Observations Supporting Decadal Predictions Roger Lukas University of

Observations Supporting Decadal Predictions Roger Lukas University of Hawaii Climate Research Committee Forum on Decadal Predictability 12/2/2008

Observations Supporting Decadal Predictions • Focus here is on decadal prediction initialization, but we must also consider prediction validation and model improvement • Assimilation formalisms provide initial state estimates from observations and a framework for assessing model errors relative to observational errors

Assimilated and with-held observations Assimilated observations XBT SST Altimeter Forecast Analysis or or T/S BGF Moorings Validation Altimeter Forecast XBT T/S SST • Observing System Experiments (OSEs) – Assimilate real observations – Systematically with-hold observation types Oke et al. 3 Evaluation/ GODAE Final Symposium, 12 – 15 November 2008, Nice, France Moorings

Overview • Focus on ocean state initialization – Identified source of decadal predictability in some models – Oceanic impacts (fisheries, coastal inundation, shipping) • Status of ocean components of climate observing system • Identify some gaps and weaknesses • Discuss strategy for enhanced observational impacts on decadal time scales • A few requirements

Ocean observing systems have advanced in the last decade – In situ surface meteorology + scatterometer + SST – Global sea level through altimetry and in situ gauges – Tropical moored arrays » TAO, Triton, Pirata implemented in Pacific and Atlantic » Under development in Indian Ocean – Argo float array – RAPID/MOCHA 26°N AMOC – Ocean. SITES (time-series stations) » NSF: HOT + BATS » NOAA: Atlantic and Pacific ORS moorings

M. Johnson, NOAA/OAR

Initial Global Ocean Observing System for Climate Status against the GCOS Implementation Plan and JCOMM targets Total in situ networks 60% February 2008 87% 100% 62% 81% 100% 24% 48% 79% 43% Milestones Drifters 2005 Argo 2007

October 2008 Volunteer Observing Ship surface met on GTS – in decline

October 2008 drifting buoy and moored buoy surface met on GTS Filling VOS gaps

Sea Level • Tide gauges + altimeters – – ENSO Integral constraints on heat content global rise rate Marginally eddy-resolving • Understanding decadal sea level variations is problematic – spatial variability is large • Salinity contributions not well constrained • Deep water mass variations may be important • Eddies may be important feedbacks on WBCs (~50% of decadal fluctuations), but can’t be initialized

Decadal change of N. Atl. MOC at 26 N estimated by an ECCO-GODAE product (Wunsch & Heimbach 2006) • Complex vertical structure: Weakening northward transport above 1000 m Strengthening southward transport of NADW Strengthening northward transport of abyssal water • No significant decrease of northward heat transport (upper-ocean warming enhances vertical temperature gradient to offset weakening of upper MOC). • Opposite trends of MOC strength at 26 N & 50 N.

Argo floats March 2008 Note dots are larger than mesoscale eddies

Interior ocean float coverage good, boundaries not so good September 2008

Polar in situ observations

Increasing Argo float lifetimes

Increasing demand on float power; duration and reliability • Biogeochemical sensors (e. g. oxygen, fluorometer, …) • Sampling upper few meters requires additional sensors • Deeper profiling • Ice detection

More Argo floats needed Smith et al. (2007) “… improvements in De. Pre. Sys relative to No. Assim on decadal time scales result mainly from initializing H. ” “Furthermore, a substantial increase in the number of subsurface ocean observations through the Argo program should substantially improve our ability to initialize the ocean in future …” Signal/noise requires more profiles in space/time to reduce aliasing noise

AMOC • Great progress with AMOC (RAPID/MOCHA array) • Need more measurements to partition effects of AMOC constituent variations Church (2007, Science)

Energetic high frequency variations LF ~mass balance Array concept works Kanzow et al. (2007)

AMOC is difference between two large noisy numbers Cunningham et al. (2007)

A complete AMOC observing system would include: • The Nordic Sea overflows • Production and export of dense waters from the Labrador Sea • The time varying strength of the AMOC in the subpolar North Atlantic following vertical entrainment and mixing processes • The time varying strength of the AMOC in the subtropical North Atlantic (e. g. , RAPID). • The time varying strength of the AMOC in the subtropical South Atlantic. US CLIVAR report 2008 -1

Multivariate Time-Series Sites

Some Obvious Observational Weaknesses (NOT prioritized) • AMOC meridional structure, deep convection regions • Boundary currents – US (NSF/OOI and NOAA/IOOS) – South Atlantic; W. Pacific – LLWBCs • Surface salinity (and global rainfall) • Deep thermohaline structure (>2000 m) • High latitude time-series generally – NSF/OOI plans: Station PAPA, Irminger Sea, SP (55 S/90 W)

Strategy for Observations supporting Decadal Predictions – We can’t get many more new realizations so need to consider observational strategy – Paleo can help extend temporal coverage, but need observations to calibrate proxies – Limited Do. F in time may be overcome to some extent by Do. F in space – Need multivariate time-series for validation Do. F = degrees of freedom

Requirements for Observations supporting Decadal Predictions • Need to improve surface forcing estimates going forward, and reanalyses • Consistent, accurate instrument calibrations are crucial • Need more integral constraints – e. g. surface salinity is an integral constraint and errors don’t directly feedback onto atmosphere (Aquarius satellite mission + in situ) – e. g. regional tomography array for deep convection regions • Careful observing system experiments e. g. Oke et al. and Lee et al. (Nov 2008, GODAE Final Symposium)

Oke et al. : OSEs and OSSEs Simulated Assimilated “observations” observations Forecast Analysis or or BGF Forecast • Observing System Experiments (OSEs) – Assimilate real observations – Systematically with-hold observation types 26 Evaluation/ Validation Observing System Simulation Experiments (OSSEs) § Assimilate pretend “observations” … from a model § Systematically include different observation types … including future observation types GODAE Final Symposium, 12 – 15 November 2008, Nice, France

Some Conclusions • We are not oversampling the ocean • Prioritizing gaps relative to decadal prediction requires better understanding and models • Shouldn’t neglect decadal signals that could add prediction skill – Decadal variations arising from tropics? – Decadal variations of midlatitude Pacific? • Harder to sustain/improve existing observing systems than to start new ones – research funding is entrepreneurial – transition from research to operations • Must build multi-decadal time-series for the future