Professional Documents
Culture Documents
Inazu 2009
Inazu 2009
(Received 19 September 2008; in revised form 8 December 2008; accepted 16 January 2009)
An accurate prediction of ocean tides in southeast Alaska is developed using a re- Keywords:
gional, barotropic ocean model with a finite difference scheme. The model skill is ⋅ Bathymetry,
verified by the observational tidal harmonics in southeast Alaska including Glacier ⋅ bottom friction,
Bay. The result is particularly improved in Glacier Bay compared to the previous ⋅ Glacier Bay,
⋅ tidal energy
model described by Foreman et al. (2000). The model bathymetry dominates the model
dissipation.
skill. We re-estimate tidal energy dissipation in the Alaska Panhandle and suggest a
value for tidal energy dissipation of 3.4 GW associated with the M2 constituent which
is 1.5 times the estimation of Foreman et al. (2000). A large portion of the M2 energy
budget entering through Chatham Strait is dissipated in the vicinity of Glacier Bay.
Moreover, it is shown that the developed model has the potential to correct the ocean
tide loading effect in geodetic data more efficiently than the model of Foreman et al.
(2000), especially around Glacier Bay.
335
(Skagway) are 188.2 (203.5) cm and 270.1° (283.9°), re-
spectively, while the modeled amplitude and phase are
250.8 (376.3) cm and 276.2° (79.6°), respectively. On the
other hand, Foreman et al. (2000, henceforth F00) devel-
oped an accurate ocean tide model and estimated tidal
energy dissipation due to bottom friction in the northeast
Pacific Ocean including the Alaska Panhandle; however,
the F00 model was not validated in Glacier Bay. Larsen
et al. (2004) conducted ocean tide observations in the
vicinity of Glacier Bay, and the observed tidal harmonics
do not show good agreement with the F00 model; the
observed M2 tidal elevation amplitude and phase at Com-
posite Island are 204.3 cm and 299.7°, respectively, while
the F00 model shows the amplitude and phase to be 176.9
cm and 286.6°, respectively. Therefore, the model result
of F00 may be inappropriate for estimating the tidal en-
ergy dissipation rate around Glacier Bay and removing
the ocean tide components from the geodetic data. In this
study we develop a more accurate ocean tide model, esti-
mate the tidal energy dissipation in the Alaska Panhandle
including Glacier Bay, and show the contribution of the Fig. 1. Computational domain, tide gauges, an ocean bottom
developed ocean tide model to the geodetic aspect. The pressure gauge, GPS stations used in this study, and bottom
topography in meters drawn from the GINA bathymetric
primary purpose is the accurate ocean tide modeling at
data. Open circles, a black circle, and diamonds denote the
the regional scale rather than smaller scales of each harbor
tide gauges, the ocean bottom pressure gauge, and the GPS
and inlet. stations, respectively. An example of the model grids around
Glacier Bay is shown in upper right of the figure.
2. Model Description
59 N
58 N
Latitude
57 N
56 N
55 N
M2 K1
accuracy, rather than the dissipation parameters and the not expected to be drastically improved compared to the
open boundary condition of the tidal elevation. previous F00 model. We aspire to achieve the accuracy
of the modeled tidal harmonics within several centimeters
2.3 Tidal elevation condition for each constituent in the Alaska Panhandle. The advan-
The open boundary condition of the model domain tage of merging the F00 and F93 models for the open
must be as accurate as possible. The F00 model covers boundary condition is shown quantitatively in Subsec-
the northeast Pacific Ocean, and seems to be appropriate tion 3.2.
in its representation of the tidal elevation thanks to the
assimilation of satellite altimeter data in the open sea. 3. Model Validation
Foreman et al. (1993, henceforth F93) computed the ocean
tides surrounding the Queen Charlotte Islands with spa- 3.1 Observed tidal harmonics
tial resolutions finer than those used in F00. We compare Ocean tides have been observed to obtain tidal har-
both the results and ascertain that the F93 model is more monics at permanent and temporary coastal sites, and an
accurate in Dixon Entrance than the F00 model (Table ocean bottom pressure gauge (OBPG). The coastal sta-
1). The F93 model shows the M2 constituent more accu- tions are administered by NOAA or the University of
rately than the F00 model, while the accuracy of the K1 Alaska Fairbanks (UAF, Larsen et al., 2004). The OBPG
constituent modeled by F00 is comparable to that given is under Tohoku University, Japan. The tidal harmonics
by F93. The result of F00 is thus used for the northeast at the sites shown in Fig. 1 are used for model validation
Pacific region and the result of F93 for Dixon Entrance. since some inlets are insufficiently represented in the
These tidal elevation conditions are linearly connected model due to its spatial resolution. At the UAF stations,
around the meridian 134°E for the eight major constitu- the sea level was recorded during periods of less than one
ents (M2, S2, K1, O1, N2, K2, P1, and Q1), which forces month (Larsen et al., 2004), and the tidal harmonics of
the computational boundary. Root-mean-square differ- the four major constituents (M2, S2, K1, and O1) were
ences (the definition is stated below, Eq. (3)) between calculated by a tidal analysis program, BAYTAP-G
the Foreman’s models and the observed tidal harmonics (Tamura et al., 1991). This program selects the optimal
of M 2 are several centimeters at coasts outside the Alaska grouping of tidal constituents based on the Bayesian in-
Panhandle. In addition, the boundary condition of the tidal ference (Akaike, 1980), and preferentially calculates the
elevation does not include the non-linear, shallow water dominant constituent within the group with reduced con-
tides which are concerned at the head of Dixon Entrance tamination by the other minor constituents (Ishiguro et
by F93. Therefore, the model accuracy in this study is al., 1983). We use long tide gauge time series to test sta-
bility of the tidal harmonics calculated by BAYTAP-G in bathymetry and the tidal elevation at the computational
various durations of the time series. Differences in the boundary have been described in the previous section.
time series duration from one year to two weeks have Here the contribution of the improved boundary condi-
been found to yield small errors within 3% in amplitude tions to the model skill is evaluated as well as the de-
and 3° in phase to the calculated tidal harmonics of the pendence of the model skill on the dissipation parameters.
four major constituents. The record length of the OBPG Figure 3 shows the model skill in cases of tidal elevation
is one year. Thus the tidal harmonics obtained at the fol- conditions and coefficients of the horizontal eddy viscos-
lowing 15 sites are used to validate the modeled tidal ity using the improved bathymetric condition as a func-
harmonics: the 12 stations of NOAA, the 2 temporary sta- tion of the bottom friction coefficient. The total rms dif-
tions of UAF in Glacier Bay, and the OBPG (Table 2). ference is smallest in the case of the tidal elevation con-
dition obtained by merging the F00 and F93 models with
3.2 Sensitivity of the model skill to boundary conditions the dissipation parameters γb = 0.02 and AH = 1 m2 sec –1.
and dissipation parameters The model dependence is apparent on the bottom friction
We compare the tidal harmonics obtained by the tide coefficient and is fairly slight on the horizontal eddy vis-
gauge observations with those of the simulations in this cosity coefficient. The model skill worsens when the F00
subsection. The model skill is evaluated by calculating result is used for the tidal elevation condition. This model
the total root-mean-square (rms) difference between the dependence on the tidal elevation condition is caused by
observed and the modeled tidal harmonics of the four the model accuracy at Clarence Strait. If the original GINA
major constituents at the 15 sites. The rms difference of bathymetry is applied to the model, the total rms differ-
each tidal constituent at each tide gauge and OBPG is ence is around 0.8–1.0, which indicates that the model
defined as: skill is absolutely sensitive to the model bathymetry. The
dependence on the spatial resolution is also an important
{( A
issue. The F00 model employs a finite element grid sys-
)
2
Dc, s = c, s cos Bc, s − ac, s cos bc, s tem, and its spatial resolution is partially finer than our
model, especially in the heads of channels and bays. Nev-
)}
2 12
(
+ Ac, s sin Bc, s − ac, s sin bc, s , (3) ertheless, the original calculation of F00 yields the model
skill of 0.276 which is less accurate than the simulation
using the F00 condition. The discrepancy between the two
where Ac,s (ac,s) and Bc,s (bc,s) are the modeled (observed) is mainly caused by the difference of the model skill in
amplitude and phase of each tidal constituent (c) at each Glacier Bay (not shown), which possibly arises from the
tide gauge (s). The sum of the relative error (Dc,s/ac,s) of difference of the model bathymetry noted in Subsection
the four major constituents at the 15 sites is used to cal- 2.2. The model skill probably depends more strongly on
culate the total rms difference, the model skill. the model bathymetry than on the spatial resolution. Since
The accurate conditions of the multibeam-derived the model dependence on the bottom friction coefficient
M2 K1
Observed Inazu model F00 Observed Inazu model F00
Amplitude Phase Amplitude Phase D D Amplitude Phase Amplitude Phase D D
Site (cm) (deg. UTC) (cm) (deg. UTC) (cm) (cm) (cm) (deg. UTC) (cm) (deg. UTC) (cm) (cm)
1 Skagway 203.5 283.9 205.6 282.0 7.0 8.0 52.7 265.4 54.7 264.5 2.1 1.2
2 Juneau 199.5 282.5 194.5 279.7 10.7 9.8 52.0 265.0 53.2 263.7 1.7 0.8
3 Lynn Canal (OBPG) 194.3 281.6 194.3 281.6 0.0 0.1 52.1 264.5 53.7 264.2 1.6 1.9
4 Composite Island 204.3 299.7 199.3 301.7 8.7 51.1 53.6 273.8 54.7 275.0 1.6 5.1
5 Willoughby Island 216.7 298.5 196.0 301.7 23.5 58.8 59.0 273.2 54.5 275.0 4.9 9.0
6 False Bay 181.2 283.2 184.3 281.1 7.3 7.9 50.5 263.4 52.7 263.9 2.2 2.8
7 Warm Spring Bay 164.3 279.3 165.2 278.6 2.3 8.7 48.7 262.3 50.6 262.9 1.9 3.3
8 Petersburg 193.0 285.5 190.1 279.0 21.8 25.8 50.4 268.1 52.9 263.7 4.7 3.1
9 Entrance Island 175.4 279.7 170.3 278.1 6.8 9.4 48.1 265.8 51.3 263.2 4.0 2.9
10 Port Alexander 123.3 276.1 120.1 274.9 4.1 3.0 47.2 264.1 47.4 261.8 1.9 0.8
11 Thoms Point 197.9 272.3 198.9 269.5 9.6 10.0 52.0 261.7 52.2 261.1 0.5 1.3
12 Ketchikan 188.2 270.1 185.4 267.7 8.2 12.7 51.3 261.1 51.1 260.5 0.6 2.1
13 Elfin Cove 124.9 276.1 124.5 276.6 1.2 3.3 47.8 263.9 47.7 263.9 0.2 1.2
14 Sitka 110.6 273.8 108.7 273.8 1.9 3.9 46.2 259.9 46.3 262.4 2.0 2.1
15 Craig 113.5 268.1 117.3 268.1 3.8 4.1 44.7 262.2 46.7 259.4 3.0 1.9
Average 7.8 14.4 2.2 2.6
Fig. 4. Computed M2 (a) amplitude and (b) phase. Contour intervals are (a) 10 cm and (b) 5°.
Fig. 5. Computed K 1 (a) amplitude and (b) phase. Contour intervals are (a) 2 cm and (b) 2°.
shows total rms differences around 0.2–0.3, the model connections of the straits. The simulation model with the
skill seems to be sensitive to the model bathymetry, the finite difference grid system would not reasonable to in-
bottom friction coefficient, and the spatial resolution in vestigate the model dependence on the resolution while
this order. retaining the connection of the straits. The present study
The grid spacing of 1′ × 1′ employed in the present therefore does not demonstrate the model dependence on
study is expected to be a minimal resolution to represent the spatial resolution which will need to be investigated
the connections of the narrow straits in the Alaska Pan- with a simulation model using the finite element grid sys-
handle, and coarser resolution is insufficient to retain the tem. In the context of the ocean tide modeling with the
finite element grid, Lefèvre et al. (2000) described that Laurent and Garrett, 2002; St. Laurent et al., 2002). The
accurate model bathymetry is essential in order to im- evaluation requires the spatial distribution of the buoy-
prove the ocean tide model over the Yellow and East China ancy frequency at ocean bottom. However, water tem-
Seas. They also showed the importance of tuning the bot- perature and salinity data derived from World Ocean At-
tom friction coefficient and a minor effect of the mesh las (Levitus and Boyer, 1994; Levitus et al., 1994) in-
refinement. This case is similar to the ocean tide modeling volves an insufficient spatial resolution and incorrect
in the present study. The discussions above will be help- depth for the Alaska Panhandle, and the effect of the in-
ful in general debate on accurate coastal ocean modeling. ternal tides is not quantitatively evaluated in the present
As mentioned above, the most accurate ocean tide study.
representation is achieved with γ b = 0.02 and A H = 1 Table 3 shows the model accuracy at each station
m2 sec–1 under the improved conditions of the boundary for the M2 and K1 constituents of the Inazu and the F00
tidal elevation and the model bathymetry. This simulated models. The model skill is highly improved in Glacier
result of the ocean tide model is called the Inazu model Bay compared to F00 for both the M2 and K1 constitu-
hereafter. The coefficient γb = 0.02 is set as the optimal ents. A slight improvement is found at Ketchikan. We can
coefficient of the bottom friction of our model, since both not clarify the reasons for the differences of the model
the semidiurnal and diurnal constituents show the small- skill between the Inazu and the F00 models, but the dif-
est rms differences around γb = 0.02 and the bottom fric- ference of the model bathymetry may be one reason even
tion coefficient second largely dominates the model skill though the spatial resolution of our model in the Alaska
after the model bathymetry. The model skill hardly de- Panhandle is partly coarser than that of F00. The improve-
pends on the coefficient of the horizontal eddy viscosity, ment at Ketchikan is probably due to the use of the F93
and the coefficient AH = 1 m2 sec–1 is employed for the result instead of the F00 result in Dixon Entrance. The
model based on the Richardson’s four thirds law. The observed M2 and K1 phase at Petersburg lags relatively
optimal bottom friction coefficient larger than a typical behind the modeled phase as well as F00. Although addi-
value (0.003) is possibly due to wall friction (Colbo, tional local dissipation seems to be needed, it is not in-
2006), form drag (Edwards et al., 2004), internal tide vestigated in detail here. A large rms difference is also
generation (e.g., Gargett, 1976; Pawlowics, 2002) in the found at Willoughby Island, which possibly arises from
complex channel/strait system, and effects of tidewater the short record length of only two weeks (Table 2).
glaciers (e.g., Prinsenberg, 1988) in the northern part of Nevertheless, the model accuracy in the overall Alaska
the Alaska Panhandle. Recently, Tanaka et al. (2007) Panhandle is achieved with the rms difference within sev-
quantitatively evaluated the drag due to internal tides eral centimeters for each major constituent. These fea-
within the Kuril Islands by a simple parameterization (St. tures of the model improvements are similar for the S2
0.09
0.11
D
effect called the body tide which is the surface deforma-
tion of the solid Earth by the Newtonian attraction due to
displacement for the M2 and O1 constituents. The phase shows the local phase lag behind the equilibrium tide at each site. D
Table 4. Observed and predicted tidal signals in the vertical displacement at the GPS sites. The amplitude indicates the vertical
293.3
293.4
(deg.)
−6.2
−7.1
−7.3
0.0
due to the oceanic tide loading (OTL). The OTL effect in
O1
Amplitude
5.73
5.73
5.38
0.78
0.77
(cm)
0.08
0.06
187.4
186.8
(deg.)
−8.2
−10.9
−10.2
0.0
1.66
1.63
4.03
2.41
2.44
(cm)
0.17
0.17
D
−7.6
−7.6
0.0
Amplitude
5.71
5.71
5.33
0.82
0.82
(cm)
0.09
0.18
tide. The computation of the body tide and the OTL ef-
D
−17.5
−16.4
−10.7
0.0
1.23
1.32
1.19
3.92
2.77
2.75
dle.
The observed and predicted tidal signals at the GPS
sites are compared in Table 4. The vertical component of
Inazu model
Inazu model
Body tide
F00
F00