def syn_sum_amp_iso_clvd(parameter1, parameter2):
Gt_Z = parameter1[0]
Gt_E = parameter1[1]
Gt_N = parameter1[2]
d_z = parameter1[3]
d_e = parameter1[4]
d_n = parameter1[5]
index_left = parameter1[6]
index_right = parameter1[7]
list_input = parameter1[8]
MTparameter = parameter2
m_rr, m_tt, m_pp, m_rp, m_rt, m_tp = functions.nmtensor(
MTparameter[0], MTparameter[1], MTparameter[2], MTparameter[3],
MTparameter[4])
m_tmp = np.array([m_rr, m_tt, m_pp, m_rt, m_rp, m_tp])
#print(i, j, k)
### assemble the greens functions to synthetics
d_syn_z = np.dot(Gt_Z, m_tmp)
d_syn_e = np.dot(Gt_E, m_tmp)
d_syn_n = np.dot(Gt_N, m_tmp)
cc = 0.
for _i, stid in enumerate(list_input):
cc1 = functions.L1_L2misfit_amp(
d_z[0][index_left[_i]:index_right[_i]],
d_syn_z[index_left[_i]:index_right[_i]])
cc2 = functions.L1_L2misfit_amp(
d_e[0][index_left[_i]:index_right[_i]],
d_syn_e[index_left[_i]:index_right[_i]])
cc3 = functions.L1_L2misfit_amp(
d_n[0][index_left[_i]:index_right[_i]],
d_syn_n[index_left[_i]:index_right[_i]])
cc += cc1 + cc2 + cc3
corr_value = cc / len(list_input) / 3
#print(corr_value)
return corr_value
def syn_sum_fem_fpm(strike, dip, rake, iso, clvd, Gt_Z, Gt_E, Gt_N, d_z, d_e,
d_n, index_left, index_right):
m_rr, m_tt, m_pp, m_rp, m_rt, m_tp = functions.nmtensor(strike, dip, rake)
m_tmp = np.array([m_rr, m_tt, m_pp, m_rt, m_rp, m_tp])
#print(i, j, k)
### assemble the greens functions to synthetics
d_syn_z = np.dot(Gt_Z, m_tmp)
d_syn_e = np.dot(Gt_E, m_tmp)
d_syn_n = np.dot(Gt_N, m_tmp)
cc = 0.
for _i, stid in enumerate(list_input):
cc1 = functions.TF_fem(d_z[0][index_left[_i]:index_right[_i]],
d_syn_z[index_left[_i]:index_right[_i]])
cc2 = functions.TF_fem(d_e[0][index_left[_i]:index_right[_i]],
d_syn_e[index_left[_i]:index_right[_i]])
cc3 = functions.TF_fem(d_n[0][index_left[_i]:index_right[_i]],
d_syn_n[index_left[_i]:index_right[_i]])
cc1 += functions.TF_fpm(d_z[0][index_left[_i]:index_right[_i]],
d_syn_z[index_left[_i]:index_right[_i]])
cc2 += functions.TF_fpm(d_e[0][index_left[_i]:index_right[_i]],
d_syn_e[index_left[_i]:index_right[_i]])
cc3 += functions.TF_fpm(d_n[0][index_left[_i]:index_right[_i]],
d_syn_n[index_left[_i]:index_right[_i]])
cc += cc1 + cc2 + cc3
corr_value = cc / len(list_input) / 3
return corr_value
def calculate_misfits(strike_store, dip_store, rake_store, iso_store,
clvd_store, Gt_Z, Gt_E, Gt_N, d_z, d_e, d_n, index_left,
index_right):
from tqdm import tqdm
import multiprocessing as mp
pool = mp.Pool(processes=8)
print(mp.cpu_count())
parameter = [[
strike_store[i], dip_store[j], rake_store[k], iso_store[m],
clvd_store[n], Gt_Z, Gt_E, Gt_N, d_z, d_e, d_n, index_left, index_right
] for m in tqdm(np.arange(0, iso_store.size, 1, dtype=int))
for n in np.arange(0, clvd_store.size, 1, dtype=int)
for i in np.arange(0, strike_store.size, 1, dtype=int)
for j in np.arange(0, dip_store.size, 1, dtype=int)
for k in np.arange(0, rake_store.size, 1, dtype=int)]
print(parameter)
outs = pool.map_async(syn_sum_amp_iso_clvd, tqdm(parameter))
pool.close()
pool.join()
output = outs.get()
data = np.array(output)
data_reshape = data.reshape(iso_store.size, clvd_store.size,
strike_store.size, dip_store.size,
rake_store.size)
index = np.unravel_index(np.argmin(data_reshape, axis=None),
data_reshape.shape)
m_rr, m_tt, m_pp, m_rp, m_rt, m_tp = functions.nmtensor(
strike_store[index[2]], dip_store[index[3]], rake_store[index[4]],
iso_store[index[0]], clvd_store[index[1]])
fault = np.array(
[strike_store[index[2]], dip_store[index[3]], rake_store[index[4]]])
m_inverse = np.array([m_rr, m_tt, m_pp, m_rt, m_rp, m_tp])
d_syn = np.dot(Gt, m_inverse)
return data, m_inverse, d_syn
def syn_sum_L2_iso_clvd_weight_abs_M0(parameter1, parameter2):
Gt_Z = parameter1[0]
Gt_E = parameter1[1]
Gt_N = parameter1[2]
d_z = parameter1[3]
d_e = parameter1[4]
d_n = parameter1[5]
weight = parameter1[9]
index_left = parameter1[6]
index_right = parameter1[7]
list_input = parameter1[8]
MTparameter = parameter2
m_rr, m_tt, m_pp, m_rp, m_rt, m_tp = functions.nmtensor(
MTparameter[0], MTparameter[1], MTparameter[2], MTparameter[3],
MTparameter[4])
#### here I inserted a amplification for the relative moment scalar for inverse the M0
m_tmp = MTparameter[5] * np.array([m_rr, m_tt, m_pp, m_rt, m_rp, m_tp
]) #print(i, j, k)
### assemble the greens functions to synthetics
d_syn_z = np.dot(Gt_Z, m_tmp)
d_syn_e = np.dot(Gt_E, m_tmp)
d_syn_n = np.dot(Gt_N, m_tmp)
cc = 0.
for _i, stid in enumerate(list_input):
cc1 = functions.L2misfit_abs(d_z[0][index_left[_i]:index_right[_i]],
d_syn_z[index_left[_i]:index_right[_i]])
cc2 = functions.L2misfit_abs(d_e[0][index_left[_i]:index_right[_i]],
d_syn_e[index_left[_i]:index_right[_i]])
cc3 = functions.L2misfit_abs(d_n[0][index_left[_i]:index_right[_i]],
d_syn_n[index_left[_i]:index_right[_i]])
cc += weight[_i] * weight[_i] * (cc1 + cc2 + cc3)
corr_value = cc / len(list_input) / 3
#print(corr_value)
return corr_value
