sandbox/bugs/two_layer_minimum.h
A solver for the Two-layer equations
The Saint-Venant equations can be written in integral form as the hyperbolic system of conservation laws \displaystyle \partial_t \int_{\Omega} \mathbf{q} d \Omega + \int_{\partial \Omega} \mathbf{f} ( \mathbf{q}) \cdot \mathbf{n}d \partial \Omega + \int_{\Omega} \mathbf{F} d \Omega= 0 where \Omega is a given subset of space, \partial \Omega its boundary and \mathbf{n} the unit normal vector on this boundary. For conservation of mass and momentum in the shallow-water context, \Omega is a subset of bidimensional space and \mathbf{q} and \mathbf{f} are written \displaystyle \mathbf{q} = \left(\begin{array}{c} h\ \ h u_x\ \ h u_y h_s\ \ h_s u_s_x\ \ h_s u_s_y \end{array}\right), \;\;\;\;\;\; \mathbf{f} (\mathbf{q}) = \left(\begin{array}{cc} h u_x & h u_y\ \ h u_x^2 + \frac{1}{2} g h^2 & h u_x u_y\ \ h u_x u_y & h u_y^2 + \frac{1}{2} g h^2\ \ h_s u_s_x & h_s u_s_y\ \ h_s u_s_x^2 + \frac{1}{2} gh_s^2 & h_s u_s_x u_s_y\ \ h_s u_s_x u_sy & h_s u_s_y^2 + \frac{1}{2} g h_s^2 \end{array}\right), \;\;\;\;\;\; \mathbf{F} = \left(\begin{array}{c} 0\ \ g h \frac{\partial}{\partial x} \left( h_s + z_b \right)\ \ g h \frac{\partial}{\partial y} \left( h_s + z_b \right)\ \ 0\ \ g h_s \frac{\partial}{\partial x} \left( z_b + \frac{\rho}{\rho_s} h \right)\ \ g h_s \frac{\partial}{\partial y} \left( z_b + \frac{\rho}{\rho_s} h \right) \end{array}\right), where h the water depth, h_s is the landslide thickness, \mathbf{u} is the velocity vector of the water, \mathbf{u_s} is the velocity vector of the landslide, and z_b the height of the topography. See also Popinet, 2011 for a more detailed introduction.
User variables and parameters
The primary fields are the water depth h, the landslide thickness h_s, the bathymetry z_b and the flow speeds of water and landslide respectively \mathbf{u} and \mathbf{u_s}. \eta is the water level i.e. z_b + h + h_s. Note that the order of the declarations is important as z_b needs to be refined before h_s, before h, before \eta.
scalar zb[], hs[], h[], eta[];
vector us[], u[];
The only physical parameter is the acceleration of gravity G
and the densities of the two fluids. Cells are considered “dry” when the water depth is less than the dry
parameter (this should not require tweaking).
double G = 1.;
double dry = 1e-10;
double RHOratio = 0.5;
Time-integration
Setup
Time integration will be done with a generic predictor-corrector scheme.
#include "predictor-corrector.h"
The generic time-integration scheme in predictor-corrector.h needs to know which fields are updated.
scalar * evolving = NULL;
We need to overload the default advance function of the predictor-corrector scheme, because the evolving variables (h, $, h_s and \mathbf{u_s}) are not the conserved variables h, $h, h_s and h_s\mathbf{u_s}.
static void advance_two_layer (scalar * output, scalar * input,
scalar * updates, double dt)
{
// fprintf(stderr,"advance_two_layer\n");
// recover scalar and vector fields from lists
scalar hi = input[0], ho = output[0], dh = updates[0];
scalar hsi = input[1], hso = output[1], dhs = updates[1];
vector uo = vector(output[2]), uso = vector(output[2+dimension]);
/*
Order within the scalar lists e.g. input, output, updates, evolving, [h, hs, u, us]
Note that this differs from what I previously did... But this should all occur within this so as long as it is self consistent it should be fine
*/
// new fields in ho[], uo[]
foreach() {
double hold = hi[];
double hsold = hsi[];
ho[] = hold + dt*dh[];
hso[] = hsold + dt*dhs[];
eta[] = ho[] + hso[] + zb[];
if (hso[] > dry){
// fprintf(stderr,"HS wet");
vector usi = vector(input[2+dimension]);
vector dhus = vector(updates[2+dimension]);
foreach_dimension()
uso.x[] = (hsold*usi.x[] + dt*dhus.x[])/hso[];}
else{
// fprintf(stderr,"HS dry");
foreach_dimension()
uso.x[] = 0.;}
if (ho[] > dry){
// fprintf(stderr," H wet\n");
vector ui = vector(input[2]);
vector dhu = vector(updates[2]);
foreach_dimension()
uo.x[] = (hold*ui.x[] + dt*dhu.x[])/ho[];}
else{
// fprintf(stderr," H dry\n");
foreach_dimension()
uo.x[] = 0.;}
}
// fprintf(stderr,"boundary\n");
//ho.prolongation = refine_linear;
boundary ({hso, ho, eta, uso, uo});
// fprintf(stderr,"done\n");
}
When using an adaptive discretisation (i.e. a quadtree)., we need to make sure that \eta is maintained as z_b + h + h_s whenever cells are refined or coarsened.
if TREE
static void refine_eta (Point point, scalar eta) { foreach_child() eta[] = zb[] + h[] + hs[]; }
static void coarsen_eta (Point point, scalar eta) { eta[] = zb[] + h[] + hs[]; } #endif
Computing fluxes
Various approximate Riemann solvers are defined in riemann.h.
#include "riemann.h"
double update_two_layer (scalar * evolving, scalar * updates, double dtmax)
{
We first recover the currently evolving fields (as set by the predictor-corrector scheme). NOTE change in order as mentioned above
scalar h = evolving[0];
scalar hs = evolving[1];
vector u = vector(evolving[2]);
vector us = vector(evolving[2+dimension]);
Fh
, Fhs
Fq
and Fqs
will contain the fluxes for h, h_s, h\mathbf{u} and h_s\mathbf{u_s} respectively and S1
and S2
are necessary to store the asymmetric topographic source terms.
face vector Fh[], Fhs[], S1[], S2[];
tensor Fq[], Fqs[];
The gradients are stored in locally-allocated fields. First-order reconstruction is used for the gradient fields.
vector gh[], ghs[], geta[];
tensor gu[], gus[];
for (scalar s in {gh, ghs, geta, gu, gus}){
s.gradient = zero;
#if TREE //Is this where the problem is? Do I need to refine some other way?
s.prolongation = refine_linear;
#endif
}
gradients ({h, hs, eta, u, us}, {gh, ghs, geta, gu, gus});
//fprintf(stdout,"%% updates\n");
The faces which are “wet” on at least one side are traversed. First we see whether “wet” in bottom fluid - if so look for lake at rest solution h_s+z_b=C_0 h=C If hs is dry look for lake at rest solution h_s=0 h+z_b=C_0
int hswet;
foreach_face (reduction (min:dtmax)) {
// First the bottom layer
double hi = hs[], hn = hs[-1,0];
if (hi > dry || hn > dry) {
// fprintf(stderr,"HS wet");
hswet=1;
Left/right state reconstruction
The gradients computed above are used to reconstruct the left and right states of the primary fields h, \mathbf{u}, z_b. The “interface” topography z_{lr} is reconstructed using the hydrostatic reconstruction of Audusse et al, 2004
double dx = Delta/2.;
double zi = eta[] - hi - h[];
double zl = zi - dx*(geta.x[] - ghs.x[]- gh.x[]);
double zn = eta[-1,0] - hn - h[-1,0];
double zr = zn + dx*(geta.x[-1,0] - ghs.x[-1,0] - gh.x[-1,0]);
double zlr = max(zl, zr);
double hl = hi - dx*ghs.x[];
double up = us.x[] - dx*gus.x.x[];
double hp = max(0., hl + zl - zlr);
double hr = hn + dx*ghs.x[-1,0];
double um = us.x[-1,0] + dx*gus.x.x[-1,0];
double hm = max(0., hr + zr - zlr);
Riemann solver
We can now call one of the approximate Riemann solvers to get the fluxes.
double fh, fu, fv;
kurganov (hm, hp, um, up, Delta*cm[]/fm.x[], &fh, &fu, &dtmax);
fv = (fh > 0. ? us.y[-1,0] + dx*gus.y.x[-1,0] : us.y[] - dx*gus.y.x[])*fh;
Topographic source term
In the case of adaptive refinement, care must be taken to ensure well-balancing at coarse/fine faces (see notes/balanced.tm).
#if TREE
if (is_prolongation(cell)) {
hi = coarse(hs,0,0,0);
zi = coarse(zb,0,0,0);
}
if (is_prolongation(neighbor(-1,0))) {
hn = coarse(hs,-1,0,0);
zn = coarse(zb,-1,0,0);
}
#endif
double sl = G/2.*(sq(hp) - sq(hl) + (hl + hi)*(zi - zl+dx*RHOratio*gh.x[]));
double sr = G/2.*(sq(hm) - sq(hr) + (hr + hn)*(zn - zr-dx*RHOratio*gh.x[-1,0]));
Flux update
Fhs.x[] = fm.x[]*fh;
Fqs.x.x[] = fm.x[]*(fu - sl);
S2.x[] = fm.x[]*(fu - sr);
Fqs.y.x[] = fm.x[]*fv;
//fprintf(stdout,"2 %g %g %g %g %g\n",x,y,fh,fu,fv);
}
else {//h_s is dry - Note that h_s is not necessarily dry in the neighbouring cell...
Fhs.x[] = Fqs.x.x[] = S2.x[] = Fqs.y.x[] = 0.;
hswet=0;
// fprintf(stderr,"HS dry");
}
Now we must calculate fluxes for h. If h_s>dry then we must satisfy h_s+z_b=C_0 h=C; If h_s=0 then we must satisfy h+z_b=C_0
Left/right state reconstruction
The gradients computed above are used to reconstruct the left and right states of the primary fields h, \mathbf{u}, z_b. The “interface” topography z_{lr} is reconstructed using the hydrostatic reconstruction of Audusse et al, 2004
double dx = Delta/2.;
double zi = eta[] - hi;
double zl = zi - dx*(geta.x[] - gh.x[]);
double zn = eta[-1,0] - hn;
double zr = zn + dx*(geta.x[-1,0] - gh.x[-1,0]);
double zlr = max(zl, zr);
double hl = hi - dx*gh.x[];
double up = u.x[] - dx*gu.x.x[];
double hp = max(0., hl + zl - zlr);
double hr = hn + dx*gh.x[-1,0];
double um = u.x[-1,0] + dx*gu.x.x[-1,0];
double hm = max(0., hr + zr - zlr);
Riemann solver
We can now call one of the approximate Riemann solvers to get the fluxes.
double fh, fu, fv;
kurganov (hm, hp, um, up, Delta*cm[]/fm.x[], &fh, &fu, &dtmax);
fv = (fh > 0. ? u.y[-1,0] + dx*gu.y.x[-1,0] : u.y[] - dx*gu.y.x[])*fh;
Topographic source term
In the case of adaptive refinement, care must be taken to ensure well-balancing at coarse/fine faces (see notes/balanced.tm).
#if TREE
if (is_prolongation(cell)) {
hi = coarse(h,0,0,0);
zi = coarse(zb,0,0,0) + hswet?coarse(hs,0,0,0):0.;
}
if (is_prolongation(neighbor(-1,0))) {
hn = coarse(h,-1,0,0);
zn = coarse(zb,-1,0,0) + hswet?coarse(hs,-1,0,0):0.;
}
#endif
double sl = G/2.*(sq(hp) - sq(hl) + (hl + hi)*(zi - zl));
double sr = G/2.*(sq(hm) - sq(hr) + (hr + hn)*(zn - zr));
Flux update
Fh.x[] = fm.x[]*fh;
Fq.x.x[] = fm.x[]*(fu - sl);
S1.x[] = fm.x[]*(fu - sr);
Fq.y.x[] = fm.x[]*fv;
//fprintf(stdout,"1 %g %g %g %g %g\n",x,y,fh,fu,fv);
}
else {// dry
Fh.x[] = Fq.x.x[] = S1.x[] = Fq.y.x[] = 0.;
// fprintf(stderr," H dry\n");
}
}
boundary_flux ({Fh, Fhs, S1, S2, Fq, Fqs});
Updates for evolving quantities
We store the divergence of the fluxes in the update fields. Note that these are updates for h and h\mathbf{u} (not \mathbf{u}).
scalar dh = updates[0], dhs = updates[1];
vector dhu = vector( updates[2]), dhus = vector(updates[2+dimension]);
foreach() {
dh[] = (Fh.x[] + Fh.y[] - Fh.x[1,0] - Fh.y[0,1])/(cm[]*Delta);
dhs[] = (Fhs.x[] + Fhs.y[] - Fhs.x[1,0] - Fhs.y[0,1])/(cm[]*Delta);
foreach_dimension(){
dhu.x[] = (Fq.x.x[] + Fq.x.y[] - S1.x[1,0] - Fq.x.y[0,1])/(cm[]*Delta);
dhus.x[] = (Fqs.x.x[] + Fqs.x.y[] - S2.x[1,0] - Fqs.x.y[0,1])/(cm[]*Delta);
}
We also need to add the metric terms. They can be written (see eq. (8) of Popinet, 2011) \displaystyle S_g = h \left(\begin{array}{c} 0\ \ \frac{g}{2} h \partial_{\lambda} m_{\theta} + f_G u_y\ \ \frac{g}{2} h \partial_{\theta} m_{\lambda} - f_G u_x \end{array}\right) with \displaystyle f_G = u_y \partial_{\lambda} m_{\theta} - u_x \partial_{\theta} m_{\lambda}
double dmdl = (fm.x[1,0] - fm.x[])/(cm[]*Delta);
double dmdt = (fm.y[0,1] - fm.y[])/(cm[]*Delta);
double fG1 = u.y[]*dmdl - u.x[]*dmdt;
double fG2 = us.y[]*dmdl - us.x[]*dmdt;
dhu.x[] += h[]*(G*h[]/2.*dmdl + fG1*u.y[]);
dhu.y[] += h[]*(G*h[]/2.*dmdt - fG1*u.x[]);
dhus.x[] += hs[]*(G*hs[]/2.*dmdl + fG2*us.y[]);
dhus.y[] += hs[]*(G*hs[]/2.*dmdt - fG2*us.x[]);
}
return dtmax;
}
We use the main time loop (in the predictor-corrector scheme) to setup the initial defaults.
event defaults (i = 0){
evolving =(scalar *) {h, hs, u, us};
foreach()
for (scalar s in evolving)
s[]=0.;
boundary(evolving);
for (int ii=0; ii<6; ii++) {
printf (" evolving[ %d ]: %s ", ii, _attribute[evolving[ii].i].name);
}
printf ("\n");
We overload the default ‘advance’ and ‘update’ function of the predictor-corrector scheme and setup the refinement and coarsening methods on quadtrees.
advance = advance_two_layer;
update = update_two_layer;
#if QUADTREE
for (scalar s in {h,zb,us,u,eta}) {
s.refine = s.prolongation = refine_linear;
}
#endif
}
The event below will happen after all the other initial events to take into account user-defined field initialisations.