ELSEVIER
Palaeogeography, Palaeoclimatology, Palaeoecology 149 (1999) 213–226
Gradient analysis of diatom assemblages in a Puget Sound salt marsh:
can such assemblages be used for quantitative paleoecological
reconstructions?
B.L. Sherrod *
Department of Geological Sciences, Box 351310, University of Washington, Seattle WA 98195, USA
Received 22 January 1997; revised version received 10 July 1997; accepted 8 June 1998
Abstract
Taphonomy is important to coastal paleoecologists because processes acting on diatom thanocoenoses tend to work
towards obscuring original ecological relationships between diatom assemblages and the environment. The purpose of
this paper is to briefly describe diatom taphonomy and present a method for quantitative reconstruction of environmental
parameters from salt marsh diatom assemblages. The main hypothesis for this study is that major environmental and
taphonomic processes (e.g., tides) act in predictable ways to distribute living and dead diatoms along environmental
gradients. To test this hypothesis, a modern transect was established across a large salt marsh in southwestern Puget Sound
for the purpose of determining modern species=environment gradients and calibrating species assemblages to environmental
variables. Canonical correspondence analysis (CCA) relates modern species assemblages to environmental gradients, and
weighted averaging calibration is used to develop transfer functions for predicting environmental information. CCA
showed that the effect of salinity and elevation on the species distributions is significant, indicating that environmental
processes control the distribution of sedimentary diatoms across the salt marsh surface in predictable ways. Salinity was
strongly correlated with CCA Axis 1 and elevation with Axis 2. The calibration results indicate that, although mixing
of allochthonous and autochthonous diatoms does occur, salt marsh diatom assemblages reflect major environmental
gradients in Puget Sound salt marshes and can be effectively used for quantitative reconstructions of former environmental
conditions.  1999 Elsevier Science B.V. All rights reserved.
Keywords: taphonomy; salt marsh; diatoms; ordination; calibration; canonical correspondence analysis
1. Introduction
The relationships between salt marsh organisms
and the environmental gradients that control their
distribution offer a critical link to determining past
environmental events in coastal areas. By targeting
‘keystone’ environmental factors operating in salt
Ł E-mail:
[email protected]
marshes, we can begin to understand the fundamental controls on the marsh ecosystem, and how species
and assemblages respond to changes in the ecosystem (Holling, 1992). Diatom thanatocoenoses (all of
the diatoms, both autochthonous and allochthonous
remains, present at a particular place in a sediment)
in salt marshes are unique in that they are the product
of ecological processes that affect the distribution of
living diatoms along environmental gradients, and
0031-0182/99/$ – see front matter  1999 Elsevier Science B.V. All rights reserved.
PII: S 0 0 3 1 - 0 1 8 2 ( 9 8 ) 0 0 2 0 2 - 8
214
B.L. Sherrod / Palaeogeography, Palaeoclimatology, Palaeoecology 149 (1999) 213–226
taphonomic processes that alter the distributions of
dead diatoms after deposition. The purpose of this
paper is to briefly describe taphonomic processes
that alter diatom assemblages and present a method
for reconstructing quantitatively environmental parameters from sedimentary diatom thanatocoenoses.
Taphonomy is important to coastal paleoecologists because processes acting on estuarine diatoms
tend to work towards obscuring original ecological
relationships between the diatom assemblages and
the environment (Sherrod et al., 1989; Denys, 1994).
In some cases, taphonomic processes can so drastically alter the species composition of a diatom
assemblage that the original ecological signals reflected by the in situ assemblage are either obscured
or obliterated. Species can be selectively removed
from an assemblage by breakage or dissolution, or
added by tidal current transport. It is the role of the
paleoecologist to decipher which factor is important
for each assemblage and determine how this factor
has affected the species composition. However, in the
absence of detailed studies on taphonomic and ecological gradients, this task often leads to speculation.
Reliance on the presumed autecology of individual species without giving due regard to processes
that shape diatom thanatocoenoses can lead to erroneous interpretations (Denys, 1994). If taphonomy is
a major problem in salt marsh diatom assemblages,
then the fidelity of environmental information that
diatom assemblages convey to the paleoecologist
is compromised. The method I am proposing relies on the relationship between species composition
and environmental gradients to infer quantitatively
missing environmental information. Canonical correspondence analysis (CCA) is an ideal solution for
this problem since the output can be used both for
ordination (determining how species are distributed
along environmental gradients) and calibration (Ter
Braak, 1995a,b).
One approach to quantitative paleoenvironmental
reconstruction is calibration. Calibration is the process of expressing an environmental variable as a
function of community composition data (Ter Braak,
1995b). This approach is more robust for salt marsh
reconstructions than indicator or autecological approaches because it relies on calibrating the occurrence of thanatocoenoses with modern environmental conditions, rather than individual species. The
resulting transfer functions are an empirical basis for
reconstructing former salt marsh environments.
The interaction between environmental factors affecting the distributions of living diatoms and the
taphonomic factors controlling the deposition and
preservation of dead diatoms has not been studied; however, it is important because it ultimately
controls species composition and hence influences
paleoecological reconstructions. Quantitative reconstructions require that: (1) the responses of modern
sedimentary diatom assemblages to environmental
and taphonomic variables are modeled using a training set of diatom assemblages; and (2) the model
results are used to infer missing (or past) environmental variables from samples where direct measurement of environmental variables is not possible
(e.g., fossil samples) (Birks et al., 1990; Ter Braak,
1995b). The present paper examines this problem
using a set of modern diatom samples to determine
how species composition in the surface sediments is
related to salinity and elevation.
Studies of fossil salt marshes require accurate reconstructions of key environmental parameters such
as elevation and salinity. Paleoseismic studies in
coastal areas often rely on diatoms to provide assessments of past salinity and elevation changes
for inferring earthquake histories (Hemphill-Haley,
1995a,b, 1995c). Therefore, determining how sedimentary diatom assemblages are distributed relative
to primary environmental gradients in salt marshes
is a critical problem for accurate reconstructions.
Salinity is often cited as a control on living diatoms
assemblage, while elevation integrates many environmental factors in salt marshes but primarily is a
reflection of tidal inundation (Nelson and Kashima,
1993).
2. Statement of hypothesis and study area
The principal concept behind this study is that
‘keystone’ environmental processes control the distribution of diatom thanatocoenoses across a salt
marsh surface. The main hypothesis is that major environmental and taphonomic processes (e.g.,
tides) act in predictable ways to distribute living
and dead diatoms along environmental gradients. A
corollary to the first hypothesis is that ‘keystone’
B.L. Sherrod / Palaeogeography, Palaeoclimatology, Palaeoecology 149 (1999) 213–226
215
diatom thanatocoenoses occurring in the salt marsh
deposits reflect major ecological processes and can
be used for accurate quantitative reconstructions.
2.1. Hypothesis testing
I propose to use direct gradient analysis to establish that diatoms occur in Puget Sound salt marshes
along environmental continua. To test the main hypothesis, I statistically related observed sedimentary
diatom assemblages in surficial sediments to measurements of selected environmental factors using
Canonical Correspondence Analysis (CCA). CCA
is special case of multivariate regression that utilizes weighted averaging to arrange both samples
and species along environmental gradients (Palmer,
1993). This statistical treatment defines ‘keystone’
assemblages if they exist. If the first hypothesis is
accepted, transfer functions based on a ‘training’ set
of diatom assemblages can then be established. I
tested the second hypothesis using linear regression
techniques on predicted and observed values for the
training set of samples. Large coefficients of determination (r 2 ) and residual plots are a measure of the
usefulness of a transfer function.
2.2. Study area
A modern transect was established at Lynch Cove,
a large salt marsh in southwestern Puget Sound,
Washington, for the purpose of determining modern species=environment gradients and calibrating
species assemblages to environmental variables. The
marsh is located at the northern end of Hood Canal,
a fjord-like body of marine water along the east
side of the Olympic Mountains (Fig. 1). The transect
extended from a barren mudflat eastward to the margin of a willow–alder swamp (Fig. 2). The waters
of Hood Canal have an approximate average salinity of 27–30‰, and a mean tidal range of 3.6 m.
Approximately 1000 years ago, the site was uplifted
approximately 2–4 m during an earthquake, resulting
in major biostratigraphic changes (Bucknam et al.,
1992). The modern transect was established at this
site as part of ongoing geological studies.
Fig. 1. (A) Map of United States showing location of Puget
Sound region in the State of Washington. (B) Enlarged map of
Puget Sound region showing location of Lynch Cove and various
geographic features.
3. Diatom taphonomy and environmental
gradients
Previous research has documented several taphonomic processes that act on diatoms in salt marshes.
The main taphonomic processes are transportation,
216
B.L. Sherrod / Palaeogeography, Palaeoclimatology, Palaeoecology 149 (1999) 213–226
Fig. 2. Idealized cross-section of Lynch Cove transect, showing major vegetation zones and marsh subenvironments. Salinity (A) and
elevation (B) profiles for the stations on the transect are shown in graphs below cross-section.
breakage, and dissolution. The processes can act
alone or in concert to alter the original in situ (autochthonous) diatom assemblages of a salt marsh.
This study primarily addresses transportation from
adjacent environments as a major contributor to
taphonomy of salt marsh diatom assemblages.
B.L. Sherrod / Palaeogeography, Palaeoclimatology, Palaeoecology 149 (1999) 213–226
3.1. Transportation
Transportation of diatom frustules from adjacent
environments is of major importance in diatombased paleoecological studies. In some areas, the
influx of allochthonous (transported) diatoms far outnumbers the production of autochthonous diatoms
(Sherrod et al., 1989; Vos and De Wolf, 1993). Epipsammic diatoms (those that live attached to sand
and mud grains) such as Achnanthes delicatula and
Opephora parva are a common autochthonous component in tidal flat assemblages but can be scoured
from the substrate by currents and transported in
great numbers into tidal channels and onto marsh
surfaces (De Jonge, 1985; Vos and De Wolf, 1993).
Vos and De Wolf (1988, 1993) proposed a classification system that divided coastal diatoms into
broad ecological groups related to specific environments. Their method presents criteria for assessing
which diatoms are allochthonous and for the interpretation of sedimentary environments. While this
method provides an excellent basis for determining
allochthonous components in a diatom assemblages,
it does not convey quantitative information concerning environmental factors that is required for detailed
paleoecological reconstruction.
217
reworking or the remnants of an assemblage heavily
modified by dissolution.
3.3. Environmental gradients and diatom
distributions
A study in the Yaquina Estuary of Oregon found
that diatom distributions in the estuary were controlled by chemical and physical gradients such
as salinity, desiccation, and seasonal light changes
(McIntire, 1978). This study suggested that species
distribution along the environmental gradients is a
continuum rather than a series of discrete, isolated
assemblages except in areas where the gradients
break down (McIntire, 1978; Sullivan, 1982). Sullivan (1982) also found relationships between environmental gradients and species composition, with
elevation and vascular plant canopy height as the
most important environmental variables. Laird and
Edgar (1992) found statistically significant non-uniform density distributions of diatoms in the surface
sediments of a New England salt marsh but did not attempt to correlate these distributions to causal factors.
4. Methods
3.2. Fragmentation and dissolution
4.1. Field and laboratory methods
Selective removal of diatom taxa from the sedimentary record is also a problem in paleoecological
studies. In many cases, delicate taxa are broken
during transportation and diagenesis (Nelson and
Kashima, 1993) resulting in an impoverished assemblage. Other processes such as chemical leaching,
compaction, and sample preparation also lead to
fragmentation and dissolution of diatom valves (Vos
and De Wolf, 1993).
Fragmentation and dissolution play a role in enhancing the relative abundance of certain robust
types. Paralia sulcata is a heavily silicified, chainforming diatom that is often dominant in coastal
deposits (Hemphill-Haley, 1995b). The dominance
of P. sulcata is due in part to a greater resistance
to breakage and dissolution (Sherrod et al., 1989).
As pointed out by Hemphill-Haley (1995b), samples
containing only a few valves of P. sulcata should be
treated with caution, as these may be the result of
A 260 m transect was established on the marsh
surface, with sampling stations marked off every 10
m. The transect extended out over the marsh surface
from the beginning point at 0 m in the upland marsh
to the last station on the barren mudflat at 260 m
(Fig. 2). Visual estimation of vascular plant species
cover employed a 1 m2 quadrat. Surface sediment
samples and environmental measurements were collected in June 1995. Scraping of surface sediments
with a steel spoon or spatula ensured collection of
only the upper 1 cm or less of sediment (depth
measured with a ruler), with the samples placed
immediately into a sample bag after collection and
sealed. Preservation of each sample using formalin
occurred within 48 hours after collection (when the
samples reached the laboratory). Sediment accumulation rates are estimated at about 0.5 to 1 mm per
year using radiocarbon dates from a stratigraphic
section at the site, with each sample representing
218
B.L. Sherrod / Palaeogeography, Palaeoclimatology, Palaeoecology 149 (1999) 213–226
about 10–20 years of sediment accumulation (Bucknam et al., 1992). The surface sediment samples thus
contained a mixture of both living epipelic diatoms
and the remains of diatoms that died at the sampling
site or were transported from adjacent environments.
Samples of surface sediment from each station
were wrapped in filter paper and squeezed in a syringe to collect interstitial pore water. Pore water
salinity measurements used an optical salinity refractometer, calibrated prior to each measurement
with distilled water (accurate to within š1‰). Random replicate samples were also analyzed in the
field for salinity to test the accuracy of the measurements (differences within limits of instrument
error). Surveying utilized an automatic level to relate
each station to a local datum, of which the elevation
relative to Mean Higher High Water (MHHW) was
known 1 .
Diatom sample preparation followed the procedures of Patrick and Reimer (1966). Digestion of
organic matter employed 30% hydrogen peroxide,
which was followed by several distilled water rinses.
Permanent slide preparation used evaporation trays
to randomly settle the diatoms onto round cover
slips, with Naphrax (refractive index D 1.7) employed as a mounting medium. Approximately 400–
600 valves for each sample were counted at magnifications of ð787 and ð1250, with lower counts
for two samples (0 and 20 m) that had low diatom
densities. Diatom identifications followed published
taxonomic monographs. Percentages of each taxon
are based on a sum of total diatoms.
4.2. Data analysis
My data analysis and results presentation generally follow procedures described in Ter Braak
(1987–1992) and Pan and Stevenson (1996). Calculations of descriptive statistics for each environmental variable and linear regression analyses for
calibration used SPSS version 6.1. Summary statistical calculations of the species assemblage data used
PC-ORD version 2.0 (McCune and Mefford, 1995).
1 Tidal elevations are from the February 1978 NOAA tidal survey
conducted at the Lynch Cove Dock (Tidal Station 944 5441).
Mean Higher High Water determined by NOAA is 3.691 m
relative to Mean Lower Low Water (MLLW).
Canonical correspondence analyses (CCA) employed the software package CANOCO version
3.1 (Ter Braak, 1987–1992). This technique relates
species composition to measured environmental factors. CCA ordinations utilize log transformed environmental variables. An unrestricted Monte Carlo
permutation procedure with 999 permutations tested
the significance of the first two ordination axes.
Calibration of environmental variables was
achieved by weighted averaging techniques (Ter
Braak, 1987–1992), using untransformed environmental variables. This technique produces transfer
functions for predicting environmental variables in
unknown samples based on the species composition. Coefficients of determination (r 2 ) between calculated (inferred) environmental variables and observed values of environmental variables allow evaluation of the transfer function’s predictive power.
5. Results
Salinity of the interstitial water ranges from a
high of 28‰ at the mudflat to a low of 0‰ at
the alder–willow–cattail swamp (Table 1; Fig. 2).
Salinity declined in an almost linear manner over
the transect, except for a sharp break at about 80 m
that corresponded to the change from the lower high
marsh to upper high marsh. Standing water in the
upland areas of the marsh had salinities of 0–1‰.
The vascular plant cover observed along the transect
corresponded to changes in salinity, with salt-tolerant
taxa near the tideflats and salt-intolerant taxa in the
upland areas.
Elevation changed almost 2 m over the length of
the transect (Table 1; Fig. 2). The interface between
the barren mudflat and the salt marsh is marked by an
80 cm tall erosional cut in the peat. The distal end of
the transect (between 240 and 260 m on the mudflat)
Table 1
Descriptive statistics of salinity measurements from interstitial
water of surface (0–1 cm) sediments, and elevation of sampling
stations relative to MLLW
Variable
N
Mean
Std. Dev.
Minimum
Maximum
Salinity
Elevation
22
22
17.4
363.0
8.9
42.8
0
261.1
28
437.1
B.L. Sherrod / Palaeogeography, Palaeoclimatology, Palaeoecology 149 (1999) 213–226
219
Fig. 3. Shannon indices (H 0 , calculated using natural logarithm) for diatom assemblages in Lynch Cove surficial sediments. Station
numbers correspond to distance on transect.
occurs well below Mean High Water (MHW D 338
cm). The low marsh occurs between 200 and 230
m at approximately 350 cm in elevation. The lower
high marsh lies between ¾80 and ¾200 m, is fairly
flat, and lies approximately 20 cm above MHHW.
The upper high marsh occurs at the landward end
of the transect between ¾80 and 0 m. A 60 cm
rise in elevation occurs from the beginning of the
upper high marsh at ¾80 m and the end of the
transect, representing the transition from salt marsh
environments to upland environments. The highest
elevation observed was 437 cm at 0 m. This is below
the highest water level of 465 cm observed by NOAA
during the 30-day tidal survey in 1978, showing that
extreme tides inundate the entire transect.
5.1. Diatom assemblages and gradient analysis
A total of 191 diatom taxa were identified from 22
surface sediment samples. The assemblages represent both autochthonous and allochthonous diatoms;
no attempt was made to separate the two groups.
Shannon indices (H 0 ) were calculated (using natural
logarithms) for the samples with PC-ORD (McCune
and Mefford, 1995). Diversity values ranged from
3.633 in the lower high marsh to 2.121 in the upper high marsh, with a mean value of 3.087. As
shown in Fig. 3, the lower high marsh samples have
larger H 0 values than upper high marsh samples (0,
20, and 60 m) and tideflat samples (240, 250, and
260 m). Species richness (Fig. 3) was highest in the
lower high marsh samples and ranged from 47 to 62.
Richness was much lower in the upper high marsh
(28–32) and mudflat (26–40).
Diatoms in each sample were well preserved. Observations of long, slender taxa, especially members
of the genera Gyrosigma and Nitzschia, suggested
minimal breakage. Lightly silicified diatoms (e.g.,
Skeletonema costatum) were also well preserved and
showed minimal evidence of breakage or dissolution.
Ninety-four diatom taxa were used in the CCA
gradient analysis, representing all taxa with relative
abundance ½1% in one or more of the 22 samples
(Table 2). Eigenvalues for the first three ordination
axes are ½1 D 0:558, ½2 D 0:219 and ½3 D 0:377; the
sum of all canonical eigenvalues is 0.776. Eigenvalues for CCA Axis 1 and Axis 2 represent 25.1% of
the cumulative variance in the species data. Species–
environment correlations were high for both axes
(Axis 1 D 0.94; Axis 2 D 0.84). Monte Carlo permutation tests indicated that the effect of both salinity
and elevation on diatom species composition along
the transect was significant (Psalinity D 0:001 and
Pelevation D 0:053).
220
B.L. Sherrod / Palaeogeography, Palaeoclimatology, Palaeoecology 149 (1999) 213–226
Table 2
List of diatom taxa in the CCA analysis
Code
Species name
Sites
Average
abundance
Occurrence in marsh
subenvironments
ACHAFF
ACHBIV
ACHDEL
ACHGRI
ACHGEM
ACHHAU
ACHLAN
ACHLEM
ACHMIN
ACHSPP
AMPCOF
AMPPED
AMPSTR
AMPVEN
BACPAX
AMPRUT
CALBAC
COCPEL
COCPLA
COCSCU
COSPUS
CYMAFF
DENSUB
DIMMIN
DIPELL
DIPINT
DIPSMI
ENTALA
EUNPEC
FRAFAS
FRASPP
FRAVIR
FRUCRU
GOMPAR
GOMANG
GOMCLE
GOMSUB
GYREXI
GYRSPE
HANAMP
HYASCO
LUTMUT
MELMON
MELNUM
MERCIR
NAVAGN
NAVCIN
NAVCRT
NAVCRY
NAVCRY1
NAVDIS
NAVGEL
Achnanthes affinis Grun.
A. brevipes Ag.
A. delicatula (Kütz.) Grun. in Cl. and Grun.
A. grimmei Krasske
A. sp. 1 (A. cf. grimmei)
A. delicatula spp. hauckiana Grun.
A. lanceolata (Breb.) Kütz.
A. lemmermannii Hust.
A. minutissima Kütz.
A. sp.
Amphora coffeaeformis A. Meyer
A. pediculus (Kütz.) Grun. in A. Sm.
A. strigosa Hust.
A. ventricosa Greg.
Bacillaria paxillifer (O. Muller) Hend.
Berkeleya rutilens (Trent ex Roth) Grun.
Caloneis bacillum (Grun.) Cl.
Cocconeis peltoides Hust.
C. placentula Ehrenb.
C. scutellum Ehrenb.
Cosmoneis pusilla W. Sm.
Cymbella affinis Kütz.
Denticula subtilis Grun.
Dimeregramma minor (Greg.) Ralfs in Pritch.
Diploneis elliptica (Kütz.) Cl.
D. interrupta (Kütz.) Cl.
D. smithii (Breb. ex W.Sm.) Cl.
Entomoneis alata (Ehrenb.) Ehrenb.
Eunotia pectinalis (O. Mull.) Raben.
Fragilaria fasciculata (Ag.) Lange-Bert.
F. sp.
F. virescens Ralfs
Frustulia creuzburgensis (Krasske) Hust.
Gomphonema parvulum (Kütz.) Kütz.
G. angustatum (Kütz.) Raben.
G. clevei Fricke in A. Schmidt
G. subclavatum (Grun. in Schn.) Grun. in Van Heurck
Gyrosigma eximium (Thwaites) Boyer
G. spenceri (W. Sm.) Cl.
Hantzschia amphioxys (Ehrenb.) Grun.
Hyalodiscus scoticus (Kütz.) Grun.
Luticola mutica (Kütz.) Mann
Melosira moniliformis (O. Muller) Ag.
M. nummuloides Ag.
Meridion circulare (Grev.) Ag.
Navicula agnita Hust.
N. cincta (Ehrenb.) Ralfs in Pritch.
N. cryptotenella Lange-Bert.
N. cryptocephala Kütz.
N. cryptocephala var. exilis (Kütz.) Grun. in Van Heurck
N. distans (W. Sm.) Ralfs in Pritch.
N. gelida Grun.
3
18
10
4
1
20
10
7
5
2
19
7
1
14
13
5
5
11
7
21
17
3
17
18
17
11
9
10
2
12
1
13
6
1
1
1
1
14
3
8
10
19
9
14
2
2
7
15
18
1
6
1
0.2
1.0
1.8
2.0
1.1
5.3
3.2
0.3
0.3
0.2
0.9
0.2
0.1
0.7
0.3
0.5
0.1
0.6
0.2
2.3
2.1
0.1
1.7
1.8
0.7
0.3
0.3
0.3
0.1
0.4
0.1
2.5
0.4
0.2
0.1
0.1
0.1
0.7
0.1
0.2
0.2
3.4
0.8
2.0
0.4
0.3
0.7
0.9
1.1
0.1
0.2
0.1
UHM, LHM
LHM, LM, MF
LHM, LM
LHM, LM
LHM
UHM, LHM, LM, MF
UHM, LHM, LM, MF
LHM, LM, MF
LHM, MF
UHM, MF
LHM, LM, MF
LHM, LM
UHM
LHM, LM, MF
LHM, LM
LHM, LM
LHM
LHM, LM, MF
LHM, LM, MF
UHM, LHM, LM, MF
UHM, LHM, LM
LHM
UHM, LHM, LM
UHM, LHM, LM, MF
UHM, LHM, LM
LHM, LM, MF
LHM, LM
LHM, LM, MF
UHM
LHM, LM
UHM
UHM, LHM, LM
UHM, LHM
UHM
UHM
UHM
UHM
LHM, LM
LHM
LHM, LM
UHM, LHM, LM, MF
UHM, LHM, LM
UHM, LHM
LHM, LM
UHM
LHM, MF
UHM, LHM
LHM, LM, MF
UHM, LHM, MF
LM
LHM, LM, MF
LM
B.L. Sherrod / Palaeogeography, Palaeoclimatology, Palaeoecology 149 (1999) 213–226
221
Table 2 (continued)
Code
Species name
Sites
Average
abundance
Occurrence in marsh
subenvironments
NAVGRE
NAVIND
NAVOCC
NAVPER
NAVPHY
NAVPRM
NAVRHY
NAVSAL
NAVSEM
NAVSLE
NAVSPP
NAVVEN
NAVVIR
NITCON
NITDUB
NITFAS
NITFON
NITFRU
NITLIM
NITLIT
NITPAL
NITREC
NITSPP
NITTEN
OPEPAC
OPEPAC
OPEPAR
PARSUL
PARSUL1
PARSUL2
PINAPP
PINLAG
PINMIC
PINSPP
PINSUB
RHOMUS
SKECOS
SURBRE
SYNRUM
SYNULN
SYNTAB
TABFEN
N. gregaria Donk.
N. indifferens Hust.
N. occuliformis Hust.
N. peregrina (Ehrenb.) Kütz.
N. phyllepta Kütz.
N. perminuta Grun. in Van Heurck
N. rhynchocephala Kütz.
N. salinarum Grun. in Cl. and Grun.
N. seminulum Grun.
N. slesvicensis (Grun.) Van Heurck
N. sp.
N. veneta Kütz.
N. viridula (Kütz.) Ehrenb.
Nitzschia constricta (Kütz.) Ralfs in Pritch.
Nitz. dubia W. Sm.
Nitz. fasciculata (Grun.) Grun. in Van Heurck
Nitz. fonticola Grun. in Van Heurck
Nitz. frustulum (Kütz.) Grun. in Cl. and Grun.
Nitz. limnicola (Ag.) W. Sm.
Nitz. littoralis (Grun.) in Cl. and Grun.
Nitz. palea (Kütz.) W. Sm.
Nitz. recta Hantzsch ex Raben.
Nitz. sp.
Nitz. tenuis W. Sm.
Opephora marina (Grun.) Petit
O. pacifica (Grun.) Petit
O. parva (Van Heurck) Krasske
Paralia sulcata (Ehren.) Cl. (undiff.)
P. sulcata (Ehren.) Cl. (large)
P. sulcata (Ehren.) Cl. (small)
Pinnularia appendiculata (Ag.) Cl.
P. lagerstedii (Cl.) A. Cl.-Euler
P. microstauron (Ehren.) Cl.
P. sp.
P. subcapitata Greg.
Rhopalodia musculus (Kütz.) O. Muller
Skeletonema costatum (Grev.) Cl.
Surirella brebissonii Krammer and Lange-Bert.
Synedra rumpens Kütz.
S. ulna (Nitzsch) Ehrenb.
Tabularia sp.
Tabellaria fenestrata (Lyngb.) Kütz.
16
12
10
15
12
10
6
17
2
8
7
7
1
14
3
6
5
13
2
17
22
7
6
17
4
17
20
21
20
16
3
4
1
1
2
14
8
10
2
6
6
3
2.2
0.9
0.4
0.7
0.4
0.6
1.1
2.0
0.1
0.5
0.3
0.7
0.1
0.3
0.1
0.1
0.4
1.3
0.1
1.3
2.1
0.7
0.2
1.2
0.3
1.4
11.8
3.7
7.4
0.7
0.8
0.3
0.1
0.2
0.1
1.6
0.5
0.3
0.2
0.2
0.2
0.2
UHM, LHM, LM
UHM, LM, MF
LHM, LM, MF
UHM, LHM, LM
LHM, LM
LHM, LM
UHM, LHM, LM
UHM, LHM, LM
LHM
UHM, LHM
UHM, LHM, LM, MF
LHM
UHM
LHM, LM, MF
LHM, LM
LHM, LM
LHM, LM
LHM, LM, MF
LHM, LM
UHM, LHM, LM, MF
UHM, LHM, LM, MF
UHM, LHM, LM, MF
UHM, LHM, LM, MF
UHM, LHM, LM
LM, MF
UHM, LHM, LM, MF
UHM, LHM, LM, MF
UHM, LHM, LM, MF
UHM, LHM, LM, MF
LHM, LM, MF
LHM
UHM, LHM
UHM
UHM
UHM
LHM, LM
UHM, LHM, LM
UHM, LHM, LM
UHM
UHM,LHM, LM, MF
UHM, LHM, LM, MF
LHM
Included for each taxon is its taxonomic authority, number of sites at which occurred, average abundance in all samples, and marsh
subenvironments that each taxon was observed (according to Fig. 2, UHM D upper high marsh, LHM D lower high marsh, LM D low
marsh, and MF D mudflat). Codes are shown for certain taxa in Fig. 4.
Canonical coefficients, t-values, and interset correlations (Table 3) represent different statistics for testing the relationship of the environmental variables
to the ordination axes (Ter Braak, 1987–1992; Pan
and Stevenson, 1996). Canonical coefficients indicate
the magnitude of a particular environmental variable’s
contribution to an ordination axis, with large canonical coefficients being associated with a higher degree
of contribution to a given axis (Pan and Stevenson,
1996). The canonical coefficient of salinity is highest
222
B.L. Sherrod / Palaeogeography, Palaeoclimatology, Palaeoecology 149 (1999) 213–226
Fig. 4. Ordination diagrams for sites (A) and species (B) based on canonical correspondence analysis of the Lynch Cove diatom data. The environmental variables salinity and
elevation are indicated by arrows. Species codes are presented in Table 2 and site numbers correspond to stations along transect. General groupings of sites and species in each
marsh subenvironment are indicated. See text for interpretation of ordination diagrams.
B.L. Sherrod / Palaeogeography, Palaeoclimatology, Palaeoecology 149 (1999) 213–226
223
Table 3
Canonical coefficients, approximate t-values, and interset correlations of environmental variables
Variable
Salinity
Elevation
Canonical coefficients
t-values of canonical coefficients
Interset correlations
Axis 1
Axis 1
Axis 1
0.723
0.042
Axis 2
0.316
0.552
10.7
0.6
Axis 2
3.9
6.7
for Axis 1 ( 0.723), while the canonical coefficient
of elevation is highest for Axis 2 ( 0.552).
In CCA, computed t-values are useful for exploratory purposes to determine if a particular environmental variable contributes more to the fit of the
species data relative to other environmental variables
in the analysis. This is accomplished by comparing
the computed t-values from the CCA output to a
critical value, which in this case is 2.08. Environmental variables with t-values that exceed the critical
t-value are interpreted as having a unique contribution to the fit of the species data (Ter Braak, 1987–
1992). For the Lynch Cove data, salinity had a high
canonical coefficient (Table 3) for Axis 1. Interset
correlations are the correlation coefficients between
the species axes (ordination axes) and the environmental variables (Ter Braak, 1987–1992). Based on
interset correlations, the first ordination axis inferred
a salinity gradient (interset correlation D 0.947),
while the second axis inferred an elevation gradient
(interset correlation D 0.69).
The results of the CCA are summarized in two
biplots (Fig. 4) that provide a graphical means for
representing the relationship between the species,
sites, and environmental variables. The length of
the arrows for salinity and elevation in Fig. 4 are
proportional to the importance of the environmental
variable in the biplot, with the arrows pointing in
the direction of increasing values for each environmental variable. The angle between the arrow and an
ordination axis indicates how well the environmental
variable is correlated with an ordination axis (the
smaller the angle, the stronger the correlation). A
perpendicular line drawn from an arrow through a
site or species point indicates the relative location
of that site (Fig. 4A) or species (Fig. 4B) along an
environmental gradient (Palmer, 1993).
As previously noted, salinity is strongly correlated
with CCA Axis 1 and elevation with Axis 2. Species
0.947
0.542
Axis 2
0.040
0.690
that fall on the left side of the diagram are more
abundant in areas of higher salinity, while those on
the right side occur more frequently in areas of
fresher water. Likewise, species in the lower part of
the diagram are indicative of higher elevations, while
those in the upper part are more abundant at lower
elevations. The sites also follow a similar distribution
pattern in relation to the environmental variables on
the diagram. The biplots indicate that there is a
strong relationship between species distribution and
the environmental variables.
The ordination resulted in the separation of the
sites and species into three main groups representing the main marsh subenvironments (Fig. 4A, B).
Species were distributed in a continuum along the environmental gradients, with most species occurring in
several marsh subenvironments. For the species biplot, certain associations of species were representative of each main marsh subenvironment. Mudflat sites had the highest salinity and lowest elevation and were best distinguished by Opephora
parva, Achnanthes lemmermanii, Achnanthes delicatula spp.hauckiana, and Paralia sulcata. Low marsh
sites did not show a clear species pattern that was recognizable from mudflat and high marsh sites. Lower
high marsh sites had the highest diversity (H 0 ) and
were characterized by species of the genus Navicula, Denticula subtilis, Luticola mutica, and others.
Upper high marsh sites had the lowest salinity and
highest elevation and were characterized by Pinnularia lagerstedtii, Navicula rhynchocephala, Meridion circulare, Eunotia pectinalis, Synedra rumpens,
and several species of the genus Gomphonema.
5.2. Calibration
CCA output can be used for weighted-averaging
calibration with a minor amount of post processing (Ter Braak, 1987–1992). Transfer functions were
224
B.L. Sherrod / Palaeogeography, Palaeoclimatology, Palaeoecology 149 (1999) 213–226
developed for both salinity and elevation using the
Lynch Cove species data. Linear regression analysis
allows a comparison of diatom-inferred environmental variables with observed values of environmental
variables for the same set of sampling locations.
This provides a means for evaluating the predictive
ability and bias of the transfer functions (Birks et
al., 1990). Coefficients of correlation (r 2 ) provide
a means of assessing the predictive power of the
transfer functions. A plot of diatom-inferred salinity
and observed values yielded an r 2 of 0.84 (Fig. 5A).
Residual salinity was within š5‰ of the observed
values in all but three cases (Fig. 5B). Slight systematic trends in residual salinity may indicate that
the current transfer function does not explain a certain amount of variance (Pan and Stevenson, 1996).
A plot of diatom-inferred elevation and observed
values (Fig. 5C) yielded an r 2 of 0.72. Residual elevation was within 50 cm of the observed value in
all but two cases (Fig. 5D). For both environmental
variables, the predictive ability of the transfer function was greater for samples in the middle of the
transect than for samples falling at the extremes of
the environmental gradients.
Fig. 5. Graphs of calibration results. (A) Scatter diagram of diatom-inferred salinity and observed salinity with fitted linear regression
line. (B) Residuals for diatom-inferred salinity from regression analysis. (C) Scatter diagram of diatom-inferred elevation and observed
elevation with fitted linear regression line. (D) Residuals for diatom-inferred elevation from regression analysis.
B.L. Sherrod / Palaeogeography, Palaeoclimatology, Palaeoecology 149 (1999) 213–226
6. Discussion
The hypothesis that major environmental and
taphonomic processes (e.g., tides) act in predictable
ways to distribute living and dead diatoms along
environmental gradients is accepted based on the results of the CCA. Monte Carlo permutation tests
indicated the effect of salinity and elevation on
the species distributions was significant, suggesting
that environmental processes control diatom thanatocoenoses across the salt marsh surface. The ordination diagram showed that distinct diatom thanatocoenoses occur in the upper marsh subenvironment
where the environmental gradients tend to break
down, while diatom thanatocoenoses tend to occur
as a continuum in other marsh subenvironments.
These observations are consistent with earlier work
in the Yaquina Estuary, Oregon. McIntire (1978)
also found large changes in the diatom flora of the
Yaquina Estuary wherever a breakdown of the environmental gradient occurred. Amspoker and McIntire (1978) also observed a sharp discontinuity in
the diatom flora where the salinity dropped below
5‰ and attributed the floristic change to differences
in osmotic regulation between fresh- and brackishwater diatoms. Similar changes in species composition were observed by Snoeijs (1995) while studying
epiphytic diatoms in the Baltic Sea. These studies, combined with the present results, suggest that
the distribution of diatom thanatocoenoses in salt
marshes are controlled in a large part by salinity.
Diversity was highest in the high marsh samples
and lowest in the upland and tideflat areas, perhaps
in part a response to a higher degree of mixing of
autochthonous and allochthonous diatoms in the salt
marsh samples. The distribution patterns observed
are likely related to two factors; salinity tolerance of
individual taxa and spatial patterns in the distribution
of allochthonous diatoms by tidal currents. Laws
(1988) found similar relationships between diatom
assemblages in surficial sediments of San Francisco
Bay and environmental gradients. Laws (1988) attributed overriding control on diatom distributions to
salinity and depth, and also cited higher species diversity in areas where mixing of brackish and freshwater diatoms occur. Similar patterns are apparent at
Lynch Cove, where offshore taxa are carried into the
marsh during high tides and deposited, with the up-
225
per limit to the distribution of allochthonous diatoms
primarily controlled by the elevation of MHHW.
Epipsammic and tychoplanktonic diatoms dominate the mudflat and lower marsh subenvironments,
with high proportions of these diatoms in the marsh
environments likely being allochthonous. Earlier
work by De Jonge (1985) found that loosely attached
diatoms living on mudflats and in channels may
be scoured off the substratum by tidal currents and
transported into adjacent environments. This process
of stripping epipsammic diatoms and transporting
the frustules into adjacent environments is often
cited as a principal taphonomic process that controls
the composition of salt marsh diatom taphocoenoces
(Sherrod et al., 1989; Hemphill-Haley, 1995b). The
results of the present study show that while transportation alters diatom thanatocoenoses, it does so
in predictable ways and results in assemblages that
may in some cases reflect environment gradients.
The hypothesis that diatom thanatocoenoses occurring in salt marsh deposits reflect major ecological processes and can be used for accurate quantitative reconstructions is also accepted based on
the limited testing presented in this study. The calibration results indicate that, although mixing of
allochthonous and autochthonous diatoms does occur, salt marsh diatom assemblages reflect major
environmental gradients and thus can be effectively
used for quantitative reconstructions of former environmental conditions. This suggests that the mixing
of allochthonous and autochthonous diatoms occurs
in a predictable manner along major environmental
gradients. The present study indicates that salinity
can be predicted within š5‰ and elevation to within
š50 cm. By broadening this study and adding additional sites, it may be possible to improve on the
predictive power of the calibration model and investigate the second hypothesis further. CANOCO
allows passive samples (samples added to an existing
ordination) in the ordination analysis, so that fossil
samples can be directly related to modern training
sets (Ter Braak, 1995a). Shennan et al. (1996) used
a similar technique to estimate the magnitude of
submergence (elevation changes) accompanying past
earthquakes along the Pacific Northwest coast but
did not take into account the role of salinity or other
environmental variables. Estimated error in the elevation changes indicated by Shennan et al. (1996)
226
B.L. Sherrod / Palaeogeography, Palaeoclimatology, Palaeoecology 149 (1999) 213–226
were 0.5 m, consistent with the error of the diatombased elevation predictions for the present study.
The results of the present study suggest diatoms
can be used as the basis for quantitative reconstructions of both salinity and elevation in paleoecological
studies. The method presented is equally applicable
to studies based on other types of microfossils, especially Foraminifera and vascular plant seeds (Nelson
and Kashima, 1993; Sherrod, unpubl. data).
Acknowledgements
The author wishes to thank Estella Leopold and
Robert Bucknam for many helpful suggestions and
comments during this study. Michael Sullivan and
Richard Laws provided constructive reviews of the
manuscript. Support from the U.S. Geological Survey is gratefully acknowledged.
References
Amspoker, M.C., McIntire, C.D., 1978. Distribution of intertidal diatoms associated with sediments in Yaquina Estuary,
Oregon. J. Phycol. 14 (4), 387–395.
Birks, H.J.B., Line, J.M., Juggins, S., Stevenson, A.C., Ter
Braak, C.J.F., 1990. Diatoms and pH reconstruction. Philos.
Trans. R. Soc. London B 327, 263–278.
Bucknam, R.C., Hemphill-Haley, E., Leopold, E.B., 1992.
Abrupt uplift within the past 1700 years at southern Puget
Sound, Washington. Science 258, 1611–1614.
De Jonge, V.N., 1985. The occurrence of ‘epipsammic’ diatom
populations: A result of interaction between physical sorting
of sediment and certain properties of diatom species. Estuarine
Coastal Shelf Sci. 21, 607–622.
Denys, L., 1994. Diatom assemblages along a former intertidal
gradient: A paleoecological study of a Subboreal clay layer
(western coastal plain, Belgium). Neth. J. Aquat. Ecol. 28 (1),
85–96.
Hemphill-Haley, E., 1995a. Distribution and taxonomy of diatoms (Bacillariophyta) in surface samples and a two-meter
core from Winslow Marsh, Bainbridge Island, Washington.
U.S. Geol. Surv. Open File Rep. 95-833, 105 pp.
Hemphill-Haley, E., 1995b. Diatom evidence for earthquake-induced subsidence and tsunami 300 yr ago in southern
coastal Washington. Geol. Soc. Am. Bull. 107 (3), 367–378.
Hemphill-Haley, E., 1995c. Intertidal diatoms from Willapa Bay,
Washington: Application to studies of small scale sea-level
changes. Northwest Sci. 69 (1), 29–45.
Holling, C.S., 1992. Cross-scale morphology, geometry, and dynamics of ecosystems. Ecol. Monogr. 62, 447–502.
Laird, K., Edgar, R., 1992. Spatial distribution of diatoms in the
surficial sediments of a New England salt marsh. Diatom Res.
7 (2), 267–279.
Laws, R., 1988. Diatoms (Bacillariophyceae) from surface sediments in the San Francisco Bay estuary. Proc. Calif. Acad.
Sci. 45 (9), 133–254.
McCune, B., Mefford, M.J., 1995. PC-ORD. Multivariate analysis of ecological data, version 2.0. MjM Software Design,
Gleneden Beach, Oregon, USA.
McIntire, C.D., 1978. The distribution of estuarine diatoms along
environmental gradients: a canonical correlation. Estuarine
Coastal Mar. Sci. 6, 447–457.
Nelson, A.R., Kashima, K., 1993. Diatom zonation in southern
Oregon tidal marshes relative to vascular plants, Foraminifera,
and sea level. J. Coastal Res. 9 (3), 673–697.
Palmer, M.W., 1993. Putting things in even better order: the
advantages of canonical correspondence analysis. Ecology 74
(8), 2215–2230.
Pan Yangdon, Stevenson, J., 1996. Gradient analysis of diatom
assemblages in western Kentucky wetlands. J. Phycol. 32,
222–232.
Patrick, R., Reimer, C., 1966. The diatoms of the United States
exclusive of Alaska and Hawaii. Volume 1. Monogr. Acad.
Nat. Sci. Philadelphia 13, 688 pp.
Shennan, I., Long, A.J., Rutherford, M.M., Green, F.M., Innes,
J.B., Lloyd, J.M., Zong, Y., Walker, K.J., 1996. Tidal marsh
stratigraphy, sea-level change and large earthquakes, I: A 5000
year record in Washington, U.S.A. Quat. Sci. Rev. 15, 1023–
1059.
Sherrod, B.L., Rollins, H.B., Kennedy, S.K., 1989. Subrecent
intertidal diatoms from St. Catherines Island, Georgia: Taphonomic implications. J. Coastal Res. 5 (4), 665–677.
Snoeijs, P., 1995. Effects of salinity on epiphytic diatom communities on Pilayella littoralis (Phaeophyceae) in the Baltic Sea.
Ecoscience 2 (4), 382–394.
Sullivan, M.J., 1982. Distribution of edaphic diatoms in a Mississippi salt marsh: a canonical correlation analysis. J. Phycol.
18, 130–133.
Ter Braak, C.J.F., 1987–1992. CANOCO — a FORTRAN program for Canonical Community Ordination. Microcomputer
Power, Ithaca, New York, USA.
Ter Braak, C.J.F., 1995a. Ordination. In: Jongman, R.H.G., Ter
Braak, C.J.F., Van Tongeren, O.F.R. (Eds.), Data Analysis in
Community and Landscape Ecology. Cambridge Univ. Press,
pp. 91–173.
Ter Braak, C.J.F., 1995b. Calibration. In: Jongman, R.H.G., Ter
Braak, C.J.F., Van Tongeren, O.F.R. (Eds.), Data Analysis in
Community and Landscape Ecology. Cambridge Univ. Press,
pp. 78–90.
Vos, P.C., De Wolf, H., 1988. Methodological aspects of paleoecological diatom research in coastal areas of the Netherlands.
Geol. Mijnbouw 67, 31–40.
Vos, P.C., De Wolf, H., 1993. Diatoms as a tool for reconstructing sedimentary environments in coastal wetlands; methodological aspects. Hydrobiologia 269–270, 285–296.