A Multivariate Reduced-order Optimal Interpolation Method and its Application
303
(a)
(b)
Fig. 15.8 50-m current velocities in (a) 1/8°, 31-level and (b) 1/4°, 19-1evel configurations
in the Crete-Rhodes area on Oct. 22, 1993. In both cases, the model was last corrected on
Oct. 15 using the same altimeter data.
Proven<;:al-Catalan current (Echevin et al., 2000). In the GCM ensemble mns, initial conditions have been randomly perturbed along the dominant vertical EOF and
an ensemble of 21 model mns have been conducted to forecast time. Not too surprisingly, the largest forecast error variances in the Western bas in were found in the
high variability areas such as the Aigerian current. The ensemble method shows
barotropic error growth in the Aigerian Current and in the Alboran gyres (a result
compatible with the results shown in this chapter), and baroclinic/tracer error
growth in the Lyons gyre where deep convection occurs in late winter 1993. The
method is very costly, but is very general in exploring the validity of the simple
hypotheses usually assumed in OI.
Acknowledgments
Support of the authors under European Union contracts MAS3-CT96-0051,
ENV4-CT98-0744, and MAS3-CT98-0171 is acknowledged.
303
(a)
(b)
Fig. 15.8 50-m current velocities in (a) 1/8°, 31-level and (b) 1/4°, 19-1evel configurations
in the Crete-Rhodes area on Oct. 22, 1993. In both cases, the model was last corrected on
Oct. 15 using the same altimeter data.
Proven<;:al-Catalan current (Echevin et al., 2000). In the GCM ensemble mns, initial conditions have been randomly perturbed along the dominant vertical EOF and
an ensemble of 21 model mns have been conducted to forecast time. Not too surprisingly, the largest forecast error variances in the Western bas in were found in the
high variability areas such as the Aigerian current. The ensemble method shows
barotropic error growth in the Aigerian Current and in the Alboran gyres (a result
compatible with the results shown in this chapter), and baroclinic/tracer error
growth in the Lyons gyre where deep convection occurs in late winter 1993. The
method is very costly, but is very general in exploring the validity of the simple
hypotheses usually assumed in OI.
Acknowledgments
Support of the authors under European Union contracts MAS3-CT96-0051,
ENV4-CT98-0744, and MAS3-CT98-0171 is acknowledged.