notes

uni notes
git clone git://popovic.xyz/notes.git
Log | Files | Refs

chap4.tex (8690B)


      1 \section{Modeling the 2004 Tsunami}
      2 \subsection{Description}
      3 On the 26. December 2004, time 7:58 a powerful earthquake generated a tsunami
      4 killing more than 275000 people and leaving millions homeless. The
      5 hypocenter of the earthquake was 30 km under the floor of the Indian Ocean,
      6 100 km away from Sumatra, an island in Indonesia. The earthquake displaced
      7 an enormous amount of water, sending tsunami waves westwards across the
      8 Indian Ocean to Sri Lanka and India and eastwards across the Andaman Basin.
      9 to Thailand and Indonesia. The Earthquake occurred over 10 minutes along a
     10 1000km long roughly straight line. Thereby we can model the corresponding
     11 fluid mechanics as 2 dimensional where in Cartesian coordinates the
     12 propagation of the tsunami wave is in $x$ direction and the $z$ direction
     13 pointing upwards perpendicular to the flat ocean surface. However the
     14 modeling assumption for two dimensions for the region outside the Bay of
     15 Bengal is not valid, since the diffraction around islands and reflection from
     16 steep shores pays a major role in the influence of the wave mechanics. Coming
     17 back to the $2004$ tsunami, which raised the ocean floor a few meters to the
     18 west and lowering it a few meters to the east, displacing the tectonic pates.
     19 The tsunami waves featured westwards a wave of elevation, meaning a wave of
     20 high amplitude, followed by a wave of depression, a wave of long wavelength
     21 hitting the coastal areas of Sri Lanka and India in roughly three hours,
     22 propagating a distance of approximately $1600\ \text{km}$. On the other hand
     23 eastwards featuring a first a wave of depression following a wave of
     24 elevation, propagating $700\ \text{km}$ in roughly two hours with a maximal
     25 amplitude of $10\ \text{m}$. Observations tell us that as the tsunami waves
     26 reached the shore the shape of the initial disturbance was not altered, which
     27 is supported by measurements by a radar altimeter two hours after the
     28 earthquake showing first an wave of elevation and then a wave of depression
     29 westwards and respectively vice versa eastwards. The conclusion is made that
     30 the shape of the tsunami remained approximately constant. Additionally it
     31 should be mentioned that the tsunami waves reach very high amplitudes due to
     32 the diminishing depth effect as they approach the shore, yet at open sea are
     33 barely noticeable. A boat on open sea positioned at high depth in the region
     34 of the tsunami during which the tsunami waves passed, captured the raise
     35 from $\pm 0.8\ \text{m}$ of the boat over a period of $10\ \text{min}$.
     36 This means that the wavelength of the tsunami wave was about $100\
     37 \text{km}$.
     38 \subsection{Long Wave, Shallow Water}
     39 At weakly nonlinear levels dispersion balances linearity/nonlinearity in
     40 certain regimes, such balance is found in the KdV equation. Among the KdV
     41 equation, the Carissa-Holm equation (CH) also features such balance since it
     42 arises as a high order approximation to KdV. Further there is also the
     43 regularized long wave equation usually called Benjamin–Bona–Mahony equation, or simply
     44 BBM. It should be noted that KdV is a solitary wave while BBM is not.
     45 Localized disturbances of flat water surfaces propagating without change of
     46 form need to be two dimension waves of elevation symmetric about the crest.
     47 The linear theory does not provide any approximation to solitary waves, only
     48 nonlinear or weakly linear approximations. These are KdV, BBM. The KdV
     49 equation is orbitaly stable, meaning the shape and form of the profile is
     50 stable under small perturbations (also CH \& BBM). Each solitary wave retains
     51 is local identity, where large waves are faster than small ones. Further the
     52 solution of the KdV equation evolves into a set of solitary waves, with
     53 tallest in from front followed by an oscillatory tail. The KdV is the
     54 proper equation, for our modeling purposes of tsunami waves. The main
     55 question arises if KdV enters the regime of validity, in our case for the
     56 $2004$ tsunami in the Andaman Basin. Or in other words are the involved
     57 geophysical scales leading to time and space scales, compatible with
     58 the KdV weak nonlinearity balance.
     59 \subsection{Governing equations}
     60 The last two section show the derivation of the modeling equation for fluid
     61 mechanics, following with nondimensionalisation and rescaling. The scaling
     62 takes the same form as in section \ref{sec:nondim}, we only introduce the
     63 parameter $\alpha = h_0\sqrt{gh_0} $ and get the following equations
     64 \begin{align}
     65     \begin{drcases}
     66     u_x + w_z = 0\\
     67     u_t + \varepsilon(uu_x + wu_z) = -p_x\\
     68     \delta^2\left(w_t + \varepsilon(w w_x + w w_z)\right) = -p_y\\
     69     u_z - \delta^2 w_x = 0
     70     \end{drcases}
     71 \end{align}
     72 on $(x, z) \in \mathbb{R}\times [0, 1+\varepsilon \eta(x, t)]$, with boundary
     73 conditions
     74 \begin{align}
     75     \begin{drcases}
     76         p = \eta \\
     77         w = h_t + \varepsilon u h_x
     78     \end{drcases}
     79     \text{on}\;\; z = 1+\varepsilon \eta(x, t)\\
     80     w = 0 \quad \text{on}\;\; z = 0.
     81 \end{align}
     82 The KdV validity arises in the region $\varepsilon = O(\delta^2)$, we are
     83 going to thereby rescale $\delta$ in favour of $\varepsilon = \beta
     84 \delta^2$, where $\beta = O(1)$ as in equation \ref{eq:epsdelta}:
     85 \begin{align}\label{eq:epsdelta}
     86     x \rightarrow \frac{\delta}{\sqrt{\varepsilon} }x, \quad t
     87     \rightarrow \frac{\delta}{\sqrt{\varepsilon} }t\quad,
     88     w \rightarrow \frac{\sqrt{\varepsilon} }{\delta}w.
     89 \end{align}
     90 This opens up the possibility to prove that provided a suitable length- \&
     91 time-scales for some $\delta$ the KdV will arise as a valid approximation for
     92 the evolution of the free surface waves. Given some $\varepsilon>0$ there
     93 exists a time in the such that the KdV balance holds, where we introduce the
     94 variables $\xi = x- t$ and $\tau = \varepsilon t$ with equation for the wave profile
     95 \begin{align}
     96     \eta_\tau - \frac{3}{2} \eta_{\xi\xi\xi} + \frac{1}{6} \eta \eta_\xi = 0,
     97 \end{align}
     98 for $\xi \in \mathbb{R}$ and $\tau > 0 $ and the boundary condition $\eta(\xi
     99 ,0)$ of the initial profile at $\tau = t = 0$. On the basis of satellite
    100 measurements for the Bay of Bengal we have
    101 \begin{align}
    102     a = 1\ \text{km}, \quad \lambda = 100\ \text{km},\quad h_0 = 4\
    103     \text{km}.
    104 \end{align}
    105 giving us
    106 \begin{align}
    107     \varepsilon = \frac{a}{h_0} = 25 \cdot 10^{-5}\\
    108     \delta = \frac{h_0}{\lambda} \simeq 4*10^{-2}
    109 \end{align}
    110 giving us a $\beta \simeq 6,4 = O(1)$ for $\varepsilon = \beta \delta^2$.
    111 The main issue is if the KdV balance can occur within the geophysical scales
    112 .The conditions $x-t = O(1)$ and $\tau = O(1)$ give
    113 \begin{align}
    114     \frac{x - t \sqrt{gh_0} }{\lambda} = O(1),\quad \frac{\varepsilon t
    115     \sqrt{gh_0} }{\lambda} = O(1).
    116 \end{align}
    117 Combining the above equations, we have
    118 \begin{align}
    119    & \frac{x}{\lambda} = O(\varepsilon^{-1})\\
    120     &x  = O(\varepsilon^{-1}\lambda)
    121 \end{align}
    122 For the tsunami wave in the Bay of Bengal westwards towards India and Sri
    123 Lanka we have
    124 \begin{align}
    125     \lambda = 10\ \text{km}, \quad \varepsilon = 25 \cdot 10^{-5},
    126 \end{align}
    127 therefore a propagation distance of $x \simeq 4 \cdot 10^{5}$ is needed for
    128 the KdV to enter the range of validity. This is however not the case since
    129 the tsunami waves propagated $1600\ \text{km}$ westwards.
    130 
    131 For the wave in the Andaman Basin towards Indonesia and Thailand we have
    132 \begin{align}
    133     h_0 = 1\ \text{km}, \quad a = 1\ \text{m}, \quad \lambda = 100\
    134     \text{km}.
    135 \end{align}
    136 Giving us the parameters
    137 \begin{align}
    138     \varepsilon  = 10^{-3}, \quad \delta = 10^{-2}.
    139 \end{align}
    140 this also does not satisfys the range of validity  $\varepsilon =
    141 O(\delta^2)$, with the requiring length scale $x = 10^{5}$.
    142 
    143 Setting $h_0 = 4\ \text{km}$ for the Bay of Bengal and $h_0 = 1\ \text{km}$
    144 for the Andaman Basin we have that the westwards tsunami propagated at speed
    145 $\sqrt{gh_0}  = 712 \frac{\text{km}}{\text{h}}$ hitting the shore in about
    146 $2h\ 10min$, while the tsunami waves eastwards propagated at speed $356
    147 \frac{\text{km}}{\text{h}}$ hitting the coast in $1h\ 57min$. These
    148 predictions align with the observations. As the waves approach the shore, the
    149 tsunami mechanics enter the region of long waves over variable depth. In this
    150 case dispersion and their front steepness play an important role, where
    151 faster wave fronts catch up to slower ones and result in large amplitudes
    152 with devastating effects.
    153 
    154 For the tsunami in 2004 the propagation distance across the Indian Ocean in
    155 both directions to Sri Lanka, India and Thailand was too short for nonlinear
    156 KdV dynamics take place. The linear wave equation model discribes the tsunami
    157 clear across the Indian Ocean untill it reaches costal regions. So for short
    158 times the Tsunami waves are governed by the linear wave equation $\eta_{x x }
    159 = \eta_{t t}$.
    160 
    161 
    162 
    163 
    164 
    165 
    166 
    167