Transcript L4 and the CPR ( ppt 4M)

On the use of long term observations for
evaluating a shelf sea ecosystem model
Examples from the
• The Western Channel Coastal Observatory
• The Continuous Plankton Survey
Icarus Allen (PML),
Katy Lewis (PML), Jason Holt (POL), John Siddorn (Met Office), Anthony
Richardson (SAHFOS/CISRO)
Marine System Model: ERSEM
ERSEM
Forcing - key features
Ecosystem
Carbon based
process
Cloud Cover
model
Wind Stress
Atmosphere
O2
CO2
DMS
Irradiation
Phytoplankton
CoccoPico-f
liths
Flagell
-ates
Functional groupHeat
approach
Flux
Resolves microbial loop
Physics
and POM/DOM dynamics
Si
NO3
Diatoms
Particulates
DIC
NH4
Bacteria
PO4
Rivers and boundaries
0D of nutrients
Complex suite
Dissolved
Includes benthic
system
1D
Heterotrophs
Explicit decoupled cycling
of C, N, P, Si and Chl.
Adaptable: DMS, CO2/pH,
phytobenthos, HABs.
3D
Consequently flexible
and applicable to a wide
UK
GOTM
range
of global
MO
POLCOMS
ecosystems.
Micro-
Meso-
Consumers
Suspension
Feeders
D
e
t
r
i
t
u
s
Oxygenated
Layer
Aerobic
Bacteria
Meiobenthos
Anaerobic
Bacteria
Deposit
Feeders
Redox
Discontinuity
Layer
Reduced
Layer
N
u
t
r
I
e
n
t
s
N
u
t
r
i
e
n
t
s
Shelf seas ecosystem hindcast – forecast modelling
Met Forcing NWP
Met Office POLCOMS
12 km Atlantic Margin
Model
Met Office 1/3o
Atlantic FOAM model
T, S,
U, V
7km Western Channel
POLCOMS-ERSEM
PML-delayed 7 day
Hindcast 2002-pres
7km MRCS
POLCOMS-ERSEM
Met Office 7 day
hindcast 2002-pres
T, S, U, V
T, S, U, V
ERSEM
POL/PML
hindcast 1988/89
Western Channel Coastal Observatory
Overall Aims and Purpose:
Our purpose is to integrate in situ measurements made
at stations L4 and E1 in the western English Channel
with ecosystem modelling studies and Earth
observation.
1. What is the current state of the ecosystem?
2. How has the ecosystem changed?
3. Short term forecasts of the state of the ecosystem.
4. The WCO as a National Facility for EO algorithm
development, calibration and validation:
Western Channel Coastal Observatory
Western English Channel:
• boundary region between oceanic and neritic
waters;
• straddles biogeographical provinces;
• both boreal / cold temperate &
• warm temperate organisms
• considerable fluctuation of flora and fauna since
records began.
Southward et al. (2005) Adv. Mar. Biol., 47
Station L4
• Situated 10nm south of Plymouth
• Sampled weekly for physical, biological and chemical
data since 1992.
• Hydrodynamically complex
• Average depth of 50m;
• Classified as a well-mixed tidal station but it exhibits
weak seasonal stratification in summer and is
influenced by the outflow from the River Tamar.
• On some occasions it represents the margin of the
tidal front characteristic of this region (Pingree,
1978).
Complex system so a good test of the model dynamics
Station L4
•
•
•
•
•
•
•
•
The thermohaline structure of the water column was determined with a CTD
probe developed from the Undulating Oceanographic Recorder (UOR) (Aiken &
Bellan, 1990).
Water samples (10m depth) analysed for nitrate, phosphate and silicate
concentrations using standard laboratory colorimetric methods (Woodward &
Rees, 2001).
Chlorophyll-a concentrations, fluorometric analysis with a Turner Design 1000R
fluorometer after extraction in 90% acetone overnight. (Rodriguez et al., 2000)
Phytoplankton is collected at 10m depth and preserved with 2% Lugol’s iodine
solution (Holligan & Harbour, 1977).
Between 10 and 100ml of sample, depending on cell density, were settled and
species abundance was determined using an inverted microscope.
Cell volume and carbon estimates for the microplankton were derived from the
volume calculations of Kovala & Larrance (1966) and the cell volume and carbon
estimations of Eppley et al. (1970).
Zooplankton samples are collected by vertical net hauls (WP2 net, mesh 200μm;
UNESCO, 1968) from the sea floor to the surface and stored in 5% formalin.
(Bacteria and picophytoplankton (the combination of synecoccus bacteria and
picoeukaryotes)) determined using a flow cytometer.
Model –Data Misfit
Model –Data Misfit
Model –Data Misfit
Assessment of overall model
performance.
Taylor plot - L4
7
Dinoflagellates
6
S (surface)
σD/σM
5
4
3
2
Chlorophyll
0
0
Silicate
S (dmean) Flagellates
1
0.1
0.2
Diatoms
Picophytoplankton
0.3
Bacteria
0.4
Nitrate
Phosphate
0.5
0.6
0.7
T (dmean)
0.8
T(surface)
r2
0.9
1
Ba
t
D
ia
Si
it
N
Ph
os
t
er
ia
os
Pi
c
t
o
Ph
y
D
in
Fl
ag
D
ia
Si
ct
er
ia
os
t
o
-7
Ba
-50
-75
-100
-125
Pi
c
-25
N
it
0
Ph
y
25
Ph
os
50
D
in
75
Ch
l
100
Fl
ag
125
f)
ur
(s
n)
m
ea
f)
ur
)
f)
Ch
l
ur
(s
n)
-5
ct
(d
Sa
l
l
Sa
p
ea
n
(s
(d
m
Te
m
p
D  M 

Pbias 
*100
D

Te
m
% Bias
l
)
f)
ur
(s
ea
n
(d
m
ea
Sa
l
Sa
p
m
(d
Te
m
p
Te
m
Model Effciency
1
-1
-3
2

 D  M 
ME  1 
  D  D 
2
-9
-11
-13
Phytoplankton Seasonal Succession
L4 Climatology
L4 Climatology
Mesozooplankton
Multivariate Analysis
all analysis's performed using PRIMER 6
•
MDS (multi-dimensional scaling)
Cluster analysis allows us to check the adequacy and mutual consistency of
both the model and the in-situ data.
A multi-variate ordination technique which can be used to reflect
configurations in the model and in-situ data
A non-metric MDS algorithm constructs MDS plots iteratively by as closely
as is possible satisfying the dissimilarity between samples; dissimilarities
between pairs of samples, derived from normalised Euclidean-distance
matrices, are turned into distances between sample locations on a map.
•
RELATE Test.
A test of ‘no relationship between distance matrices’, essentially a test for
concordance in multivariate pattern.
A correlation between corresponding elements in each distance matrix was
calculated using Spearman’s rank correlation, adjusted for ties (Kendall,
1970).
The significance of the correlation was determined by a Monte Carlo
permutation procedure, using the PRIMER program RELATE.
For the ideal model r =1.
MDS
Data
MDS constructed from
temperature, salinity,
chlorophyll, nitrate,
phosphate silicate, diatom
biomass, flagellate
biomass, dinoflagellate
biomass.
RELATE TEST
r = 0.44, p=0.0001
T,S and Nutrients only
r = 0.55, p = 0.0001
Model
Correlations between variables at L4
Model
model tempm n1p
t
m n1p
-0.457847
mn3n
-0.570095
mn4n
0.046861
m n5s
-0.402593
mchl
0.130871
mp1c
-0.303375
mp2c
0.359215
mp3c
0.245126
mp4c
0.24224
mn3n
mn4n
m n5s
mchl
0.954959
0.529493
0.906356
-0.698287
-0.055811
-0.766679
-0.605957
-0.634571
-0.114804
0.905404
-0.614281
-0.197708
-0.809386
-0.600442
-0.7684
0.280813
-0.183459
0.115604
-0.260506
-0.005784
0.119805
-0.795268
-0.200407
-0.773454
-0.697641
-0.734496
n1p
n3n
n4n
n5s
chl
-0.337629
-0.465216
0.546979
-0.051787
-0.106652
-0.127074
-0.218995
0.212952
0.084967
0.396891
0.140648
0.299217
-0.03518
-0.033792
0.006721
-0.467903
-0.424347
-0.537803
0.634643
-0.328287
-0.262189
-0.196063
-0.461161
-0.508357
-0.090878
-0.054339
-0.06749
0.08752
-0.078324
-0.069027
0.024203
0.111345
-0.104425
-0.425008
-0.368589
mp1c
mp2c
mp3c
mp4c
Nitrate control in model to strong?
0.372556
0.735123 0.480541
0.405182 0.054098 0.485834
0.734557 0.597562 0.832742 0.552918
Data
temp
temp
n1p
n3n
n4n
n5s
chl
p1c
p2c
p3c
p4c
p1c
p2c
p3c
p4c
0.887541
0.175688 0.02293
0.119782 0.016338 -0.049943
0.511074 0.425981 0.148858 0.847533
RELATE test between these data sets indicates a statistically significant similarity between the matrices
r = 0.53 p=0.012
i.e. model explains ~ 28% of observed correlations
Summary
Model does well reproducing temperature and has some
skill for nutrients, but phytoplankton must be
improved before any confidence can be had in the
model ability to forecast.
The model does not accurately simulate the timing of the
spring bloom and further work is required to assess
whether the causes of this are hydrodynamic, optical
or physiological.
Issues with model
a) Salinity and hence water column structure /
turbulence
b) Grazing pressure
c) Nitrogen dynamics in phytoplankton
d) Dinoflagellate dydnamic incorrect (lack of motility /
heterotrophy?)
e) Optics
Qualitative Validation
Model validation with
plankton abundance
K. Lewis et al., Error quantification of a high resolution coupled hydrodynamicecosystem coastal-ocean model: Part3, validation with Continuous Plankton Recorder
data, Journal of Marine Systems (2006), doi:10.1016/j.jmarsys.2006.08.001.
Continuous Plankton
Survey
www.sahfos.ac.uk
The aim of the CPR Survey is to monitor the near-surface plankton of the
North Atlantic and North Sea on a monthly basis, using Continuous
Plankton Recorders on a network of shipping routes that cover the area.
Resolving shifts
in species distributions
Zooplankton species
geographical shift
We need to be able to model this to
understand how climate will affect
marine bioresources
•
Simulated ‘tows’ were performed by extracting biomass data from
archived model
•
Due to the semi-quantitative nature of the CPR, data for each individual
tow of both the CPR and corresponding model output were standardised
to a mean of zero and a unit standard deviation (σ) of the relevant data
to produce a dimensionless z-score.
•
This allows a direct qualitative comparison of model biomass with
discrete survey counts.
Domain-wide daily mean values
for all CPR samples and
corresponding model output were
used to compare the magnitude
and timing of the behaviour of
the biological variables over the
two-year period.
Summary seasonal cycles
Total Phytoplankton
Total Copepods
Spatial Distribution of Errors
Total Phytoplankton
% Model results month by month that differ from the CPR
samples by less that 0.5 SD from the mean in 1988
Summary
•
Simple linear regression and absolute error maps provide a qualitative
evaluation of spatio-temporal model performance
•
z-scores indicate model reproduces the main pelagic seasonal features
•
good correlation between magnitudes of these features with respect to
standard deviations from a long-term mean.
•
The model is replicating up to 62% of the mesozooplankton seasonality
across the domain, with variable results for the phytoplankton.
•
There are, however, differences in the timing of patterns in plankton
seasonality.
•
The spring diatom bloom in the model is too early, suggesting the need
to reparameterise the response of phytoplankton to changing light
levels in the model.
•
Errors in the north and west of the domain imply that model turbulence
and vertical density structure need to be improved to more accurately
capture plankton dynamics.
General Conclusions
• Long-term time series observations are important
resources for the assessment of model performance;
they can be used to highlight errors in model
hindcasts, which can subsequently be improved.
• These types of analysis are only possible because of
the existence of large self-consistent data sets.
Unfortunately, such data sets are relatively rare and
a concerted effort is required to collate existing
data sets into model friendly formats, collect new
ones and make them readily available.
• L4 is situated in a hydrographically complex region
therefore it provides a substantial test of model
ability, however for the model to be evaluated more
extensively it is essential to perform these tests
over a wider spatial scale.
Advances in Marine Ecosystem Modelling Research
• Workshop on ‘validation of global ecosystem models
(4-6th Feb 2007)
• Workshop on ‘Ocean Acidifcation’ (11-13th Feb 2007)
• Both workshops to be held in Plymouth, register
online at www.amemr.info by 17th November.
• AMEMR II is scheduled for June 2008.
A coherent ecosystem approach.
In-Situ Data
Earth Observation
Meteorological Station
3D Ecosystem Modelling
Web based data delivery systems