest and most signiŽ cant linear relationship existed
between percentage sand and ATM bands 9 (1.55–1.75 mm) (r 0.51) and 10
2 5
(2.08–2.35 mm) (r 0.61). In addition, these bandwidths had signiŽ cantly greater
2 5
coe cients of variance (%CV) along the transects sampled, in comparison to the
other bands, indicating that they are potentially more responsive to changes in
sediment conditions. Table 3 shows that the average spectral variation within bands
9 and 10 was greater along transects perpendicular to the main channel, i.e. the
direction of maximum change in sediment grain size (%CV 37.0, 53.6), than along
5
the parallel transects (%CV 19.7, 25.95). Along the perpendicular transect sampled
5
at Warton bank in 1996, a positive relationship between bands 9 and 10 and
percentage sand was observed before tidal inundation (0.52 and 0.45, p 0.01).
<
3.1.1. In uence of interstitial moisture on the sediment–re ectance relationship
In contrast to the above, a signiŽ cant negative relationship was identiŽ ed between
Õ
Õ
ATM bands 9/10 and percent sand ( 0.6 and 0.6, p 0.01) immediately following
<
tidal inundation, as shown in Ž gure 2. This suggests that when intertidal sediments
are completely saturated, the sand has a signiŽ cantly lower re ectance than the mud,
which is the reverse to what was observed when the sediments were relatively dry.
This re ectance relationship with grain size, which was only observed once during
the Ž eld investigations, was better described when percentage moisture was included
in the multiple regression equation below (r increased from 0.36 to 0.67),
2
[
]
[
sand(%) 0.00265
Ö
]
moisture(%)
Õ
ATM9 0.302 0.00182
Õ
(1)
5
Ö
Table 3. Spectral variation along the perpendicular (A ) and parallel (B , C and D
transects sampled during the 1996 study.
)
96
96 96
96
Transect
Orientation to main channel CV% (ATM9)
CV% (ATM10)
Transect A (Dry)
Transect A (Wet)
Transect B (Wet)
Transect C (Wet)
Transect D (Wet)
perpendicular
perpendicular
parallel
parallel
parallel
35.6
38.4
12.1
28.1
19.0
51.8
55.4
19.0
32.7
26.2

Spectral characteristics of intertidal sediments
3031
Figure 2. Transect A at Warton Bank sampled in 1996 demonstrating (a) topography,
(b) re ectance before high tide and (c) re ectance after high tide.

3032
M. P. Rainey et al.
As shown in equation (1), the eŒect of interstitial moisture on the spectral response
of the sediment is greater than the in uence of percentage sand concentration, and
any remaining noise may be associated with surface moisture.
Similarly, within intertidal sediments exposed for only short periods of time to
strong drying conditions, and which consequently have relatively high interstitial
moisture contents, only weak relationships, if any, were observed between ATM
bands 9 and 10 and percentage sand/clay. For example, no signiŽ cant relationship
between sediment grain size and ATM bands 9 and 10 re ectance were observed
along transects parallel to the main channel sampled shortly after tidal inundation
at Warton Bank in 1996. Equally, after a short period of exposure at Warton Bank
in 1997, no signiŽ cant relationship was observed along the transects sampled. For
example, a negative though insigniŽ cant correlation was observed (transect B,
Õ
r
positive though insigniŽ cant relationship (transect C, r
0.41, p 0.1) between simulated ATM band 10 and percentage sand, and a
<
5
Õ
0.41, p 0.1) was observed
<
5
between simulated ATM band 9 and percentage sand. Very similar observations
were also made at Lytham St Anne’s in 1997 (Ž gure 1) between ATM band 10 and
percentage sand (r 0.38, p 0.1), conŽ rming that this was not a site-speciŽ c eŒect.
5
<
No consistent correlation was observed on these wet sediments, demonstrating the
importance of water on spectral re ectance and the inability to characterize them
under such wet conditions.
3.1.2. In uence of surface water on the sediment–re ectance relationship
Surface water, in the form of small pools and saltmarsh creeks, was found along
some of the transects sampled. As expected, these pools signiŽ cantly attenuated the
overall sediment re ectance, masking the relationship between sediment grain size
and re ectance. Consequently, the spectral variability within sediment domains
containing pools and/or creeks is predominantly caused by surface water attenuation.
For example, along transect A at Warton Bank, sampled before tidal inundation in
1996, a large pool of water dramatically lowered the sediment re ectance, increasing
the spectral variability (%CV 37.5) within the sediment domain (Ž gure 2(b)).
5
3.1.3. Results of PCA of in situ spectra: spectral decorrelation
From the scatterplot of PCA bands 1 and 2, it was found that pure sand and
mud, which are exposed for a long period of time to rapid drying conditions, are
spectrally distinct in ‘feature space’ as seen in the scatterplot (Ž gure 3). Although no
clear clustering exists, a general trend from dry sand to dry mud is recognizable
along PC1, which contains 73% of the spectral variance. This is underlined when
the endmember sand and mud spectra are examined individually as in Ž gure 4. It is
also evident from this scatterplot that grain size is not the dominant determinant of
saturated sediment spectra distribution in feature space. Although some of the
saturated sand spectra are clustered, the rest of the spectra, including the
remaining saturated sand spectra, are randomly distributed. The saturated sediment
spectra are predominately located below the solid line in Ž gure 3.
3.2. L aboratory characterization
The re ectance of the representative sand and silt samples (Ž gure 5) and the
215–125 mm and 63–125 mm samples (Ž gure 6), approximate a positive linear relation-
ship with loss of moisture. The strongest relationships exist between ATM bands 9
and 10 and percentage moisture (table 4). Although the association between band 9
and percentage moisture for these samples is approximated as being linear there is

Spectral characteristics of intertidal sediments
3033
Figure 3. Scatterplot of PC1 against PC2, with the location of the main endmembers labelled.
Figure 4. Example ATM-simulated endmember spectra, representing a dry mud and a dry
sand sample identiŽ ed through PCA.
an indication of an additional in uence as the relationship is better characterized by
a slightly sinuous line rather than a straight one. The spectral contrast between the
coarser samples does not change signiŽ cantly with drying, in band 9, due to their
similar response to drying.
In contrast, the representative mud and the 63 mm samples have a nonlinear
<

3034
M. P. Rainey et al.
Figure 5. Relationship between ATM band 9 re ectance and percentage moisture for typical
(a) sand, (b) silt and (c) mud samples from the Ribble Estuary.
Figure 6. Relationship between ATM band 9 re ectance and percentage moisture for three
<
sediment size fractions, (a) 215–125 mm, (b) 125–63mm and (c) 63 mm.
Table 4. Coe cient of variance for each ATM band, and the correlation and regression
values between re ectance and loss of moisture for the sand sample of the laboratory
experiments.
2
R
ATM band
Coe cient of variance
Pearson correlation (signiŽ cance level)
Õ
Õ
Õ
Õ
Õ
Õ
Õ
Õ
Õ
Õ
<
ATM 1
ATM 2
ATM 3
ATM 4
ATM 5
ATM 6
ATM 7
ATM 8
ATM 9
ATM 10
29.9
30.4
29.0
28.9
29.3
29.6
29.7
29.8
37.9
52.4
0.95 ( 0.01)
0.89
0.87
0.89
0.90
0.90
0.91
0.91
0.92
0.96
0.93
<
0.93 ( 0.01)
<
0.95 ( 0.01)
<
0.95 ( 0.01)
<
0.95 ( 0.01)
<
0.96 ( 0.01)
<
0.96 ( 0.01)
<
0.96 ( 0.01)
<
0.98 ( 0.01)
<
0.96 ( 0.01)
re ectance response to drying. The mud sample has two distinct spectral stages
related to the loss of moisture. For example, the ATM band 9 value initially remains
reasonably constant during the loss of the Ž rst 75% of moisture, however following
this re ectance increases rapidly from 10 DN to 34 DN (Ž gure 5(c)). Sample 3
(
<
63 mm) has at least three distinct spectral stages during drying. In ATM band 9
for example (Ž gure 6(c)), re ectance increases from 2.5 to 15 DN (moisture 100% to
95%), before decreasing from 15 to 9 DN (moisture 95% to 50%) and lastly, in
stage 3, re ectance increases rapidly from 9 to 32.5 DN (moisture 50% to 0%). At
absolute dryness the mud sample has a much lower maximum band 9 value than

Spectral characteristics of intertidal sediments
3035
the sand and silt samples, however at saturation the diŒerence is negligible. The
maximum spectral contrast between the sand sample and the mud sample, in ATM
band 9, exists when the sediments are at 25% saturation.
3.2.1. L aboratory spectra PCA results
It is evident from the scatterplot of the laboratory PCA results (Ž gure 7) that
the spectral eŒect of sediment grain size becomes more obvious with loss of moisture.
The spectra of the sand, silt and mud samples when they were saturated are all
relatively closely located in feature space. As the sediment moisture declines, the
position of their spectra gradually moves towards the left of the scatterplot, i.e. from
positive to negative along PC1 which contains 96.8% of the variance. The spectral
contrast between the mud sample and the coarser samples also increases as the
sediments get drier. However, the spectral contrast between the silt and the sand
samples remains minimal with loss of moisture.
Discussion
It has been determined that a strong positive relationship exists between percent-
age sand and ATM re ectance along transects that have been exposed to rapid
drying conditions for a long period of time (i.e. 3 h). The relationship between
>
re ectance and percentage sand is strongest in ATM bandwidths 9 and 10. Since
this association is not seen in saturated sediments, it can be concluded that it is the
loss of interstitial moisture that enables the grain size–re ectance relationship to
exist. This is supported by the laboratory results which show the spectral contrast
between muds and sands increases as the moisture content decreases to 25%, which
is the minimum moisture level expected in the majority of intertidal sediments.
Similarly, the PCA results of the in situ and laboratory spectra clearly show that
dry sands and muds are spectrally distinct in feature space, and the level of spectral
contrast is dependent on moisture content. These results also show that there is a
Figure 7. Scatterplot of PC1 against PC2, with the location of the main endmembers labelled.

3036
M. P. Rainey et al.
minimal increase in the spectral diŒerence between sands and silts with drying, in
contrast to that between the coarse sands and the mud sample.
The 1996 results demonstrate that generally weak and very variable relationships
exist between ATM bands 9/10 and sediment grain size when the intertidal area has
not been exposed to drying conditions for a long period of time. The 1997 results
show that under these conditions the grain size–re ectance relationships are not
stable and more often insigniŽ cant. This is evident in the laboratory experiments,
which demonstrate that a negligible spectral contrast exists between sand, silt and
mud before drying. This is due to the predominance of the spectral signal of water
resulting in the re ectance characteristics of the sediment being unidentiŽ able. In
the scatterplot of the in situ spectra PCA results, it is also evident that the location
of saturated sediment spectra in feature space is not determined by grain size.
Similarly, the spectra of all the saturated laboratory samples are shown to be closely
located in feature space.
The laboratory experiments show that the spectral contrast between sand and
mud is primarily a product of mud’s low spectral response during the loss of the
initial 75% of moisture. As an explanation of this spectral behaviour, it is suggested
that the
(table 1), acts as a control on the re ectance response. In the laboratory experiment
63 mm component, which is largest in the representative mud sample
<
the re ectance of the 63 mm sample lowered signiŽ cantly as the moisture levels
<
changed from 90% to 40%. It is proposed that this spectral eŒect limits the response
of mud in the early stages of drying. The silt and sand samples have insigniŽ cant
quantities of the 63 mm component (table 1), and hence their spectral response
to drying is not noticeably constrained.
<
Due to the signiŽ cant in uence of moisture, the relationship between re ectance
and sediment grain size is often determined by intertidal topography and drying
conditions. Although it is not immediately evident in these results, it is accepted that
sloped intertidal regions present in Warton Bank initially lose their moisture faster
than  at regions like Lytham St Anne’s. Consequently the spectral in uence of
interstitial moisture is relatively less on the sloped areas, which may allow the grain
size of the sediments to be more accurately mapped. In addition, where topographic
lows exist in the intertidal zones pools of surface water develop and, as demonstrated
in the 1996 and 1997 results, these dramatically attenuate sediment re ectance,
masking any relationship with grain size. Due to a lack of topographical data for
this study area it was impossible to accurately identify or compensate areas where
topography may have indirectly in uenced the ability to diŒerentiate sediment grain
size. The in uence of seasonal drying conditions on the sediment–re ectance relation-
ship is also apparent in this study. The best relationships observed were at Warton
Bank in August 1995 after the sediments had been exposed for a long period of time
to strong summer drying conditions. Due to winter drying conditions insu ciently
reducing the sediment interstitial moisture content at both Lytham St Anne’s and
Warton Bank in February 1997, no signiŽ cant correlation existed along the sampled
transects before or after exposure. It can be concluded from this that topography
and seasonality are important considerations in any project that attempts to spec-
trally diŒerentiate estuarine sediments through remote sensing.
Finally, it has been demonstrated that the spectral contrast between Ž ne and
coarse intertidal sediments is primarily a product of their very diŒerent spectral
responses to moisture loss. This is enhanced in the Ž eld, as coarser sediments tend
to lose moisture more rapidly than mud, which has greater retention capabilities.

Spectral characteristics of intertidal sediments
3037
Since the majority of intertidal mud retains a minimum of 25% of its original
moisture content during exposure, it is suggested that the re ectance of these
sediments also remains relatively constant whilst that of sand and silt increases
linearly. Consequently, it is suggested that the unique spectral stability of mud may
make it the easiest intertidal cover type to map from remotely sensed imagery.
This is important, as this sediment type contains the highest percentage clay and
consequently the highest levels of particle reactive pollutants.
5. Summary and implications for future remote sensing campaigns
It is possible to spectrally distinguish between diŒerent sediments in an intertidal
environment that has been exposed to drying conditions. The relationship between
ATM re ectance and percentage sand or clay is shown to be strongest for bands 9
and 10. The in situ and laboratory experiments have shown that only weak negative
or positive relationships, if any, exist between grain size and re ectance when the
sediments retain high moisture contents. The level of spectral contrast between
intertidal sediments is clearly related to the length of time since inundation and the
seasonal drying eŒects. However, quantitative in situ measurement of interstitial
moisture was found to be di cult and unreliable, consequently it was impossible to
deŽ ne a direct relationship between exposure and moisture contents.
The in uence of interstitial and surface moisture on the spectral contrast between
sediments of diŒerent size fractions has important implications on any remote sensing
campaign attempting to map intertidal sediment domains. Consequently, the accu-
racy of any estuarine intertidal sediment classiŽ cation based on ATM data is inher-
ently linked to the timing of image collection. It is evident from the results of this
study that the image collected after the longest period of intertidal exposure, i.e.
when the interstitial and surface moisture is minimized, is most likely to provide the
best basis for diŒerentiating sediment grain sizes. This is also inherently reliant on
the imagery being collected during the summer when the drying conditions, i.e.
evaporation rates are likely to be more extreme. This conclusion was applied to the
timing of the subsequent collection of ATM imagery of the Ribble Estuary in 1997.
In addition, this conclusion highlights the need for temporally  exible remotely
sensed data and consequently the unsuitability of most satellite datatypes for
monitoring estuarine surface sediment distributions.
Acknowledgments
This work was funded the by the University of Stirling and Westlakes ScientiŽ c
Consulting Ltd, in conjunction with BNFL. The authors would like to acknowledge
NERC EPFS for equipment support throughout this study. The authors would also
like to acknowledge Bill Jamieson and David Aitchison who drafted a number of
Ž gures used in this paper.
References
Asrar, G., 1989, T heory and Applications of Optical Remote Sensing (Chichester: Wiley).
Bartholdy, J., and Folving, S., 1986, Sediment classiŽ cation and surface type mapping in
the Danish Wadden Sea by remote sensing. Netherlands Journal of Sea Research,
20, 337–345.
Black, K. S., and Paterson, D. M., 1997, Measurement of the erosion potential of cohesive
marine sediments: a review of current in situ technology. Journal of Marine
Environmental Engineering, 26, 43–83.

3038
Spectral characteristics of intertidal sediments
Brown, J., 1997, T he Behaviour of Radionuclides in the Ribble Estuary, NW England.
Unpublished PhD Thesis, University of Reading, UK.
Bryant, R. G., Tyler, A. N., Gilvear, D. J., McDonald, P., Teasdale, I., and Ferrier, G.,
1996, A preliminary investigation into the VNIR spectral characteristics of inter-tidal
estuarine sediments. International Journal of Remote Sensing, 17, 405–412.
Burton, D. J., West, J. R., Horsington, R. W., and Randle, K., 1995, Modelling transport
processes in the Ribble Estuary. Environmental International, 21, 131–141.
Doerffer, R., and Murphy, D., 1989, Factor analysis and classiŽ cation of remotely sensed
data for monitoring tidal  ats. Helgolander Meeresunters, 43, 275–293.
Folving, S., 1984, The Danish Wadden Sea: thematic mapping by means of remote sensing.
Folia Geographica Danica, 15, 1–56.
Goss-Custard, J. D., and Yates, M. G., 1992, Towards predicting the eŒect of salt-marsh
reclamation on feeding bird numbers on the Wash. Journal of Applied Ecology, 29,
330–340.
Populus, J., Moreau, F., Coquelet, D., and Xavier, J. P., 1995, An assessment of environ-
mental sensitivity to marine pollutions: solutions with remote sensing and
Geographical Information Systems (GIS). International Journal of Remote Sensing,
16, 3–15.
Smith, M. O., Johnson, P. E., and Adams, J. B., 1985, Quantitative determination of mineral
types and abundances from re ectance spectra using Principle Component Analysis.
15th Proceedings of the L unar Planet Science Conference (Washington: American
Geophysical Union), Volume 2, pp. 797–804.
Tyler, A. N., Bryant, R., Gilvear, D., McDonald, P., Teasdale, I., Brown, J., and
Ferrier, G., 1995, The spectral characterisation of radionuclide bearing sediments in
the Ribble Estuary. Proceedings of the 21st Annual Conference of the Remote Sensing
Society (Nottingham: Remote Sensing Society), pp. 433–440.
Tyler, A. N., Sanderson, D. C. W., Scott, E. M., and Allyson, J. D., 1996, Accounting for
spatial variability and Ž elds of view in environmental gamma ray spectrometry. Journal
of Environmental Radioactivity, 33, 213–235.
Yates, M. G., Jones, A. R., McGrorty, S., and Goss-Custard, J. D., 1993, The use of satellite
imagery to determine the distribution of intertidal surface sediments of The Wash,
England. Estuarine, Coastal and Shelf Science, 36, 333–334.