Remote Sensing of Environment 115 (2011) 1034–1042

Contents lists available at ScienceDirect

Remote Sensing of Environment

j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / r s e

Spectral response of benthic diatoms with different sediment backgrounds

Laurent Barillé a,⁎, Jean-Luc Mouget c, Vona Méléder a, Philippe Rosa a, Bruno Jesus b,d
a
Université de Nantes, Mer Molécules Santé EA 2160, Faculté des Sciences et des Techniques, B.P. 92208, 44322 Nantes cedex 3, France
b
Centro de Oceanografia da Faculdade de Ciências da Universidade de Lisboa, Campo Grande, 1749-016, Lisboa, Portugal
c
Université du Maine, Mer Molécules Santé EA 2160, Faculté des Sciences et Techniques, Ave O. Messiaen, 72085 Le Mans cedex 9, France
d
Centro de Biodiversidade, Genómica Integrativa e Funcional (BioFIG), Faculdade de Ciências, Universidade de Lisboa, Campo Grande, 1749-016 Lisboa, Portugal

a r t i c l e i n f o a b s t r a c t

Article history: Remote sensing of terrestrial vegetation uses a wide range of vegetation indices (VIs) to monitor plant
Received 3 September 2010 characteristics, but these indices can be very sensitive to canopy background reflectance. This study
Received in revised form 8 December 2010 investigated background influences on VIs applied to intertidal microphytobenthos, using a synthetic spectral
Accepted 11 December 2010
library constituted by a spectral combination of three contrasting types of sediment (sand, fine sand, and
Available online 15 Janaury 2011
mud) and reflectance spectra of benthic diatom monospecific cultures obtained in controlled conditions. The
Keywords:
spectral database exhibited, for the same biomass range (3–182 mg chlorophyll a m− 2), marked differences
Background reflectance in albedo and spectral contrast linked to sediment variability in water content, grain size, and organic matter
Intertidal flats content. Several VIs were evaluated, from ratios using visible and near infrared wavelengths, to hyperspectral
Microphytobenthos indices (derivative analysis, continuum removal). Among the ratios, the Normalized Difference Vegetation
NDVI Index (NDVI) appeared less sensitive to background effects than VIs with soil corrections such as the
Remote sensing Perpendicular Vegetation Index (PVI), the Soil-Adjusted Vegetation Index (SAVI), the Modified second Soil-
Vegetation index Adjusted Vegetation Index (MSAVI2) or the Transformed Soil-Adjusted Vegetation Index (TSAVI). The lower
efficacy of soil-corrected VIs may be explained by the structural differences and optical behavior of soil vs.
canopies compared to sediment vs. microphytobenthos biofilms. The background effects were minimized
using Modified Gaussian Model indices at 632 nm and 675 nm, and the second derivative at 632 nm, while
poor results were obtained with the red-edge inflection point (REIP) and the second derivative at 675 nm. The
least sensitive index was the Phytobenthos Index which is very similar to the NDVI, but uses a red wavelength
at 632 nm instead of 675 nm, to account for the absorption by chlorophyll c. The modified NDVI705, where the
705 nm wavelength replaces the red band, showed moderate background sensitivity. Moreover, the NDVI705
and the Phytobenthos Index have the additional relevant property of being less sensitive to the index
saturation response with increasing biomass. Unfortunately, these VIs cannot be applied to broad-band
multispectral satellite images, and require sensors with a hyperspectral resolution. Nevertheless, this study
showed that the background influence was not a limitation to applying the ubiquitous NDVI to map intertidal
microphytobenthos using multispectral satellite images.
© 2010 Elsevier Inc. All rights reserved.

1. Introduction microphytobenthos (Carrère et al., 2004; Jesus et al., 2006; Kromkamp

Vegetation indices (VIs) are used as proxies in a wide range of
scientific applications to monitor the biophysical and biochemical
properties of vegetation. They are calculated from reflectance spectra
(the ratio of upwelling radiance to downwelling irradiance) at various
scales from ground-based to satellite measurements. Numerous
studies in the literature deal with VIs developed for terrestrial
vegetation to characterize crop or canopy biophysical properties (e.g.
Broge & Mortensen, 2002; Gitelson et al., 2002; Sims & Gamon, 2002;
Vescovo & Gianelle, 2008; Zhao et al., 2007) but only a limited variety
of indices have been applied to marine vegetation, and particularly to
⁎ Corresponding author. Tel.: + 33 2 51 12 56 55; fax: + 33 2 51 12 56 68.
E-mail address: laurent.barille@univ-nantes.fr (L. Barillé).
et al., 2006; Méléder et al., 2003a,b; Murphy et al., 2005a,b; Paterson
et al., 1998). Composed of algae and photosynthetic bacteria that
colonize benthic substrata at low tide, microphytobenthos (MPB) is
crucial to the functioning of estuarine and coastal ecosystems
(MacIntyre et al., 1996; Underwood & Kromkamp, 1999). Information
about MPB biomass spatial distribution is essential in many ecological
studies but, due to the sometimes dangerous and highly inaccessible
nature of mudflats, it is difficult to collect at the tidal flat scale. Thus,
there is a growing interest in developing remote sensing methods that
can replace, or complement, the field data (Carrère et al., 2004; Jesus
et al., 2006; Kromkamp et al., 2006; Méléder et al., 2003a,b; Murphy
et al., 2005a,b; Paterson et al., 1998).
Only a few studies have tested a broad range of indices, from
multispectral to hyperspectral, to retrieve information from microphy-
tobenthic reflectance spectra (Combe et al., 2005; Murphy et al., 2005a),

0034-4257/$ – see front matter © 2010 Elsevier Inc. All rights reserved.
doi:10.1016/j.rse.2010.12.008
 L. Barillé et al. / Remote Sensing of Environment 115 (2011) 1034–1042 1035

while many have analyzed the spectral properties of surface sediment
with ratios combining discrete spectral bands. The most widely used ratio
throughout all research fields remains the Normalized Difference
Vegetation Index (NDVI) which combines information contained in two
spectral bands, the red and near infrared (NIR). It is commonly used as an
index to assess microphytobenthos biomass (Coelho et al., 2009; Jesus
et al., 2006; Kromkamp et al., 2006; Serôdio et al., 2009). In spite of its
ubiquitous use, the NDVI has some limitations, the most important of
which is its sensitivity to soil background influence (Escadafal & Huete,
1991; Huete, 1988; Huete et al., 2002) and its non-linear relationship with
biomass for terrestrial vegetation and benthic microphytes (Gitelson et al.,
2002; Gitelson, 2004; Méléder et al., 2003a). This explains why many
other indices have been developed for terrestrial applications, either based
on ratio calculations integrating a soil correction (Huete, 1988; Jiang et al.,
2008; Qi et al., 1994) or taking into account the whole reflectance
spectrum such as VIs based on derivative analysis or spectral continuum
removal, two techniques that reduce interference from background
reflectance (Demetriades-Shah et al., 1990).
Microphytobenthos reflectance spectra are the result of a complex
interplay between biotic (photo-autotrophic assemblages with an array
of photosynthetic and accessory pigments and the constitution of a
biofilm created by microalgal mucilaginous secretions) and abiotic
factors associated with the sediment itself (e.g. sediment type and water
content). Although a microphytobenthic biofilm is structurally different
from an angiosperm plant canopy, light transmitted through it is
similarly reflected on the substratum and back through the biofilm again
(Combe et al., 2005). However, scattering properties in the NIR are
radically distinct (Kazemipour et al., 2011). For a similar fractional cover,
it is therefore likely that sediment influence on microphytobenthos
reflectance is much greater than soil influence on terrestrial vegetation.
Sediment contribution depends mainly on grain size, water content, and
organic matter (Méléder et al., 2010; Rainey et al., 2000; Verpoorter,
2009). Grain size has an effect on light scattering, which increases with
decreasing particle size, but Murphy et al. (2005a) indicated that this
was not the most consistent factor explaining the differences between
pigments as possible. Sediments were then analyzed for organic
matter content by loss on ignition, for their water content, for particle
size distribution using GRADISTAT (Blott & Pye, 2001), and for
pigment concentration by High-Performance Liquid Chromatography
(HPLC). For each variable and sediment type, eight replicates were
measured, except in the pigment analysis for which three samples
were processed. A detailed account of sediment processing can be
found in Méléder et al. (2005). Reflectance spectra of each sediment
sample were obtained with a field portable spectroradiometer ASD
FieldSpec 3FR, measuring the radiance (mW cm− 2 nm− 2 sr− 1)
between 350 and 2500 nm with a spectral sampling interval of 1.4 nm
between 350 and 1050 nm and a spectral resolution from 3 to 10 nm
(Fig. 1A). Data were collected using the ASD contact probe with its
halogen light source. The reflectance probe was kept at a constant
distance from the sample and from the reference panel by using the
probe distance ring, designed for this effect. Spectral responses were
studied in the visible-near infrared (VNIR) range between 400 and
900 nm, since photosynthetic and accessory pigments have their
diagnostic absorption features in this area and absorption by water is
strong above 900 nm. Surface reflectance was determined by first
measuring the light reflected by a ~ 99% reflective Spectralon®
reference panel, and then light reflected by the sediment surface.
The instrument was calibrated to subtract the dark-current for each
radiance measurement. At least ten spectra were averaged to calculate
a mean reflectance of each sample.
2.2. Synthetic spectral library
A synthetic spectral library was constructed mixing the reflectance
from the three types of sediment with the reflectance spectra from
0.5
A sand
0.4
Reflectance

fine sand
sediment spectral characteristics. Rainey et al. (2003) suggested that for 0.3
intertidal mudflats, variations in surface and interstitial moisture were
the main sources of heterogeneity in sediment reflectance spectra. 0.2 mud
Water significantly decreases sediment average brightness or albedo,
while sediment dryness modifies the spectral contrast, increasing the
0.1
slope steepness between visible and NIR reflectance, characteristic of
sandy sediments. For terrestrial vegetation, albedo and slope variations
are known to affect VIs independently of the amount of vegetation 0
400 500 600 700 800 900
(Elvidge & Lyon, 1985; Elvidge & Chen, 1995). The influence of these
features however has not yet been fully elucidated for microphyto- 0.5
B
benthos, and the indices tested and proposed by Murphy et al. (2005a) sand
on a mudflat relied on an intertidal environment dominated by 0.4
filamentous macroalgae, very different from benthic diatom biofilms.
Reflectance

The main objective of this study was to investigate the effects of fine sand
0.3
sediment background reflectance on benthic diatom reflectance spectra,
using a synthetic spectral library constituted by a spectral combination of
three types of sediment and reflectance spectra of benthic diatom 0.2
mud
monospecific cultures obtained in controlled conditions. A range of
multispectral and hyperspectral VIs were tested for their sensitivity to
0.1
background influence and to saturation with increasing biomass.

2. Data and methods 0

400 500 600 700 800 900
Wavelength (nm)
2.1. Sediment collection and analysis
Fig. 1. (A) Sediment reflectance spectra used in this study to construct the synthetic spectral
Three types of sediment from sand to mud were collected in the library. Sand, fine sand, and mud were named following GRADISTAT granulometric analysis
intertidal area of Bourgneuf Bay (47°04.02′ N, 2°01.20′ W) located (Blott & Pye, 2001). (B) Synthetic spectral library of benthic diatom biomass variations for
south of the Loire river, France. Bare sediment (visual assessment) three types of substrata. Monospecific culture spectra (Méléder et al., 2003a), with chlorophyll
a biomass ranging from 3 to 182 mg Chl a m− 2, were combined with the sediment reflectance
without any visible microphytobenthic biofilms at the sediment (see Data and methods). For the same amount of biomass, the reflectance dataset obtained
surface was chosen, and sterilized by autoclavation at 120 °C for with the different substrata exhibits albedo and slope variations due to the background
20 min to remove as many traces of photosynthetic and accessory influence.
1036 L. Barillé et al. / Remote Sensing of Environment 115 (2011) 1034–1042
benthic diatom monospecific cultures, i.e. Navicula ramossissima and
Entomoneis paludosa, measured with a GER3700 spectroradiometer
(Méléder et al., 2003a). These authors simulated microalgal biofilms by
slow filtration of the cultures on fiberglass filters. Pigment composition
and biomass were estimated by HPLC, and a wide range of Chl a
concentration was obtained (3–182 mg Chl a m− 2). The library was
created following the method proposed by Combe et al. (2005): each
diatom spectrum was first divided by the fiberglass reflectance spectrum
to remove its background influence (characterized by a negative slope
over the VNIR wavelength range; Méléder et al. (2003a)), then it was
multiplied by the three sediment reflectance spectra respectively, after a
spectral degradation of the ASD data to match GER spectral resolution
using a square band pass function. Three series of reflectance spectra with
the same range of biomass were obtained, without any rescaling, but with
distinct overall shapes (continuum) due to the contribution of the three
types of sediment (Fig. 1). The artificial biofilms (cells deposited on a
fiberglass filter) used in this study have been recently analyzed by
scanning electronic microscopy showing the presence of extracellular
polymeric substances that are typically present on microphytobenthic
biofilms (Kazemipour et al., 2011). Retrieving the optical properties of
diatom biofilms, these authors have shown that these artificial biofilms
were optically very similar to natural biofilms when the latter are found
forming thick layers on top of the sediment.
2.3. Spectral indices
Vegetation indices (VIs) tested in this study were selected for their
property of reducing or removing soil/sediment background influence
and because they are representative of the three main types of VI:
ratio indices, derivative and continuum removal (Table 1). Three
ratios belong to the Normalized Difference Vegetation Index (NDVI)
family, including the widely used NDVI calculated from a red band
associated with the chlorophyll a absorption at 675 nm and a near
infrared (NIR) band positioned on the NIR plateau. The Phytobenthos
Index (PI), specifically designed for diatoms, uses the reflectance at
635 nm associated with chlorophyll c, a diagnostic absorption feature
for this class and for dinoflagellates (Forster & Jesus, 2006; Méléder
et al., 2003a). This index is particularly useful on estuarine intertidal
flats, which are strongly dominated by benthic diatoms (MacIntyre
et al., 1996; Paterson et al., 1998). The third NDVI-like index tested was
proposed by Gitelson and Merzlyak (1994). This uses the reflectance at
Table 1
Ratio and hyperspectral (derivative analysis and continuum removal) vegetation indices tested in this study. L is a soil adjustment factor (Huete, 1988); a and b are soil line
parameters (see Data and methods).
Ratio vegetation index
Normalized Difference Vegetation Index (NDVI)
Phytobenthos Index (PI)
NDVI705
R562/R647
Indices with soil correction
Perpendicular Vegetation Index (PVI)
Soil-Adjusted Vegetation Index (SAVI)
Transformed Soil-Adjusted Vegetation Index (TSAVI)
Modified second Soil-Adjusted Vegetation Index (MSAVI2)
Enhanced Vegetation Index 2 (EVI2)
Derivative analysis
First derivative measure (δ)
Wavelength of the Red Edge Inflection Point (REIP)
Second derivative measure (δδ)
Derivative at 632, 675 nm
Continuum removal
Modified Gaussian Model (absorption band strength at 632, 675 nm)
705 nm, on the right edge of the main chlorophyll a 675 nm absorption
band, designed to be less sensitive to this absorption band saturation at
high biomass. A ratio including two bands at 562 and 647 nm proposed
by Murphy et al. (2005a) for microphytobenthos was also tested. These
four VIs are supposed to be the most sensitive to background effects,
since they do not include any soil/sediment corrections. They were thus
compared with five ratio indices with a soil correction factor: the
Perpendicular Vegetation Index (PVI) (Richardson & Wiegand, 1977),
the Soil-Adjusted Vegetation Index (SAVI) proposed by Huete (1988),
the Transformed Soil-Adjusted Vegetation Index (TSAVI) (Baret et al.,
1989), the Modified second Soil-Adjusted Vegetation Index (MSAVI2)
proposed by Qi et al. (1994), and the Enhanced Vegetation Index 2
(EVI2) developed by Jiang et al. (2008). Other ratio soil-corrected VIs are
available, but these five were selected because they can be applied to
broad-band satellite without a blue band such as SPOT HRVIR, and three
of them, SAVI, MSAVI2, EVI2, do not require the calculation of the soil
line parameters in the red-NIR space (Broge & Mortensen, 2002). For the
PVI and TSAVI, the soil line was estimated from a September 2009 SPOT
5 image of Bourgneuf Bay intertidal flat, NIR = 2.5235 red− 0.2142,
r2 = 0.99. This line represents the linear relationship observed between
soil reflectance without vegetation at two wavelengths, in the visible
and NIR. It is estimated from a two dimensional diagram, where all the
pixels of an image are displayed according to these two spectral bands.
The SAVI was calculated with a range of soil adjustment factor L: 0.25,
0.5 and 1 (Huete, 1988). Among ratio indices (with and without soil
corrections), only the NDVI and the soil-corrected VIs are applicable to
multispectral satellite sensors which possess red and NIR broad-bands,
since the others are based on discrete VNIR bands, corresponding to
individual wavelengths without broad-band satellite equivalence.
The derivation of reflectance spectra has been applied in reflectance
spectroscopy to remove background signals and resolve overlapping
spectral features (Demetriades-Shah et al., 1990). This hyperspectral
processing uses spectrally continuous data, and provides different types of
information according to the level of derivation (Tsai & Philpot, 1998). VIs
based on first- and second-order derivatives have already been applied to
microphytobenthic reflectance spectra to quantify biomass variations and
infer taxonomical composition (Jesus et al., 2006; Murphy et al., 2005a,
2008). The Red Edge Inflection Point (REIP) is widely used in terrestrial
vegetation studies to quantify biomass variations (Broge & Mortensen,
2002) and corresponds to the wavelength position of the main peak in a
first-order derivative spectrum (Horler et al., 1983). Derivative-based
Calculation Reference
(R750 − R675)/(R750 + R675) Rouse et al. (1973)
(R750 − R635)/(R750 + R635) Forster and Jesus (2006)
(R750 − R705)/(R750 + R705) Gitelson and Merzlyak (1994)
R562/R647 Murphy et al. (2005a,b)
(R750 − a R675 − b)/(a2 + 1)0.5 Richardson and Wiegand (1977)
(1 + L)[(R750 − R675)/(R750 + R675 + L)] Huete (1988)
a(R750 − a R675 − b)/(R750 + a R675 − ab + 0.08(a2 + 1)) Baret et al. (1989)
0.5[2(R750 + 1) − (2(R750 + 1)2 − 8(R750 − R675))0.5] Qi et al. (1994)
2.5 [(R750 − R675)/(R750 + 2.4R675 + 1)] Jiang et al. (2008)
λREIP Horler et al. (1983)
δδR632, δδR675 Jesus et al. (2006)
MGM632, MGM675 Combe et al. (2005)
 L. Barillé et al. / Remote Sensing of Environment 115 (2011) 1034–1042
indices are less sensitive to background effects (Demetriades-Shah et al.,
1990), thus it was hypothesized that the REIP and second derivative VI
would perform better than the NDVI with MPB biofilms. The two second-
order derivative values used correspond to peaks in the second derivative
function associated with the absorption features of the main diatom
pigments, chlorophyll a (675 nm) and chlorophyll c (632 nm). A
smoothing spline function with 60 nodes was adjusted to the reflectance
spectra before calculating the derivatives.
The continuum represents the overall shape of a reflectance
spectrum and its removal enables the absorption features of interest
(pigment absorption bands for vegetation studies, mineral absorption
bands for geological applications) to be isolated from other contribu-
tions (Clark & Roush, 1984; Combe et al., 2005). In intertidal sediments,
the continuum is mainly influenced by water content and grain size
(Carrère et al., 2004), but also by unknown processes that contribute to
the overall shape of reflectance spectra. Continuum removal-based VIs
should therefore be less sensitive to sediment background influences
compared to the NDVI. The continuum was removed with the Modified
Gaussian Model approach, which provides the depth or strength of
continuum-removed absorption bands in logarithmic reflectance
(Combe et al., 2005). From the seven bands used for the spectral
deconvolution of benthic diatom biofilms by Barillé et al. (2007), two
band centers were tested in this study, positioned at 632 and 675 nm, to
account for the absorptions by chlorophyll c and a, respectively.
2.4. Statistical analysis
After verification of normality, sediment characteristics were
compared using a one-way analysis of variance (ANOVA). To compare
sediment background influence between VIs, for a biomass range
estimated by increasing concentration of chlorophyll a (Chl a), a non-
linear model VI= a + b [1− exp(− c. Chla)], was adjusted for each VI,
using all the data from the three sediments. The same model was used to
fit all the relationships and the coefficient of determination was used to
assess the variability generated by the different substrata. The VI
sensitivity to the signal saturation process caused by increasing biomass
was described after standardization of regression models to 1.
3. Results
3.1. Sediment characteristics and synthetic spectral library
The three types of sediment had different particle-size distributions,
discriminating a sand composed of medium (55%) and fine (45%)
particles, a sand mainly composed of fine particles (88.8% comprised
between 125 and 250 μm and 1.2% of mud) hereafter called fine sand,
and a mud (97% of particles below 63 μm). They also showed significant
differences in water content (WC) and organic fraction (OF) (ANOVA,
P b 0.01). The mud had the highest water content and organic fraction,
58.4% (S.D. 0.9) and 9.2% (S.D. 0.5) respectively, compared to the fine
sand (WC = 30.3%, S.D. = 4.8; OF = 2.7%, S.D. = 1.2) and the sand
(WC = 7.9%, S.D. = 0.7; OF = 0.5%, S.D. = 0.0). The mud exhibited a
spectral shape similar to that of the fine sand, but with a lower albedo
related to its higher water and organic matter contents, while the sand
spectral response was characterized by both a steeper slope in the VNIR
and the highest albedo (Fig. 1A). In spite of the initial sediment cleaning
treatment, a dip in the reflectance at 675 nm, indicative of chlorophyll a
absorption, was still present for the mud and the fine sand. The cleaned
sand showed the lowest chlorophyll a concentration (0.18 mg m− 2, S.
D.= 0.01) compared to the cleaned mud (0.30 mg m− 2, S.D. = 0.02)
and the cleaned fine sand (0.47 mg m− 2, S.D. = 0.01). Nevertheless, it
was considered that the chlorophyll contribution from the sediment
itself was negligible in comparison with the range of biomass analyzed
(3–182 mg Chl a m− 2). The synthetic spectral library combining the
spectral responses of sediments and diatom monospecific cultures
exhibited, for the same biomass range, a sharp distinction in the NIR
1037
linked to the three types of sediment (Fig. 1B). In the visible, the main
difference was the steeper slope of diatom reflectance spectra with a
sand substratum.
Among the ratio indices, the NDVI-like indices appeared less
sensitive to background effects for equivalent vegetation amounts
than VIs with soil corrections (Fig. 2). For the latter, the influence of
the three different substrata was clearly detectable, with darker
sediment producing lower VI responses, for the SAVI (Fig. 2D),
MSAVI2 (Fig. 2E), and EVI2 (Fig. 2F). The same background influence
was observed with the three L factors tested for the SAVI. The TSAVI
and PVI (not shown) had a similar distribution of data points. For
NDVI-like indices, the sediment brightness influence was the reverse,
with darker sediment producing higher VI responses than brighter
sediment (e.g. NDVI, Fig. 2A). For the hyperspectral indices, a lower
background influence was obtained with indices calculated after
continuum removal, compared to those retrieved after derivative
analysis (Fig. 3). The second derivative at 675 nm was characterized
by a high variability among each type of sediment (Fig. 3B). The
sediment brightness influence was similar for all hyperspectral
indices, with darker sediment producing higher VI responses.
The results from the non-linear regression analysis of chlorophyll a
on vegetation indices indicated that the Phytobenthos Index was the
least sensitive to background influences (r2 = 0.97), followed by the
NDVI (r2 = 0.95), then the two indices obtained after continuum
removal, MGM 675 and MGM 632 (r 2 = 0.94), and the NDVI 705
(r2 = 0.93) (Table 2). VIs with soil corrections had r2 values ranging
from 0.86 to 0.82, while the REIP was even more sensitive to
background influence. The second derivative at 675 nm was not a
good biomass estimator in this study, while the ratio R562/R647 could
not be used at all to assess benthic diatom biomass variations. The VI
sensitivity to the saturation process caused by increasing biomass is
described in Fig. 4, after standardization of regression models to 1.
Most of the models showed a steep increase at low biomass levels and
then an asymptotic leveling off beyond 100 mg Chl a m− 2, except the
PI and NDVI705, which were less sensitive to saturation.
The synthetic library was used to obtain quantitative estimations
of benthic diatom biomass by inversion of the relationships described
in Figs. 2 and 3. Two VIs, PI and NDVI, were chosen for their low
sensitivity to background to illustrate the relationships that can be
obtained from reflectance data to infer diatom biomass. The data from
fine sand and mud, which gave the most similar results (Fig. 2A, B),
were pooled and fitted with the following non-linear models: Chl
a = 1009.9 PI3.134 (r2 = 0.97, p b 0.001) and Chl a = 0.048 exp(9.106
NDVI)
(r2 = 0.95, p b 0.001), (Fig. 5). The relationship obtained with the
NDVI showed higher sensitivity to biomass saturation, as expected,
but remains useful to map microphytobenthos biomass with satellite
or air-borne images, if it does not exceed NDVI values of 0.6.
4. Discussion
4.1. NDVI vs. soil-adjusted indices
The Normalized Difference Vegetation Index has been extensively
used in terrestrial remote sensing. This index, as well as spectral
vegetation indices combining red and near infrared wavelengths, has
been found to be well correlated with plant biomass, pigment
concentrations, and canopy cover (Sellers, 1985). Naturally, it has
thus been applied in microphytobenthos studies seeking to quantify
benthic microalgal biomass by optical approaches (Jesus et al., 2006;
Kromkamp et al., 2006; Méléder et al., 2003a,b; Murphy et al., 2005a,
b; Serôdio et al., 2009). However, the NDVI can be significantly
influenced by the spectral properties of soils and background surfaces
(Huete, 1988; Huete & Jackson, 1987; Huete et al., 1985), with
variations, for the same vegetation type, canopy structure and amount
of vegetation, from 0 to 0.2 units on a scale of 0 to 1. It was thus
hypothesized that these variations, mainly caused by soil brightness
1038 L. Barillé et al. / Remote Sensing of Environment 115 (2011) 1034–1042

A B
0.6 0.6

0.4 0.4
NDVI

PI
0.2 0.2

0 0
0 50 100 150 200 0 50 100 150 200

0.4

C D
0.2 0.3
NDVI705

SAVI
0.2

0.1

0.1

0 0
0 50 100 150 200 0 50 100 150 200

0.5

0.4 E F
0.4

0.3
0.3
MSAVI2

EVI2

0.2
0.2

0.1 0.1

0 0
0 50 100 150 200 0 50 100 150 200
Chl a (mg m-2) Chl a (mg m-2)

Fig. 2. Relationships between ratio indices and benthic diatom chlorophyll a concentrations (Chl a). For each vegetation index (VI), the same biomass range was tested with three
different substrata: sand (white circles), fine sand (gray circles) and mud (black circles), and one non-linear model was fitted for the three substrata. (A) NDVI, (B) PI, (C) NDVI705,
(D) SAVI, (E) MSAVI2, and (F) EVI2. VI acronyms and regression analysis are provided in Tables 1 and 2, respectively.

differences, may also occur in intertidal sediments. Intertidal
environments exhibit considerable sediment heterogeneity linked to
grain size, mineralogy, interstitial moisture, organic matter and
sediment brightness, which can all be very variable (Bryant et al.,
1996; Carrère et al., 2004; Rainey et al., 2000, 2003). Surprisingly, in
the present study, the NDVI was slightly less sensitive to sediment
background variations than VIs with soil corrections or based on
hyperspectral processing, known to reduce or remove the influence of
background. The least sensitive index was the Phytobenthos Index,
which is very similar to the NDVI but uses a red wavelength at 632 nm
instead of 675 nm, to account for the absorption by chlorophyll c
present in diatom biofilms in strong concentrations. The lower
efficacy of soil-corrected VIs may be explained by the structural
differences and optical behavior of soil and canopies compared to
sediment and benthic diatoms. The main distinction probably arises in
the NIR, as in the visible the absorption processes associated with
photosynthetic and accessory pigments are fundamentally identical
between terrestrial angiosperms and marine unicellular autotrophs.
Terrestrial vegetation is characterized by both transmission and
scattering processes in the NIR region, with significant fluxes
scattered by foliar structure (Jacquemoud & Baret, 1990), whereas
diatoms are only characterized by transmission at these wavelengths
(Kazemipour et al., 2011). Indices such as the SAVI or MSAVI are built
to account for the different optical depth penetrations of red and NIR
 L. Barillé et al. / Remote Sensing of Environment 115 (2011) 1034–1042 1039

0.8 5
A B
4
0.6

δδ 675 (x 103)
δδ 632 (x 103)

0.4

0.2
1

0 0
0 50 100 150 200 0 50 100 150 200

1.2 1.2
C D
MGM 632 (x -1)

MGM 675 (x-1)

0.8 0.8

0.4 0.4

0 0
0 50 100 150 200 0 50 100 150 200
Chl a (mg m-2) Chl a (mg m-2)

Fig. 3. Relationships between hyperspectral vegetation indices and benthic diatom chlorophyll a concentrations (Chl a). For each vegetation index (VI), the same biomass range was
tested with three different substrata: sand (white circles), fine sand (gray circles) and mud (black circles), and one non-linear model was fitted for the three substrata. Second
derivative at 632 nm (A) and 675 nm (B). MGM band depth at 632 nm (C) and 675 nm (D) after continuum removal by a Modified Gaussian Model.

radiations through the canopy to the soil surface (Huete, 1988).
Sediment–diatom interactions probably resemble those of low
vegetation cover, for which red and NIR penetration through the
canopy are nearly equal, with most of the downwelling flux reaching
the sediment coming from solar irradiance, since the diatoms do not
themselves contribute through NIR scattering.
4.2. Background influence on VIs
The brightness of the background material is known to have a
significant effect on VIs (Elvidge & Chen, 1995). However, for low
canopy covers, normalized difference VIs are less sensitive to
brightness effects (Huete & Jackson, 1987), which may explain why,

Table 2 1.0
Non-linear regression analysis of chlorophyll a on vegetation indices. For each
vegetation index (VI), the data obtained with the three different substrata have been
0.8
pooled. The same model, VI = a + b(1 − exp(− c. Chla)), has been used to fit all the
Vegetation index

relationships represented in Figs. 2 and 3. VIs are detailed in Table 1.

Vegetation index a b c r2 p-value

0.6

PI 0.131 0.448 0.016 0.97 b 0.001

NDVI 0.148 0.474 0.028 0.95 b 0.001 0.4
MGM675 0.489 0.652 0.032 0.94 b 0.001
MGM632 0.362 0.599 0.022 0.94 b 0.001
0.2
NDVI705 0.050 0.176 0.012 0.93 b 0.001
δδ632 0.029 0.533 0.025 0.91 b 0.001
TSAVI − 0.020 0.450 0.034 0.86 b 0.001 0.0
PVI − 0.013 0.179 0.037 0.86 b 0.001 0 50 100 150 200
EVI 0.093 0.253 0.033 0.86 b 0.001
Chl a (mg m-2)
MSAVI2 0.090 0.238 0.033 0.83 b 0.001
SAVI 0.090 0.207 0.036 0.82 b 0.001
Fig. 4. Regression models standardized to 1, between vegetation indices and benthic
REIP 689.86 10.61 0.01 0.80 b 0.001
diatom chlorophyll a concentrations (Chl a). Data from Table 2; dashed line = NDVI705,
δδ675 0.348 2.744 0.091 0.66 b 0.001
dotted line = Phytobenthos Index, solid lines = other vegetation indices (not all lines
R562/R647 0.549 0.302 4.446 0.01 0.803
shown for clarity).
1040 L. Barillé et al. / Remote Sensing of Environment 115 (2011) 1034–1042
200
150
Chl a (mg m-2)
100
50
0 the initial file parameter, prior to the Gaussian deconvolution (Barillé
0 0.1 0.2 0.3 0.4 0.5 0.6
Vegetation index
Fig. 5. Inverse relationships between the NDVI (gray squares), Phytobenthos Index (PI,
white squares) and chlorophyll a concentrations (Chl a) calculated from pooled data for
two types of substrata, fine sand and mud.
in this study, the PI, NDVI and NDVI705 showed few variations
associated with the three types of sediment tested. These variations
were consistent with the effects previously described for terrestrial
vegetation, with darker sediment (mud) producing higher indices
than lighter sediment (sand) (Elvidge & Lyon, 1985; Huete & Jackson,
1987; Huete et al., 1985). This phenomenon is explained by a
disproportionate decrease in the NDVI denominator due to sediment
darkening (Rocha & Shaver, 2009). An opposite trend was observed
for soil-corrected VIs, with darker sediment producing lower indices.
In this study, the SAVI, MSAVI2, TSAVI, EVI2 and PVI all exhibited the
highest values with the sandy substratum, under the influence of
sediment brightness. The spectral contrast corresponds to the
magnitude of the slope between the visible and the NIR wavelength
range, and is a secondary source of variation among substrata, related
to soil/sediment composition associated with a different color
(Escadafal & Huete, 1991; Huete et al., 1985). In this study, the sand
is characterized by a marked VNIR slope, while the fine sand and the
mud have flat spectral profiles (Fig. 1). The slope remains clearly
visible in the synthetic library for the sand substratum and this
spectral contrast, inherent in the sediment itself, is known to increase
VIs artificially, with or without soil corrections (Escafadal & Huete,
1991; Elvidge & Chen, 1995; Murphy et al., 2005). Non-linear mixing
occurs with benthic diatom biofilms (Combe et al., 2005), and this
effect acts like the spectral contrast, raising VI values over bright
backgrounds (Elvidge & Chen, 1995; Huete et al., 1985). This
phenomenon may also explain the results obtained with the soil-
corrected VIs, arranged largely in order of brightness. These sources of
variation had little effect on normalized difference VIs tested in this
experiment, particularly the PI and NDVI, which appeared as the best
VIs in terms of background reflectance sensitivity. This agrees with the
finding of Jesus et al. (2006) who showed that the NDVI was a good
predictor of benthic diatom biomass in the Tagus estuary, but
contrasts with that of Murphy et al. (2005a), who described it as a
poor predictor of benthic chlorophyll a in a mudflat characterized by a
range of sediment, from fine mud to coarse sand. However, the
intertidal environment studied by Murphy et al. (2005a) was
dominated by filamentous macroalgae, which can be intimately
mixed within the sediment, and whose optical properties must be
quite different from those of diatom biofilms. Differences in structure
and pigment composition may explain why the VI they proposed,
R562/R647, could not describe the diatom biomass variations for any of
the three substrata. The REIP was sensitive to background effects,
which confirms that it is not an appropriate VI to estimate benthic
chlorophyll a (Murphy et al., 2005a). Second derivatives produced
contrasting results, with the derivative at 632 nm showing a limited
background influence, while that at 675 nm displayed a high
variability and could not reproduce biomass variations. This may
result from a double-peak, which can appear in the second-order
derivate spectra in the region of the chlorophyll a absorption band,
due to chlorophyll fluorescence at 683 nm (Serôdio et al., 2009). In
fact, these authors indicated that fluorescence-independent VIs
should be used instead of any VIs based on the chlorophyll a red
absorption band, and that the Phytobenthos Index should be
preferred to the NDVI. Nevertheless, the hyperspectral VIs obtained
after continuum removal, MGM632 and MGM675, gave satisfactory
results, removing most of the background influences. The double-
peak feature was not generated with the index MGM675, and the
variability was lower to that observed with the second derivative at
675 nm, because a single band center position at 675 nm was set up in
0.7
et al., 2007).
4.3. Saturation of signal at high biomass values
A non-linear relationship between VIs and biomass is a major
concern in terrestrial vegetation remote sensing (Huete et al., 2002). The
NDVI applied to terrestrial angiosperm reflectance spectra is known to
saturate asymptotically with increasing vegetation (Gitelson et al.,
1996); this has also been observed in benthic diatom biofilms (Méléder
et al., 2003a). Extended linearity has therefore been an important
attribute in the selection of indices for the remote qualification of
vegetation (Gitelson, 2004). Most of the VIs tested in this study were
sensitive to saturation, in particular soil-corrected VIs and hyperspectral
indices. Only two VIs displayed a wider dynamic range, namely the PI
(Forster & Jesus, 2006) and the NDVI705 (Gitelson & Merzlyak, 1994). Of
these two, the latter was the least sensitive to saturation, but was
slightly more influenced by background reflectance. The Phytobenthos
Index (PI) can therefore be described as the most robust index
considering both background and saturation sensitivity. It is based on
the chlorophyll c absorption band at 635 nm, which is less sensitive to
saturation than the 675 nm chlorophyll a absorption (Méléder et al.,
2003b), and has the useful property of being a typical diatom pigment
(Millie et al., 2002), which are frequently the dominant group on
intertidal estuarine sediments (MacIntyre et al., 1996; Paterson et al.,
1998). Unfortunately, this VI cannot be applied to satellite images
acquired with broad-band multispectral sensors. It requires a hyper-
spectral sensor capturing the chlorophyll c absorption band, such as
ROSIS (Verpoorter, 2009). However, the non-linearity of VIs over the
high biomass conditions tested in this study may not be a critical issue
for benthic diatom mapping. In fact, the biomass range obtained with
the monospecific cultures is very likely greater than the biomass present
in the intertidal flat photic zone (Barillé et al., 2007). In mudflats, photic
depth estimations well below 1 mm have been proposed: for example,
750 μm (Jesus et al., 2006) or 200 μm (Serôdio et al., 1997), suggesting
that the upwelling radiance detected by the sensor would be related to
the most superficial sediment layers. Combe et al. (2005) thus estimated
the maximum biomass of epipelic diatom biofilms at 35 mg Chlorophyll
(Chl) a m − 2 using DAIS hyperspectral data, far below the
100 mg Chl a m− 2 threshold found in this study. Similarly, Méléder
et al. (2003b) described diatom assemblages within an NDVI range of 0
to 0.3, using SPOT multispectral images, corresponding to the linear
portion of the NDVI–biomass relationship (Fig. 2A). Van der Wal et al.
(2008) identified microphytobenthos in a comparable NDVI range, from
0.1 to 0.4, using hyperspectral sensors, still below the threshold
saturation. This suggests that, although globally sensitive to saturation,
the NDVI may be used to map the microphytobenthic biomass spatial
distribution present in the photic zone, which can be useful for
multispectral images.
4.4. Multi- vs. hyperspectral data for microphytobenthos mapping
This study showed that the NDVI had a limited sensitivity to
background influences compared to soil-corrected VIs. Thus, the
relationship obtained by the inversion of the spectral library (Fig. 5)
can be used, after a spectral re-sampling, to infer benthic diatom
 L. Barillé et al. / Remote Sensing of Environment 115 (2011) 1034–1042
biomass, using multispectral sensors with red and NIR broad-bands.
The main limitation of multispectral data is, however, related to the
case of mixed pixels, and the estimation of microphytobenthos
fractional cover (Combe et al., 2005). Microphytobenthos spatial
distribution is known to be highly patchy (e.g. Jesus et al., 2005;
Murphy et al., 2008), and this variability occurs at a subpixel scale
(b1 m2). With low spatial resolution satellite images, pixel upwelling
radiance thus often results in spectral mixtures. Linear mixing can be
estimated for multispectral data by algorithms retrieving vegetation
fractional cover using ratio indices (Gitelson et al., 2002). Neverthe-
less, Combe et al. (2005) highlighted the fact that microphytobenthos
biofilms are also characterized by non-linear processes associated
with intimate mixtures between biofilms and substrata, which cannot
be described by classical linear unmixing models. Jiang et al. (2006)
proposed a scaled NDVI to infer vegetation fraction, overcoming non-
linearity and scaling effects over heterogeneous surfaces. However,
this VI corrects the non-linearity between the NDVI and the
vegetation fractional cover but relies on the assumption of spectral
mixture linearity of soils and vegetation components. Nevertheless,
since this VI can be applied to broad-band sensors, it would be
interesting to estimate the error associated with the violation of this
assumption when retrieving microphytobenthos fractional cover or
biomass. The mixed pixel constraints are obviously strong for
multispectral data, but they will certainly be reduced for future
multispectral sensors with a spatial resolution of less than 1 m, such
as Pléiades-HR.
There is no doubt that hyperspectral data have an intrinsic greater
potential to retrieve a more complete set of variables to describe benthic
vegetation. This study suggests, however, that the relevance and
accuracy of the NDVI, applied at the ecosystem level using multispectral
satellite data, should be thoroughly examined and compared with
synchronous hyperspectral data. The Phytobenthos Index showed the
most interesting properties in mapping benthic diatom biomass: low
sensitivity to background, biomass saturation, and algal fluorescence.
Nevertheless, the widely used NDVI, sometimes referred to as a
continuity index for its ability to describe time-series, still yields very
accurate results in the study of intertidal microphytobenthos biofilms
(van der Wal et al., 2010).
Acknowledgements
This work was financially supported by the Ministère Français des
Affaires Etrangères through a Partenariat Hubert Curien (PHC)-PESSOA
research program. B. Jesus was funded by an FCT postdoctoral grant (POCI/
BPD/20993/2004). The ASD spectroradiometer was provided by the
Laboratoire de Planétologie et Géodynamique of the University of Nantes.
The authors wish to thank the CNES (Centre National d'Etudes Spatiales)
for the ISIS program regarding the use of SPOT satellite products.
References
Baret, F., Guyot, G., & Major, D. (1989). TSAVI: A vegetation index which minimizes soil
brightness effects on LAI or APAR estimation. 12th Canadian Symposium on Remote
Sensing and IGARSS'90, Vancouver, Canada 10–14 July.
Barillé, L., Méléder, V., Combe, J. -P., Launeau, P., Rincé, Y., Carrère, V., et al. (2007).
Comparative analysis of field and laboratory spectral reflectances of benthic
diatoms with a modified Gaussian model approach. Journal of Experimental Marine
Biology and Ecology, 343(2), 197−209.
Blott, S. J., & Pye, K. (2001). Gradistat: A grain size distribution and statistics package for
the analysis of unconsolidated sediments. Earth Surface Processes and Landforms, 26,
1237−1248.
Broge, N. H., & Mortensen, J. V. (2002). Deriving green crop area index and canopy
density of winter wheat from spectral reflectance data. Remote Sensing of
Environment, 81, 45−57.
Bryant, R., Tyler, A., Gilvear, D., McDonald, P., Teasdale, I., Brown, J., et al. (1996). A
preliminary investigation into the spectral characteristics of inter-tidal estuarine
sediments. International Journal of Remote Sensing, 17, 405−412.
Carrère, V., Spilmont, N., & Davoult, D. (2004). Comparison of simple techniques for estimating
chlorophyll a concentration in the intertidal zone using high spectral-resolution field-
spectrometer data. Marine Ecology Progress Series, 274, 31−40.
1041
Clark, R. N., & Roush, T. L. (1984). Reflectance spectroscopy: Quantitative analysis
techniques for remote sensing applications. Journal of Geophysical Research, 89(B7),
6329−6340.
Coelho, H., Viera, S., & Serôdio, J. (2009). Effects of desiccation on the photosynthetic
activity of intertidal microphytobenthos biofilms as studied by optical methods.
Journal of Experimental Marine Biology and Ecology, 381, 98−4104.
Combe, J. -P., Launeau, P., Carrère, V., Despan, D., Méléder, V., Barillé, L., et al. (2005).
Mapping microphytobenthos biomass by non-linear inversion of visible-infrared
hyperspectral images. Remote Sensing of Environment, 98, 371−387.
Demetriades-Shah, T. H., Steven, M. D., & Clark, J. (1990). High resolution derivatives
spectra in remote sensing. Remote Sensing of Environment, 33, 55−64.
Elvidge, C. D., & Chen, Z. (1995). Comparison of broad-band and narrow-band red and
near-infrared vegetation indices. Remote Sensing of Environment, 54, 38−48.
Elvidge, C. D., & Lyon, R. J. P. (1985). Influence of rock-soil spectral variation on the
assessment of green biomass. Remote Sensing of Environment, 17, 265−279.
Escadafal, R., & Huete, A. (1991). Etude des propriétés spectrales des sols arides
appliquée à l'amélioration des indices de végétation obtenus par télédétection.
Comptes Rendus de l'Académie des Sciences Paris, 312(II), 1385−1391.
Forster, R. M., & Jesus, B. (2006). Field spectroscopy of estuarine habitats. International
Journal of Remote Sensing, 27(17), 3657−3669.
Gitelson, A. A. (2004). Wide dynamic range vegetation index for remote quantification
of biophysical characteristic of vegetation. Journal of Plant Physiology, 161, 165−173.
Gitelson, A. A., Kaufman, Y. J., & Merzlyak, M. N. (1996). Use of a green channel in
remote sensing global vegetation from EOS-MODIS. Remote Sensing of Environment,
58, 289−298.
Gitelson, A. A., Kaufman, Y. J., Stark, R., & Rundquist, D. (2002). Novel algorithms for
remote estimation of vegetation fraction. Remote Sensing of Environment, 80,
76−87.
Gitelson, A., & Merzlyak, M. N. (1994). Spectral reflectance changes associated with
autumn senescence of Aesculus hippocastanum L. and Acer platanoides L. leaves.
Spectral features and relation to chlorophyll estimation. Journal of Plant Physiology,
143, 286−292.
Horler, D. N., Dockray, M., & Barber, J. (1983). The red edge of plant leaf reflectance.
International Journal of Remote Sensing, 4, 273−288.
Huete, A. R. (1988). A soil-adjusted vegetation index (SAVI). Remote Sensing of 624
Environment, 25, 295−309.
Huete, A., Didan, K., Miura, T., Rodriguez, E. P., Gao, X., & Ferreira, L. G. (2002). Overview
of the radiometric and biophysical performance of the MODIS vegetation indices.
Remote Sensing of Environment, 83, 195−213.
Huete, A. R., & Jackson, R. (1987). Suitability of spectral indices for evaluating vegetation
characteristics on arid rangelands. Remote Sensing of Environment, 83, 195−213.
Huete, A. R., Jackson, R., & Post, D. (1985). Spectral response of a plant canopy with
different soil backgrounds. Remote Sensing of Environment, 17, 37−53.
Jacquemoud, S., & Baret, F. (1990). PROSPECT: A model leaf optical properties spectra.
Remote Sensing of Environment, 34, 75−91.
Jesus, B., Brotas, V., Marani, M., & Paterson, D. M. (2005). Spatial dynamic of microphyto-
benthos determined by PAM fluorescence. Estuarine, Coastal and Shelf Science, 65, 30−42.
Jesus, B., Mendes, C. R., Brotas, V., & Paterson, D. M. (2006). Effect of sediment type on
microphytobenthos vertical distribution: Modelling the productive biomass and
improving ground truth measurements. Journal of Experimental Marine Biology and
Ecology, 332, 60−74.
Jiang, Z., Huete, A. R., Chen, J., Chen, Y., Li, J., Yan, G., et al. (2006). Analysis of NDVI and
scale difference vegetation index retrievals of vegetation fraction. Remote Sensing of
Environment, 101, 366−378.
Jiang, Z., Huete, A. R., Didan, K., & Miura, T. (2008). Development of a two-band
enhanced vegetation index without a blue band. Remote Sensing of Environment,
112, 3833−3845.
Kazemipour, F., Méléder, V., & Launeau, P. (2011). Optical properties of microphytobenthic
biofilms: Biomass retrieval implication. Journal of Quantitative Spectroscopy and Radiative
Transfer, 112, 131−142.
Kromkamp, J. C., Morris, E. P., Forster, R. M., Honeywill, C., Hagerthey, S., & Paterson, D. M.
(2006). Relationship of intertidal surface sediment chlorophyll concentration to
hyperspectral reflectance and chlorophyll fluorescence. Estuaries and Coasts, 29(2),
183−196.
MacIntyre, H., Geider, R., & Miller, D. (1996). Microphytobenthos: The ecological role of
the “secret garden” of unvegetated, shallow water marine habitats. I. Distribution,
abundance and primary production. Estuaries, 19, 186−201.
Méléder, V., Barillé, L., Launeau, P., Carrère, V., & Rincé, Y. (2003a). Spectrometric
constraint in analysis of benthic diatom biomass using monospecific cultures.
Remote Sensing of Environment, 88(4), 386−400.
Méléder, V., Barillé, L., Rincé, Y., Morançais, M., Rosa, P., & Gaudin, P. (2005). Spatio-temporal
changes in microphytobenthos structure analysed by pigment composition in a
macrotidal flat (Bourgneuf Bay, France). Marine Ecology Progress Series, 29, 83−99.
Méléder, V., Launeau, P., Barillé, L., & Rincé, Y. (2003b). Cartographie des peuplements
du microphytobenthos par télédétection spatiale visible-infrarouge dans un
écosystème conchylicole. Comptes Rendus de l'Académie des Sciences- Biologie,
326, 377−389.
Méléder, V., Launeau, P., Barillé, L., Combe, J.-P., Carrère, V., Jesus, B., & Verpoorter, C. (2010).
Hyperspectral imaging for mapping microphytobenthos in coastal areas. In: Mohamed
Maanan and Marc Robin (Eds.), Geomatic solutions for coastal environment, Nova Science
Publishers, Inc., pp. 71–139.
Millie, D. F., Schofield, O. M. E., Kirkpatrick, G. J., Johnsen, G., & Evens, T. J. (2002). Using
absorbance and fluorescence spectra to discriminate microalgae. European Journal
of Phycology, 37, 313−322.
Murphy, R. J., Tolhurst, M. G., Chapman, M. G., & Underwood, A. J. (2005a). Estimation of
surface chlorophyll-a on an emersed mudflat using field spectrometry: Accuracy of
1042 L. Barillé et al. / Remote Sensing of Environment 115 (2011) 1034–1042
ratios and derivative-based approaches. International Journal of Remote Sensing, 26
(9), 1835−1859.
Murphy, R. J., Tolhurst, M. G., Chapman, M. G., & Underwood, A. J. (2008). Spatial
variation of chlorophyll on estuarine mudflats determined by field-based remote
sensing. Marine Ecology Progress Series, 365(1), 45−55.
Murphy, R. J., Underwood, A. J., Pinkerton, M. H., & Range, P. (2005b). Field
spectrometry: New methods to investigate epilithic micro-algae on rocky shores.
Journal of Experimental Marine Biology and Ecology, 325(1), 111−124.
Paterson, D. M., Wiltshire, K. H., Miles, A., Blackburn, J., Davidson, I., Yates, M. G., et al.
(1998). Microbiological mediation of spectral reflectance from intertidal cohesive
sediments. Limnology and Oceanography, 43, 1207−1221.
Qi, J., Chehbouni, A., Huete, A. R., Kerr, Y. H., & Sorooshian, S. (1994). A modified soil
adjusted vegetation index. Remote Sensing of Environment, 48, 119−126.
Rainey, M. P., Tyler, A. N., Bryant, R. G., & Gilvear, D. J. (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, 3025−3038.
Rainey, M. P., Tyler, A. N., Gilvear, D. J., Bryant, R. G., & McDonald, P. (2003). Mapping
intertidal estuarine sediment grain size distributions through airborne remote
sensing. Remote Sensing of Environment, 86, 480−490.
Richardson, A. J., & Wiegand, C. L. (1977). Distinguishing vegetation from soil
background information. Photogrammetric Engineering and Remote Sensing, 43,
1541−1552.
Rocha, A. V., & Shaver, G. R. (2009). Advantages of two band EVI calculated from solar
and photosynthetically active radiation fluxes. Agricultural and Forest Meteorology,
149, 1560−1563.
Rouse, J. W., Haas, R. H., Schell, J. A., & Deering, D. W. (1973). Monitoring vegetation
systems in the Great Plains with ERTS. Proceedings of the third ERTS Symposium,
NASA, SP-351, vol.1 (pp. 309−317). Washington, DC: NASA.
Sellers, P. C. (1985). Canopy reflectance, photosynthesis and transpiration. International
Journal of Remote Sensing, 6, 1335−1372.
Serôdio, J., Cartaxana, P., Coelho, H., & Viera, S. (2009). Effects of chlorophyll
fluorescence on the estimation of microphytobenthos biomass using spectral
reflectances indices. Journal of Experimental Marine Biology and Ecology, 113,
1760−1768.
Serôdio, J., da Silva, J. M., & Catarino, F. (1997). Nondestructive tracing of migratory
rhythms of intertidal benthic microalgae in vivo using chlorophyll a fluorescence.
Journal of Phycology, 33, 542−553.
Sims, D. A., & Gamon, J. A. (2002). Relationships between leaf pigment content and spectral
reflectance across a wide range of species, leaf structures and developmental stages.
Remote Sensing of Environment, 81, 337−354.
Tsai, F., & Philpot, W. (1998). Derivative analysis of hyperspectral data. Remote Sensing
of Environment, 66, 41−51.
Underwood, G., & Kromkamp, J. (1999). Primary production by phytoplancton and
microphytobenthos in estuaries. Advances in Ecological Research, 29, 93−153.
Van der Wal, D., Herman, P. M., Foster, R. M., Ysebaert, T., Rossi, F., Knaeps, E., et al.
(2008). Distribution and dynamics of intertidal macrobenthos predicted from
remote sensing: Response to microphytobenthos and environment. Marine Ecology
Progress Series, 367, 57−72.
Van der Wal, D., Wielemaker-van den Dool, A., & Herman, P. M. (2010). Spatial
synchrony in intertidal benthic algal biomass in temperate coastal and estuarine
ecosystems. Ecosystems, 13, 338−351.
Verpoorter, C. (2009). Télédétection hyperspectrale et cartographie des faciès
sédimentaires en zone intertidale. Application à la Baie de Bourgneuf. Ph.D. thesis,
Nantes University, 423 pp. http://tel.archives-ouvertes.fr/tel-00391184/fr
Vescovo, L., & Gianelle, D. (2008). Using MIR bands in vegetation indices for the
estimation of grassland biophysical parameters from satellite remote sensing in the
Alps region of Trentino (Italy). Advances in Space Research, 41, 1764−1772.
Zhao, D., Huang, L., Li, J., & Qi, J. (2007). A comparative analysis of broadband and
narrowband derived vegetation indices in predicting LAI and CDD of cotton canopy.
ISPRS Journal of Photogrammetry and Remote Sensing, 62, 25−33.