Use of ocean colour (GlobColour) data for operational oceanography Rosa Barciela, NCOF, Met Office Thanks to Matt Martin (Met Office) and John Hemmings (NOCS)

The Talk Coupled physical-biogeochemical operational models Use of ocean colour data: validation and data assimilation - What are the aims? - What tools are we using? - What have we developed so far? - Assimilation of satellite-derived chlorophyll - What will we be doing next? - What can we do as champion user of GlobColour ?

The Talk What are the aims? What tools are we using? What have we developed so far? Assimilation of satellite-derived chlorophyll What will we be doing next? What can we do as champion user of GlobColour?

What are the aims? This work is part of the Centre for observation of Air-Sea Interactions and fluXes (CASIX), a UK project. The primary goal of CASIX is to quantify accurately the global air-sea fluxes of carbon dioxide. More accurate knowledge of the ocean biology is also required for: water clarity predictions. improvement of light attenuation estimates: SST, MLD, sea-ice. the Royal Navys ability to minimise risks to the maritime environment when deploying active sonar systems. supplying boundary conditions for the Shelf Seas system.

The Talk What are the aims? What tools are we using? What have we developed so far? Assimilation of satellite-derived chlorophyll What will we be doing next? What can we do as champion user of GlobColour?

What tools are we using? –FOAM Forecasting Ocean Assimilation Model –HadOCC Hadley Centre Ocean Carbon Cycle Model Coupling together two models …

Forecasting the open ocean: the FOAM system Operational real-time deep- ocean forecasting system Daily analyses and forecasts out to 6 days Low resolution global to high resolution nested configurations Relocatable system deployable in a few weeks Hindcast capability (back to 1997) Assimilates T and S profiles, SST, SSH, sea-ice concentration FOAM = Forecasting Ocean Assimilation Model Real-time data Obs QC Analysis Forecast to T+144 NWP 6 hourly fluxes Automatic verification Product delivery Input boundary data Output boundary data

Operational configurations 12km (1/9º) Mediterranean 6km (1/20º) North East Atlantic 36km (1/3º) North Atlantic and Arctic 12km (1/9º) North Atlantic 1º Global 36km (1/3º) Indian Ocean 12km (1/9º) Arabian Sea 27km (1/4º) Antarctic All configurations run daily in the operational suite

Hadley Centre Ocean Carbon Cycle model HadOCC is a NPZD (plus DIC and alkalinity) biogeochemical model used at the Hadley Centre for climate studies. HadOCC has been coupled (on-line) within the FOAM system. Initial tests have been run with 1˚ global, 1/3˚ NA and Arctic and 1/9˚ NA FOAM configurations. Palmer, J.R. & Totterdell, I.J. (2001). Deep-Sea Research I, 48,

The Talk What are the aims? What tools are we using? What have we developed so far? Assimilation of satellite-derived chlorophyll What will we be doing next? What can we do as champion user of GlobColour?

FOAM-HadOCC at 1º & 1/3 º resolutions, Mar 27 th 2003 pCO 2 (ppm)Chlorophyll (mg m -3 ) 1º Global 1/3º NA & Arctic

Validation of FOAM-HadOCC results Validation of surface chlorophyll against SeaWiFS data Daily mean North Atlantic fields for 20 th April º Global 1/3º North Atlantic & Arctic 1/9º North Atlantic SeaWiFS 5-day composite

The Talk What are the aims? What tools are we using? What have we developed so far? Assimilation of satellite-derived chlorophyll What will we be doing next? What can we do as champion user of GlobColour?

Observations SeaWiFS data processed at the University of Plymouth: derived chl (GSM) For each observation, an estimate of the error is also provided. Data assimilation schemes generally assume observations to have Gaussian error statistics. However, chlorophyll obs do not have this property. To get around this problem, the data is converted into observations of log10(Chl) which has been shown to then have approximately Gaussian behaviour.

3D analysis Chlorophyll data assimilation scheme A 2D analysis of log 10 (Chl) is performed using the same method as for SST (OI-type scheme). This uses the error statistics described in the previous slide. The output from this is a field of surface log 10 (Chl) increments. These can then be converted into surface phytoplankton increments using the models N:Chl ratio. In order to start the model from a balanced state, increments to the other ecosystem model variables are calculated using a scheme jointly developed by NOCS and Met Office (next slide). The analysed ecosystem model variables are then used directly as the starting conditions for the next model forecast. 2D analysis of log(Chl) 2D analysis of P ΔNΔN ΔPΔP ΔZΔZ ΔDΔD Δalk ΔDIC Model forecast N:Chl Observations

Chlorophyll data assimilation scheme Two stage analysis scheme: Model chl vs. satellite obs: increments (ACS) Balancing increments to biogeochemical variables Increments constrained to conserve total nitrogen & carbon at each grid point (if sufficient nitrogen is available) Surface increments applied to mixed layer. Nutrient-profile correction increments below mixed layer. Hemmings, Barciela and Bell (2007). Accepted by JMS. Increments to other pools (N, Z, D, DIC, Alk) depend on the likely contributions to phytoplankton error from errors in growth and loss

Phytoplankton (mmol N/m 3 )Zooplankton (mmol N/m 3 ) Detritus (mmol N/m 3 )Nutrients (mmol N/m 3 ) Control - truthAssimilation - truth 3-D Twin experiments: daily mean RMS errors in the North Atlantic Total DIC (mmol C/m 3 ) Air-sea exchange of CO 2 significantly improved after assimilating ocean colour data Joint assimilation of Medspiration SST and ocean colour is desirable as carbon solubility is strongly dependent on temperature Free run BDA run

Real world experiments – annual mean No biological assimilation With biological assimilation Phytoplankton Nutrients

Real world experiments Global average RMS (solid lines) and mean (dashed lines) errors compared to the satellite chlorophyll data. Green – no data assimilation Black – with physical data assimilation Red – physical and biological assimilation

FOAM-HadOCC run from Jan 2003 to Jan 2005 Inter-annual variability N 27.5 W 47.5 N 27.5 W Chl da has large impact on chl and other biological compartments Chl da wipes out seasonal variability Smoothing in chl assimilation or variability not present in obs? Red: Chlorophyll Blue: Nutrient Solid line: physical da only Dashed line: chl + physical da

An ocean colour data assimilation scheme has been designed and implemented within FOAM-HadOCC. Initial identical twin experiments seem to indicate that the scheme has potential. Real-world experiments show that the scheme is able to improve the chlorophyll – is difficult to verify other biological fields but some work is underway in this area. Further work needed to explore the lack of seasonal variability in oligotrophic regions: - smoothing of assimilation? - absence of variability in satellite data? Summary of ocean colour assimilation work

The Talk What are the aims? What tools are we using? What have we developed so far? Assimilation of satellite-derived chlorophyll What will we be doing next? What can we do as champion user of GlobColour?

What will be doing next? - Operational pre-operational status from January Climate 10-year re-analysis of FOAM-HadOCC with/without chlorophyll and physical assimilation. biological assimilation scheme to be assessed for implementation in Hadley Centre Carbon Cycle Data Assimilation System (CCDAS) – IPCC report

The Talk What are the aims? What tools are we using? What have we developed so far? Assimilation of satellite-derived chlorophyll What will we be doing next? What can we do as champion user of GlobColour?

What can we do as champion user of GlobColour? Met Office has developed the capability for the simulation of surface and deep ocean biogeochemistry in NRT unique operational system fully coupled (on-line!) to an ecosystem and carbon cycle model state of the art data assimilation scheme for ocean colour/derived chl hindcast capability back to 1997, which makes possible the quantification of impact of GlobColour products on variables of climate interest: air-sea CO 2 flux, carbon sequestration, acidity, PP, chl, etc. well positioned to add value to the merged data by ensuring suitability for use for both operational oceanography and climate research transitioning of R&D product into operations However: development work will be required funding

Rosa Barciela

Experiments – identical twin set-up Start from a spun-up model state, then run the model forced by 6 hourly NWP fluxes for 1 year, with physical (T, S, SST) data assimilation. This is called the true run. Observations of Chl are taken from this true model state once a day. The ecosystem model variables are initialised using the biological fields from March 2003, with the physical fields taken from the true run. Starting from these new initial conditions, the model is run from April 2003 without (control) and with (assim) the Chl observations assimilated.

Real world experiments – on 1 st July 2003 Log(chl) from model with no biological assimilation Log(chl) observations Log(chl) from model with biological assimilation

GlobCOLOUR/Ocean Colour Operational User Requirements Specific requirements for GlobCOLOUR - L2 Global Area Coverage of chl a plus quantified errors from merged and individual sensors - Best possible accuracy: essential to decrease errors in derived chl below 35% - Extensive product quality control: include quantified errors and quality flags - Validation against in situ data and across biogeochemical regions. - Product format: WMO GRIB or netCDF - Delivery method: FTP - Large biases in the merged product corrected by in situ data - Spatial resolution: 4 Km spacing (highest resolution models have) - Bias information from individual sensors

Phytoplankton background error before the first analysis. Phytoplankton analysis error after the first analysis, with data everywhere. Phytoplankton errors (mmolN/m 3 ) Assimilation of Derived Chlorophyll Results from 3-D twin experiments

GlobCOLOUR/Ocean Colour Operational User Requirements Joint GlobCOLOUR/Medspiration products would be an advantage: - single file format - single file delivery - reduced data processing time - diagnostic data set applied to GlobCOLOUR data For operational purposes … Long-term provision of quality-controlled products in a timely (within 1 day) manner. - sustainability is key as lots of investment required to use the data - stable formats and delivery: (very) high availability and reliability NW European Shelf (NOOS) user requirements may need to be gathered

Future Plans To use GHRSST-PP data operationally from next year (development work required)

Future plans To transition the FOAM-HadOCC system into pre-operational state by 2008 (assimilation of ocean colour products)