def geo_filter(input_sr_data):
'''
Given a list of SnowRadar datafiles (.mat, .h5, .nc), filter out
any files whose bounding geometry intersect with land
Landmask is based on NaturalEarth 1:10m Cultural v4.1.0 (Canada, Greenland, and USA)
http://www.naturalearthdata.com/
Arguments:
input_sr_data: list of supported SnowRadar data files
Output:
subset of input_sr_data where no land intersections occur
'''
# Drop all data that intersects with land features
land = gpd.read_file('/vsizip/' + str(
Path(__file__).parent / 'data' / 'natearth' /
'ne_10m_admin_0_countries_northamerica.zip'))
# Load the datafiles in 'meta' mode to just scrape the simplified track line
sr_meta = [SnowRadar(sr, 'meta') for sr in input_sr_data]
sr_gdf = gpd.GeoDataFrame(data={'file': [sr.file_path for sr in sr_meta]},
geometry=[sr.line for sr in sr_meta],
crs={'init': 'epsg:4326'})
sr_gdf = sr_gdf.drop(
gpd.sjoin(sr_gdf, land, how='inner', op='intersects').index)
if len(sr_gdf) < 1:
LOGGER.warning('No suitable datafiles left after geospatial filtering')
return []
return sr_gdf.file.tolist()
import numpy as np
from pathlib import Path
from pySnowRadar import SnowRadar
from pySnowRadar.algorithms import Wavelet_TN, GSFC_NK, NSIDC, Peakiness
# adapted from Data_20160419_04_010.mat
sr = SnowRadar(Path('__file__').parent.parent / 'pySnowRadar' / 'data' / 'sr' /
'Data_20160419_04_010.mat',
l_case='full')
sr.surf_bin, sr.surface = sr.get_surface()
bnds = sr.get_bounds(5)
# only use the middle trace
middle = np.round(sr.data_radar.shape[1] / 2).astype(int)
data = sr.data_radar[bnds[1]:bnds[0], middle]
n2n = 0.20186025505333335
dfr = 0.012975781596299215
# density set to 0.3 kg/m^3
n_snow = np.sqrt((1 + 0.51 * 0.300)**3)
def test_wavelet_tn():
results = Wavelet_TN(data, n2n, dfr, n_snow, 1, 10)
assert len(results) == 2
airsnow, snowice = results
assert type(airsnow) == np.int64
assert type(snowice) == np.int64
assert snowice > airsnow
def test_gsfc_nk():
results = GSFC_NK(data)
def extract_layers(data_path,
picker=algorithms.Wavelet_TN,
params=None,
dump_results=False):
'''
For a given SnowRadar datafile, estimate the air-snow and snow-ice interfaces
using the supplied picker and snow density
Arguments:
data_path: file path to input SnowRadar data file
picker: Which picker algorithm to apply (default is algorithms.Wavelet_TN)
params: A dictorary of expected parameters to pass to the picker
dump_results: whether or not to save the dataframe to a local csv file under ./dump/
Output:
A pandas dataframe with the following columns:
'src': the name of the source SnowRadar data file
'picker': the name of the picker algo
'lat': latitude of trace
'lon': longitude of trace
'n_snow': the refractive index used during layer picking
'b_ref': the reference bin considered as 0 in the origional file
'b_as': the picked air-snow interface layer
'b_si': the picked snow-ice interface layer
'snow_depth': estimated snow depth based on picked layers
'params': a dict of all input and generated params
'''
if dump_results:
outpath = Path('./dump')
outname = Path(data_path).stem + '.csv'
outfile = outpath / outname
if outfile.exists():
LOGGER.warning('File exists for %s. Skipping processing....',
Path(data_path).name)
result = pd.read_csv(str(outfile), index_col=0)
return result
# Check that the picker passed exists
if not (picker in algorithms.available_pickers()):
raise ValueError('Invalid picker name:' % picker.__name__)
# TODO: Modify to allow refractive index (n_snow) as an alternative
if 'snow_density' not in params:
raise ValueError(
'Snow density or refractive index input required for all pickers')
elif (not (0.1 <= params['snow_density'] <= 0.4)):
raise ValueError(
'Invalid snow density passed: %.3f (Must be between 0.1 and 0.4)' %
params['snow_density'])
# Load radar data
radar_dat = SnowRadar(data_path, 'full')
radar_dat.surf_bin, radar_dat.surface = radar_dat.get_surface()
radar_dat.calcpulsewidth()
# Subset radar traces to reduce computational load
# TODO: Should we allow the subset bounds to be user defined?
lower, upper = radar_dat.get_bounds(m_above=5)
radar_sub = radar_dat.data_radar[upper:lower, :]
# Calc or init other necessary params
params['n_snow'] = np.sqrt((1 + 0.51 * params['snow_density'])**3)
params['null_2_space'] = radar_dat.n2n
params['delta_fast_time_range'] = radar_dat.dfr
# Apply the picker to the file, trace by trace
try:
airsnow, snowice = np.apply_along_axis(picker,
axis=0,
arr=radar_sub,
**params)
except:
# We catch and log the exception
errtype, errval, _ = sys.exc_info()
LOGGER.error('%s with picklayers on file %s: %s' %
(errtype, radar_dat.file_name, errval))
# Set interfaces to NaN if anything goes wrong
airsnow = np.array([np.nan] * radar_dat.lat.shape[0])
snowice = np.array([np.nan] * radar_dat.lat.shape[0])
# Calc snow depth and remove back picks (ie negative snow depth)
snow_depth = (snowice - airsnow) * radar_dat.dfr / params['n_snow']
# trick to get around the invalid-value runtime warnings
# props to Jaime: https://stackoverflow.com/a/25346972
mask = ~np.isnan(snow_depth)
mask[mask] &= snow_depth[mask] < 0
snow_depth[mask] = np.nan
# Add SnowRadar source file
data_src = np.array([radar_dat.file_name] * radar_dat.lat.shape[0])
result = pd.DataFrame({
'src': data_src,
'picker': picker.__name__,
'lat': radar_dat.lat,
'lon': radar_dat.lon,
'n_snow': params['n_snow'],
'b_as': upper + airsnow,
'b_si': upper + snowice,
'snow_depth': snow_depth,
})
if dump_results:
outpath.mkdir(parents=True, exist_ok=True)
result.to_csv(str(outfile), na_rep='nan')
return result
def test_bad_file():
with pytest.raises(IOError):
bad_file = SnowRadar(TEST_NOT_A_FILE, 'meta')
def test_invalid_load_case():
with pytest.raises(ValueError):
wrong_l_case = SnowRadar(str(OIB_TEST_FILE), l_case='not_a_real_l_CASE')
def nsidc_meta():
'''Meta load class for NSIDC matfile IRSNO1B_20160419_04_006_deconv.nc'''
return SnowRadar(str(NSIDC_TEST_FILE), l_case='meta')
def test_file_missing():
with pytest.raises(FileNotFoundError):
fake_file = SnowRadar('this-file-definitely-does-not-exist-on-anyones-computer-probably.jpeg.gif.tif.ogg.mp4', l_case='full')
def nsidc_full():
'''Full load class for NSIDC matfile IRSNO1B_20160419_04_006_deconv.nc'''
return SnowRadar(str(NSIDC_TEST_FILE), l_case='full')
def awi_meta():
'''Meta load class for AWI matfile Data_20170410_01_006.mat'''
return SnowRadar(str(AWI_TEST_FILE), l_case='meta')
def awi_full():
'''Full load class for AWI matfile Data_20170410_01_006.mat'''
return SnowRadar(str(AWI_TEST_FILE), l_case='full')
def oib_meta():
'''Meta load class for OIB matfile Data_20160419_04_010.mat'''
return SnowRadar(str(OIB_TEST_FILE), l_case='meta')
def oib_full():
'''Full load class for OIB matfile Data_20160419_04_010.mat'''
return SnowRadar(str(OIB_TEST_FILE), l_case='full')