Enhancement of primary production at greater resolved scalesResults from the Greenseas project

W McKiver, M Vichi, T Lovato, A Storto, S Masina

45th International Liège Colloquium 13 – 17 May 2013

Liège, Belgium

The greenseas project

9 partners, led by NERSC, Bergen

GreenSeas employs a combination of observation data, numerical simulations and a cross-disciplinary synthesis to develop a high quality, harmonized and standardized plankton and plankton ecology long time-series, data inventory and information service

Questions and aims

• Global plankton data are sparse and the usage of biogeochemical data for model assessment raises several questions:

• Do in situ data have enough signal to allow extrapolation?• What are the limits of model assessment given the available

data?• Are rates and process measurements more evanescent than

stock data?• Can we efficiently use mesoscale features from satellite

products?• On the modelling side, advances in computational technology has led to more complicated models at greater spatial and temporal resolution. More details can be produced but the production costs are high.• Here we focus on a direct comparison between a OBGCM used at two different resolutions: LO-RES – 2 degree resolutionHI-RES – ¼ degree resolution

The Model: PELAGOS

PELAgic Biogeochemistry for Global Ocean Simulation (Vichi et al., 2007a,b; Vichi and Masina 2009)

Global ocean implementation of a coupling betweem: Biogeochemical Flux Model (BFM): Biomass based continuum description of lower trophic levels through a set of differential equations that solves the dynamical stoichiometry of fluxes of C, N, P, Si and Fe among selected biological functional groups NEMO Ocean Model (v3.4): Primitive equations for momentum, temperature, salinity with LIM2 Seaice model, (Madec et al., 1998)

Model is implemented on ORCA grid at both 2 degree and ¼ degree resolution with the same biogeochemical parameterizations

• Stoichiometric biomass-based model, with a unified theory built on the concept of Chemical Functional Families

• Allows to describe lower trophic levels by implementing any number of functional groups and constituents

• Standard pelagic setting:• C,N,P,Si,Fe,O,Alk• 3 phytoplankton

groups• 3 zooplankton groups• 1 bacterioplankton

•Open source code online:http://bfm-community.eu

Biogeochemical Flux Model (BFM)http://bfm-community.eu

Biogeochemical Flux Model (BFM)http://bfm-community.eu

Some theory: Vichi et al. 2007 (JMS)

ORCA2 vs ORCA025

Horizontal Grid 182 x 149 1442 x 1021Vertical Levels 31 50Time Step 96 mins 18 mins

• BFM: 57 Pelagic state variables with full diagnostics• Large computing power ~ 900 cores. Large memory requirements for HI-RES, 850 GB• Large storage requirement, approx 18 GB output per time unit

Experimental setup • Simulations performed at ORCA2 (LO-RES) and ORCA025 (HI-RES) resolution with same atmospheric forcings from ERA-interim (on-line interpolation)

• LO-RES model run for 30 years

• HI-RES physics run for 4 years, then coupled to the biogeochemical model using the biogeochemical variables interpolated from the LO-RES experiment to initialize. January and June initializations.

• Then both LO- and HI-RES models are run for 6 months storing every 8 days

• Our results will mainly focus on the Atlantic and Southern Ocean which is are well-known regions with large model biases

• Present first the physical drivers and then examine their impact on the marine biogeochemical system (and related problems!)

Momentum is put into the system through the wind-stress.

Both the LO-RES and HI-RES cases converts a certain amount of this into kinetic energy.

Here we show the ratio of the total average Turbulent Kinetic Energy and the Wind Stress for both the LO-RES (blue) and HI-RES (red) experiments.

The HI-RES experiment has much higher TKE.

Physical Drivers: Mean EKE/Wind Stress

Ratio of Vertical to Horizontal motion

As a qualitative indicator of the relative importance of the vertical motions we show the averaged vertical-to-horizontal velocity ratios.

Overall vertical motions are much stronger in HI-RES case as the mesoscale is resolved

LO-RES HI-RES

Mixed Layer Depth: Seasonal cycle

LO-RES --- de Boyer Montegut et al. 2004 --- HI-RES --- Hosoda et al (Argo) 2010 ---

MLD is much deeper in the HI-RES Particularly strong difference in the Southern Ocean Overall the physics is very different in the two cases, how does this

impact the biology?

North Atlantic Tropical Atlantic Southern Ocean

Seawifs

HI-RES

LO-RES (initial for HR)

LO-RES

Evolution after 3 months (March)

SST [degC] (March)

Net Phytoplankton Growth Rate [mg C/m3/d] (March)

Overall enhancement of coastal production

Surface Chl [mg/m3] (March)

Overall enhancement of coastal chl

Temperature section [degC] (March)

MLD

Net production [mg C/m3/d] and Chlorophyll [mg/m3]

Note the

change of

units!

This decoupling is a consequence of variable chl:C!

Example timeseries (summer to winter) on the APF

LO-RES – 2 deg HI-RES – ¼ deg

Iron cycle [umol/m3] is not balanced at Hi-Res scales

Scale is 5 times larger!

Summary • Performed simulations at two different resolutions• Clearly higher resolution enhances the mesoscale features having an impact on the marine plankton• Enhanced growth in coastal regions, while growth is first enhanced and then suppressed in the Southern Ocean• Not all parameterizations are resolution dependent as biogeochemical features get (visually) better (more physics, more details). However, “loophole” parameterizations as in the Fe case are more dangerous as remineralization parameters are scale dependent

• This is just the modelling side. Assessment phase has started. • Chain of restarts: what happens in the HI-RES beyond the adjustment phase? Is there convergence?• How much resolved scales are needed? Can some intermediate resolution address the cost issue while providing better physics?

Future work

Thanks