seasonal sea ice forecasting - · pdf filecourtesy f. massonnet ... go for seasonal sea ice...
TRANSCRIPT
Seasonal Sea Ice Forecasting
Brief overview - challenges for YOPP
Matthieu Chevallier (CNRM/GAME, Météo France)With valuable help from and discussions with: Ed Blanchard-Wrigglesworth, Jonny Day, Virginie Guémas, François Massonnet.
YOPP Summit13-15 July 2015, WMO, Geneva
Arctic sea ice predictability
Coupled atmospheric-sea ice-ocean processes (dynamic/thermodynamic)Importance of information on sea ice thickness (melt ponds) ➙ observations?
thickness
concentration
SnowMeltponds
Blanchard et al., 2011Chevallier and Salas-Mélia, 2012Day et al., 2013Kapsch et al., 2014Schröder et al., 2014
•September sea ice extent: OK fromJuly, probably OK from May-June usingmelt ponds/radiative fluxes, potentiallyOK from March using ice thickness…
•March sea ice extent: OK fromJanuary, potentially OK earlier from icethickness, upper ocean…
Initial states:Initialize non-observable quantities (ice thickness)?Initialize the whole system (atmosphere, ocean, land surface…)?Sample initial state uncertainty (ensemble generation)?
How to build a sea ice forecasting system
MODEL
Initial state Forecast ObservationsObservations
Model:Statistical: long time series of observational data required, non-stationarity problemsPhysical: what complexity? Resolution? Uncertainty due to errors in model parameters?
Operational systemsHindcast: May 1 ➙ September (5 months)Mean September Arctic sea ice area
Do all these systems together provideuseful information?
➡YES!
Merryfield et al., 2013Chevallier and Guémas, 2015, in prep
Why are the predictions so different?-initialization?-model physics?
Guémas et al., 2014, QJRMS
➡Still unclear…
+ GFDL, UK Met Office…
Coupledmodels
Initialization issues
thickness
heatcontent
concentrationmixed layer heat content
SnowMeltpond
s
Courtesy F. Massonnet
➡ How to initialize the whole state with incomplete observations?➡ How about reanalyses?
Initialization issues
Average March sea ice thickness (2003-2007) in reanalysesChevallier et al., 2015, Clim DynGODAE OceanView/CLIVAR-GSOP Ocean ReAnalyses Intercomparison Project
Some systems assimilate concentration (some don’t…)
No thickness data assimilation
Differences in model, setups, datasets, DA methods…
Relevant diagnostics for end-users
+ Spatial patterns (ice edge) + Probabilistic information+ …?
Date (julian day) of ice clearance for May 1st 2008 hindcast with CNRM-CM5.1 (May 1st= day #120). Ensemble max and min.
Max MinDate of ice clearance/appearance
Sea
ice
area
ano
mal
iesRegional sea ice area Nov ➙ March forecasts
with CNRM-CM
Barents Sea
Recommendations for YOPPYOPP is a unique opportunity for better understanding the coupled processes at the ocean‐sea ice‐atmosphere‐land interfaces playing a role in sea ice predictability from hours to seasons and for improving our coupled forecasting systems and their uses.
Have sea ice models and diagnostics ready in all seasonal forecasting systems!
Coordinated sea ice forecasting experiments for all seasons Hindcast mode: systematic multi-model approach Sensitivity tests (initial conditions, new model parameterizations, ensemble generation) Dedicated experiments during YOPP using specific data Understanding the drivers of certain YOPP events
Use YOPP IOP to intensify links between modellers and observationalists and developcoupled data assimilation
Think beyond the September pan-Arctic sea ice extent (end-users diagnostics)!
Go for seasonal sea ice forecasting in the Southern Ocean!
Recommendations for YOPPYOPP is a unique opportunity for better understanding the coupled processes at the ocean‐sea ice‐atmosphere‐land interfaces playing a role in sea ice predictability from hours to seasons and for improving our coupled forecasting systems and their uses.
Have sea ice models and diagnostics ready in all seasonal forecasting systems! Météo-France subseasonal/seasonal forecasting systems
Coordinated sea ice forecasting experiments for all seasons Hindcast mode: systematic multi-model approach Sensitivity tests (initial conditions, new model parameterizations, ensemble generation) Dedicated experiments during YOPP using specific data Understanding the drivers of certain YOPP events
Joint effort Météo-France/IC3 (ANR), S2S, Sea Ice Prediction Network
Use YOPP IOP to intensify links between modellers and observationalists and developcoupled data assimilation
French ‘Chantier Arctique’: link with iAOOS, Copernicus Marine Service
Think beyond the September pan-Arctic sea ice extent (end-users diagnostics)! Sea Ice Prediction Network
Go for seasonal sea ice forecasting in the Southern Ocean!