-
Notifications
You must be signed in to change notification settings - Fork 0
/
Fig(3,5) Two_layered_analytical_results.py
204 lines (160 loc) · 10.3 KB
/
Fig(3,5) Two_layered_analytical_results.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
#Infiltration in a two-layered initially-dry soil (Combined)
#Mohammad Afzal Shadab and Marc Hesse
#Date modified: 08/05/21
from supporting_tools import *
#parameters
simulation_name = f'two_layered'
m = 3 #Cozeny-Karman coefficient for numerator K = K0 (1-phi_i)^m
n = 2 #Corey-Brooks coefficient krw = krw0 * se(sw)^n
s_wr = 0.0 #Residual water saturation
s_gr = 0.0 #Residual gas saturation
phi_U = 0.5; phi_L = 0.2 #Porosity of upper and lower layer
theta_L = s_gr*phi_L #setting the saturation in domain
zsurface = 0 #dimensionless surface height
z0 = 1.0 #dimensionless location of jump
R = 0.64 #Array of dimensionless rainfall rates
#spatial discretization
zbottom = 2
Nz = 1000
zc = np.linspace(zsurface,zbottom,Nz) #depth array
phi = phi_U*np.ones((len(zc),1)) #porosity vector: upper layer
phi[zc>z0] = phi_L #porosity vector: upper layer
#temporal discretization
tmax = 20 #time scaling with respect to z0/fc
Nt = 20000#time steps
t = np.linspace(0,tmax,Nt+1) #time array
#time stamps of interest: [beginning, stage1, stage2: just after saturation (ts), stage2, stage2: ponding time (tp), stage3]
t_interest = [0,0.3,0.6249999999999999 +0.005,0.7,0.871335520204737,1.0]
#############################################################
#Functions for two-layered soils
#############################################################
#Flux in the saturated region qs
def qs(x):
return (x-1)/(x/f_Cm(np.array([phi_U]),m,phi_U) - 1/f_Cm(np.array([phi_L]),m,phi_U))
#To convert dimensionless rainfall to moisture content at the surface
def funtheta_L2phil_U(theta_U):
return (f_Cm(np.array([phi_U]),m,phi_U)[0]*f_Cn(np.array([theta_U]),np.array([phi_U]),s_gr,s_wr,n))[0] - R
theta_U = opt.fsolve(funtheta_L2phil_U,0.01)[0]
#Stage 2: ODEs for upper and lower shock locations y[0] and y[1] respectively: Equations (26) and (27) from paper
def rhs_stage2(t, y):
return [((y[0]-y[1])/(y[0]/f_Cm(np.array([phi_U]),m,phi_U)-y[1]/f_Cm(np.array([phi_L]),m,phi_U))-f_Cm(np.array([phi_U]),m,phi_U)*f_Cn(np.array([theta_U]),np.array([phi_U]),s_gr,s_wr,n))[0]/(phi_U*(1-s_gr)-theta_U), \
((y[0]-y[1])/(y[0]/f_Cm(np.array([phi_U]),m,phi_U)-y[1]/f_Cm(np.array([phi_L]),m,phi_U))-f_Cm(np.array([phi_L]),m,phi_U)*f_Cn(np.array([theta_L]),np.array([phi_L]),s_gr,s_wr,n))[0]/(phi_L*(1-s_gr)-theta_L)]
#Stage 3: ODEs for upper and lower shock locations y[0] and y[1] respectively: Equations (26) and (27) from paper
def rhs_stage3(t, y):
return [0, \
((y[0]-y[1])/(y[0]/f_Cm(np.array([phi_U]),m,phi_U)-y[1]/f_Cm(np.array([phi_L]),m,phi_U))-f_Cm(np.array([phi_L]),m,phi_U)*f_Cn(np.array([theta_L]),np.array([phi_L]),s_gr,s_wr,n))[0]/(phi_L*(1-s_gr)-theta_L)]
print("############################################################# \n")
print("2 (top) Volume fraction with depth")
print("############################################################# \n")
# First set up the figure, the axis
fig,([ax1,ax2,ax3,ax4,ax5,ax6]) = time_sequenced_figure(zbottom,zsurface,0,1,t_interest)
#Relation (32) - (35) to find the shock speed ratio k
aa = phi_U/phi_L*(1-theta_U/phi_U-s_gr)/(1-theta_L/phi_L-s_gr)*(1- f_Cm(np.array([phi_L]),m,phi_U) * f_Cn(np.array([theta_L]),np.array([phi_L]),s_gr,s_wr,n) / f_Cm(np.array([phi_U]),m,phi_U))
bb =-phi_U/phi_L*(1-theta_U/phi_U-s_gr)/(1-theta_L/phi_L-s_gr)*(1- f_Cm(np.array([phi_L]),m,phi_U) * f_Cn(np.array([theta_L]),np.array([phi_L]),s_gr,s_wr,n) / f_Cm(np.array([phi_L]),m,phi_U)) - (1 - f_Cm(np.array([phi_U]),m,phi_U) * f_Cn(np.array([theta_U]),np.array([phi_U]),s_gr,s_wr,n) / f_Cm(np.array([phi_U]),m,phi_U))
cc = 1 - f_Cm(np.array([phi_U]),m,phi_U) * f_Cn(np.array([theta_U]),np.array([phi_U]),s_gr,s_wr,n) / f_Cm(np.array([phi_L]),m,phi_U)
k = (-bb-np.sqrt(bb**2-4*aa*cc))/(2*aa)
#Calculating the shock speeds
s_U_analy =((qs(k) - f_Cm(np.array([phi_U]),m,phi_U)*f_Cn(np.array([theta_U]),np.array([phi_U]),s_gr,s_wr,n))[0]/(phi_U*(1-s_gr-theta_U/phi_U)))
s_L_analy =((qs(k) - f_Cm(np.array([phi_L]),m,phi_U)*f_Cn(np.array([theta_L]),np.array([phi_L]),s_gr,s_wr,n))[0]/(phi_L*(1-theta_L/phi_L-s_gr)))
Sf = ((f_Cm(np.array([phi_U]),m,phi_U)[0]*f_Cn(np.array([theta_U]),np.array([phi_U]),s_gr,s_wr,n))[0] / (phi_U*(theta_U/phi_U - s_wr))) #initial front speed
ts = (z0 - zsurface)/Sf #dimensionless time of saturation
tp = ts + (zsurface-z0)/s_U_analy #dimensionless time of ponding
xf = zsurface + Sf*t_interest[0]
S_w_analy_int = (1-s_gr)*np.ones((Nz,1))
S_w_analy_int[zc<=xf] = theta_U/phi_U
S_w_analy_int[zc>=xf] = theta_L/phi_L
S_w_analy_int_combined = []
S_w_analy_int_combined = S_w_analy_int.copy()
ax1.fill_betweenx(zc,1, facecolor=red,label=r'$\phi_g$')
ax1.fill_betweenx(zc,(1-phi+phi*S_w_analy_int)[:,0], facecolor=blue,label=r'$\phi_w$')
ax1.fill_betweenx(zc,(1-phi)[:,0], facecolor=brown,label=r'$\phi_s$')
xf = zsurface + Sf*t_interest[1]
S_w_analy_int = (1-s_gr)*np.ones((Nz,1))
S_w_analy_int[zc<=xf] = theta_U/phi_U
S_w_analy_int[zc>=xf] = theta_L/phi_L
S_w_analy_int_combined = np.hstack([S_w_analy_int_combined,S_w_analy_int])
ax2.fill_betweenx(zc,1, facecolor=red,label=r'$\phi_g$')
ax2.fill_betweenx(zc,(1-phi+phi*S_w_analy_int)[:,0], facecolor=blue,label=r'$\phi_w$')
ax2.fill_betweenx(zc,(1-phi)[:,0], facecolor=brown,label=r'$\phi_s$')
res = solve_ivp(rhs_stage2, (0, 0.005), [-1e-14,1e-15],t_eval=[0.005])
S_w_analy_int = (1-s_gr)*np.ones((Nz,1))
S_w_analy_int[zc<=res.y[0,0]+z0] = theta_U/phi_U
S_w_analy_int[zc>=res.y[1,0]+z0] = theta_L/phi_L
S_w_analy_int_combined = np.hstack([S_w_analy_int_combined,S_w_analy_int])
ax3.fill_betweenx(zc,1, facecolor=red,label=r'$\phi_{gas}$')
ax3.fill_betweenx(zc,(1-phi+phi*S_w_analy_int)[:,0], facecolor=blue,label=r'$\phi_{water}$')
ax3.fill_betweenx(zc,(1-phi)[:,0], facecolor=brown,label=r'$\phi_{soil}$')
res = solve_ivp(rhs_stage2, (0, t_interest[3]-ts), [-1e-14,1e-15],t_eval=[t_interest[3]-ts])
S_w_analy_int = (1-s_gr)*np.ones((Nz,1))
S_w_analy_int[zc<=res.y[0,0]+z0] = theta_U/phi_U
S_w_analy_int[zc>=res.y[1,0]+z0] = theta_L/phi_L
S_w_analy_int_combined = np.hstack([S_w_analy_int_combined,S_w_analy_int])
ax4.fill_betweenx(zc,1, facecolor=red,label=r'$\phi_g$')
ax4.fill_betweenx(zc,(1-phi+phi*S_w_analy_int)[:,0], facecolor=blue,label=r'$\phi_w$')
ax4.fill_betweenx(zc,(1-phi)[:,0], facecolor=brown,label=r'$\phi_s$')
res = solve_ivp(rhs_stage2, (0, tp-ts), [-1e-14,1e-15],t_eval=[tp-ts])
S_w_analy_int = (1-s_gr)*np.ones((Nz,1))
S_w_analy_int[zc<=res.y[0,0]+z0] = theta_U/phi_U
S_w_analy_int[zc>=res.y[1,0]+z0] = theta_L/phi_L
S_w_analy_int_combined = np.hstack([S_w_analy_int_combined,S_w_analy_int])
ax5.fill_betweenx(zc,1, facecolor=red,label=r'$\phi_{gas}$')
ax5.fill_betweenx(zc,(1-phi+phi*S_w_analy_int)[:,0], facecolor=blue,label=r'$\phi_{water}$')
ax5.fill_betweenx(zc,(1-phi)[:,0], facecolor=brown,label=r'$\phi_{soil}$')
ytop = res.y[0,0]
ybot = res.y[1,0]
res = solve_ivp(rhs_stage3, (0, t_interest[5]-tp), [ytop,ybot],t_eval=[t_interest[5]-tp])
S_w_analy_int = (1-s_gr)*np.ones((Nz,1))
S_w_analy_int[zc<=res.y[0,0]+z0] = theta_U/phi_U
S_w_analy_int[zc>=res.y[1,0]+z0] = theta_L/phi_L
S_w_analy_int_combined = np.hstack([S_w_analy_int_combined,S_w_analy_int])
ax6.fill_betweenx(zc,1, facecolor=red,label=r'$\phi_{gas}$')
ax6.fill_betweenx(zc,(1-phi+phi*S_w_analy_int)[:,0], facecolor=blue,label=r'$\phi_{water}$')
ax6.fill_betweenx(zc,(1-phi)[:,0], facecolor=brown,label=r'$\phi_{soil}$')
ax3.set_title(r'''%.2f ($t_s'$)'''%(ts+0.005), fontsize='medium')
ax5.set_title(r'''%.2f ($t_p'$)'''%tp, fontsize='medium')
ax1.legend(loc='lower left', shadow=False, fontsize='medium')
plt.xlabel("Volume fractions $\phi$", fontsize='medium')
plt.savefig(f"./Figures/{simulation_name}_CL{theta_U}CR{theta_L}_phiL{phi_U}phiR{phi_L}.pdf")
print("############################################################# \n")
print("2 (bottom) Saturation with depth")
print("############################################################# \n")
fig,([ax1,ax2,ax3,ax4,ax5,ax6]) = time_sequenced_figure(zbottom,zsurface,-0.05,1.05,t_interest)
ax1.plot(S_w_analy_int_combined[:,0],zc , c = 'k',linestyle='--')
ax2.plot(S_w_analy_int_combined[:,1],zc , c = 'k',linestyle='--')
ax3.plot(S_w_analy_int_combined[:,2],zc , c = 'k',linestyle='--')
ax4.plot(S_w_analy_int_combined[:,3],zc , c = 'k',linestyle='--')
ax5.plot(S_w_analy_int_combined[:,4],zc , c = 'k',linestyle='--')
ax6.plot(S_w_analy_int_combined[:,5],zc , c = 'k',linestyle='--')
ax3.set_title(r'''%.2f ($t_s'$)'''%(ts+0.005), fontsize='medium')
ax5.set_title(r'''%.2f ($t_p'$)'''%tp, fontsize='medium')
plt.xlabel("Water saturation $s_w$", fontsize='medium')
plt.savefig(f"./Figures/swvsZpanelshock_{simulation_name}_CL{theta_U}CR{theta_L}_phiL{phi_U}phiR{phi_L}.pdf")
print("############################################################# \n")
print("3 Dimensionless infiltration rate with dimensionless time")
print("############################################################# \n")
print('R/fc:',R,''', tp':''',tp,''', zs':''',z0,''', ts':''',ts,'\n')
res = solve_ivp(rhs_stage2, (0, tp-ts), [-1e-14,1e-15],t_eval=[tp-ts])
t0 = np.linspace(0,ts,1000)
qs0 = (f_Cm(np.array([phi_U]),m,phi_U)*f_Cn(np.array([theta_U]),np.array([phi_U]),s_gr,s_wr,n))* np.ones_like(t0)
t1 = np.linspace(ts,tp,1000)
qs1 = (f_Cm(np.array([phi_U]),m,phi_U)*f_Cn(np.array([theta_U]),np.array([phi_U]),s_gr,s_wr,n))*np.ones_like(t1)
t2 = np.linspace(tp,10,10000)
res= solve_ivp(rhs_stage3, (t2[0],t2[-1]), [ytop,ybot],t_eval=t2)
y = res.y
qs2 = (y[0]-y[1])/(y[0]/f_Cm(np.array([phi_U]),m,phi_U) - y[1]/f_Cm(np.array([phi_L]),m,phi_U))
T = np.concatenate([t0,t1,t2])
QS = np.concatenate([qs0,qs1,qs2])
fig = plt.figure(figsize=(8,8) , dpi=100)
plot = plt.plot(T,QS/f_Cm(np.array([phi_U]),m,phi_U),c='r')
plt.hlines(f_Cm(np.array([phi_L]),m,phi_U), np.min(T), np.max(T), colors=gray, linestyles='--')
plt.vlines(ts, 0, qs0[0]/f_Cm(np.array([phi_U]),m,phi_U), colors=gray, linestyles='--')
plt.vlines(tp, 0, qs0[0]/f_Cm(np.array([phi_U]),m,phi_U), colors=gray, linestyles='--')
plt.hlines(f_Cm(np.array([phi_L]),m,phi_U), np.min(T), np.max(T), colors=gray, linestyles='--')
plt.ylabel(r'''$I(t')/f_c$''', fontsize='medium')
plt.xlabel(r'''$t'$''', fontsize='medium')
plt.xticks(fontsize='medium')
plt.yticks(fontsize='medium')
plt.xlim([np.min(T), np.max(T)])
plt.ylim([0,round(np.max(QS),1)+0.1])
plt.tight_layout(pad=0.4, w_pad=0.5, h_pad=1.0)
plt.savefig(f"./Figures/ICvsT_shock_{simulation_name}_CL{theta_U}CR{theta_L}_phiL{phi_U}phiR{phi_L}.pdf")