def ResampleWind():
tt = np.load(coast_root + 'Left coast for each row.npy')
d = rd.SentinelData()
d.Get_List_NetCDF(ocn_root)
for indice in range(len(d.series)):
temp = rd.Data_Level2(d.series[indice], d.FList[indice])
temp.Get_WindData()
temp.CalWindVector()
mask = temp.windX.mask.reshape(-1)
Ind = np.where(mask == 0)[0]
lon = temp.lon.data.reshape(-1)
lat = temp.lat.data.reshape(-1)
U = temp.windX.data.reshape(-1)
V = temp.windY.data.reshape(-1)
lon, lat, U, V = lon[Ind], lat[Ind], U[Ind], V[Ind]
r, c, u, v = md2.GridLL_Wind(lon, lat, U, V)
ind = [xx for xx in range(len(r)) if tt[r[xx]] > c[xx]]
r_new = [r[xx] for xx in ind]
c_new = [c[xx] for xx in ind]
u_new = [u[xx] for xx in ind]
v_new = [v[xx] for xx in ind]
r, c, u, v = r_new, c_new, u_new, v_new
del r_new, c_new, u_new, v_new, ind
# md2.save_fig(d.series[indice], r, c, u, v)
if len(r) >= 4:
u10, v10 = Lc.InterpolationWinds(r, c, u, v)
np.save(level2_root + 'Interp/' + d.series[indice] + '-u10.npy',
u10)
np.save(level2_root + 'Interp/' + d.series[indice] + '-v10.npy',
v10)
u10 = np.zeros([15, 19], dtype=np.float16)
v10 = np.zeros_like(u10)
du = {}
dv = {}
for ii in range(len(r)):
r1, c1 = r[ii] // 150, c[ii] // 150
if (r1, c1) in du:
du[r1, c1].append(u[ii])
dv[r1, c1].append(v[ii])
else:
du[r1, c1] = [u[ii]]
dv[r1, c1] = [v[ii]]
for rr, cc in du.keys():
u10[rr, cc] = np.mean(du[rr, cc])
v10[rr, cc] = np.mean(dv[rr, cc])
np.save(level2_root + 'NotInterp/' + d.series[indice] + '-u10.npy',
u10)
np.save(level2_root + 'NotInterp/' + d.series[indice] + '-v10.npy',
v10)
return None
def MakeAGrid(name, dir):
dat = rd.Data_Level1(name, dir)
dat.OneStep()
GCP = dat.GCPs
SigmaNaught = dat.denoiseNRCS
y_lim, x_lim = SigmaNaught.shape
del dat, SigmaNaught
n = 10
gm.AllLonLat('Base', x_lim, y_lim, GCP, n)
return None
def eliminate(sss):
# N_list=list(range(38, 72))
# N_list.remove(39)
# N_list.remove(58)
# N_list.remove(69)
N_list = list(range(36, 67))
temp = rd.Grid_data()
temp.LoadData(ty=sss)
wspeed_cds = np.sqrt(temp.u_cds * temp.u_cds + temp.v_cds * temp.v_cds)
wspeed_s2 = np.sqrt(temp.u_s2 * temp.u_s2 + temp.v_s2 * temp.v_s2)
wd_cds = np.arctan2(-temp.u_cds, -temp.v_cds) * 180 / np.pi
wd_s2 = np.arctan2(-temp.u_s2, -temp.v_s2) * 180 / np.pi
NRCS = temp.NRCS
del temp
tt = np.zeros(15, dtype=np.int8)
ss = np.zeros_like(tt)
ss[:] = 18
for r in range(15):
for c in range(19):
if wspeed_s2[42, r, c] > 0:
tt[r] = max(tt[r], c)
ss[r] = min(ss[r], c)
nums = len(N_list)
ws2 = np.zeros([nums, 15, 19], dtype=np.float_)
wcds = np.zeros_like(ws2)
ds2 = np.zeros_like(ws2)
dcds = np.zeros_like(ws2)
sigma = np.zeros_like(ws2)
for r in range(15):
if ss[r] <= tt[r]:
for c in range(ss[r], tt[r] + 1):
slist = Lc.Extract_NRCS(r, c, NRCS)
sigma[:, r, c] = slist
now = 0
for n in N_list:
ws2[now, r, c] = wspeed_s2[n, r, c]
wcds[now, r, c] = wspeed_cds[n, r, c]
ds2[now, r, c] = wd_s2[n, r, c]
dcds[now, r, c] = wd_cds[n, r, c]
now += 1
# ds2 = np.where(ds2<0, ds2+360, ds2)
for n in range(len(N_list)):
for r in range(15):
for c in range(19):
if np.logical_not(np.isnan(ds2[n, r, c])):
if ds2[n, r, c] < 0:
ds2[n, r, c] += 360
if np.logical_not(np.isnan(dcds[n, r, c])):
if dcds[n, r, c] < 0:
dcds[n, r, c] += 360
# dcds = np.where(dcds<0, dcds+360, dcds)
return sigma, ws2, wcds, ds2, dcds, ss, tt
def Resample(n, mode='uniform'):
d = rd.SentinelData()
d.Get_List(file_root)
m = 1 # resize factor
for indice in range(n, len(d.series)):
stt = time.time()
temp = rd.Data_Level1(d.series[indice], d.FList[indice])
temp.OneStep()
GCP = temp.GCPs
SigmaNaught = temp.denoiseNRCS
name = 'Layer' + str(indice) + '-' + d.series[indice]
del temp
y1, x1 = SigmaNaught.shape
sub_img, up_left = cod.N_sub(tempN, SigmaNaught)
del SigmaNaught, up_left
jo.NMatrix_save(sub_img, 'SigmaNaught', temp_root)
del sub_img
md2.AllLonLat(name, x1, y1, GCP, m)
img, pixel_num = md2.GridLL_SigmaNaught(mode)
jo.Matrix_save(img, name, layer_root)
jo.Matrix_save(pixel_num, name + '-distribution', layer_root)
print('1 image:', (time.time() - stt) / 60, 'mins')
print('Done!\n')
return None
def Merge(root):
s1 = datetime(2017, 1, 11)
s2 = datetime(2019, 5, 25)
datelist = [s1]
nows = s1
while nows != s2:
nows = nows + timedelta(days=12)
datelist.append(nows)
data = rd.winddata()
data.GetData(datelist, root, ty='resize')
np.save(root + 'u10.npy', data.u)
np.save(root + 'v10.npy', data.v)
# data.GetData(datelist, root)
# np.save(root + 'u10-interp.npy', data.u)
# np.save(root + 'v10-interp.npy', data.v)
return None
def MatchCDS():
tt = np.load(coast_root + 'Left coast for each row.npy')
data = rd.layers()
data.LoadData()
for indice in range(len(data.t)):
lon, lat, u, v = md2.ReadCDSData(data.t[indice][:-1])
rows, cols = u.shape
lon_N = [lon[c] for r in range(rows) for c in range(cols)]
lat_N = [lat[r] for r in range(rows) for c in range(cols)]
lon_N = np.array(lon_N)
lat_N = np.array(lat_N)
u, v = u.reshape(-1), v.reshape(-1)
r, c, u, v = md2.GridLL_Wind(lon_N, lat_N, u, v)
# ind = [xx for xx in range(len(r)) if tt[r[xx]] > c[xx]]
# r_new = [r[xx] for xx in ind]
# c_new = [c[xx] for xx in ind]
# u_new = [u[xx] for xx in ind]
# v_new = [v[xx] for xx in ind]
# r, c, u, v = r_new, c_new, u_new, v_new
# del r_new, c_new, u_new, v_new, ind
# md2.save_fig(data.t[indice][:-1], r, c, u, v, cds_root)
# if len(r)>=4:
# u10, v10 = Lc.InterpolationWinds(r, c, u, v)
# np.save(cds_root + 'Interp/' + data.t[indice][:-1] + '-u10.npy', u10)
# np.save(cds_root + 'Interp/' + data.t[indice][:-1] + '-v10.npy', v10)
# u10 = np.zeros([15,19], dtype = np.float16)
# v10 = np.zeros_like(u10)
du = {}
dv = {}
for ii in range(len(r)):
r1, c1 = r[ii] // 150, c[ii] // 150
if (r1, c1) in du:
du[r1, c1].append(u[ii])
dv[r1, c1].append(v[ii])
else:
du[r1, c1] = [u[ii]]
dv[r1, c1] = [v[ii]]
# for rr, cc in du.keys():
# u10[rr, cc] = np.mean(du[rr, cc])
# v10[rr, cc] = np.mean(dv[rr, cc])
u10, v10 = Lc.InterpolationCDSwind(du, dv)
np.save(cds_root + 'NotInterp/' + data.t[indice][:-1] + '-u10.npy',
u10)
np.save(cds_root + 'NotInterp/' + data.t[indice][:-1] + '-v10.npy',
v10)
return None
def OtherLayer(d, n):
for indice in range(n, len(d.series)):
temp = rd.Data_Level1(d.series[indice], d.FList[indice])
temp.OneStep()
GCP = temp.GCPs
SigmaNaught = temp.denoiseNRCS
name = 'Layer' + str(indice) + '-' + d.series[indice]
del temp
y1, x1 = SigmaNaught.shape
gm.AllLonLat(name, x1, y1, GCP, 5)
start = time.time()
SigmaNaught = Lc.Resize_SigmaNaught(SigmaNaught, 5)
print('Time for resize SigmaNaught:', (time.time() - start) / 60,
'mins')
Sigma_N, nums_arr = gm.GridSigmaNaught(SigmaNaught)
print('Time for resampling 1 SAR image:', (time.time() - start) / 60,
'mins')
jo.Matrix_save(Sigma_N, name, layer_root)
return None
def MeanValue():
ll = rd.layers()
ll.LoadData()
ll.CalMean()
return None
if __name__ == '__main__':
s1 = datetime(2017,1,11)
s2 = datetime(2019,5,25)
datelist = [s1]
nows = s1
while nows != s2:
nows = nows + timedelta(days=12)
datelist.append(nows)
dates = [datetime.strftime(s, '%Y%m%d') for s in datelist]
Nlist = list(range(36,67))
temp = rd.Grid_data()
temp.LoadData()
wspeed_cds = np.sqrt(temp.u_cds*temp.u_cds+temp.v_cds*temp.v_cds)
wspeed_s2 = np.sqrt(temp.u_s2*temp.u_s2+temp.v_s2*temp.v_s2)
wd_cds = np.arctan2(-temp.u_cds, -temp.v_cds)*180/np.pi
wd_s2 = np.arctan2(-temp.u_s2, -temp.v_s2)*180/np.pi
tt = temp.mask[0]
mask = temp.mask[1].astype(np.float_)
del temp
spd = wspeed_s2 - wspeed_cds
wdd = wd_s2 - wd_cds
del wspeed_cds, wspeed_s2, wd_cds, wd_s2
mask[mask == 0] = np.nan
nums, rows, cols = spd.shape
for n in range(36,nums):
spd[n, :, :] = spd[n, :, :] * mask