def save_rate(outdir, xvec, yvec, rate, grdnaming):
os.makedirs(outdir, exist_ok=True)
grd_io.write_grd(xvec,
yvec,
rate,
'%s/rate_mm_yr.grd' % outdir,
naming=grdnaming)
def save_ts(outdir, xvec, yvec, dates, ts, grdnaming):
os.makedirs(outdir, exist_ok=True)
for i in range(len(dates)):
grd_io.write_grd(xvec,
yvec,
ts[i],
'%s/ts_mm_%04d.grd' % (outdir, dates[i]),
naming=grdnaming)
def check_ts(ts, data, igram_ids, xvec, yvec, unw_check_threshold, grdnaming):
#predict each interferogram from the timeseries, and compare it to the actual
os.makedirs('ts_check', exist_ok=True)
for i, pair in enumerate(igram_ids):
synth_igram = ts[pair[1], :, :] - ts[pair[0], :, :]
resid = synth_igram - data[i, :, :]
grd_io.write_grd(xvec,
yvec,
resid,
'ts_check/resid_%d.grd' % i,
naming=grdnaming)
#mask any data that exceed the threshold
mask = ~np.isnan(resid)
mask[mask] = np.abs(resid[mask]) > unw_check_threshold
#print(i)
#print('non-nan before masking: %d'%np.count_nonzero(~np.isnan(data[i])))
data[i, mask] = np.nan
#print('non-nan after masking: %d'%np.count_nonzero(~np.isnan(data[i])))
return data
def check_data(xvec, yvec, data, nt, igram_ids, unw_check_threshold,
grdnaming):
# function to check all loops and identify pixels where the loop value is close to zero. Mark these pixels as valid for each igram in the triplet
# create a NaN matrix the same size as data (a 3D matrix containing all of the interferograms)
validdata = np.zeros(np.shape(data))
# output directory
os.makedirs('unw_check', exist_ok=True)
#iterate over loop list
for loop in itertools.combinations(range(nt), 3):
#get interferogram IDs for each loop
try:
i0 = igram_ids.index((loop[0], loop[1]))
i1 = igram_ids.index((loop[1], loop[2]))
i2 = igram_ids.index((loop[0], loop[2]))
except ValueError:
#one of the interferograms does not exist, skip this loop
continue
#compute the loop residual
cycle = data[i0, :, :] + data[i1, :, :] - data[i2, :, :]
#get zero if the pixel has a large residual, otherwise 1
#cyclevalid checks if the values are less than the threshold. Any NaN values in the loop will just result in zero vote
mask = ~np.isnan(cycle)
mask[mask] = np.abs(cycle[mask]) < unw_check_threshold
cyclevalid = 1 * mask
#note, testing 'less than' on NaN throws a RuntimeWarning, but still gives the correct answer:
#cyclevalid=1*(np.abs(cycle)<threshold)
#validdata tracks the number of total 'votes' for a good pixel for each interferogram.
validdata[i0, :, :] += cyclevalid
validdata[i1, :, :] += cyclevalid
validdata[i2, :, :] += cyclevalid
loopname = '%d_%d_%d' % (loop[0], loop[1], loop[2])
grd_io.write_grd(xvec,
yvec,
cycle,
'unw_check/loop_%s.grd' % loopname,
naming=grdnaming)
grd_io.write_grd(xvec,
yvec,
cyclevalid,
'unw_check/cyclevalid_%s.grd' % loopname,
naming=grdnaming)
# mask pixels with too few votes.
# note, it's possible for a bad pixel to still get 1 vote, if there is an opposing unwrapping error in another ifg.
minvalid = 2
data[validdata < minvalid] = np.nan
# TODO: get the interferogram directory and write out valid / invalid / (number of votes)
for i in range(len(validdata)):
grd_io.write_grd(xvec,
yvec,
validdata[i, :, :],
'unw_check/valid_%d.grd' % i,
naming=grdnaming)
return data
args = parser.parse_args()
xvec, yvec, insardat = grd_io.read_grd(args.insarfile)
print('read %s, dimensions (%d,%d)' %
(args.insarfile, len(xvec), len(yvec)))
X, Y = np.meshgrid(xvec, yvec)
gpsdat = np.loadtxt(args.gpsfile)
print('read %s, length %d' % (args.gpsfile, len(gpsdat)))
G, d = detrend_ts.construct_gpsconstraint(0, insardat, gpsdat, X, Y)
#select the specified number of trend parameters
G = G[:, 0:3]
#invert for a trend that matches the value around given GPS points
m = np.dot(np.linalg.pinv(G), d)
print(m)
#reconstruct the model - a bit ugly here dealing with unknown number of trendparams
onemat = np.ones(np.shape(X))
model = detrend_ts.reconstruct_model_nparams(3, m, X, Y, onemat)
insar_detrend = insardat - model
grdnaming = grd_io.read_naming(args.insarfile)
outfile = '%s_fitgps.grd' % os.path.splitext(args.insarfile)[0]
print('writing %s' % outfile)
grd_io.write_grd(xvec, yvec, insar_detrend, outfile, naming=grdnaming)
#convert to NetCDF-3 file type
subprocess.call('nccopy -k classic %s temp_nc3.grd' % args.infile,
shell=True)
datfile = 'temp_nc3.grd'
else:
datfile = args.infile
if args.lonlat:
grdnaming = 'lonlat'
else:
grdnaming = 'xy'
#read input file
x, y, z = grd_io.read_grd(datfile, naming=grdnaming)
if args.ncconvert:
#remove temporary NetCDF-3 file
os.remove('temp_nc3.grd')
#do nearest neighbor interpolation
interpdata = nneigh_interp(x, y, z)
#write output file
grd_io.write_grd(x, y, interpdata, args.outfile, naming=grdnaming)
if args.plots:
import matplotlib.pyplot as plt
plot_grid(z, 'input data')
plot_grid(interpdata, 'interpolated data')
plt.show()