Fróðskaparrit - 01.01.2007, Page 144

142
VALIDATION OF THE ECMWF ANALYSIS WAVE DATA
FOR THE AREA AROUND THE FAROE ISLANDS
T is generally larger than the bias of
Tm()r This suggests that the wave-period-
bias is not primarily due to shortcomings
in the wind fíeld, but rather due to underes-
timation of the energy in the high fre-
quency part of the spectrum by the EW4
model. The EW4 underestimation of the
energy content in the higher frequencies is
also apparent in the spectral time series
(Figure 6).
The fact that the seasonal under predic-
tion is not clear in the F8 wind series but is
present in the WVD-4 H 0 series, could in-
dicate that the F8 station does not always
represent the 10 m wind correctly.
It was observed that the spectral peak
had a tendency to be broader than the ob-
served peak (Figures 6 and 7). This is a
commonly observed mismatch between
operational wave models and observed
spectra, and can to a large extent be attrib-
uted to the defíciencies in the DIA approx-
imation (van Vledder, 1990; Forristall and
Greenwood, 1998).
It was also noted that some swell inci-
dences (see example in Figure 6) were
smooth and diffuse in the EW4 results. The
missing swell event in the EW4 data, such
as the one on the 25’th August 2000, can
be caused by different causes. Previous in-
spections have found that shortcomings in
operational wind models are a common
reason for underestimation of wave events
(Komen et al., 1994; Cardone et al., 1996).
Detailed investigations omitted here, indi-
cate that the swell generating area is esti-
mated to be some 1700 Km west to south-
west ofWVD-4, that is roughly to 60 EW4
model grid lengths of0.25°, so propagation
effects might also play a role in this case.
The arriving swell is also observed to be
quite narrow in frequency and direction
(Figure 7), which again could make the
representation and advection of the swell
more difficult within the limitations of the
EW4 resolution and numerical propaga-
tion. It has previously been mentioned e.g.
by Wingeart et al. (2001), that underre-
ported swell events by operational WAM
models can be due to the diffusive lst order
propagation used in WAM. In this particu-
lar case inspected here, the missing swell
energy is so evident that the primary cause
must be due to an underestimation in the
forcing wind fíeld.
Recent model improvements at ECMWF
At ECMWF, there is a sustained effort at
improving model performances and assim-
ilation techniques. For this reason, it is
necessary to give an outline of the most
important model upgrades introduced into
the operational model after the time peri-
ods inspected here.
In March 2004, a scheme for the inclu-
sion of the effect of unresolved bathymet-
ric features on wave propagation was in-
troduced. The impact of this scheme is bet-
ter model predictions downwind froin sub
grid features such as the Faroe Islands
(Janssen et al., 2005). A re-formulation of
the wave dissipation source term was im-
plemented in April 2005. The impact of
this re-formulation on spectral wave pa-
rameters such as mean wave periods was
highly beneficial (Janssen et al., 2005 and
Bidlot et al, 2007). Finally in February
2006, the atmospheric model spectral reso-