Section 8.3: Multi varia te Analysis
157
model, and two GeMs of increasing resolution (see Frankignoul et al., 1994,
for details of the various models), in the same context of 1982-1984 thermodine depth fluctuations. The approximate error bars on the misfits are based
on the non-central T 2 , and they should only be taken into ac count once in
the intercomparison since the observation and the forcing errors are the same
in each case. Although Ho is rejected in all the cases, the differences in model
performance are significant, and consistent with our expectations that more
complex models perform generally better than simpler ones.
157
model, and two GeMs of increasing resolution (see Frankignoul et al., 1994,
for details of the various models), in the same context of 1982-1984 thermodine depth fluctuations. The approximate error bars on the misfits are based
on the non-central T 2 , and they should only be taken into ac count once in
the intercomparison since the observation and the forcing errors are the same
in each case. Although Ho is rejected in all the cases, the differences in model
performance are significant, and consistent with our expectations that more
complex models perform generally better than simpler ones.