doc = 'Convert FEWS flood forecast data to hazard layers for InaSAFE'
parser = argparse.ArgumentParser(description=doc)
parser.add_argument('filename', type=str,
help='NetCDF filename from FEWS')
parser.add_argument('--hours', metavar='h', type=int, default=24,
help='Number of hours to use from forecast')
parser.add_argument('--regions', metavar='regions', type=str,
help=('Administrative areas to be flagged as '
'flooded or not'))
args = parser.parse_args()
print args
print
tif_filename = convert_netcdf2tif(args.filename, args.hours,
verbose=True)
# Tag each polygon with Y if it contains at least one pixel
# exceeding a specific threshold (e.g. 0.3m).
if args.regions is not None:
print 'Tagging %s as "affected" or not' % args.regions
polygons = read_layer(args.regions)
grid = read_layer(tif_filename)
res = tag_polygons_by_grid(polygons, grid,
threshold=0.3,
tag='affected')
# Keep only those that are affected (speeds things up a lot,
# but will reduce overall bounding box for buildings under
# consideration)
# geom = res.get_geometry()
def test_tag_regions_by_flood(self):
"""Regions can be tagged correctly with data from flood forecasts.
"""
threshold = 0.3
label = 'affected'
tif_filename = convert_netcdf2tif(self.nc_filename, 24, verbose=False)
region_filename = os.path.join(TESTDATA, 'rw_jakarta_singlepart.shp')
grid = read_layer(tif_filename)
polygons = read_layer(region_filename)
res = tag_polygons_by_grid(polygons, grid,
threshold=threshold,
tag=label)
os.remove(tif_filename)
geom = res.get_geometry()
data = res.get_data()
# Check correctness of affected regions
affected_geom = []
affected_data = []
for i, d in enumerate(data):
if d[label]:
g = geom[i]
affected_geom.append(g)
affected_data.append(d)
assert len(affected_geom) == 37
assert len(affected_data) == 37
# Check that every grid point exceeding threshold lies inside
# one of the polygons marked as affected
P, V = grid.to_vector_points()
flooded_points_geom = []
flooded_points_data = []
for i, point in enumerate(P):
val = V[i]
if val > threshold:
# Point that is flooded must be in one of the tagged polygons
found = False
for polygon in affected_geom:
if is_inside_polygon(point, polygon):
found = True
msg = ('No affected polygon was found for point [%f, %f] '
'with value %f' % (point[0], point[1], val))
verify(found, msg)
# Collected flooded points for visualisation
flooded_points_geom.append(point)
flooded_points_data.append({'depth': val})
# To generate files for visual inspection.
# See
# https://raw.github.com/AIFDR/inasafe/master/files/flood_tagging_test.png
# https://github.com/AIFDR/inasafe/blob/master/files/flood_tagging_test.tgz
tmp_filename = unique_filename(prefix='grid', suffix='.tif')
grid.write_to_file(tmp_filename)
#print 'Grid written to ', tmp_filename
tmp_filename = unique_filename(prefix='regions', suffix='.shp')
res.write_to_file(tmp_filename)
#print 'Regions written to ', tmp_filename
tmp_filename = unique_filename(prefix='flooded_points', suffix='.shp')
v = Vector(geometry=flooded_points_geom, data=flooded_points_data)
v.write_to_file(tmp_filename)
def test_convert_netcdf2tif(self):
"""NetCDF flood forecasts can be converted to tif.
"""
# First check that input file is as expected
from Scientific.IO.NetCDF import NetCDFFile
fid = NetCDFFile(self.nc_filename)
x = fid.variables['x'][:] # Longitudes
y = fid.variables['y'][:] # Latitudes
inundation_depth = fid.variables['Inundation_Depth'][:]
T = inundation_depth.shape[0] # Number of time steps
M = inundation_depth.shape[1] # Steps in the y direction
N = inundation_depth.shape[2] # Steps in the x direction
assert T == 71
assert M == 162 # Latitudes
assert N == 160 # Longitudes
assert len(x) == N
assert len(y) == M
# Pick a max value in an area known to have flooding
# (approximately picked with ncview)
max_136_51 = max(inundation_depth[:, 136, 51])
assert numpy.allclose(max_136_51, 1.58)
#print max_136_51
#print 'y[136]', y[136] # Lat
#print 'x[51]', x[51] # Lon
assert numpy.allclose(x[51], 106.7777)
assert numpy.allclose(y[136], -6.124634)
# Run script over all hours
all_hours_tif = convert_netcdf2tif(self.nc_filename, T, verbose=False)
msg = 'Expected file %s did not exist' % all_hours_tif
assert os.path.isfile(all_hours_tif), msg
# Read resulting layer and check
L = read_layer(all_hours_tif)
D = L.get_data()
os.remove(all_hours_tif)
# Check point taking up-down flip into account
assert numpy.allclose(D[-136 - 1, 51], max_136_51)
# Run script for one hour and check first band
one_hour_tif = convert_netcdf2tif(self.nc_filename, 1, verbose=False)
D = read_layer(one_hour_tif).get_data()
assert numpy.allclose(max(D.flat), 0.74) # Checked band 1 with QGIS
# Characterisation test of location of max inundation
assert D[28, 53] == max(D.flat)
assert numpy.allclose(y[28], -6.3199)
assert numpy.allclose(x[53], 106.781)
os.remove(one_hour_tif)
return
parser.add_argument('--hours',
metavar='h',
type=int,
default=24,
help='Number of hours to use from forecast')
parser.add_argument('--regions',
metavar='regions',
type=str,
help=('Administrative areas to be flagged as '
'flooded or not'))
args = parser.parse_args()
print args
print
tif_filename = convert_netcdf2tif(args.filename, args.hours, verbose=True)
# Tag each polygon with Y if it contains at least one pixel
# exceeding a specific threshold (e.g. 0.3m).
if args.regions is not None:
print 'Tagging %s as "affected" or not' % args.regions
polygons = read_layer(args.regions)
grid = read_layer(tif_filename)
res = tag_polygons_by_grid(polygons,
grid,
threshold=0.3,
tag='affected')
# Keep only those that are affected (speeds things up a lot,
# but will reduce overall bounding box for buildings under
# consideration)
def process_flood_event(netcdf_file=None, hours=24):
"""A function to process this_netcdf_file to a forecast file.
:param netcdf_file: The netcdf file. If it's None the download it.
:param hours: Positive integer determining how many bands to use.
:type hours: int
"""
print 'Start flood forecasting'
if netcdf_file is None:
# retrieve data from the web
netcdf_file = download_file_url(netcdf_url, forecast_directory)
else:
netcdf_file = download_file_url(netcdf_url,
name=netcdf_file,
download_directory=forecast_directory)
print 'Do flood forecasting for %s ...' % netcdf_file
## check if a forecasting file has been created or not
# is_exist, polyforecast_filepath = get_result_file_name(this_netcdf_file,
# hours)
#
#if is_exist:
# print 'Current flood forecasting has been already created.'
# print 'You can look it at %s' % polyforecast_filepath
# return
# convert to tif
# tif_file = polyforecast_filepath.replace('_regions.shp', '.tif')
tif_filename = convert_netcdf2tif(netcdf_file,
hours,
verbose=False,
output_dir=flood_directory)
print 'tif_file', tif_filename
tif_file = read_layer(tif_filename)
# check if there is another file with the same name
# if so, do not do the forecasting
polyforecast_filepath = tif_filename.replace('.tif', '_regions.shp')
zip_filename = polyforecast_filepath.replace('.shp', '.zip')
if os.path.isfile(zip_filename):
print('File %s is exist, so we do not do the forecasting' %
zip_filename)
else:
polygons = read_layer(polygons_path)
result = tag_polygons_by_grid(polygons,
tif_file,
threshold=0.3,
tag='affected')
new_geom = result.get_geometry()
new_data = result.get_data()
date = os.path.split(netcdf_file)[-1].split('_')[0]
v = Vector(geometry=new_geom,
data=new_data,
projection=result.projection,
keywords={
'category':
'hazard',
'subcategory':
'flood',
'title': ('%d hour flood forecast regions '
'in Jakarta at %s' % (hours, date))
})
print 'polyforecast_filepath', polyforecast_filepath
v.write_to_file(polyforecast_filepath)
print 'Wrote tagged polygons to %s' % polyforecast_filepath
# zip all file
if os.path.isfile(zip_filename):
print 'Has been zipped to %s' % zip_filename
else:
zip_shp(polyforecast_filepath,
extra_ext=['.keywords'],
remove_file=True)
print 'Zipped to %s' % zip_filename
def processFloodEvent(netcdf_file=None, hours=24):
"""A function to process netcdf_file to a forecast file.
"""
print 'Start flood forecasting'
if netcdf_file is None:
# retrieve data from the web
netcdf_file = download_file_url(netcdf_url, forecast_directory)
else:
netcdf_file = download_file_url(netcdf_url, name=netcdf_file,
download_directory=forecast_directory)
print 'Do flood forecasting for %s ...' % netcdf_file
# # check if a forecasting file has been created or not
# is_exist, polyforecast_filepath = get_result_file_name(netcdf_file, hours)
#
# if is_exist:
# print 'Current flood forecasting has been already created.'
# print 'You can look it at %s' % polyforecast_filepath
# return
# convert to tif
# tif_file = polyforecast_filepath.replace('_regions.shp', '.tif')
tif_filename = convert_netcdf2tif(netcdf_file, hours,
verbose=False, output_dir=flood_directory)
print 'tif_file', tif_filename
tif_file = read_layer(tif_filename)
# check if there is another file with the same name
# if so, do not do the forecasting
polyforecast_filepath = tif_filename.replace('.tif', '_regions.shp')
zip_filename = polyforecast_filepath.replace('.shp', '.zip')
if os.path.isfile(zip_filename):
print ('File %s is exist, so we do not do the forecasting'
% zip_filename)
else:
my_polygons = read_layer(polygons_path)
my_result = tag_polygons_by_grid(my_polygons, tif_file, threshold=0.3,
tag='affected')
new_geom = my_result.get_geometry()
new_data = my_result.get_data()
date = os.path.split(netcdf_file)[-1].split('_')[0]
v = Vector(geometry=new_geom, data=new_data,
projection=my_result.projection,
keywords={'category': 'hazard',
'subcategory': 'flood',
'title': ('%d hour flood forecast regions '
'in Jakarta at %s' % (hours,
date))})
print 'polyforecast_filepath', polyforecast_filepath
v.write_to_file(polyforecast_filepath)
print 'Wrote tagged polygons to %s' % polyforecast_filepath
# zip all file
if os.path.isfile(zip_filename):
print 'Has been zipped to %s' % zip_filename
else:
zip_shp(polyforecast_filepath, extra_ext=['.keywords'],
remove_file=True)
print 'Zipped to %s' % zip_filename