Improving GODAE products for IMBER applications (Summary of break-out group #1)

1. Improving GODAE products for IMBER applications (Summary of break-out group #1) IGST-XII meeting, St. John’s, Newfoundland, Aug. 7-9, 2007 Tony Lee NASA Jet Propulsion Laboratory, California Institute of Technology Three proposed actions: • 1. Evaluate the quality of GODAE-derived vertical fluxes by using GODAE products to drive IMBER models. • 2. Examine the impact of volume conservation in GODAE products on ecosystem variables. • 3. Assess the reliability of forcing fields estimated by GODAE systems.

2. 1. Evaluate the quality of GODAE-derived vertical fluxes by using GODAE products to drive IMBER models • Background: • Abrupt temporal “shocks” caused by sequential assimilation methods result in spurious convective adjustment of the model state. • Some evidence that this problem cause unrealistic vertical fluxes, which degrade the simulation of IMBER variables.

3. An example of spurious vertical exchange due to sequential update of model state: problems may be hidden in the physical state, but emerges in biogeochemical applications. CO2 flux during ENSO inferred from a Kalman filter estimation (physically inconsistent) is unrealistically large (left), but that based on Kalman filter-smoother (physically consistent) is reasonable (right). Kalman Filtered estimate Kalman-filter/smoother estimate McKinley et al., 2002

4. Is this a common problem for systems based on sequential assimilation methods? Prompt for a systematic evaluation of the vertical fluxes in GODAE systems in the context of IMBER: • GODAE provides 3D velocity, temperature, salinity, and vertical diffusivity on a common grid to drive IMBER models offline. • Prior states (before assimilation) also needed to evaluate impact of assimilation. • All products should be accompanied by meta data info. • NETCDF format preferred. • Regional high-res evaluation + global coarse-res evaluations. • IMBER interested in offline tracer codes of GODAE systems (potential issues: advection and grid schemes).

5. A web-based tool for forward and adjoint tracer simulation

6. Horizontal distribution of source waters for the Pacific cold tongue as a function of years backward in time estimated using ECCO adjoint tracer (Fukumori et al. 2004) IMBER need such tools

7. Strategy to improve vertical fluxes in GODAE systems • Ad-hoc approach: spread out the assimilation correction over a certain time window as opposed to at a single-time step. • Use a smoother in addition to sequential filter. • Use assimilation methods that allow estimated state to satisfy model equations (e.g., Green’s function, 4D-VAR).

8. 2. Examine the impact of volume conservation in GODAE products on ecosystem variables • Background: • Many GODAE products do not satisfy volume conservation. • May impact diagnostic studies of IMBER. Proposed approach to evaluate impact on IMBER: Analyze budget imbalance on IMBER variables.

9. 3. Assess the reliability of forcing fields estimated by GODAE systems Background: IMBER scientists need forcing fields that are consistent with the GODAE state estimates to perform diagnostic studies. Issues: Only 2 of the GODAE systems have forcing fields that are consistent with the state estimation (ECCO and Japan K-7). Potential drawback: These forcing estimates may be model dependent.