The Western (English) Channel Observatory as a test bed for improving ecosystem forecasts.

Icarus Allen, Tim Smyth et al. PML

www.westernchannelobservatory.org.uk

‘Growing concern about human influence on marine ecosystems conflicts with our inability to separate man-made from ‘natural’ change. This limitation results from lack of adequate baselines and uncertainty as to whether observed changes are local or on a broad scale. Long-term monitoring programmes should be able to solve both these deficiencies’ (Duarte et al, 1992. Nature)

Sustained Observations in the Western English Channel: past, present and future.

Plymouth time series since 1900

Geographical Region

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

Overall aims and purpose Our purpose is to integrate in situ measurements made at stations L4, L5, E1 and adjacent coasts in the western English Channel (see Fig 3) with ecosystem modelling studies and Earth observation; this will be facilitated by web-based GIS technology. which allows the following science questions to be addressed at a range of temporal and spatial scales:

•What is the current state of the ecosystem?

•How has the ecosystem changed? •Improve short term forecasts of the state of the ecosystem.

• A national facility for EO algorithm development, calibration and validation.

Remote Observatory

Virtual Observatory

In-situ sampling (L4, E1, L5, buoy, etc.)

long-term time-series (linked to non WCO series via MECN)

scientific investigation

Remote Sensing

SST, Ocean Colour

Other sensors

Modelling

ERSEM

Met Office (NCOF)

Data

Database (SQL)

Web (Webmap server)• each element has strengths and weaknesses – synergy.

MECN: Knowledge Transfer / policy advice

In-Situ Sampling

In situ sampling

i) Marine measurements:

OPERATIONS: weekly sampling @ L4; fortnightly @ E1

• The Observatory consists of the core measurements:

• Hydrography (CTDf);

• Nutrients;

• Optics;

• Pigments;

• Zooplankton and phytoplankton

ii) Atmospheric measurements:

• meteorological stations (PML, Rame Head)

• sun photometric aerosol retrievals (PML)

Latest HydrographyThe year so far (L4) …

latitude: 50.6750.67°°NNlongitude: 4.584.58°°WW

start: 19881988sample frequency: weeklyweekly

abundance: 1988- on going1988- on goingbiomass: 1993-19981993-1998copepods eggs production: copepods eggs production: 1992-20051992-2005

L4L4

L4 - Zooplankton time-seriesL4 - Zooplankton time-series

vertical net hauls: from the sea floor (~55m) to the surfaceWP2 net: mesh 200µm (UNESCO 1968)

samples are stored in 5% formalin

taxonomic identification

zooplankton database

data analysis

Composition of the “Top20” dominant species at L4 (Plymouth) 1988-2006

Sagitta sp.

90%90%

Array of autonomous moorings• An observatory needs to directly observe something;

• Need real-time data (rather than just RS / modelling);

• Capital bid successful: currently specifications out to tender …

• Have permissions for moorings at L4 and E1.

E1:Moored (profiling?) buoy

CTDf, Optics, Nutrients …

L4:Moored profiling buoy

CTDf, Optics, Nutrients … expandable for visitors?

Rame Head:(shore node)

Met. Station; Aerosols;

PML:(base node)

Met. Station; Aerosols;

Remote Observatory

Remote Sensing

• SST (1981 - ) and ocean colour (1997 - ) to provide synoptic overview of WCO domain.

• Delivered via web and Web Map Server technology.

Tidal mixing

frontal region

stratified

frontal region

case I

case IIAutumn bloom

AVHRR SST26 Aug - 01 Sep 02-08 Sep 9-15 Sep

16-22 Sep 23-29 Sep

NRT data provided in partnership with NEODAAS PML.

MODIS chlorophyll-a26 Aug - 01 Sep 02-08 Sep 9-15 Sep

16-22 Sep 23-29 Sep

NRT data provided in partnership with NEODAAS PML.

Virtual Observatory

Western English Channel Model

7km Western Channel POLCOMS-ERSEMPML-delayed 7 day Hindcast 2002-pres

2km under-development, 500m regional sub-model in vicinity of L4/E1 proposed

Hetero-trophs

Bacteria

Meso-Micro-

Particulates

Dissolved

Phytoplankton

Consumers

Pico-fDiatoms

Flagell-ates

NO3

PO4

NH4

Si

DIC

Nutrients

Cocco-liths

Meio-benthos

AnaerobicBacteria

AerobicBacteria

DepositFeeders

SuspensionFeeders

Detritus

NutrIents

OxygenatedLayer

Reduced Layer

RedoxDiscontinuity

Layer

AtmosphereO2 CO2 DMS

3D

IrradiationWind Stress

Heat Flux

0D

Cloud Cover

Riv

ers

and

boun

darie

s

1D

Forcing

Marine System Model: ERSEM

Ecosystem

Physics

GOTM

POLCOMS

UK

MO

Hetero-trophs

Bacteria

Meso-Micro-

Particulates

Dissolved

Phytoplankton

Consumers

Pico-fDiatoms

Flagell-ates

NO3

PO4

NH4

Si

DIC

Nutrients

Cocco-liths

Meio-benthos

AnaerobicBacteria

AerobicBacteria

DepositFeeders

SuspensionFeeders

Detritus

NutrIents

OxygenatedLayer

Reduced Layer

RedoxDiscontinuity

Layer

AtmosphereO2 CO2 DMS

Hetero-trophs

Bacteria

Meso-Micro-

Particulates

Dissolved

Phytoplankton

Consumers

Pico-fDiatoms

Flagell-ates

NO3

PO4

NH4

Si

DIC

Nutrients

Cocco-liths

Meio-benthos

AnaerobicBacteria

AerobicBacteria

DepositFeeders

SuspensionFeeders

Detritus

NutrIents

OxygenatedLayer

Reduced Layer

RedoxDiscontinuity

Layer

AtmosphereO2 CO2 DMS

3D

IrradiationWind Stress

Heat Flux

0D

Cloud Cover

Riv

ers

and

boun

darie

s

1D

Forcing

Marine System Model: ERSEM

Ecosystem

Physics

GOTM

POLCOMS

UK

MO

TruthT

OResidualf(O-P)

Observation

Prediction

PredictiveError

Observational Error

Validation and verificationRelationships between model and data(adapted from the ideas of Dan Lynch)

Predictiveuncertainty(e.g. numerical error, parameter uncertainty)

Observationalaccuracy, (e.g. measurement error, range of replicates etc.)

Data assimilation isthe art of reducing this distance

P

TruthT

OResidualf(O-P)

Observation

Prediction

PredictiveError

Observational Error

Validation and verificationRelationships between model and data(adapted from the ideas of Dan Lynch)

Predictiveuncertainty(e.g. numerical error, parameter uncertainty)

Observationalaccuracy, (e.g. measurement error, range of replicates etc.)

Data assimilation isthe art of reducing this distance

P

Development of Model Metrics

x

xx

x

xx

x

x

x

xx

x

x

Model

Data

xx

x

x

x

x

x

t1

t1

TN

TP

FP

FN

x

xx

x

xx

x

x

x

xx

x

x

Model

Data

xx

x

x

x

x

x

t1

t1

TN

TP

FP

FN

Dsicrimination Analysis Taylor Diagram

How well does ERSEM capture the seasonal succession at L4?

-100

-80

-60

-40

-20

0

20

0 100 200 300 400 500 600 700

Day 2003-04

PC

1

L4 Model

Multivariate Validation: Comparison of PC1 for modelled and observed phytoplankton

0

1

2

3

4

5

6

7

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

r2

σD

/σM

T(surface)

Dinoflagellates

S (surface)

PicophytoplanktonDiatoms

Chlorophyll

S (dmean) Flagellates

Bacteria

Silicate Phosphate Nitrate T (dmean)

Assessing short term forecast skillLike with Like comparison

Log Chlorophyll

Monthly Mean Chlorophyll

Data Delivery

http://www.npm.ac.uk/rsg/projects/mceis/zx/metoffice_model/?navbar_Nav=open&navbar_Current=open&navbar_CurSat=open&navbar_Cur1AVHRR=closed&navbar_Cur1SeaWiFS=closed&navbar_Cur1MODIS_Aqua=open&navbar_Cur1MODIS_Aqua_NASA=closed&navbar_Cur1MERIS=closed&navbar

Hosting of Ecosystem parameters (NCOF box)

Summary

IPCP

CPPPV

If the model predicts a bloom what’s the probability it is correct?

Positive Predictive Value A. PPV for threshold > mean chlorophyll, B. PPV for threshold > 1.5 mean chlorophyll for that

pixel, Allen et al Harmful Algae (in press)