Center for Observations, Modeling and Prediction at Scripps (COMPAS)

One line of COMPAS research focuses on processes that occur at small scales, in particular turbulent mixing of momentum, heat, dissolved gases, nutrients, and pollutants. Though turbulent mixing occurs at the smallest scales in the ocean, it governs the distribution of mass and momentum at all scales. In much of the open ocean, internal tides (internal waves generated by tidal flow over topography) supply a large fraction of energy ultimately available for turbulent mixing The global pattern of mixing produced by breaking of the internal tide is controlled both by the geography of internal-tide generation sites and the complex and often subtle nonlinear processes that drain energy from internal tides as they propagate across ocean basins. Recent numerical work by MacKinnon and Winters (2005, 2006) has shown that within a certain latitude range, a particular nonlinear interaction known as Parametric Subharmonic Instability (PSI) is particularly efficient at transferring energy from the internal tide to highly sheared smaller-scale waves of half the frequency (and ultimately turbulent mixing).

This numerical study directly led to a successful NSF-funded fieldwork program to study PSI in the North Pacific in the spring of 2006 [Alford et al 07]. Closer to the surface, mixing in the shallowest layer of the ocean mediates exchange between ocean and atmosphere. Recent observational evidence suggests that Langmuir Circulation (wind-driven 'rolls' that efficiently mix tracer properties throughout the surface boundary layer) may be sensitive to the `background' internal wave field. The convergences an divergences associated with internal waves may have a vortex stretching effect on the Langmuir cells, enhancing their transport ability. Jerome Smith (MPL/SIO) and MacKinnon are working on an NSF-funded project to develop a new Large Eddy Simulation (LES) numerical model that will be used to explore this interaction. Two postdocs (A. Tejada-Martinez, now faculty at the University of South Florida, and J. Polton, SIO) are or have been employed by this project. The first paper, describing details of the new numerical techniques, is in preparation.

Work by SIO graduate student H. Seo, with advisors A. Miller, J. Roads and M. Kanamitsu has gone toward developing a high-resolution regional coupled model using the Regional Spectral Model (RSM) atmosphere and the Regional Ocean Modeling System (ROMS) ocean, the first of its kind. This model is useful in investigating small to mesoscale ocean-atmosphere interactions throughout the the global ocean. Thus we address many interesting aspects of the coupled feedback arising from the ocean eddies and the synoptic atmospheric events.

Four results emerged: 1) The coupled model reproduces aspect of the observed atmospheric response to the mesoscale eddies and the filamentous structures of the upwelling system in the CCS. 2) The model confirmed the observational evidences that the central American gap winds can induce thermocline responses in the eastern Pacific warm pool, and in turn punch a permanent dry hole in the summertime mean ITCZ. 3) The model reveals the observed response of the atmospheric boundary layer to the tropical instability waves (TIWs) in the equatorial oceans. These experiments revealed that the ocean-atmosphere coupling due to the TIWs damps the energy of the TIWs. This was the natural questions that people begin to ask from the observations: how do the atmospheric responses feed back on to the ocean eddies? 4) When focusing on climate biases exhibited by the current CGCMs and ask how these small-scale feedback can give rise to the large-scale climate variability in the tropical Atlantic Ocean. This indicates that both the mesoscale SST features in the ocean and the synoptic-scale atmospheric easterly waves originating from Africa have to be realistically resolved in the coupled climate models in order to generate improved mean and seasonal variability of the large-scale SST and the ITCZ.

In another effort to understand ocean climate and variability, models and observations can be brought together to give a description of the ocean that is better than either alone. Detlef Stammer brought the ECCO project (http://www.ecco-group.org) to SIO, producing 10-year and longer hindcasts of global ocean state synthesized from most of the observations available, including remote sensing and in situ data (Stammer et al 04). Several assimilation projects have remained behind after his departure. These projects include a local grid by SIO PhD Student Sarah Zedler. In each case, the assimilation adjusts the MITgcm to fit the available observations using control parameters which include initial conditions, temperature, salinity, and velocity on open boundaries and time-dependent surface fluxes of heat, salt, and momentum. In particualar Sarah Zedler is studying a regional version of the MITgcm at high horizontal (1/6 degree) and vertical (40 layers) resolution, driven by idealized hurricane forcing. The work has two goals: (1) to describe deep response to a hurricane and (2) to assess regional differences in the upper and deep ocean temperature response by comparing fields for different initial temperature fields and latitude ranges representative of mid-latitudes in the North Atlantic, the Caribbean, and the eastern Pacific.

The model grid has dimensions of 420x240x40, and requires 4 hours of run time on 60 processors to generate one 10 day integration with a 1-hour timestep. For Zedler's thesis work, several thousand CPU-hours of computing time have already been used. Without the cluster, this topic would probably not have been practical for a PhD. As found in previous studies, the deep response to a hurricane consists of Ekman upwelling throughout the depth of the entire water column during the storm-forced part of the response. During the post storm response, for storms with translational speeds greater than the fastest moving internal gravity wave, there is generation of a wake of internal waves at nearly twice the inertial frequency, and an adjusting geostrophic wake. The temperature response for a hurricane consists of cooling on the right hand side of the storm forced predominantly by shear-instability mixing, and secondarily by latent heat flux release to the atmosphere. The deep response consists of Ekman upwelling during the storm forced period, and vertical pumping during the post-storm period. In addition, postdoctoral researcher Sangwook Parke conducted a sensitivity study of the large scale oceanic response to atmospheric aerosols emanating from southeast Asia. Resulting from diminished short wave radiation reaching the ocean surface were changes in temperature, salinity, sea level, and velocity, especially in the Indian and western Pacific Ocean.

PhD student Elisabeth Douglass is studying interannual variability of the circulation in the Northeast Pacific Ocean through a joint analysis of expendable bathythermograph (XBT) and expendable conductivity-temperature-depth (XCTD) data, satellite altimetry, and output from the MITgcm constrained by (assimilating) ocean data. XBT temperature profiles with high spatial resolution are available in the eastern North Pacific along two repeated transects. Geostrophic velocities from XBT data are compared with geostrophic velocities from model output as well as the full model velocity fields. Correlations in variability on interannual time scales between transport in the subpolar gyre and in the subtropical gyre are present in both model output and data (douglassetal06). The nature of the variability, and its relation to the changes of the strength of the North Pacific Current (NPC), which supplies the water for both gyres, are focuses of this study. Interannual variability in gyre transport is found to be related to both the bifurcation of the NPC, resulting in an anticorrelation in transport between the two gyres, and to variations in NPC strength, resulting in simultaneous changes in the two gyres. The dominant signal is found to be a long-term increase in the NPC, which results in a strengthening of the subtropical gyre. Possible connections with local-scale wind stress changes and with the ENSO are also explored. A 50 year ocean reanalysis, recently produced as part of the ECCO project, is also being analyzed in similar ways.

Douglas is now running the model in the Pacific Ocean north of 26S, to improve the fit to data at higher resolution and with better spatial covariances than the original ECCO global assimilation. The regional assimilation should achieve a better match between the data and the model, and allow for better interpretation and understanding of the processes identified by the observations. Access to a dedicated computer is crucial, since rapid turnaround is necessary when making a new application work. The goal of the assimilation is a closed heat and salt budget for the North Pacific, including interannual variation, part of understanding the meridional heat transport in the North Pacific.

Return to Research

Hydrologic simulations conducted by Hugo Hidalgo, with Mike Dettinger and Dan Cayan are using the VIC distributed, large basin hydrologic model have been focused on providing better understanding of the processes controlling spring-summer drought and their sensitivities to changes in precipitation and temperature. Preliminary results have shown that winter precipitation is an important determinant of spring-summer drought not only in the snow-covered areas, but also in a large part of arid and semi-arid regions of the western United States. The effect on the land surface of temperature fluctuations, which bears on how severe will be the impacts of climate warming, appears to vary considerably over the western region, depending on whether a region has high actual evapotranspiration (AET) to potential evapotranspiration (PET) ratios and thus has energy limited evapotranspiration or whether it has low AET/PET ratios and thus has water limited evapotranspiration. The range and magnitude of these different shades of climate sensitivity need still to be worked out, by exploring how the model responds to ensembles of climate variability and climate change simulations,

We are using “COMPAS” to develop the new Experimental Climate Prediction Center (ECPC) Coupled Prediction Model (ECPM), investigate its skill and apply it to seasonal forecasting. The ECPM includes the Jet Propulsion Laboratory (JPL) version of the Massachusetts Institute of Technology (MIT) ocean model coupled to the ECPC version of the National Centers for Environmental Research (NCEP) Atmospheric Global Spectral Model (GSM). The adjoint and forward versions of the MIT model forced with the NCEP atmospheric analyses are routinely used at JPL for ocean state assimilation. An earlier version of the GSM was used for the NCEP/DOE Reanalysis-2 project and for operational seasonal forecasts at NCEP.

COMPAS has provided computing power for several key features of the Scripps Experimental Climate Prediction (ECPC) research and applications, spearheaded by J. Roads and M. Kanamitsu. On a routine basis we make 2-tier (uncoupled) seasonal forecasts for the International Research Institute (IRI) and National Centers for Environmental Prediction (NCEP) as well as coupled seasonal forecasts, using the COMPAS facility. These forecasts are critically examined in order to not only to eventually develop better forecasts and forecast modules for IRI and NCEP but also to better understand how to make better predictions. In that regard, we are developing a number of increased resolution regional models that can down scale

the global predictions and thus provide further information for local applications. Not only are the seasonal 2-tier predictions being down-scaled on the COMPAS cluster, we are also participating in state and national programs interested in down scaling global projections of potential changes due to increased greenhouse gases and other anthropogenic changes. Both coupled and uncoupled down scaling simulations are being developed and explored. The experimental seasonal forecast is providing 12 member 7-month forecast to IRI and NCEP, once a month, and has been judged to be one of the best and most useful such forecasts. The experimental coupled forecast is providing single member 1 year forecast to the public, via http://ecpc.ucsd.edu/COUPLED/CM/coupled.html. Many historical hindcasts were performed to provide dataset to evaluate forecast skill and for a seasonal predictability study. The COMPAS cluster machine is extensively utilized to optimize the forecast model and to test new physical processes. Model runs formed the basis for several studies, including the effective use of ensemble forecasts, the importance of land surface processes on the seasonal forecasts, cloud parameterizations, the importance of SST for the historical reanalysis, mechanisms involved in the Indian Dipole.

Another set of runs performed on COMPAS is aimed at getting more statistically robust results, especially for individual predictions. We are currently performing 5 member ensemble predictions starting each November and May of the recent 14 years period (the JPL ocean state analysis is updated monthly and is available from 1993-present). A single initial condition is used for the ocean initial condition since only single oceanic initial conditions are available each month. However, the multiple initial conditions for the atmosphere ensemble will be extracted from R-2 from every 12-hour initial state nearest to the beginning of each month. ECPC coupled models appear to be a useful contribution to a multiple model prediction, and lead to increased multiple model prediction skill. In addition, these global coupled model simulations and forecasts are beginning to be used as boundary conditions for regional coupled model simulations and forecasts. In particular, we are beginning to develop a corresponding regional coupled atmosphere-ocean model that can be used in coastal regions.

COMPAS was instrumental in the California Energy Security Project, conducted by D. Pierce and T. Barnett. This project evaluated the ability of climate forecasting to generate forecasts of use to the energy industry. In conjunction with workers at Cal-ISO, SDG&E, and Pacificorp (all of who produce and/or coordinate the transmission of energy in Calilfornia), experimental forecasts were produced of the California Delta Breeze, peak load days in San Diego, and summer irrigation pump loads in Idaho. COMPAS facilities supported computations and visualization effort.

COMPAS was also involved in a recent study of the upper tropospheric water vapor, which compared IPCC AR4 model representations of upper tropospheric water vapor to observations from the AIRS satellite system. A difficult problem with satellite sensing of atmospheric quantities is that of sampling. For instance, the AIRS satellite whose data we used cannot sample highly cloudy scenes. In order to determine what effect this might have on the reported humidity field, we used the output of an existing CCSM3 run that had been performed on COMPAS and had saved cloud and humidity data every 20 minutes.

Return to Research

Model runs made using the COMPAS facility were used by Pierce and Barnett to study the predictability of low-frequency variability in the Pacific Ocean. Decadal climate variability in the North Pacific ocean is associated with strong climatic signals over North America, with either warmer or colder winters, depending on the phase of the variability. Idealized model experiments were conducted to examine the decline of predictability with time over the North Pacific ocean. Model analysis showed there is a region with enhanced potential predictability over the North Pacific, but the idealized model experiments did not show that beginning with the same ocean initial state, and perturbed atmospheric conditions, would lead to useful prediction on the interannual time scale.