Intercomparison and integration of plankton data

1
Intercomparison and integration of plankton
data Sonia Batten
2
Calibration means comparing against a known
standard. Not possible in the plankton
world! What we really want to do is compare and
integrate plankton data from different
samplers Scale is a major issue
3
Size zooplankton range from micro to mega
(ciliates to jellies), no one sampler can sample
this range.
4
If we just consider the meso zooplankton there
are numerous ways to sample. The 2000 ICES
zooplankton methodology manual lists Simple
nets e.g. WP2, SCOR, NORPAC Multiple nets -
CPR, LHPR, Gulf V, BIONESS, MOCNESS,
RMT18 Electronic optical/acoustic OPC, VPR,
ADCP to name just a few..
5
Scale of distribution - Zooplankton exist in
4D Some samplers, deployed repeatedly, can
sample 3D Changes in the horizontal plane and
through time can be dealt with by increasing
deployments (if you have money) Changes in the
vertical plane are more complex.
6
There are samplers that operate at a fixed
depth, over an integrated depth, and
depth-resolving samplers But, abundance of a
species and abundance of species BOTH change with
depth.

7
In the Southern Ocean, species richness increases
with depth and so the CPR did not sample many
species found in the top 150 m
From Hunt and Hosie, 2003
8
Within a species there are ontogenetic and diel
changes in depth distribution. Some species have
stage-specific depth distributions
From Kobari and Ikeda, 2001
9
Plus, species may also have a distribution
related to chlorophyll maxima, thermoclines or
the surface mixed layer which are not at a
constant depth and may be hard to predict So,
integrating data from samplers which sample
different depths will be difficult Nevertheless
.
10
Several studies have taken place which have
compared the CPR to other zooplankton
samplers CPR replacements Vertical
profilers Standard nets
11
CPR Comparison studies
Study Area Published UORCPR North Sea
Aiken et al.,(77) Williams Lindley
(80) UTowCPR Baltic Batten et al
(03) LHPRCPR N. Atlantic Richardson et al
(04) WP2CPR North Sea Clark et al
(01) WP2CPR English Channel John et al
(01) WP2CPR Baltic Batten et al
(03) NORPACCPR S Ocean Hunt Hosie
(03) BongoCPR N Pacific Batten (unpub)
12
Were comparison criteria met in these
studies? Study Mesh Depth Time/space UO
RCPR, North Sea ? X ? UTowCPR,
Baltic X ? LHPRCPR, N. Atlantic ? ? X WP2
CPR, North Sea X ? X WP2CPR, English Channel
X ? ? WP2CPR, Baltic X ? ? NORPACCPR, S
Ocean ? ? ? BongoCPR, N Pacific X X ?
13
Most studies concluded that the CPR under
samples, but that this is SPECIES DEPENDENT
Example 1 - Richardson et al., 2004
Time series of abundance from the N Atlantic
14
Example 2. Batten et al., unpub.
CPR v Bongo for adjacent samples in NE pacific. A
11 line is also shown
15
Not a mesh size effect (larger species were
undersampled) Thought to be depth or avoidance
related.
16
On the positive side, these studies also showed
that 1. Seasonal cycles were well replicated
John et al., 2001 Clarke et al., 2001 English
Channel North Sea
17
Example 2, Richardson et al., 2004 NE Atlantic
2. Interannual patterns were also generally
well-replicated
18
3. And despite differences in absolute abundance,
often the same community is sampled
Hunt and Hosie, 2003, Southern ocean
19
The 20 most abundant taxa in the sampling area
(west coast Vancouver Island) according to each
sampling device (12 in common)
20
Re-cap on comparisons Each sampler has
biases. Need to account for mesh and depth
differences of sampler Need to consider each
species on a case by case basis Comparisons of
seasonal cycles or long term trends may be more
robust
21
Data Integration New samplers MUST be
quantitative We need to work on data from
existing samplers to define sources of error and
convert to quantified data
22
For each sampler quantification depends on
knowing Volume sampled Avoidance
characteristics active (depends on species) or
passive (depends on sampler) Counting
methodology e.g. sub-sampling
23
Quantifying CPR data abundance traditionally
quoted as per sample and assumed 100
efficiency 3m3 But, clogging of silk reduces
efficiency (and effective mesh size) EM
flowmeters fitted and showed great variability
0.4-2.8 m3 on one tow (Walne et al, 1998)
24
However, John et al. 2002, showed that the mean
was 3.11m3 and when large numbers of samples
averaged, effect of clogging on filtration rate
not large. Previous results still stood and
could be recalculated based on multiple
regression of plankton abundances v flow But,
data integration may use small number of samples
so do need to know volume sampled
25
Jonas et al (2003) found flow decreased when ship
speed increased (which it has over the
years) as did Hunt and Hosie (in press)
in SO study. They also concluded that the
flowmeter itself may have caused this
relationship (heavy tail end)
26
In any case effect of clogging was more
pronounced than ship speed (60 reduction in flow)
Hunt and Hosie in press Predicted (100
efficiency) Actual (flowmetre data)
27
They call for small flowmeters to be fitted to
all CPRs Nets also have issues with recording
flows e,g McKinnel and Mackas (2003) If
wire-out and wire-angle are used to estimate flow
can be affected by bad weather (when vertical tow
becomes more horizontal) Flowmeters in the mouth
are more accurate but are sometimes mis-read, or
spin in the wind before/after tow
28
Avoidance Hard to quantify most parameters not
known Each sampler has its own bias
hydrodynamics, aperture size, colour all
influence avoidance Each species has a different
escape response, not necessarily related to its
size Day/night and ambient light levels will
also influence avoidance, so even within a
sampler there may not be consistency
29
Counting methodology CPR analysis is done
on-silk (except in Southern Ocean) Small,
transparent organisms are harder to see (e.g.
Oithona) and may be undercounted How to compare
with off-silk analysis? Category system of
counting is accurate for large areas, less so on
an individual sample basis.
30
Making recommendations for new samplers Recognis
e that people will usually not interrupt an
existing time series to switch methods New
studies may be restricted by cost or equipment
availability (esp. in developing countries) so
old equipment will be used for a while to
come All things being equal, people will choose
the best sampler for the study and data from
other samplers may not be so appropriate
31
An alternative approach use Data
products e.g. Mackas, Batten and Trudel (in
prep) Net samples from Stn P and BC shelf
32
Peak biomass of Neocalanus plumchrus
33
Related to sea surface temperature
34
CPR sampling 2000-2004
35
Peak biomass of Neocalanus plumchrus from stage
composition
36
In this approach 2 datasets, 1 spatially
restricted but lengthy time series, 1 wide
geographic coverage but short time series So
long as each data set is internally consistent,
absolute abundances are irrelevant. Each
dataset complements the other and integrating the
data products gives much more information than
for one alone.
37
Conclusions It is very difficult to find
existing data that can be successfully and easily
merged Sampling depth and depth distributions
are variable Species specific responses to
sampling Unquantified data New samplers must be
quantitative and we must work with old/existing
samplers to quantify data Use of data products
should also be considered