marine ecosystem analysis and prediction task team (meap tt)godae-data/oceanview/events/... ·...
TRANSCRIPT
Co-chairs since fall 2014:
Katja Fennel & Marion Gehlen
Marine Ecosystem Analysis and Prediction
Task Team (MEAP TT)
GODAE OceanView COSS TT mtg – Madrid, Sept 20, 2018
GOV MEAP TT
Focus on developing the underpinning science and tools which
will eventually enable full integration of biogeochemistry and
(simplified representations) of ecosystems (end-2-end coupling)
in existing physical operational systems.
Mission: contribute to
improving of the biogeochemical forecasting systems,
improving and developing of assimilation schemes for
biogeochemical observations,
modelling of Essential Biogeochemical Variables (EBV),
bridging the gap with end-users
Where we came from: Status of biogeochemical (pre-)
operational applications in 2014
Table 2 from Gehlen et al., JOO (2015)
Pre-operational and operational systems with biogeochemical component
Satellite chl-a is the only observable presently used for operational assimilation and validation.
It is insufficient for constraining and validating biogeochemical models
30% error Photo-acclimation
Models are:
•highly non-linear,•heavily parameterized,•fundamentally unconstrained in most cases,•typically resolve different plankton functional groups/size classes.
There is no consensus on “right” level of complexity.
Depends on science questions/intended purpose of model application.
30% error Photo-acclimation
Attempts to address these errors:• by making remote sensing reflectances model variables for
direct assimilation,• assimilating functional phytoplankton groups derived from
ocean color
GOV MEAP-TT Workshop
Dalhousie University, Halifax, Canada
23-24 June 2015
sponsored: MEOPAR & GODAE/GOV
> 40 participants 37 abstracts submitted
GOV MEAP TT
Focus areas (refined)
● Assessment of methods for DA on biological/BGC
predictability
● Assessment of techniques for downscaling for one-way or
two-way coupling
● Assessment of dependence of model skill on biological/
biogeochemical model complexity with emphasis on model
portability and predictive skill
● Articulation of biological/biogeochemical OO products with
respect to end-user needs in specific regions
GOV MEAP TT
Focus areas (refined)
● Assessment of methods for DA on biological/BGC
predictability
● Assessment of techniques for downscaling for one-way or
two-way coupling
● Assessment of dependence of model skill on
biological/ biogeochemical model complexity with
emphasis on model portability and predictive skill
● Articulation of biological/biogeochemical OO products with
respect to end-user needs in specific regions
When validating models or performing DA, observations/products need to be quantitatively compared to models, but:
• often observed quantities are not directly comparable (then we need a way to map one onto the other), and
• uncertainties of observation/products need to be known.
State estimation examples
Assimilation of satellite-derived functional phytoplankton groups (Brewin et al. 2017, Ciavatta et al. 2018)
Direct assimilation of ocean color (Jones et al. 2016)
Assimilation of OC PFT data into models
To improve the simulation of marine ecosystem models
Assimilation of plankton functional types
Nanopl
Ddia
tom
sD
inoflag
A - Data [mg m-3] B - Data error [mg m-3] C - Model [mg m-3] D - Difference [mg m-3]
Dia
tom
s
Din
ofl
agel
late
s
Nan
op
lan
kto
n
Pic
op
lan
kto
n
To
tal
1
2
m2
0
m
Pic
opl
Brewin et al, Front. Mar. Sci., 2017 Ciavatta et al, J. Geophys. Res , in review
Ocean-colour
chlorophyll
Model
Jones et al., Biogeosciences, 2016
Addition of spectral optical model to biogeochemical model allowed direct assimilation of remote sensing reflectances (as opposed to derived chl product).
GOV MEAP TT
Focus areas (refined)
● Assessment of methods for DA on biological/BGC
predictability
● Assessment of techniques for downscaling for one-way or
two-way coupling
● Assessment of dependence of model skill on biological/
biogeochemical model complexity with emphasis on model
portability and predictive skill
● Articulation of biological/biogeochemical OO products with
respect to end-user needs in specific regions
GOV MEAP TT
Focus areas (refined)
● Assessment of methods for DA on biological/BGC
predictability
● Assessment of techniques for downscaling for one-way or
two-way coupling
● Assessment of dependence of model skill on biological/
biogeochemical model complexity with emphasis on model
portability and predictive skill
● Articulation of biological/biogeochemical OO products with
respect to end-user needs in specific regions
Joint DA & MEAP TT
2016 University of Santa Cruz
July 11-13
> 50 participants, 43 abstracts
Assessment of methods for DA on biological/biogeochemical predictability:
Study on joint physical/biological DA: Yu et al., Ocean Modelling (2018)
Ongoing in coordination with DA TT: review article being written
Example from U.K. Met Office global analysisChlorophyll fields degraded by updates to model physics
Kindly provided by David Ford, U.K. Met Office
Candidate obs. for 3D models: SSH, SST, T&S profiles, ocean color
Options:
A) Assimilate phys. obs. to correct model physicsExpectation: accurate physics should improve bgc dynamics.
B) Assimilate chl obs. to correct model bgcImplicit assumption: physics is good enough.How much information about the bgc state is contained in obs.?
D) Assimilate phys. & bgc obs. Expectation: Should be better than option C, no?
C) Assimilate phys. (or bgc) obs. to correct physics & bgc stateImplicit assumption: phys. obs. contain info about bgc state
and vice versa.Can this work?
Often not the case in practice.
Liuqian Yu’s twin experiments
free run“truth”
Free run has biased physics and biology.
B2B: chl to update NPCZB2PB: chl to update T & NPCZ
B2B+Npr: chl+N prof. to update NPCZB2PB+Npr: chl+N prof. to update T & NPCZ
P2P: SSH, SST to update TP2PN: SSH, SST to update T&N
SSH, SST, T&N prof. to update T & NPCZ
Updating density structure without updating nutrient distribution or vice versa breaks correlation and will degrade simulation.
Same applies to other bgc properties (phosphate, iron, oxygen, DIC and alkalinity distributions).
3D biogeochemical state estimation is observation-limited at present.
Physical and biogeochemical model states have to be updated simultaneously, otherwise internal property relationships, and hence overall model skill, are degraded.
The multivariate nature of assimilation algorithms allows information to propagate, e.g. themocline and nutriclinelocations, if the correlations are known.
Subsurface information and broader suite of observables is needed (surface chl is not enough).
Ensemble-based biogeochemical state estimation
2017:
• Special Session at 2017 EGU GA “Recent advances in analysis and prediction of marine biogeochemistry and ecosystems” (convenors: Gehlen, Fennel, Dutkiewicz (IOCCG))
• TT meeting during EGU GA in Vienna: explored possibility for a WS on down-scaling to be organized in 2018 or 2019 as a multi-TT event (liaise with COSS-TT).
• Continue interactions with other TT (esp. DA, IV, COSS)
MEAP Activities for Broader Community Outreach
(1) Co-chairs participated in BioArgo planning workshop and drafting of the science plan; Katja Fennel is member of the Scientific Steering Committee
(2) We are providing input on IOCCP activities around EOVs; Marion Gehlen attended the Implementation of Multi-Disciplinary Sustained Ocean Observations (IMSOO) Workshop in Miami in February 2017
(3) Actively pursuing information exchange and potential coordination with IMBER (intention of information exchange with MEAP was articulated at the last IMBER SSC meeting)
Status 3 years after JOO paper:
● biogeochemical applications to date largely focused on
marine productivity BUT limited by data streams
● ocean carbon cycle application still exploratory
ALSO limited by data streams
● applications targeting hypoxia and marine resource
management emerging
● no application targeting OA (to our knowledge)
1) Integrated physical/BGC model systems
we are moving towards:
(i) global ocean at high spatial resolution
(ii) novel biogeochemical/biological data streams for
assimilation and evaluation
Challenges:
Enhance/develop capability for biogeochemical forecasting
and develop capability for ecological forecasting: move
from productivity to food web (LMR)
Develop multi-objective assimilation capacities: state and
parameter estimation, remote sensed data and in situ (2D
and 3D)
Future challenges (next 10 years):
2) Regional applications
(i) taylored to the specific needs of the system (e.g.
complexity of biological/ecological components)
Challenges:
spatial coherence of systems: seamless interfacing,
global/regional
develop models for the land-ocean interface: nearshore
and estuarine models (connect structured and
unstructured grids?)
Future challenges (next 10 years):
3) Applications, user needs:
often regional and with high degree of specificity
Challenges:
build capacity for improved hypoxia, pH and C cycle
monitoring and forecasting
develop products for the forecasting of habitat (LMR),
avoidance of by-catch (turtles, mammals), HABs etc.,
pollution events …
develop products in support to marine ecosystem
management (MPA), risk assessment …
Future challenges (next 10 years):