A31B-01
Assimilation of satellite ocean chlorophyll data for biogeochemical state estimation - univariate and multivariate aspects
Chlorophyll data from the Sea-viewing Wide Field-of-view Sensor (SeaWiFS) is assimilated into the three- dimensional global NASA Ocean Biogeochemical Model (NOBM) for the period 1998-2004. The ensemble-based SEIK filter is applied in a multivariate configuration. It is used here with a localized analysis and simplified by the use of a constant covariance matrix. In addition, an online bias estimation algorithm is applied. The multivariate assimilation updates the four phytoplankton groups of the model as well as nutrient fields. With assimilation, the chlorophyll estimates become superior to both the free-run model and SeaWiFS data. However, the results are less clear for the nutrients. We discuss the behavior and issues involved by the multivariate assimilation process.
A31B-02
A data assimilation OSSE for assessing uncertainty and utility of soil moisture retrievals
Soil moisture is of interest for a variety of reasons, including water cycle studies and the initialization of weather and climate forecasts, but it is also difficult to observe at the global scale. Satellite retrievals of surface soil moisture are subject to large uncertainties because the physical processes that relate brightness temperature to soil moisture are difficult to parameterize, and because the necessary parameters are difficult to obtain on the global scale. An important question for the design of new satellite sensors is just how uncertain satellite retrievals can be and still add useful information to a land data assimilation system. In this paper, we address this question with an Observing System Simulation Experiment (OSSE) that is based on high-resolution (1 km) "true" soil moisture fields and associated passive microwave brightness temperatures from a long-term integration of the TOPLATS land surface model over the Red-Arkansas river basin. From the true fields, we simulate many different retrieval data sets at a typical satellite footprint scale (36 km). The various retrieval data sets reflect various realistic sources of uncertainty with different error structure and magnitude. We use the NASA Catchment land surface model (CLSM) to set up different land model scenarios that reflect varying degrees of uncertainty in model parameter and forcing data. The various simulated retrieval data sets are then assimilated into the various model scenarios with an Ensemble Kalman filter (EnKF) in a suite of data assimilation experiments. The EnKF is fitted with a simple and effective method of bias removal (cumulative distribution function matching) and an adaptive component for on-line model error estimation. The adaptive feature is critical because correct assessment of the utility of the retrievals depends on the near-optimal performance of the assimilation system for each data assimilation experiment. Earlier research has shown that off-line estimation of model error parameters outside of the cycling assimilation system is not a viable strategy. Calibration of the model parameters by repeating the each data assimilation experiment until acceptable model error parameters are found is not computationally feasible. Finally, the quality of the assimilation estimates (with respect to the synthetic truth) is compared with that of a baseline integration of the Catchment model without assimilation. This procedure permits us to quantify the maximum level of uncertainty in the satellite retrievals for which information is still added in the assimilation, depending on the application. Performance measures include the traditional absolute (RMS) error, which is important for water cycle studies, and the time series correlation coefficient. The latter measures how well (scaled) anomalies are estimated, which contain the key information for forecast initialization.
A31B-03
Assimilation of Temperature, Salinity, and Sea Surface Height Data into the GMAO Ensemble Kalman Filter and its Impact on Seasonal Hindcast Skill
A multivariate ensemble Kalman filter (EnKF) is used to assimilate in situ temperature and salinity observations and remotely sensed altimeter data into a global OGCM. To evaluate the impact of the state-dependent multivariate covariances from the EnKF, the EnKF sub-surface analysis fields for temperature, salinity, and zonal velocity are compared to the GMAO production ocean analysis obtained with a) temperature optimal interpolation (OI-T) complemented with a salinity correction to preserve the (T,S) relationship, and b) temperature and salinity optimal interpolation (OI-TS) where Argo salinity and synthetic salinity from TAO, XBT, and Pirata are assimilated. Observations from TAO, Pirata, and Argo, and independent data from ADCPs, and CTDs are used to validate the analyses. The ensemble of ocean states is then used to initialize the ocean model used in the GMAO coupled forecasting system. Three experiments are performed to evaluate the impact of different observing systems on hindcast skill. -Exp1 uses the following data: Temperature profiles from TAO, Pirata, Argo, and XBTs Salinity profiles from Argo Synthetic salinity observations estimated from temperature profiles and climatological T-S relations Sea surface height anomalies from Topex and Jason altimeter data -Exp2 excludes the salinity data from Argo -Exp3 excludes the altimeter data The hindcast skill of the forecast system initialized with the EnKF is compared to that of the production forecasting system in which univariate optimal interpolation (OI) is used to initialize the OGCM prior to coupling. Hindcast skill for sea surface temperature and sea surface height in the equatorial Pacific Nino regions are assessed.
A31B-04
The AFES-LETKF Experimental Ensemble Reanalysis: ALERA
The local ensemble transform Kalman filter (LETKF) is applied to the AFES (AGCM for the Earth Simulator) to perform an experimental reanalysis. The reanalysis is called ALERA, standing for AFES-LETKF Experimental ensemble ReAnalysis. Using the system developed by Miyoshi and Yamane (2007), real observations except satellite radiances are assimilated for more than 19 months since 1 May 2005. The AFES-LETKF data assimilation cycle has been stable for over a year. The analysis fields are compared with the NCEP/NCAR reanalysis, a de facto standard reanalysis; the verification indicates that the ALERA data actually reproduce the nature atmosphere fairly well. Moreover, ALERA ensemble spreads capture the analysis uncertainties well, especially large uncertainties in the SH and upper levels where satellite radiances play an important role.
A31B-05
High resolution ensemble forecasting for the Gulf of Mexico eddies and fronts
As oil production moves further into deeper waters, the costs related to strong current hazards are increasing accordingly, and accurate three-dimensional forecasts of currents are urgently needed. To be useful, models have to locate eddies and fronts to an accuracy of 30 km at a nowcast stage, which is almost impossible to accomplish with the use of satellite data of the same accuracy. The use of stochastic forecast allows us to give confidence of our prediction. We are using a nested configuration of the Hybrid coordinate ocean model (HYCOM), where the TOPAZ system, which covers the Atlantic and the Artic, gives lateral boundary condition to a high-resolution (5km) model of the Gulf of Mexico (GOM). TOPAZ is a real-time forecasting coupled ocean-ice model, which assimilates sea level anomaly (SLA), sea surface temperature, and sea ice concentration, with the ensemble Kalman filter. The high- resolution model assimilates SLA using the ensemble optimal interpolation, which updates accordingly the currents, salinity, temperature, and layer interface at all depths. Here, we evaluate the ensemble forecast capabilities of our high-resolution model, for eddy Extreme that has been observed from altimeters around the 15th of July. We run 6 successive ensemble runs composed of 10 members of equal likelihood. Members differ by perturbations of the initial state, of the lateral boundary conditions, and of the atmospheric boundary conditions. We have started the experiment 1 month prior to the shedding event, because it was the time necessary for perturbation of boundary conditions to spread uniformly and reach a significant level across the GOM. The ensemble reproduces well the dynamics of the eddy shedding and produces a significant spread at the boundary of the eddy, but underestimates the RMS error of the SLA. Prior to the shedding time, the error growth increase, induced by the highly non-linear growth of cyclonic eddies at the boundary of the Loop Current. Additionally, this ensemble has allowed for optimization of data assimilation parameters depending on the range of the forecast.
A31B-06
A General-Purpose Ensemble Assimilation Facility: DART
Although it is trivial to develop an ensemble data assimilation facility for an atmospheric prediction model, the
standard ensemble
algorithms have a number of shortcomings. They are subject to most
error sources that impact more traditional assimilation methods
and also to sampling error from small ensemble sizes. A general
purpose ensemble facility must provide additional adjunct algorithms
that can deal adaptively with these errors. The Data Assimilation
Research Testbed facility developed at NCAR includes a wide range
of novel algorithms. To deal with sampling error, DART includes
a hierarchical Bayesian algorithm that can automatically recommend
a multi-variate, spatially anisotropic localization. Hierarchical
Bayesian algorithms for spatially- and temporally-varying
inflation are also included. A methodology for adaptive
thinning is available to efficiently assimilate observations where they are
dense. Both stochastic and deterministic ensemble filters, as well
as novel hybrid particle/ensemble filter algorithms are available
in the facility. This poster will provide an overview of these
algorithms and examples of their application in global NWP assimilations.
http:www.image.ucar.edu/DAReS/DART
A31B-07
Strong Adjoint Sensitivities in Tropical Eddy-Permitting Variational Data Assimilation
A variational data assimilation system has been implemented for the tropical Pacific Ocean for an eddy-permitting regional implementation of the MIT general circulation model (MITgcm). The model uses realistic topography with parameterizations for the surface boundary layer (KPP) and open boundaries at the south and north, as well as in the Indonesian throughflow. The adjoint method is used to adjust the model to observations in the tropical Pacific region using control parameters which include initial temperature and salinity, temperature, salinity and horizontal velocities at the open boundaries, and twice-daily surface fluxes of momentum, heat and freshwater. The model is constrained with most of the available datasets in the tropical Pacific, including climatologies, TAO, ARGO, XBT, and satellite SST and SSH data. The forward model runs exhibit strongly growing flow instabilities in the regions of high kinetic energy and low planetary potential vorticity gradient. The growth of these perturbations is limited by nonlinearities once they reach finite size, meaning that the high linear growth rates do not apply for long time periods. This poses a technical problem for adjoint-based assimilation, which depends on the linearized sensitivities to adjust the controls. Relative to the forward model runs, increased viscosity and diffusivity terms are used in the adjoint model runs to avoid large sensitivities related to the flow instabilities present in the high-resolution model. This talk will discuss some of the technical aspects and show results for 1 year assimilation period.
A31B-08
Online Bias Correction of ARGO Temperature and Salinity Data in the GMAO Ocean Ensemble Kalman Filter
Past experiments with the online bias correction (OBC) algorithm used in the GMAO ocean ensemble Kalman fiter (EnKF) have demonstrated the importance of applying OBC when sea surface height (SSH) anomalies are assimilated. In this case, OBC is necessary to update the SSH climatology used to reconstruct the SSH field. While correcting the model climatology by means of OBC is less important when in situ temperature, salinity or current data are assimilated because these data are not ingested in their anomalous form, analyses reveal that the corresponding model fields are systematically biased, even though a standard (i.e., without OBC) assimilation can partially compensate the biases. While we were always interested in assessing the usefulness of OBC in the assimilation of temperature or salinity data, we were prevented in doing so until recently because the OBC algorithm is ineffective with sparse observations. The near exponential increase in the number of assimilate-able ARGO measurements has recently changed that reality. Hence, we examine how applying OBC in the assimilation of ARGO temperature and salinity data affects the quality of the analysis products.
A31B-09
Impact of the Extratropics EOF-Perturbation Modes in the CPTEC Ensemble Weather Forecast
The EOF-based perturbation method, applied operationally at the Center for Weather Forecast and Climate Studies (CPTEC) to produce perturbed initial conditions, is used in order to perturb the extratropics. The influence that these new perturbations cause in the CPTEC ensemble weather forecast is evaluated through statistical indexes. The results suggest that the application of the EOF-method to perturb the extratropics, simultaneously with the tropical perturbations used currently may help to improve the quality of the ensemble forecast of the CPTEC. Preliminary results about the utilization of the breeding of growing modes perturbation method are also presented.