HYCOM
419
surface wind stress, air temperature, and specific humidity (from dewpoint
temperature and sea level pressure) in addition to shortwave and longwave
radiation. Surface heat flux is calculated using NOGAPS fields and the Kara
et al. (2002) bulk parameterization of latent and sensible heat flux, which
uses model SST.
Mostly because of its simplicity, robustness, and low computational
costs, operational ocean prediction systems around the world (NLOM,
MERCATOR, FOAM, etc.) are presently using Optimal Interpolation (OI)
based data assimilation techniques. For the current 1/12
o Atlantic HYCOM
ocean forecasting system, we have adopted a similar approach by selecting
an OI technique with Cooper and Haines (1996) for downward projection of
SSH from altimetry [see Chassignet et al. (2005) for details]. Real time
satellite altimeter data (Geosat-Follow-On (GFO), ENVISAT, and Jason-1)
are provided via the Altimeter Data Fusion Center (ADFC) at
NAVOCEANO to generate the two-dimensional Modular Ocean Data
Assimilation System (MODAS) SSH (1/4
ż
) analysis (Fox et al., 2002) that is
assimilated daily. The MODAS analysis is an OI technique which is using a
complex covariance function that includes spatially varying length and time
scales as well as propagation terms derived from many years of altimetry
(Jacobs et al., 2001). The model sea surface temperature is relaxed to the
daily MODAS 1/8
ż SST analysis which uses daily Multi-Channel Sea
Surface Temperature (MCSST) data derived from the 5-channel Advanced
Very High Resolution Radiometers (AVHRR) – globally at 8.8 km
resolution and at 2 km in selected regions. The system runs once a week
every Wednesday and consists of a 10-day hindcast and a 14-day forecast.
At the present time, evaluation of the model outputs relies on systematic
verification of key parameters and computation of statistical indexes by
reference to both climatological and real time data, and, in a delayed mode,
to quality controlled observations. The accuracy of data assimilative model
products is theoretically a non-decreasing function of the amount of data that
is assimilated. A degradation caused by assimilation generally indicates
inaccurate assumptions in the assimilation scheme. While models can be
forced to agree with observations (e.g., by replacing equivalent model fields
with data), improvements with respect to independent observations are not
trivial. An assessment of model improvement (or lack of degradation) with
respect to unassimilated, independent measurements is therefore an effective
means of assessing the performance of an assimilation system. Variances of
these model-data differences serve as common measures of the estimation
accuracy.
For the evaluation of flow accuracy and water mass
characteristics, we follow the guidelines put forward by the international
GODAE metrics group as well as the validation tests commonly used at the
operational centers before official transition to operational use. In the
419
surface wind stress, air temperature, and specific humidity (from dewpoint
temperature and sea level pressure) in addition to shortwave and longwave
radiation. Surface heat flux is calculated using NOGAPS fields and the Kara
et al. (2002) bulk parameterization of latent and sensible heat flux, which
uses model SST.
Mostly because of its simplicity, robustness, and low computational
costs, operational ocean prediction systems around the world (NLOM,
MERCATOR, FOAM, etc.) are presently using Optimal Interpolation (OI)
based data assimilation techniques. For the current 1/12
o Atlantic HYCOM
ocean forecasting system, we have adopted a similar approach by selecting
an OI technique with Cooper and Haines (1996) for downward projection of
SSH from altimetry [see Chassignet et al. (2005) for details]. Real time
satellite altimeter data (Geosat-Follow-On (GFO), ENVISAT, and Jason-1)
are provided via the Altimeter Data Fusion Center (ADFC) at
NAVOCEANO to generate the two-dimensional Modular Ocean Data
Assimilation System (MODAS) SSH (1/4
ż
) analysis (Fox et al., 2002) that is
assimilated daily. The MODAS analysis is an OI technique which is using a
complex covariance function that includes spatially varying length and time
scales as well as propagation terms derived from many years of altimetry
(Jacobs et al., 2001). The model sea surface temperature is relaxed to the
daily MODAS 1/8
ż SST analysis which uses daily Multi-Channel Sea
Surface Temperature (MCSST) data derived from the 5-channel Advanced
Very High Resolution Radiometers (AVHRR) – globally at 8.8 km
resolution and at 2 km in selected regions. The system runs once a week
every Wednesday and consists of a 10-day hindcast and a 14-day forecast.
At the present time, evaluation of the model outputs relies on systematic
verification of key parameters and computation of statistical indexes by
reference to both climatological and real time data, and, in a delayed mode,
to quality controlled observations. The accuracy of data assimilative model
products is theoretically a non-decreasing function of the amount of data that
is assimilated. A degradation caused by assimilation generally indicates
inaccurate assumptions in the assimilation scheme. While models can be
forced to agree with observations (e.g., by replacing equivalent model fields
with data), improvements with respect to independent observations are not
trivial. An assessment of model improvement (or lack of degradation) with
respect to unassimilated, independent measurements is therefore an effective
means of assessing the performance of an assimilation system. Variances of
these model-data differences serve as common measures of the estimation
accuracy.
For the evaluation of flow accuracy and water mass
characteristics, we follow the guidelines put forward by the international
GODAE metrics group as well as the validation tests commonly used at the
operational centers before official transition to operational use. In the