WATER-HM/SWOT meeting CNES HQ, Paris, February 2008 Constraining coastal ocean models with altimetry Pierre De Mey, LEGOS/POC

Coastal ocean domains and scales SST 24 Aug 2003 Oregon CTZ Leeuwin Current Sea-level variance Jan 1999 North & Celtic Seas Chl

Constraining the small-scale coastal EKE Surface EPE (J/m 2 ) Surface EKE (J/m 3 ) Small-scale EKE patches and filaments Fast time scales as critical as short space scales to constrain EKE on shelves Small-scale EKE patches and filaments Fast time scales as critical as short space scales to constrain EKE on shelves (after Michel & Huthnance, 2008) Surface EPE and EKE in POLCOMS

WATER-HM/SWOT meeting CNES HQ, Paris, February Downscaling to the coastal scale Message: Downscaling of large-scale models to the coastal scale is shaping into a community. New science issues are being addressed.

GODAE Coastal and Shelf Seas Working Group

GODAE Coastal and Shelf Seas Working Group –38 projects –Varied in objectives and methods –Geographically clustered Africa: 1 Australia: 2 China Seas: 2 Japan Seas: 3 Indian Ocean: 1 North Pacific: 1 North America West Coast: 4 Gulf of Mexico: 4 North America East Coast: 2 Arctic and Nordic Seas: 3 Norwegian, Baltic, and North Seas: 7 Northeast Atlantic: 6 Mediterranean: 4 P1 P P21-22 P20 P19 P18 P17 P16 P P12 P P8 P7 P6 P5 P4 P3 P2 P P30 P29-37 P P27 P31 P36 (bold = new or updated in 2007)

GODAE CSSWG: Science issues White Paper available on GODAE web site Main interests of downscaling from GODAE solutions (WP 2.2) –Better local estimates, extend predictability –Enhance representativeness of high-resolution observations, opening the door for better data assimilation in Coastal and Shelf Seas Critical issues and requirements (WP 3.2) –Access to “adequate” large-scale solutions Importance of an adequate fine-scale vorticity field > –Jump in ocean physics and in forcing functions –Initialization –Tides and high-frequency barotropic dynamics –How do we measure impact and value in CSS models? MERSEA/GODAE legacy on metrics + impact assessment procedures in coastal models –Additional issues: data assimilation, two-way coupling, unstructured grid modelling Community in need of routine, adequate observations Bay of Biscay vorticity

WATER-HM/SWOT meeting CNES HQ, Paris, February What should we constrain? Message: We are beginning to get a quantitative idea of the errors we must aim at correcting in coastal models with the help of sea-level measurements.

Instantaneous ensemble variance: a proxy of model error variance Bay of Biscay 3D model 30m temperature ensemble variance, July 30, 2004 (after Le Hénaff & De Mey, 2008) What is this? Near-surface temperature ensemble variance in response to atmospheric forcing errors. It is a proxy of the actual model errors. Exhibits the fine time/space scales of a tracer and is a mix of shelf, shelf- break, upwelling, and mesoscale responses. Relevance to WATER-HM? We need to be able to constrain these fine error scales. What is this? Near-surface temperature ensemble variance in response to atmospheric forcing errors. It is a proxy of the actual model errors. Exhibits the fine time/space scales of a tracer and is a mix of shelf, shelf- break, upwelling, and mesoscale responses. Relevance to WATER-HM? We need to be able to constrain these fine error scales.

EOF % EOF % EOF % SLADepth-averaged velocitySurface salinityTemperature Non-local, structured errors in coastal current (Jordà et al., 2006) What is this? Ensemble multivariate EOFs in the Catalan Sea coastal current in response to coastal current inflow perturbations (mimicking downscaling errors). Relevance to WATER-HM? SLA errors are small-scale (O(40km)) and strongly correlated to fine-scale (u,v,T,S) 3-dimensional errors which we can then expect to correct if SLA is observed at sufficiently fine scales. What is this? Ensemble multivariate EOFs in the Catalan Sea coastal current in response to coastal current inflow perturbations (mimicking downscaling errors). Relevance to WATER-HM? SLA errors are small-scale (O(40km)) and strongly correlated to fine-scale (u,v,T,S) 3-dimensional errors which we can then expect to correct if SLA is observed at sufficiently fine scales.

EOF % EOF % EOF % SLADepth-averaged velocitySurface salinityTemperature Non-local, structured errors in coastal current (Jordà et al., 2006)

Activation of coherent error features by storms Ensemble EOF-3 SLA, 3D BoB model July 1August What is this? The SLA component of a particular ensemble EOF in response to atmospheric forcing errors. It is a proxy of the actual model errors. As the time series shows, it is activated during the July 7-8 storm and is characterized by a shelf-wide response, a surge response, and a mesoscale response with O(1day) time scale. Relevance to WATER-HM? Questions 1 (mesoscale), 2 (coastal) and 3 (storm-related). We expect a wide-swath altimeter to consistently constrain the fine-scale, multivariate ocean response to those fast events, and hopefully help better predict the associated phenomena. What is this? The SLA component of a particular ensemble EOF in response to atmospheric forcing errors. It is a proxy of the actual model errors. As the time series shows, it is activated during the July 7-8 storm and is characterized by a shelf-wide response, a surge response, and a mesoscale response with O(1day) time scale. Relevance to WATER-HM? Questions 1 (mesoscale), 2 (coastal) and 3 (storm-related). We expect a wide-swath altimeter to consistently constrain the fine-scale, multivariate ocean response to those fast events, and hopefully help better predict the associated phenomena.

Constraining shelf-break exchanges Ensemble-time EOF-1 16-Nov Z00 -- State vector: (SLA HF, U b, SLA IB,  ) scaled transport (Lamouroux and De Mey, 2007) SLA HF SLA IB What is this? The BT transport, high- frequency SLA, and inverted-barometer SLA components of a particular ensemble EOF in response to atmospheric forcing errors. It is a proxy of the actual model errors. As the bottom right panel shows, the situation is that of a southwesterly wind blowing towards the English Channel. The top right panel shows water piling up in the channel. The left panel shows the corresponding fine-scale exchanges through shelf-break canyons and around capes. Relevance to WATER-HM? Questions 2 (coastal) and 3 (storm-related). We expect a wide-swath altimeter to resolve fast time scales in straits and semi-enclosed seas to constrain the variability of exchanges between shelf and deep ocean. What is this? The BT transport, high- frequency SLA, and inverted-barometer SLA components of a particular ensemble EOF in response to atmospheric forcing errors. It is a proxy of the actual model errors. As the bottom right panel shows, the situation is that of a southwesterly wind blowing towards the English Channel. The top right panel shows water piling up in the channel. The left panel shows the corresponding fine-scale exchanges through shelf-break canyons and around capes. Relevance to WATER-HM? Questions 2 (coastal) and 3 (storm-related). We expect a wide-swath altimeter to resolve fast time scales in straits and semi-enclosed seas to constrain the variability of exchanges between shelf and deep ocean.

Constraining shelf-break exchanges Ensemble-time EOF-1 16-Nov Z00 -- State vector: (SLA HF, U b, SLA IB,  ) scaled transport (Lamouroux and De Mey, 2007) SLA HF SLA IB

WATER-HM/SWOT meeting CNES HQ, Paris, February How would a single wide-swath instrument do? Message: A single wide-swath instrument on a JASON-type orbit would significantly contribute to the constraint of the coastal ocean mesoscale, coastal current variability and shelf-wide sea-level changes.

Wide-swath vs. nadir in Bay of Biscay Stochastic modelling with atm. forcing perturbations in 3D BoB Top: Wide Swath (4 dof’s) Mid: WS over deep ocean (2 dof’s) Bottom: JASON (1 dof) Scaled RM spectra Array Modes -- SLA “Slosh”Meso1Meso2 (after Le Hénaff & De Mey, 2008) (left panel) What is this? The RM spectra plot on the left shows the number of degrees of freedom of model (forecast) error which can be detected by a particular array amidst observational noise. This is done by counting eigenvalues above 1. This is shown for three arrays (legend). Representer matrices are calculated by stochastic modelling with atmospheric forcing errors. Relevance to WATER-HM? Questions 1 (mesoscale) and 2 (coastal). One Wide- Swath altimeter on a JASON orbit detects 4 degrees of freedom, while one nadir instrument (JASON) detects only one. The more d.o.f.’s are detected, the more critical ocean processes will be constrained. (left panel) What is this? The RM spectra plot on the left shows the number of degrees of freedom of model (forecast) error which can be detected by a particular array amidst observational noise. This is done by counting eigenvalues above 1. This is shown for three arrays (legend). Representer matrices are calculated by stochastic modelling with atmospheric forcing errors. Relevance to WATER-HM? Questions 1 (mesoscale) and 2 (coastal). One Wide- Swath altimeter on a JASON orbit detects 4 degrees of freedom, while one nadir instrument (JASON) detects only one. The more d.o.f.’s are detected, the more critical ocean processes will be constrained.

Wide-swath vs. nadir in Bay of Biscay Stochastic modelling with atm. forcing perturbations in 3D BoB Top: Wide Swath (4 dof’s) Mid: WS on deep ocean (2 dof’s) Bottom: JASON (1 dof) Scaled RM spectra Array Modes -- SLA “Slosh”Meso1Meso2 (after Le Hénaff & De Mey, 2008) (right panel) What is this? The “array modes” of model error corresponding to the spectra to the left. For each array, mode 1 is mostly water sloshing around between shelf and deep-ocean domains; modes 2 and 3 are a mix of mesoscale & submeso response, slope current variability and shelf processes. Relevance to WATER-HM? Questions 1 (mesoscale) and 2 (coastal). As was seen on the left panel, JASON can only detect (and constrain) the “slosh” mode. One needs a wide-swath instrument to detect (constrain) all three modes + a 4th one not shown. In this way, one can objectively demonstrate that a wide-swath instrument is needed to constrain the coastal ocean mesoscale and coastal current variability. (A collaboration between LEGOS and OSU’s OST proposals has been proposed on this topic) (right panel) What is this? The “array modes” of model error corresponding to the spectra to the left. For each array, mode 1 is mostly water sloshing around between shelf and deep-ocean domains; modes 2 and 3 are a mix of mesoscale & submeso response, slope current variability and shelf processes. Relevance to WATER-HM? Questions 1 (mesoscale) and 2 (coastal). As was seen on the left panel, JASON can only detect (and constrain) the “slosh” mode. One needs a wide-swath instrument to detect (constrain) all three modes + a 4th one not shown. In this way, one can objectively demonstrate that a wide-swath instrument is needed to constrain the coastal ocean mesoscale and coastal current variability. (A collaboration between LEGOS and OSU’s OST proposals has been proposed on this topic)

Wide swath: degradation due to roll errors (after Le Hénaff & De Mey, 2008) Top: No roll errors (3 dof’s) Bottom: Along-track correlated roll errors (2 dof’s) Scaled RM spectra What is this? Scaled RM spectra for one wide-swath track in the Bay of Biscay without, and with, along-track correlated (50sec) roll errors. Relevance to WATER-HM? Questions 1 (mesoscale) and 2 (coastal). As long as we have a good idea of the along-track period of the roll errors, the effect of those errors is limited (1 detected eigenvalue drop). However the effects of (1) shorter- period roll errors and of (2) yaw errors would have to be further quantified. What is this? Scaled RM spectra for one wide-swath track in the Bay of Biscay without, and with, along-track correlated (50sec) roll errors. Relevance to WATER-HM? Questions 1 (mesoscale) and 2 (coastal). As long as we have a good idea of the along-track period of the roll errors, the effect of those errors is limited (1 detected eigenvalue drop). However the effects of (1) shorter- period roll errors and of (2) yaw errors would have to be further quantified.

Summary Downscaling of large-scale models to the coastal ocean is shaping into a community. New science issues are being addressed, some of them similar to SWOT’s. The knowledge of coastal circulation is critical to many applications (ecosystem management, water quality, disaster prevention, transportation). These applications by themselves are beyond the scope of this mission (see Washington meeting summary), but a wide-swath altimeter would provide solid scientific foundations towards those objectives. We are beginning to get a quantitative idea of the errors we must aim at correcting in coastal models with the help of sea-level measurements. A single wide-swath instrument on a JASON-type orbit would significantly contribute to the constraint of the coastal ocean mesoscale, coastal current variability and shelf-wide sea-level changes. Other comments: Long-wavelength roll errors are probably easier to separate from coastal processes than other errors such as shorter-wavelength roll and yaw errors. Swath overlap is a desired feature to maximize revisit time and better constrain the higher-frequency part of the spectrum (shelf seas).