Publications

Export 2 results:
Sort by: Author Title Type [ Year  (Desc)]
2015
Furue, R, Jia YL, McCreary JP, Schneider N, Richards KJ, Muller P, Cornuelle BD, Avellaneda NM, Stammer D, Liu CY, Kohl A.  2015.  Impacts of regional mixing on the temperature structure of the equatorial Pacific Ocean. Part 1: Vertically uniform vertical diffusion. Ocean Modelling. 91:91-111.   10.1016/j.ocemod.2014.10.002   AbstractWebsite

We investigate the sensitivity of numerical model solutions to regional changes in vertical diffusion. Specifically, we vary the background diffusion coefficient, kappa(b), within spatially distinct subregions of the tropical Pacific, assess the impacts of those changes, and diagnose the processes that account for them. Solutions respond to a diffusion anomaly, delta kappa(b), in three ways. Initially, there is a fast response (several months), due to the interaction of rapidly propagating, barotropic and gravity waves with eddies and other mesoscale features. It is followed by a local response (roughly one year), the initial growth and spatial pattern of which can be explained by one-dimensional (vertical) diffusion. At this stage, temperature and salinity anomalies are generated that are either associated with a change in density ("dynamical" anomalies) or without one ("spiciness" anomalies). In a final adjustment stage, the dynamical and spiciness anomalies spread to remote regions by radiation of Rossby and Kelvin waves and by advection, respectively. In near equilibrium solutions, dynamical anomalies are generally much larger in the latitude band of the forcing, but the impact of off equatorial forcing by delta kappa(b) on the equatorial temperature structure is still significant. Spiciness anomalies spread equator ward within the pycnocline, where they are carried to the equator as part of the subsurface branch of the Pacific Subtropical Cells, and spiciness also extends to the equator via western-boundary currents. Forcing near and at the equator generates strong dynamical anomalies, and sometimes additional spiciness anomalies, at pycnocline depths. The total response of the equatorial temperature structure to delta kappa(b) in various regions depends on the strength and spatial pattern of the generation of each signal within the forcing region as well as On the processes of its spreading to the equator.

2010
Hoteit, I, Cornuelle B, Heimbach P.  2010.  An eddy-permitting, dynamically consistent adjoint-based assimilation system for the tropical Pacific: Hindcast experiments in 2000. Journal of Geophysical Research-Oceans. 115   10.1029/2009jc005437   AbstractWebsite

An eddy-permitting adjoint-based assimilation system has been implemented to estimate the state of the tropical Pacific Ocean. The system uses the Massachusetts Institute of Technology's general circulation model and its adjoint. The adjoint method is used to adjust the model to observations by controlling the initial temperature and salinity; temperature, salinity, and horizontal velocities at the open boundaries; and surface fluxes of momentum, heat, and freshwater. The model is constrained with most of the available data sets in the tropical Pacific, including Tropical Atmosphere and Ocean, ARGO, expendable bathythermograph, and satellite SST and sea surface height data, and climatologies. Results of hindcast experiments in 2000 suggest that the iterated adjoint-based descent is able to significantly improve the model consistency with the multivariate data sets, providing a dynamically consistent realization of the tropical Pacific circulation that generally matches the observations to within specified errors. The estimated model state is evaluated both by comparisons with observations and by checking the controls, the momentum balances, and the representation of small-scale features that were not well sampled by the observations used in the assimilation. As part of these checks, the estimated controls are smoothed and applied in independent model runs to check that small changes in the controls do not greatly change the model hindcast. This is a simple ensemble-based uncertainty analysis. In addition, the original and smoothed controls are applied to a version of the model with doubled horizontal resolution resulting in a broadly similar "downscaled'' hindcast, showing that the adjustments are not tuned to a single configuration (meaning resolution, topography, and parameter settings). The time-evolving model state and the adjusted controls should be useful for analysis or to supply the forcing, initial, and boundary conditions for runs of other models.