The impact of Argo data on ocean and climate forecasting

Slides:



Advertisements
Similar presentations
© Crown copyright Met Office NEMOVAR status and plans Matt Martin, Dan Lea, Jennie Waters, James While, Isabelle Mirouze NEMOVAR SG, ECMWF, Jan 2012.
Advertisements

Tuning and Validation of Ocean Mixed Layer Models David Acreman.
OSE meeting GODAE, Toulouse 4-5 June 2009 Interest of assimilating future Sea Surface Salinity measurements.
Experiments with Monthly Satellite Ocean Color Fields in a NCEP Operational Ocean Forecast System PI: Eric Bayler, NESDIS/STAR Co-I: David Behringer, NWS/NCEP/EMC/GCWMB.
Assimilation of Sea Surface Temperature into a Northwest Pacific Ocean Model using an Ensemble Kalman Filter B.-J. Choi Kunsan National University, Korea.
Assimilating SST and Ocean Colour into ocean forecasting models Rosa Barciela, NCOF, Met Office
1 Evaluation of two global HYCOM 1/12º hindcasts in the Mediterranean Sea Cedric Sommen 1 In collaboration with Alexandra Bozec 2 and Eric Chassignet 2.
Jon Robson (Uni. Reading) Rowan Sutton (Uni. Reading) and Doug Smith (UK Met Office) Analysis of a decadal prediction system:
Building Bluelink David Griffin, Peter Oke, Andreas Schiller et al. March 2007 CSIRO Marine and Atmospheric Research.
The NCEP operational Climate Forecast System : configuration, products, and plan for the future Hua-Lu Pan Environmental Modeling Center NCEP.
Exploring strategies for coupled 4D-Var data assimilation using an idealised atmosphere-ocean model Polly Smith, Alison Fowler & Amos Lawless School of.
© Crown copyright Met Office UK report for GOVST Matt Martin GOVST-V, Beijing, October 2014.
The Global Ocean Data Assimilation System (GODAS) at NCEP
Dr Mark Cresswell Model Assimilation 69EG6517 – Impacts & Models of Climate Change.
Ensemble-variational sea ice data assimilation Anna Shlyaeva, Mark Buehner, Alain Caya, Data Assimilation and Satellite Meteorology Research Jean-Francois.
Collaborative Research: Toward reanalysis of the Arctic Climate System—sea ice and ocean reconstruction with data assimilation Synthesis of Arctic System.
Ocean Data Variational Assimilation with OPA: Ongoing developments with OPAVAR and implementation plan for NEMOVAR Sophie RICCI, Anthony Weaver, Nicolas.
The Centre for Australian Weather and Climate Research A partnership between CSIRO and the Bureau of Meteorology Summary/Future Re-anal.
Inter-comparison and Validation Task Team Breakout discussion.
Potential benefits from data assimilation of carbon observations for modellers and observers - prerequisites and current state J. Segschneider, Max-Planck-Institute.
Dataset Development within the Surface Processes Group David I. Berry and Elizabeth C. Kent.
IICWG 5 th Science Workshop, April Sea ice modelling and data assimilation in the TOPAZ system Knut A. Lisæter and Laurent Bertino.
© Crown copyright Met Office Plans for Met Office contribution to SMOS+STORM Evolution James Cotton & Pete Francis, Satellite Applications, Met Office,
EUROBRISA WORKSHOP, Paraty March 2008, ECMWF System 3 1 The ECMWF Seasonal Forecast System-3 Magdalena A. Balmaseda Franco Molteni,Tim Stockdale.
UK National Report IGST, November 2005.
Developments within FOAM Adrian Hines, Dave Storkey, Rosa Barciela, John Stark, Matt Martin IGST, 16 Nov 2005.
Sophie RICCI CALTECH/JPL Post-doc Advisor : Ichiro Fukumori The diabatic errors in the formulation of the data assimilation Kalman Filter/Smoother system.
Validation of decadal simulations of mesoscale structures in the North Sea and Skagerrak Jon Albretsen and Lars Petter Røed.
Validation of US Navy Polar Ice Prediction (PIPS) Model using Cryosat Data Kim Partington 1, Towanda Street 2, Mike Van Woert 2, Ruth Preller 3 and Pam.
2nd GODAE Observing System Evaluation Workshop - June Ocean state estimates from the observations Contributions and complementarities of Argo,
GODAE Progress on national activities France (Mercator Ocean) Eric Dombrowsky - Mercator Océan.
© Crown copyright Met Office The EN4 dataset of quality controlled ocean temperature and salinity profiles and monthly objective analyses Simon Good.
Near real time forecasting of biogeochemistry in global GCMs Rosa Barciela, NCOF, Met Office
Building Bluelink David Griffin, Peter Oke, Andreas Schiller et al. March 2007 CSIRO Marine and Atmospheric Research.
Evaluation of Tropical Pacific Observing Systems Using NCEP and GFDL Ocean Data Assimilation Systems Y. Xue 1, C. Wen 1, X. Yang 2, D. Behringer 1, A.
The Impact of Data Assimilation on a Mesoscale Model of the New Zealand Region (NZLAM-VAR) P. Andrews, H. Oliver, M. Uddstrom, A. Korpela X. Zheng and.
One float case study The Argo float ( ) floating in the middle region of Indian Ocean was chosen for this study. In Figure 5, the MLD (red line),
One-year re-forecast ensembles with CCSM3.0 using initial states for 1 January and 1 July in Model: CCSM3 is a coupled climate model with state-of-the-art.
Assimilating Satellite Sea-Surface Salinity in NOAA Eric Bayler, NESDIS/STAR Dave Behringer, NWS/NCEP/EMC Avichal Mehra, NWS/NCEP/EMC Sudhir Nadiga, IMSG.
Trials of a 1km Version of the Unified Model for Short Range Forecasting of Convective Events Humphrey Lean, Susan Ballard, Peter Clark, Mark Dixon, Zhihong.
PreSAC Progress on NEMOVAR. Overview of NEMOVAR status First NEMOVAR experiments Use of NEMOVAR analyses to initialize ocean only forecasts Missing.
Impact of TAO observations on Impact of TAO observations on Operational Analysis for Tropical Pacific Yan Xue Climate Prediction Center NCEP Ocean Climate.
1 A multi-scale three-dimensional variational data assimilation scheme Zhijin Li,, Yi Chao (JPL) James C. McWilliams (UCLA), Kayo Ide (UMD) The 8th International.
Application of HYCOM in Eddy- Resolving Global Ocean Prediction Community Effort: Community Effort: NRL, Florida State, U. of Miami, GISS, NOAA/NCEP, NOAA/AOML,
Impact of Blended MW-IR SST Analyses on NAVY Numerical Weather Prediction and Atmospheric Data Assimilation James Cummings, James Goerss, Nancy Baker Naval.
The Mediterranean Forecasting INGV-Bologna.
HYCOM/NCODA Variational Ocean Data Assimilation System James Cummings Naval Research Laboratory, Monterey, CA GODAE Ocean View III Meeting November.
Ocean Data Assimilation for SI Prediction at NCEP David Behringer, NCEP/EMC Diane Stokes, NCEP/EMC Sudhir Nadiga, NCEP/EMC Wanqiu Wang, NCEP/EMC US GODAE.
1 3D-Var assimilation of CHAMP measurements at the Met Office Sean Healy, Adrian Jupp and Christian Marquardt.
ECMWF NEMOVAR update, NEMOVAR meeting Jan ECMWF update : NEMOVAR ORA-S4 (Ocean Re-Analysis System 4) implemented operationally  Based.
Global vs mesoscale ATOVS assimilation at the Met Office Global Large obs error (4 K) NESDIS 1B radiances NOAA-15 & 16 HIRS and AMSU thinned to 154 km.
Use of high resolution global SST data in operational analysis and assimilation systems at the UK Met Office. Matt Martin, John Stark,
Assimilation of S(T) from ARGO Keith Haines, Arthur Vidard *, Xiaobing Zhou, Alberto Troccoli *, David Anderson * Environmental Systems Science Centre,
TAIYO KOBAYASHI and Shinya Minato
POAMA (Predictive Ocean Atmosphere Model for Australia)
Bruce Cornuelle, Josh Willis, Dean Roemmich
Spatial Modes of Salinity and Temperature Comparison with PDO index
Plans for Met Office contribution to SMOS+STORM Evolution
Local Ensemble Transform Kalman Filter for ROMS in Indian Ocean
Operational Oceanography Science and Services for Europe and Mediterranean Srdjan Dobricic, CMCC, Bologna, Italy on behalf of National Group of Operational.
Winter storm forecast at 1-12 h range
Stéphane Laroche Judy St-James Iriola Mati Réal Sarrazin
Ocean Sub-Surface Observing Network
A coupled ensemble data assimilation system for seasonal prediction
Y. Xue1, C. Wen1, X. Yang2 , D. Behringer1, A. Kumar1,
Progress in Seasonal Forecasting at NCEP
Assimilation of Global Positioning System Radio Occultation Observations Using an Ensemble Filter in Atmospheric Prediction Models Hui Liu, Jefferey Anderson,
ECMWF activities: Seasonal and sub-seasonal time scales
NOAA Objective Sea Surface Salinity Analysis P. Xie, Y. Xue, and A
Supervisor: Eric Chassignet
Presentation transcript:

The impact of Argo data on ocean and climate forecasting Matt Martin, Mike Bell Contents: 1. Introduction 2. Data assimilation and Argo data 3. Indirect impact of Argo data 4. Summary Mainly concentrate on results from Met Office FOAM operational system. Will start off outlining the FOAM system including assimilation system Then show examples of experiments showing the direct impact of Argo on the errors in the system Then give some examples of how Argo data can be used to improve the assimilation/model system Summary and future work

1. FOAM system FOAM = Forecasting Ocean Assimilation Model Real-time data Obs QC & processing Analysis Forecast to T+120 NWP 6 hourly fluxes Automatic verification T+24 forecast used in QC Product dissemination Operational system Hindcast system FOAM – Forecasting Ocean Assimilation Model - runs every day in the operational suite at the Met Office producing analysis and 5 day forecasts of the deep ocean. The data is obtained from the GTS (global telecommunications system), processed and quality controlled automatically The data is used with the previous 24 hour forecast to produce an analysis 6 hourly fluxes are obtained from the Met Office NWP system and they force the model during the forecast Also have ability to re-run the model/analysis in hindcast mode Operational real-time deep-ocean forecasting system Daily analyses and forecasts out to 5 days Hindcast capability (back to 1997) Relocatable high resolution nested model capability

FOAM model configurations 1° (operational since 1997) 1/9° (pre-operational since April 2002) Data available from http://www.nerc-essc.ac.uk 1/3° (operational since 2001) Various resolution models 1 degree global model has been running since 1997 1/3 N atlantic model is nested inside global model and also runs operationally Also run a 1/9 degree model with output freely available for research purposes on the web from Reading University website All models have 20 levels in the vertical – will improve the vertical resolution soon

Operational data assimilation Operational models assimilate: Temperature profiles (including ARGO data) In situ and satellite SST (2.5º AVHRR) Satellite altimeter SSH (Jason-1, GFO, ERS-2) SSMI-derived seaice data from CMC Sequential scheme based upon the Analysis Correction scheme of Lorenc et al. (1991) Operational upgrade implemented on 28th October 2003 includes: Implementation of salinity assimilation Significant developments to original system Upgraded QC of data from ENACT project Recently upgraded assimilation component of the system. Previously only assimilated temperature data (not salinity) – will show the impact of assimilating Argo salinity on the system. Only assimilated T data down to 1000m – will show the impact of assimilating at all depths Will get higher resolution SST data, maybe from GHRSST project (part of GODAE)

2. Data assimilation and Argo data (a) Impact of salinity Argo data using simple assimilation scheme. (b) Impact of Argo data in operational models. (c) Comparison between impact of withholding Argo and other data types. Show 3 different results from experiments which use Argo data. Salinity assimilation Impact on operational models Withholding data experiments otherwise known as observation system experiments (OSEs). One OSE from FOAM and one from ECMWF (European centre) seasonal forecasting system

(a) Investigation of impact of salinity data Aim: To investigate the impact of the salinity data assimilation prior to implementation 5-month runs of the operational global 1º model Running for Jan - May 2003 Forced by 6-hourly NWP surface fluxes Initial state taken from operational model Assimilating only Argo data - no other data types Experiments run: Assimilating temperature and salinity profiles Assimilating temperature profiles only Assimilating salinity profiles only Control run assimilating no data New assimilation scheme includes salinity assimilation so did some tests with the Argo data to assess its impact Used global model for 5 months. Only assimilated Argo data Did 4 experiments with the different data combinations

Results (1) RMS errors against observations that have not yet been assimilated for final month of integrations over entire globe RMS T Error (ºC) RMS S Error (PSU) 1.0 2.0 3.0 0.4 0.2 Depth (m) 200 400 600 800 No assim S assim T assim T & S assim Climatology Temperature Salinity Red lines are without data assimilation Yellow lines are with T & S assimilation Green is climatology Depth along horizontal axes RMS errors against independent observations – obs which have not yet been assimilated Assimilating S only degrades T, even worse than with no assimilation at all Assimilating T only degrades S, “” Probably because assimilating one and not the other degrades the density structure Assimilating T&S improves salinity statistics significantly over assimilating S only.

Results (2) Monthly mean salinity field differences (PSU) from Levitus climatology at 1000m for May (Levitus - model) No assim T assim S assim T & S assim -1.0 1.0 0.0 Compare No assim with S assim – large biases disappear in the regions where there are data, except for in the N atlantic. Med outlow region especially – known to be wrong in 1 degree model – model bias could be feeding back onto the salinity. With T & S assim, these biases in the N Atlantic are reduced. South pacific still has large-scale biases because no Argo floats there.

(b) Preparation for operational implementation Aim: To ensure that new operational system is working correctly and making better use of Argo data Parallel suite trial running since August 2003 Upgraded version of operational FOAM suite Global 1º and 1/3º North Atlantic models Running daily at 05:00 Accessing only real-time data Analysis and forecast cycle uses new assimilation scheme Initial state from operational models Comparing previous and current assimilation systems – making better use of Argo data in the new system Running for about 3 months in parallel Show results from 1/3 degree N Atlantic model Only assimilated real-time data Initial state for both runs was from old operational models

Impact of salinity data on operational models Salinity data assimilated in upgraded system, but not in previous operational system Salinity differences (PSU) from Levitus climatology on 15th September at 300m (Levitus - model): New operational model -1.0 0.0 1.0 In the old system, no salinity assimilation which led to large biases in salinity compared to climatology of order 1 psu. Example shown at 300m depth on rotated grid – point out landmarks. New system assimilates salinity and the simple assimilation scheme significantly reduces the large biases in the North Atlantic. Even after only about 1 and a half months of running. Previous operational model

Impact of temperature data on operational models Temperature data not assimilated below 1000m in current operational system Temperature differences (ºC) from Levitus on 15th September at 1500m (Levitus - model): In the old system, previously no T data assimilated below 1000m. Let to large drifts in T below this depth. Showing T differences from climatology at 1500m. Biases seem to have been reduced even after only about 1 and a half months of model runs. Previous operational model New operational model -2.5 0.0 2.5

(c) Data withholding experiments – impact of Argo and altimeter data The reference integration assimilates all data Other integrations withhold selected data types Details of experiment: 1/9º North Atlantic model integrations from Jan – Mar 2003 initial state from operational models rms differences calculated using profile data before their assimilation old assimilation scheme used To assess the impact of Argo data and compare it with the impact of altimeter data in FOAM. Use 1/9 degree N Atlantic model 3 month runs Used old system so only look at impact on temperature

Impact of withholding Argo and altimeter data from FOAM Differences between model and observations yet to be assimilated FOAM 12km N Atlantic model driven by 6 hourly fluxes from Jan-Mar 2003 SST, XBT and Pirata data are also assimilated old assimilation scheme Show differences from independent profile observations for whole of 3 month period over whole of N Atlantic Depth along horizontal axis All data and altimeter data are similar – altimeter doesn’t have much impact on T – uses old assimilation scheme which is being improved Withholding Argo data has large detrimental impact on the RMS errors – shows the value of the Argo data.

Impact of withholding different data types in seasonal forecasting From ECMWF seasonal forecasting system (HOPE model, OI scheme) Potential temperature RMS differences from experiment with all data assimilated, 1998-2003 Upper 300m of ocean From A. Vidard No moorings No Argo No XBT 5 year integrations from ECMWF – European Centre – provided by Arthur Vidard 3 runs withholding different types of temperature obsevations and looking at the RMS differences to a run which assimilated all data Argo data only really available for the last couple of years of the integration so not fair to Argo but does show that Argo has a large impact on temperature analysis

3. Indirect impact of Argo (a) Improved estimation of error covariances. (b) Mixed layer improvements.

(a) Model Error Covariances estimated using pairs of observed temperature profiles Use collocated observation and model forecast values to estimate covariance values – bin together to have enough statistical information Assume separability of the error covariance, i.e. horizontal and vertical correlations can be calculated separately. Assume the forecast errors arise from two distinct sources: errors in the internal model dynamics => “mesoscale” errors errors in the atmospheric forcing => “synoptic” scale errors Fit a combination of 2 SOAR functions to the (obs-f/c) covariance values to estimate the variance and horizontal correlation scales of the two forecast error components. The observation error variance is the difference between the total (obs-f/c) mean square error and the total forecast error variance.

Modelling the covariance Schematic of method Example at 30W, 40N for SSH. Mesoscale length scale = 37km Synoptic length scale = 560km Circles - (obs-forecast) covariances Dotted line - synoptic scale function Dashed line- mesoscale function Solid line - sum of the two functions

Temperature profile error covariances Variance Length Meso 1.80 47 km Synop 0.5 1060 km Variance Length Meso 0.9 59 km Synop 0.1 540 km Covariance Depth = 55 m Depth = 240 m Separation Variance Length Meso 0.6 40 km Synop 0.2 500 km All observations binned together to get meaningful results Show different covariances with depth Large synoptic scales at 55m depth Depth = 800 m

Mesoscale error variances for SSH (cm²) and SST (K²) for 1/3o FOAM Atlantic model SSH Variance SST Variance

(b) 1D Mixed Layer Assessments using Argo data Initialise T & S profiles using Argo observation Use model and surface forcing to integrate forward for 10 days Estimate error in forecast using next Argo observation Run for 1 year for many different Argo floats Use this framework to test assimilation strategy Argo ob Background Forecast error statistics Surface fluxes Analysis Depth 10 day forecast Forecast Argo ob

Assimilation in the mixed layer Timeliness of assimilation MLD rms errors (m) 20 40 60 10 days 5 days 1 day 12 hours 6 hours 1 hour No assimilation “Kraus-Turner” scheme “Large” scheme For Large et al. model, the accuracy of the forecast decreases if the increments are nudged in over more than 1 hour A large number of other factors can also be explored e.g. vertical resolution, time sampling of fluxes Equatorial undercurrent now penetrates to the eastern Pacific and surface currents are better

Summary and future work Argo salinity data assimilation improves salinity fields, even with simple scheme Further work to improve salinity assimilation, i.e. use isopycnal coordinates for analysis Both Argo and altimeter data assimilation improve the fit of the analyses to independent temperature data with Argo data having the largest impact in FOAM Argo data also has significant usage for improving the data assimilation methods used, i.e. error covariances Use Argo data to help improve the assimilation of altimeter data Compare results in FOAM with other centres to make the most of the Argo data