Since the two devastating tsunamis in 2004 (Indian Ocean) and 2011 (Great
East Japan), new findings have emerged on the relationship between tsunami
characteristics and damage in terms of fragility functions. Human loss and
damage to buildings and infrastructures are the primary target of recovery
and reconstruction; thus, such relationships for offshore properties and
marine ecosystems remain unclear. To overcome this lack of knowledge, this
study used the available data from two possible target areas (Mangokuura Lake
and Matsushima Bay) from the 2011 Japan tsunami. This study has three main
components: (1) reproduction of the 2011 tsunami, (2) damage investigation,
and (3) fragility function development. First, the source models of the 2011
tsunami were verified and adjusted to reproduce the tsunami characteristics
in the target areas. Second, the damage ratio (complete damage) of the
aquaculture raft and eelgrass was investigated using satellite images taken
before and after the 2011 tsunami through visual inspection and binarization.
Third, the tsunami fragility functions were developed using the relationship
between the simulated tsunami characteristics and the estimated damage ratio.
Based on the statistical analysis results, fragility functions were developed
for Mangokuura Lake, and the flow velocity was the main contributor to the
damage instead of the wave amplitude. For example, the damage ratio above 0.9
was found to be equal to the maximum flow velocities of 1.3 m s
Aquaculture and ecological systems provide many services and functions to
humans and are important to the global economy (Costanza et al., 1997). The
2011 Great East Japan tsunami caused devastating damage to inland and
offshore properties. Considerable economic damage resulting from the loss of
aquaculture products and the impact to ecological systems was also caused by
this tsunami. Since the 2004 Indian Ocean tsunami and the 2011 tsunami,
numerous quantitative measures of tsunami vulnerability, such as fragility
functions, have been developed for buildings (Leelawat et al.,
2014; Charvet et al.,
2015, 2017; Suppasri et al., 2013, 2016), infrastructures
(Shoji and Nakamura, 2017), and marine vessels (Suppasri et al., 2014; Muhari
et al., 2015). However, only one criterion is based on a previous study of
the 1960 Chilean tsunami, which struck the west side of Japan: damage to an
aquaculture raft (pearl) begins to occur when the tsunami flow velocity is
larger than 1 m s
To quantitatively assess such damage to marine products and marine ecosystems, the main objective of this study is to develop the fragility functions as the first step toward understanding the relationship between the tsunami characteristics and the damage. After reviewing previous works, this study comprises three main sections: (1) reproduction of the 2011 tsunami, (2) damage investigation, and (3) development of fragility functions. The first section presents a validation of the proposed source models for the 2011 tsunami and the adjustment for tsunami reproduction in the study areas. The second section presents the available damage data and damage quantification. The third section presents statistical analysis methods to develop the fragility functions using the results obtained from the first and second sections. Finally, new findings, recommendations, and the limitations of this study are discussed.
This section reviews selected previous studies related to the damage
characteristics of offshore facilities and marine plants against tsunamis.
The first attempt was based on the 1960 Chilean tsunami, which struck the west
of Japan. The damaged aquaculture rafts were plotted against the simulated
maximum water level and flow velocity (Nagano et al., 1991). As shown in
Fig. 1, the damage to the aquaculture raft (pearl) begins to occur when the
tsunami flow velocity is higher than 1 m s
After the 2011 tsunami, Suppasri et al. (2014) and Muhari et al. (2015)
developed fragility functions for fishing boats. Based on their results, the
threshold water level and flow velocity values for the complete destruction
of small boats of less than 5 t are 2 m and 1 m s
Damage criteria of the aquaculture raft based on the damage data from Kii Peninsula, western Japan, from the 1960 Chilean tsunami (Adapted from Nagano et al., 1991)
Because the size of the 2011 tsunami was extremely large, most aquaculture rafts and other marine plants were completely destroyed. There are only two well-suited locations with specific coastal geography, namely, Mangokuura Lake and Matsushima Bay in Miyagi Prefecture (Fig. 2), where the effects of the tsunami were comparatively small (Suppasri et al., 2012) and the aquaculture rafts were undamaged and the eelgrass survived (Northwest Pacific Region Environmental Cooperation Center, 2016). Mangokuura Lake has a notably narrow entrance from the Pacific Ocean through Ishinomaki Bay, and the average sea depth is as shallow as 5 m or less. Matsushima Bay is protected by almost 300 small islands around the bay front. Thus, the 2011 tsunami inundation and run-up heights in both areas were less than 1–2 m, whereas they were as high as 10 m in other nearby areas (Suppasri et al., 2012). As a result, some aquaculture rafts and other marine plants survived in these two locations, which enabled the development of fragility functions.
Study areas:
To obtain tsunami-related parameters, including the water level and flow
velocity, the 2011 tsunami was reproduced using a numerical analysis. The
2011 tsunami was numerically simulated using a set of nonlinear shallow-water
equations, which were discretized using the staggered leap-frog
finite-difference scheme (TUNAMI model; Imamura, 1996) with bottom friction
in the form of Manning's formula, similar to previous studies (Suppasri et
al., 2011; Charvet et al., 2015; Macabuag et al., 2016). Six computational
domains adopted from a previous study (Macabuag et al., 2016) which used
original data from the Geospatial Information Authority of Japan (GSI, 2015) were used as a nesting grid system of
1215 m (region 1), 405 m (region 2), 135 m (region 3), 45 m (region 4),
15 m (region 5), and 5 m (region 6). The tidal level of
Six computational areas for
Three models of fault parameters were selected to reproduce the 2011 tsunami:
model 1: Tohoku University model (Imamura et al., 2011); model 2: Satake
model (Satake et al., 2013); and model 3: Japan Nuclear Energy Safety
Organization (JNES) model (Sugino et al., 2013). The corresponding fault
parameters were used to estimate the seafloor deformation proposed by
Okada (1985), which later became the initial seafloor condition for the
tsunami numerical simulation. The simulated tsunami inundation and run-up
height with the actual measured values (Mori et al., 2012) were validated for
each area using Aida's
For Mangokuura Lake, model 3 produced the optimal values of Aida's
Aida's
Comparison of the simulated and measured tsunami heights in Mangokuura Lake and Matsushima Bay.
The hydrodynamic properties of the 2011 tsunami were reproduced based on the
model calibration and verification as mentioned above. Figure 5 shows that the
average maximum water level and flow velocity in the bay of Mangokuura Lake
are approximately 0.5 m and 1–2 m s
Simulated maximum water level and flow velocity in Mangokuura Lake and Matsushima Bay.
Damage inspection was performed using satellite images taken before and after the tsunami through a visual inspection for the aquaculture rafts and an image analysis for the eelgrass.
Based on criteria for the recovery process developed by the Japan Fisheries
Agency (2010), aquaculture raft damage can be classified into four types:
(1) complete damage (washed away), (2) major damage (70–100 % physical
damage), (3) moderate damage (30–70 % physical damage), and
(4) minor damage (less than 30 % physical damage). Because of the
limitations encountered when using satellite images, only the complete-damage
(washed away) level could be investigated using the satellite images taken
before and after the tsunami, similar to other previous studies related to
buildings (Koshimura et al., 2009; Suppasri et al., 2011). In this study,
only the long-line type of aquaculture raft (Fig. 6) had sufficient
quantities to develop the fragility function. This type of aquaculture raft
is common in the study area and is used for oyster and seaweed farming.
Examples of the visual inspection of the aquaculture rafts in the lake before
(Fig. 7a) and after the tsunami (Fig. 7b) are shown. Approximately half of
the rafts remained after the tsunami; the others were completely washed away.
The aquaculture rafts that completely disappeared were classified as complete
damage (washed away), whereas the damage levels of the remaining aquaculture
rafts ranged from no damage to major damage. This classification is used to
calculate the damage probability of complete damage in Sect. 4. Figure 7 also
shows the visual-inspection results (presented as polygons) of complete
damage versus no damage and other damage levels for the aquaculture rafts
(long-line type) in Mangokuura Lake. Many completely damaged aquaculture
rafts were found near the entrance to and in the middle of the lake. Then,
the created polygons were gridded into 5
Aquaculture raft (long-line type).
Visual damage interpretation of aquaculture rafts (long-line type)
Damage to eelgrass occurs in one of three modes: cut-off, deposition, or erosion, as shown in Fig. 8. Although the deposition and erosion can be estimated using a sediment transport model, more detailed data and surveys are required to obtain the necessary data for the model input. This pilot study considered only the tsunami itself. In addition, the erosion was controlled primarily by the flow velocity. Therefore, the cut-off and erosion were considered damage from the horizontal force of the tsunami.
Color images from an actual satellite image taken before the 2011 tsunami and
after the 2011 tsunami were analyzed (Northwest Pacific Region Environmental
Cooperation Center, 2016; Tsujimoto et al., 2016). At this stage, the areas
for land, sea, aquaculture raft, eelgrass, and mudflat were first identified.
To identify only the eelgrass area, the colored images were binarized to
binary (black and white) images using the ImageJ image analysis software,
which is being developed at the National Institutes of Health, the United
States (ImageJ, 2016). This binarization helps distinguish eelgrass and
non-eelgrass areas. Figures 9 and 10 show the eelgrass areas before and after
the 2011 tsunami in Mangokuura Lake and Matsushima Bay, respectively. Similar
to the aquaculture rafts, the eelgrass that was completely damaged could be
investigated by comparing the images taken before and after the tsunami. The
identified damaged and undamaged areas for both aquaculture rafts and
eelgrass were gridded into 5
Areas of the eelgrass
Areas of the eelgrass
A comparison of the aquaculture raft data in the cases of the 1960 Chilean
tsunami (Fig. 1) and the 2011 Japan tsunami is shown in Fig. 11. Most of the
aquaculture rafts that were not completely damaged in the 2011 tsunami were
limited to a maximum flow velocity of less than 1.5 m s
Comparison of the aquaculture raft data from the 1960 Chilean tsunami (Fig. 1) and the present study on the 2011 Japan tsunami.
Maximum water level and complete-damage probability for the
Only the simulated maximum flow velocity and damaged-eelgrass data in
Mangokuura Lake could be used to develop the fragility functions. The tsunami
fragility functions were developed by applying the classical standardized
lognormal distribution function throughout the linear regression analysis for
both aquaculture rafts and eelgrass. For Mangokuura Lake, Fig. 12 shows the
histograms of the numbers of damaged and undamaged aquaculture rafts in every
100 grids (Fig. 13a) and 0–50 % damaged and 50–100 % damaged
eelgrass in every 5000 grids (Fig. 13b) in terms of the simulated maximum
flow velocity range. Both histograms show that the damage data increase when
the flow velocity increases. A linear regression analysis was performed to
develop the fragility function. The cumulative probability
Histogram of the numbers of
Least-squares fit on lognormal probability paper for the
With the regression analysis, the parameters that best fit the fragility
functions with respect to the maximum flow velocity are shown in Table 2.
The tsunami fragility curves for the aquaculture rafts and eelgrass were
developed as shown in Fig. 15a and b, respectively. The proposed
fragility functions show that a complete-damage ratio above 0.5 corresponds
to maximum flow velocities of 0.8 m s
Parameters to create the tsunami fragility functions.
Tsunami fragility functions for completely damaged
This study was the first attempt in this field to develop fragility functions
for aquaculture rafts and eelgrass. The careful selection of the study areas
and availability of the damage data enabled this attempt. First, we
reproduced the hydrodynamic characteristics, i.e., the water level and flow
velocity of the 2011 tsunami, using the tsunami trace data for the model
calibration and verification based on the finest grid of
5 Based on the reproduced hydrodynamic characteristics of the 2011 tsunami,
Matsushima Bay was hit by a stronger tsunami than Mangokuura Bay (Fig. 5). The maximum water level is not related to the damage to aquaculture rafts and
eelgrass (Fig. 12). The threshold value (at 90 % damage probability) of the maximum flow
velocity for completely damaged aquaculture rafts and eelgrass is 1.3 and
3.0 m s The proposed fragility function for the aquaculture rafts is consistent with
the previously proposed damage criteria and can further provide the values
of the damage ratio at other flow velocities in addition to the threshold
value. This information on the tsunami damage in offshore areas is expected to be
useful for marine product and environmental damage assessment and
recommendations for aquaculture raft zoning to mitigate the effects of
tsunamis in the future.
Although this study successfully developed fragility functions for aquaculture rafts and eelgrass for the first time, certain limitations and considerations exist when applying the fragility functions, and possible improvements to be pursued in future studies are as follows.
The developed fragility functions may underestimate the economic damage
related to aquaculture rafts because the loss of marine products may occur
even when the rafts remain. For example, although the aquaculture rafts were
present in the satellite image, in some cases the marine products were
completely washed away or damaged when the rafts collided with each other. This study simulated only the hydrodynamic characteristics of the tsunami,
which can directly explain the damage caused by cut-off and erosion.
However, the damage caused by deposition was not considered. The use of the actual surveyed damage to the aquaculture rafts and eelgrass
and the application of a sediment transport model may increase the accuracy
of the fragility functions. The fragility functions for both aquaculture rafts and eelgrass may differ
based on the type of aquaculture raft and the environmental conditions of
the eelgrass. Future studies of aquaculture rafts and eelgrass in other
areas impacted by historical tsunami events may improve our understanding of
these differences and the generalizability of the fragility functions.
Source code and details of tsunami simulation were sourced from Imamura (1996). Topography and bathymetry data were obtained from previous research (Macabuag et al., 2016), which sourced the original data from the Geospatial Information Authority of Japan (GSI, 2015). Damage data were obtained from published results (NPEC, 2016). Figures 2, 3, 5, 9, and 10 were made using an open-source collection of computer software tools called Generic Mapping Tools (GMT, 2016).
The authors declare that they have no conflict of interest.
We thank the Miyagi Prefecture Fisheries Cooperative Association (Japan Fisheries Cooperatives (JF)), Ishinomaki Bay branch, for their information on the aquaculture rafts and Daisuke Sugawara (Museum of Natural and Environmental History, Shizuoka) for his help in developing the bathymetry and topography data. This study was funded through IRIDeS, Tohoku University, by the Tokio Marine & Nichido Fire Insurance Co., Ltd.; the Willis Research Network (WRN); and JSPS Grant-in-Aid for Young Scientists (B) “Applying developed fragility functions for the Global Tsunami Model (GTM)” (grant no. 16K16371). Edited by: Ira Didenkulova Reviewed by: two anonymous referees