Master

Effects

of

historical

perturbations

on
aquatic

bacterial

communities
A

microbial

journey

into

the

4th

dimension!
Martin

Andersson
Degree

project

in

biology,

Master

of

science

(2

years),

2012
Examensarbete

i

biologi

45

hp

till

masterexamen,

2012
Biology

Education

Centre

and

Limnology

Department

of

Ecology

and

Genetics,

Uppsala

University
Supervisor:

Silke

Langenheder
External

opponent:

Torsten

Jeske,

Jiazhuo

Zhang

---

Abstract

This project investigates whether historical spatial and environmental mechanisms are of
importance in determining bacterial community composition (BCC). The study was
performed in rock pools at the Swedish Baltic Sea Coast. The sampling was done in 48 hours
intervals in a time series over a 9 day period. I tested whether bacterial community
composition at the end of the sampling period correlated stronger with past environmental
conditions compared to present conditions, using partial redundancy analysis and Mantel
tests. The results showed that spatial factors were generally not important and that salinity
was the most important structuring environmental factor. Correlations between salinity and
BCC demonstrated that higher correlations were achieved with past compared to
contemporary salinity conditions. This demonstrates the existence of an interference of past
perturbation events on present bacterial communities in a natural environment. My results
also highlight the necessity of conducting sampling for bacterial community analysis in time
series in order to acquire a greater understanding of the historical effects and the timespan
during which they affect bacterial communities.

Introduction

Bacteria are key players for the functioning and stability of all ecosystems on Earth as a result
of their cosmopolitan distribution, abundance and metabolic capabilities (Whitman et al.,
1998; Hallin et al., 2009). This creates a great incentive to understand the diversity and global
biogeography of bacterial communities. However, due to the simple fact that bacteria are
microscopic (and therefore hard to observe and manipulate), our understanding of bacterial
communities is less complete compared to those of macro organisms. Technological
development and adaptation of previous metacommunity research (a metacommunity is
defined as a set of local communities that are linked by dispersal of multiple interacting
species (Wilson, 1992)) has nonetheless allowed for significant advances of our
understanding of the biogeography of bacteria communities over the last decade (Zinger et al.,
2011; Lindstrom & Langenheder, 2012).

The three main perspectives used in BCC (bacteria community composition) research that are
derived from metacommunity research are Species sorting (Leibold 1998), Mass effect
(Shmida & Wilson 1985) and Neutral theory (Hubbell 2001). Species sorting explains the
community as an effect of species competition and species adaptations to the local
environment. Essentially the species sorting perspective assumes that species have different
niche requirements and can thrive under different abiotic and biotic conditions and that this
difference is what determines if a species can exist in a locality. This was the first ecological
concept to be established for bacterial communities, beginning with Baas-Beckings famous
statement "everything is everywhere, but the environment "selects". The idea is that bacteria
are vastly abundant with a high level of plasticity and at the same time easily spread as a
result of their minute size. This was interpreted as that all species of bacteria must be spread
globally. The presence or absence of bacterial species is then explained as a pure product of
environmental selection (Baas-Becking, 1934). Despite originally being based on logic rather
than empirical studies the theory still holds a lot of recognition today (Finlay & Clarke, 1999;
Fenchel & Finlay, 2004). Species sorting is in fact the perspecive that most often significantly
explains BCC (Lindstrom & Langenheder, 2012), although most researchers nowadays
acknowledge that there are limitations in bacterial dispersal (Schauer et al., 2010) in contrast
with Baas-Beckings statement.

2

---

The Mass effect perspective predicts that declining populations of a species can be stable due
to regular inflow of migration. Regarding bacterial communities the Mass effect is similar to
species sorting in the way that it predicts that there is/can occur a vast and continuous
distribution of bacteria in the environment. The Mass effect perspective then predicts that as a
result of massive fluxes of bacteria, high abundances of bacteria commonly end up in
environments that they are poorly adapted to. According to the Mass effect perspective these
species can remain in the community as a result of constant immigration and the balance
between the decline and the rate of the immigration would then set the abundance level of that
species (Shmida & Wilson, 1985; Mouquet & Loreau, 2002). This perspective could also
explain why bacterial communities correlat so poorly with factors that other organisms are
strongly dependent on, as Mass effects decouples bacterial communities from their
environment. If bacteria can be stable in an environment that actually should cause them to
decline, it would make a researcher that samples this environment incline to believe that the
environmental parameters have no effect. Recent research has casted doubts regarding if Mass
effect is a likely event to occur in a natural bacterial communities (Logue & Lindstrom, 2010;
Lindstrom & Ostman 2011)

Neutral theory is a relatively new theory in the field of metacommunity research and has
sparked a lot of interest and controversy. Neutral theory states that the difference between
similar species on the same trophic level is irrelevant or neutral to their chance of success
(Hubbell, 2001). Furthermore it assumes that randomly occurring events in nature are the
actual determining factors for a species abundance and biodiversity in general. For example:
Assume four similar species of flowers that exist as seeds in a forest. When spring comes
three out of the four sprouts, the fourth dies as a result of landing in a Sphagnum area. Out of
the remaining three only one is able to produce seeds as the two other seeds happened to land
in a shady area and did not receive the necessary energy to produce offspring. During the
following year the two species in the shady area have grown larger and can now reach the
sunlight. The third species has now nine individuals this year. As a result the third species
now has more opportunities to create offspring this year. Hence, the third species will always
be producing more offspring from this point forward and thus become the dominating species.
Eventually, if there is no migration and so forth, the third species will outcompete the other
flower species (as an effect of it having more opportunities to succeed) and with time be the
only remaining flower in the forest. This means that the biodiversity and abundance among
the flowers in the forest is all a result of the original seed by chance being blown by the wind
to a suitable spot even though the difference between the flowers was "neutral".
Neutral theory is often referred to as a null theory to other ecological metacommunity
concepts, including mass effect and species sorting (Bell, 2001). Neutral theory has been
adapted to bacterial community research and seems to be increasingly more approved as
several studies shows patterns similar to that predicted by the neutral model (Sloan et al.,
2006; Keymer et al., 2009; Drakare & Liess, 2010; Ostman et al., 2010).

Additionally are the influences of bacterial dispersal limitation and priority effects are often
emerging in the discussions of bacterial community assembly. Dispersal limitation simply
states that bacteria are subjects of geographical limitations. These patterns have also been
observed in bacterial community studies (Schauer et al., 2010). This is of great importance for
bacterial community research as several theories are based on the fact that there is a limitation
in bacterial dispersal. Priority effects (or founder effect) are comparable to the saying "first
come, first served". The organism that can establish itself first in an emerging patch has a
competitive advantaged towards late-comers, either as result of reducing the available amount
of a limiting resource or by changing the environment in a way that makes in unsuitable for
3

---

other species. This effect has been observed among coral reef fish (Almany, 2003) and plant
communities (Fukami et al., 2005) but has not yet been investigated for bacterial
communities.

The common standpoint in current metacommunity research is that all theories mentioned are
relevant for the understanding of metacommunities but that none of them gives a complete
answer for a given community. Despite substantial progress within these concepts there is still
a great gap remaining in our understanding of the variation seen in bacterial communities,
commonly leaving 60-80% of the seen variation among locations in a metacommunity
unexplained (Langenheder & Ragnarsson 2007; Van der Gucht et al., 2007; Zinger et al.,
2011). A constant lack for all concepts mentioned above is the influence of historical events.
Historical events are frequently discussed as one of the factors that could influence the results
of metacommunity research (Leibold & Mikkelson, 2002; Langenheder & Ragnarsson, 2007)
but which have yet to be thoroughly tested for bacteria communities. The ideas behind
historical factors are similar to those of other fields of ecology, with many explanatory factors
being influenced by usually unknown events that have occurred in the past. In this study I will
look at the influences that historical events might cause for the interpretation of how
environmental and spatial factors influence bacterial community composition. Other concepts
such as Neutral theory and Priority effects are also likely to be heavily influenced by
historical factors, the specific investigation of Priority effects and Neutral theory are,
however, outside of the scope of this study.

Historical environmental effects on BCC can be fairly straightforward. For example consider
a situation where a strong but momentary change in pH has recently wiped out a significant
fraction of the species within a bacterial community. The pH quickly returns to its original
state, while lack of immigration (or other factors) prevents the bacterial community to go back
to its previous state. As a result the correlation between environmental values and BCC
becomes low, which then falsely would be interpreted as the environment having a minor
effect. This type of event will henceforth be referred to as "historical interference". In
addition to historical interference I want to mention "present interference" and delayed
response time. Present interference is when a long term community adaptation in
combination with an environmental perturbation causes a decrease in correlation between the
environment and the community. For example, consider a situation where a researcher
collects bacterial samples from two lakes with similar bacterial communities. One of the lakes
recently had a substantial drop in oxygen; the bacterial community is temporarily coping with
the stress of the low oxygen levels. When the researcher later correlates oxygen levels
towards BCC the correlation gets incorrectly low as the two lakes have similar BCC but very
different oxygen levels. Thus, whereas historical interference changes the bacterial
community but leaves no measurable trace; present interference does not leave an impact on
the bacterial community but changes the measured parameters. In reality it would not be
possible to separate the two from one another if the sampling was not done in a time series as
both results in a none-traceable decrease of correlations between environmental conditions
and BCC.

Finally is there the delayed response time of the community, which is the difference between
initial and "complete" response of a community to an environmental or spatial fluctuation.
For example, consider that there is a bacterial community with 50 000 different species within
it. A relatively minor change in environmental conditions causes 1% of the species to get
extinct, 69% of the bacteria can resist the environmental change but have to use a substantial
part of their metabolism to resist the stress while 30% of the species will not get noticeably
4

---

affected. The initial measurable response would thus be that 1% of the bacteria are lost.
However, if the environmental increase remains an increasing share out of stressed bacteria
might also get extinct. This could then get further complicated by the fact that the competitive
balance between species will begin to move as an effect of the stress and thus initiate a
species sorting process (the recovery time from a perturbation is an additional factor). This
will result in a large difference between the initial and final effect of the fluctuation. Baho et
al. (2012) found strong alterations in BCC still ongoing two weeks after a pulse of salinity
was added to the community. This shows that the gap between the initial and final response
does exist and that this effect can be ongoing a long time after the perturbation, therefore
increasing the likelihood of this being an issue when conducting field studies.

Historical spatial effects can be a large variation of events; rain, wind or temporary
immigration bridges and so forth. Historical spatial effects pose similar issues for BCC
research by possibly blurring results of correlation tests. One could argue that the testing of
mass effects is in some regards are tests of historical spatial effects (i.e. mass effect
immigrations caused by spatial events). There are, however, numerous other possible
historical spatial events, e.g. previous long term connections, mixing of multiple communities
or creating opportunities for founder effects.

There are issues verifying that historical mechanisms occur in bacterial communities in their
natural environment. A historical mechanism would be calculated as following: Impact of
historical mechanism X = X past - X present. There are few logical reasons to assume that the
past consistently and greatly would affect the bacterial community to a higher degree then the
present. Firstly, there is no specific reason to believe that the conditions that lie close in the
past would be vastly different from the current conditions, i. e. they will often be
autocorrelated. Secondly, if the community has been affected by a sudden pulse of "change"
and thereafter resumed its original state there is reason to assume that the bacterial community
would start to recover with time. This would be an effect of the present conditions being a
selecting factor on the community, thus always pushing the community towards the most
ideal adaptation to the present environmental state.

If there has not been a sudden change or if the change has remained until the present point, the
past and present are identical and there is no reason to separate the two (X =1-1 =0). As a
result of this the power of historical factors is likely to be significantly lower than the factor
itself in most cases. The only scenario where a historical effect is more important than the
present factor itself is if the event in the past has more than twice the effect as the same factor
in the present (Past = 2.05, Present = 1 (Historical factor X = 2.05 - 1 = 1.05. 1<1.05)). It is
fair to assume that this is an unlikely scenario. If historical effects are on average weaker
compared to present factors this is an issue in particularly when working with BCC in field
studies, where known explanatory factors often only explain a minor fraction of variation in
community composition. As a result of being dependent on previously known factors, the
historical aspects are not likely to revolutionize the degree of explanation we observe for each
given factor.

On the other hand does the history influence the patterns of all theories explaining BCC.
Therefore, the addition of historical aspects to a factor could contribute to a somewhat higher
degree of explanation. This would be of particular importance if several different theories are
used for explaining observed variance in a community. In most cases the historical effects are
likely to remain an unknown factor, affecting the outcome of the observed pattern to an
unidentified extent. By investigating the likelihood and magnitude of historical impacts in the
5

---

different metacommunity perspectives we can start to assess the relevance of missing
historical effects and thereby learn how to better approach the issue when conducting
research. Studies of historical factors will also contribute to other areas of understanding of
bacterial communities, such as bacterial resistance and response time to environmental
perturbations.

The possibilities and impact of historical events on BCC depend on several factors. Firstly,
bacteria must have a limitation in their dispersal ability. If bacteria species would be able to
instantly establish in any suitable environment the effect of historical events would be none.
The resilience of bacterial species would also impact the significance of historical events. If
species of bacteria can survive in harsh conditions for a period of time, or only reduce their
abundance, it would infer that temporary disturbances would have a relatively small effect on
BCC. The response rate of the bacterial community to changes is also of importance as the
time period needed for observing the effect would be decided by this variable.

In this study I implemented a field study with the aim to test if historical mechanisms occur in
bacterial communities in their natural environment. The bacterial communities were located
in rock pools, i.e. a small bedrock depression filled with either fresh or brackish water. Rock
pools are suitable objects for metacommunity research as they present an opportunity to look
at a large number of separated bacterial communities in small "areas". The rock pools are also
very heterogeneous in environmental conditions, in particular with regard to salinity and
water colour. Moreover, pool volume is varying greatly, ranging from 1m3< to <0.05 m3,
depending on the rock depression and weather conditions. Their small size implicates that
rock pools are likely to experience strong variability in environmental conditions over time
and thus makes them an ideal system to study effects of historical events. The purpose here
was to test how much of the variation in BBC among pools can be related to differences in
present environmental conditions as well as to environmental conditions measured multiple
times during a 9 day period prior to the sampling event.

Specifically, the aim of the project was to investigate the following questions:

1) Can historical environmental events affect BCC in their natural environment?

2) Can historical spatial events affect BCC in their natural environment?

3) What is the time period during which historical impacts on BCC become observable?


Material and Methods

Field sampling
Samples were taken from twenty rock pools located along the Baltic Sea coast in the
province of Uppland in central Sweden (60 29 54 N, 18 25 45 E). Each rock pool was
sampled five times, starting at August 3rd 2011 and ending at August 11th 2011, with a 48h
interval between each sampling occasion. At each sampling occasion the following factors
were measured in the field: pool width, length, depth, salinity and temperature. Length, width,
and depth of the pools were measured using a yardstick and measuring tape. Salinity and
water temperature were measured using a WTW Conductometer LF 191. Additionally, water
samples were collected for measurements of BCC, bacterial abundance, total-phosphorous
6

---

concentration, chlorophyll-a concentration, absorbance and abundances of zooplankton and
flagellates. In addition, samples for BCC and bacterial abundance were also collected from air
and rain samples with 3 samples of each at each sampling occasion. For this a set of traps was
constructed. The traps consisted of a plastic bucket with a smaller sterilized plastic container
within it. In order to prevent larger particles to get into the sample a net was placed on the top
of the bucket. In the air traps a solid lid was also placed above the net, with approximately 4
cm gap between the lid and the bucket, in order to prevent rain bacteria from contaminating
the air sample. The traps for air bacteria were filled with 100 ml Mili-Q water, whereas the
rain traps did not contain any water. Water was collected in 1L sterilized plastics bottles.
Upon arrival to the laboratory the water samples were processed further (see below).
Geographical coordinates for pools were recorded using a GPS unit. Weather data was
collected from a local weather station belonging to the Swedish Meteorological and
Hydrological Institute (SMHI).

Laboratory measurements
All measurements mentioned below were carried out 5 times per pool, resulting in a total of
100 samples/values for each environmental parameter. For the analyses of chlorophyll-a
concentration 100-500 mL water were filtered onto 47-mm glass fiber filters (Whatman
GF/C). Chlorophyll-a was then extracted by submerging the filter in 95% ethanol for 1
minute. The extracts absorbance was measured at 665nm and 770nm wavelength and the
final value was corrected for pheophytin interference. Total phosphorus concentration were
analysed according to Mezel & Corwin (1965). Water colour was obtained by measuring
absorbance at 436 nm in a 5 cm cuvette. For bacterial abundance measurements the samples
where preserved in filter-sterilized formaldehyde and stored in a 4 oC dark room. Bacteria
cells were counted using a flow cytometer (CyFlow space, Partec, Germany). To determine
flagellate abundance 2 ml formaldehyde-preserved water were filtered onto a 0.8 m
polycarbonate filter and stained with DAPI (final concentration 100 g ml-1). Counting was
then performed with an epifluorescence microscope by counting all flagellates in a pre-
determined area on the filter surface (25 mm x 100 m).

For zooplankton abundance and identification animals were collected by filtering water (2L)
through a plankton net and were then sorted and kept in a 50 mL polypropylene test tube
(Falcon) with a 50% ethanol solution and were stored in a 4oC room. The zooplankton was
counted and identified using an Olympus SZ61 microscope.

Pool volume was calculated from the length, width and depth presuming that each pool had
the shape of an inverted pyramid. Locations and distances between pools on the first day was
calculated with ArcGIS 9.2. Due to issues with the ArcGIS software the calculation of
changes in the closest neighbor distance (as pool volume and flow connection changed with
the rain) was done "manually" in Excel. Here I calculated for all sampling occasions, the
distances between the GPS points with the reduction in distance that was added as a result of
the increasing areas of the pools.

BCC was determined twice, once at the starting point (day 1) and then at the final sampling
(day 9). BCC was analysed with the T-RFLP method (terminal restriction fragment length
polymorphism) (Liu et al. 1997). DNA was extracted from the sample using the Soil DNA
isolation kit according to the instruction manual (MOBIO Laboratories). The 16S rRNA gene
was then amplified using PCR (Polymerase chain reaction) with the fluorescently labelled
bacteria-specific forward primer 8F-HEX and the universal reverse primer 519r. One 50 L
PCR reaction was carried out for each sample with the following mixture:2 M of each
7

---

primer, 2.5 mM of each dNTP, 50mM MgCl2 in 1 x NH4 buffer -and 0,5L 5 U/L Biotaq
DNA polymerase. PCR products were then concentrated and purified with a PCR purification
kit (Qiagen). Quantification of the purified products was done with the Quant-
iTTMPicoGreen(R) dsDNAReagent kit. Fluorescence measurements were performed with a
Tecan ultra evolution microplate reader (TECAN - Ultra 384). DNA was afterwards diluted
with Mili-Q water in order to achieve a concentration of 4 ng/L of purified DNA. The
restriction enzymes Hha I and Hae III (New England Biolabs, Ipswich, Massachusetts, USA)
were used for 2 separate digests, each sample having a duplicate and incubated at 37 oC for 18
hours. Restriction fragments were then separated using an ABI3730XL DNA Analyzer in the
Rudbeck laboratory in Uppsala.

The analysis of the T-RFLP data was performed using GeneMarker (Version 1.95). All peaks
smaller than 50 base pairs and less than 0.5% of the total signal were removed from the
analysis. Peaks closer than 0.5 base pairs were merged in order to account for the differences
in running time between different samples. Each peak that remained after these modifications
was considered an operational taxonomical unit (OTU). For statistical analyses, only the Hae
III-digested samples were used since there were more successful runs within those samples.

Statistical analyses
Prior to the testing all data was transformed to a logarithmic scale (log (x+1)) to achieve a
normal distribution of the data. To determine correlations between BCC and spatial and
environmental factors, partial redundancy analysis (pRDA) was used. The pRDA procedure
enables the determination of the independent effects of each explanatory factor on BCC as
well as shared effects due to co-variation. For these calculations CANOCO 4.5 was used
using Chord transformations of the species data to make it conform to a linear gradient, which
is a requirement for the use of RDA (e.g. Legendre & Birks, 2012). In all models significance
testing was done using Monte Carlo permutation tests with 999 permutations

The statistical analysis was done in following steps:
1) A standard RDA was performed to test the correlation of all environmental factors (total-P,
chl-a, absorbance, zooplankton and flagellates) and BCC at day 9 separately for all 5
sampling days. Then the same procedure was performed with the spatial factors volume and
closest neighbor values. Volume was used as spatial factor since it relates to merges of rock
pools, over-all decreases in distance between pools, and increase of water movement.
Environmental factors were significantly correlated with BCC at day 9 at all sampling
occasions while none of the spatial factors gave a significant result at any sampling point.
Hence, no further calculations were performed with the spatial data. 2) Forward selection of
the environmental factors. Forward selection of environmental variables was implemented as
described in Blanchet et al (2008) using two-cut off values to determine whether or not
environmental variables made a significant contribution to explaining variation in BCC: (a) a
p-value < 0.05 and the adjusted R2 value (Peres-Neto et al., 2006) of the global model
calculated in step 1. Hence, for all sampling days, salinity was the only environmental factor
that was included into the model. 3) pRDAs were performed using salinity at the 5 sampling
days and BCC in order to see which day had the highest degree explanation. This was also
done in order to test if the patterns of explanation would remain the same with co-variation
being included and to further to identify trends in the degree of BCC that can be explained by
salinity variations over time. For example the effect of salinity at day 1 on BCC at day 9 was
tested with salinity at day 3 as co-variable, next the effect of salinity at day 1 on BCC at day 9
was tested with day 5 salinity as co-variable and so on, with all 20 possible combinations
being tested.
8

---

Finally correlations between salinity at the different sampling dates and BCC were also tested
with partial Mantel tests, which were implemented using the Excel add-on XLSTAT.
Similarity matrices were created for all variables after they had been log10 transformed and for
the BCC matrix Bray-Curtis dissimilarity was used, for salinity data matrices Euclidian
distances were used.

Terminology

BCC always refers to BCC at day 9 (final sampling) if nothing else is mentioned.


Results

Weather data was retrieved for a 18 day period, 9 days prior to the sampling and 9 days
during the sampling (Table 1). The weather conditions had 3 distinct stages that were of
importance, the first stage being the 12 opening days, 8 days prior to the sampling plus the 4
initial days of sampling. This period was characterized by dry weather with a remarkably
constant temperature. The second stage was on day 3-5 of the sampling period, during which
the area received intensive rainfall. This event increased the average volume of the pools by
more than 100%, which in turn triggered a number of environmental responses (Table 2). The
third stage, constituting the 4 last days, was characterized again by dry weather and also by a
drop in water temperature and more homogenous conditions among pools. The substantial
change in environmental conditions that was initiated during the second stage remained.

BCC patterns among pools at days 1 and 9 were weakly, but significantly, correlated (Mantel
test, p<0.01, RM = 0.28). Only approximately half of all observed OTUs (47%) were,
however, found at only one of the two sampling occasions. With regard to samples taken for
community analyses of rain and air bacteria only two (one air and one rain) of the six samples
were successful in both the PCR and T-RLFP process. Hence no statistical testing was
possible. When the rain and air sample were included in an NMDS analysis with the pool
samples they were very remote from the cluster of pool samples (Fig. 1).




Table 1 Daily mean values and standard deviation of air and water temperature (C) and amount of precipitation
(mm). Before rain refers to the period 9 days before sampling and at day 1-3 of sampling. During rain refers
to the period between days 3-5 during the sampling period and After rain to the period between days 5-9.
Weather
Before rain
During rain
After rain
Average air temp SD
18.1 1.21
20.4 3.26
16.4 1.62
Average water temp SD
25.2 2.01
23.0 1.16
20.1 0.88
Rain (mm)
2
19.2
0.8




9

---

Table 2. Mean values standard deviations (SD) of environmental variables in the rock pools before and after
the rain period. Before: day 1and day 3. After: day 5 and day 7.
Environmental factors
Sampling before the rain period
Sampling after rain period
Salinity average SD (psu)
2.85 4.58
0.49 0.76
Absorbance average SD (436nm)
71.3 47.6
39.7 21.4
Total-P average SD (g/l)
10.3 13.6
4.73 4.86
Clh-a average SD (g/l)
8.05 14.2
4.4 7.3
Volume average m3 SD
0.14 0.11
0.34 0.25



Results from pRDAs (Table 3) and partial Mantel tests (Table 4) show that initial
environmental conditions, in particular those found at day 1, 3 and 5, often explained more of
the variation in (or were more strongly correlated to) BCC than more recent environmental
conditions, i.e. those observed at days 7 or 9. Without accounting for co-variation, salinity at
day 1 and day 3 had a higher correlation to BCC than salinity at days 5, 7 and 9 has, both in
Mantel and pRDA tests (Tables 3 and 4). With co-variation among sampling days taking into
account, the pattern remained the same although slightly harder to perceive. Salinity at days 1
and 3 had a total of 9 significant correlations with BCC (both Mantel tests and pRDAs) while
days 5, 7 and 9 only had 4 significant correlations in total. Notably, salinity at any sampling
date was never significant when co-variation with days 1 and 3 was taken into account. Days
1, 3 and 5 were on the other hand significantly correlated to BCC in 10 out of 12 cases also
when co-variation of day 7-9 was taken into account (Table 3 and Table 4). Finally, in only 3
out of the 40 tests was the salinity at a specific day significantly influencing BCC when co-
variation to salinity to a sampling point closest to it in either direction of time was included.









Table 3. Results from pRDA where salinity at the 5 different sampling points was correlated to BCC at day 9.
The values are the percentage of variation in BCC explained. The first row is the total variation in BCC
explained by salinity at the individual day. The following rows are the explained variance with co-variance with
other sampling days withdrawn from the degree of explained variance.
Salinity-BCC correlation pRDA Day 1
Day 3
Day 5
Day 7
Day 9
R2a without co-variation
0.12
0.12
0.11
0.10
0.10
R2a day 1 with co-variation
-
ns
ns
ns
0.05
R2a day 3 with co-variation
ns
-
ns
0.04
0.05
R2a day 5 with co-variation
ns
ns
-
0.04
0.04
R2a day 7 with co-variation
ns
ns
0.04
-
ns
R2a day 9 with co-variation
ns
ns
ns
ns
-





10

---