Source code for pyart.aux_io.kazr_spectra

"""
Utilities for reading ARM KAZR spectra files.

"""

import datetime
import re

import numpy as np
try:
    import xarray as xr
    _XARRAY_AVAILABLE = True
except ImportError:
    _XARRAY_AVAILABLE = False

from ..config import FileMetadata
from ..core.radar_spectra import RadarSpectra
from ..exceptions import MissingOptionalDependency
from ..io.common import _test_arguments


KAZR_SPECTRA_FIELD_NAMES = {'spectra': 'spectra'}

[docs]def read_kazr_spectra(filename, field_names=None, additional_metadata=None, file_field_names=False, exclude_fields=None, include_fields=None, **kwargs): """ Read a ARM KAZR spectra netCDF file. Parameters ---------- filename : str Name of KAZR spectra netCDF file to read data from. field_names : dict, optional Dictionary mapping field names in the file names to radar_spectra field names. Unlike other read functions, fields not in this dictionary or having a value of None are still included in the radar_spectra.fields dictionary, to exclude them use the `exclude_fields` parameter. Fields which are mapped by this dictionary will be renamed from key to value. additional_metadata : dict of dicts, optional This parameter is not used, it is included for uniformity. file_field_names : bool, optional True to force the use of the field names from the file in which case the `field_names` parameter is ignored. False will use to `field_names` parameter to rename fields. exclude_fields : list or None, optional List of fields to exclude from the radar spectra object. This is applied after the `file_field_names` and `field_names` parameters. Set to None to include all fields specified by include_fields. include_fields : list or None, optional List of fields to include from the radar spectra object. This is applied after the `file_field_names` and `field_names` parameters. Set to None to include all fields not specified by exclude_fields. delay_field_loading : bool True to delay loading of field data from the file until the 'data' key in a particular field dictionary is accessed. In this case the field attribute of the returned RadarSpectra object will contain LazyLoadDict objects not dict objects. Delayed field loading will not provide any speedup in file where the number of gates vary between rays (ngates_vary=True) and is not recommended. Returns ------- radar_spectra : RadarSpectra RadarSpectra object. """ if not _XARRAY_AVAILABLE: raise MissingOptionalDependency( "Xarray is required to read KAZR spectra files " "but is not installed!") # test for non empty kwargs _test_arguments(kwargs) # create metadata retrieval object if field_names is None: field_names = KAZR_SPECTRA_FIELD_NAMES filemetadata = FileMetadata('kazr_spectra', field_names, additional_metadata, file_field_names, exclude_fields) # read the data xrobj = xr.open_dataset(filename) # retrieve the correct time format # time format changed in later KAZR files try: times = [(datetime.timedelta(seconds=sec) + datetime.datetime.utcfromtimestamp( xrobj.base_time.values.astype("O")/1e9)) for sec in xrobj.time_offset.values] except TypeError: ts = [(i - np.datetime64('1970-01-01T00:00:00Z')) / np.timedelta64(1, 's') for i in xrobj.time_offset.values] bs = ( xrobj.base_time.values - np.datetime64( '1970-01-01T00:00:00Z')) / np.timedelta64(1, 's') times = [(datetime.timedelta(seconds=sec) + datetime.datetime.utcfromtimestamp(bs.astype("O")/1e9)) for sec in ts] times = np.array(times) time_dict = filemetadata('time') base_time = ('seconds since ' + datetime.datetime.strftime( times[0], '%Y-%m-%DT%H:%M:%SZ')) time_array = [(x-times[0]).seconds for x in times] time_dict['units'] = base_time time = xr.DataArray(np.array(time_array), attrs=time_dict, dims='time') rng_array = np.squeeze(xrobj.range.values) rng_dict = filemetadata('range') _range = xr.DataArray(rng_array, attrs=rng_dict, dims='range') #metadata = filemetadata('metadata') # Retrieve the ARM site info if 'nsa' in filename: xrobj.attrs['site_id'] = 'nsa' elif 'ena' in filename: xrobj.attrs['site_id'] = 'ena' else: xrobj.attrs['site_id'] = 'sgp' lat_dict = filemetadata('latitude') lon_dict = filemetadata('longitude') alt_dict = filemetadata('altitude') lat, lon, alt = get_kazr_location(xrobj.attrs['site_id']) latitude = xr.DataArray(np.array(lat, dtype='float32'), attrs=lat_dict) longitude = xr.DataArray(np.array(lon, dtype='float32'), attrs=lon_dict) altitude = xr.DataArray(np.array(alt, dtype='float32'), attrs=alt_dict) elevation_dict = filemetadata('elevation') elevation = xr.DataArray(np.array(90.0*np.ones(len(time_array)), dtype='float32'), attrs=elevation_dict, dims='time') azimuth_dict = filemetadata('azimuth') azimuth = xr.DataArray(np.array(360.0*np.ones(len(time_array)), dtype='float32'), attrs=azimuth_dict, dims='time') fix_agl_dict = filemetadata('fixed_angle') fixed_angle = xr.DataArray(np.array(90.0, dtype='float32'), attrs=fix_agl_dict) sweep_number_dict = filemetadata('sweep_number') sweep_number = xr.DataArray(np.array([1], dtype='int32'), attrs=sweep_number_dict) sweep_mode_dict = filemetadata('sweep_mode') sweep_mode = xr.DataArray(np.array('spectra'), attrs=sweep_mode_dict) sweep_start_ray_index_dict = filemetadata('sweep_start_ray_index') sweep_start_ray_index = xr.DataArray(np.array([0], dtype='int32'), attrs=sweep_start_ray_index_dict) sweep_end_ray_index_dict = filemetadata('sweep_end_ray_index') sweep_end_ray_index = xr.DataArray(np.array([len(times)-1], dtype='int32'), attrs=sweep_end_ray_index_dict) spectra_array = [_get_spectra(xrobj, x) for x in xrobj.time.values] spectras = np.stack(spectra_array) c = 299792458 npulses_max = xrobj.speclength.values wavelength = c/float(re.findall(r"[-+]?\d*\.\d+|\d+", xrobj.radar_operating_frequency)[0])*1e-9 scan_type = 'vpt' metadata = xrobj.attrs metadata['wavelength'] = wavelength velocity_dict = xrobj.velocity_bins.attrs bins = xrobj.velocity_bins.values velocity_bins = xr.DataArray(bins, attrs=velocity_dict, dims='npulses_max') xrobj.close() return RadarSpectra(time, _range, spectras, metadata, scan_type, latitude, longitude, altitude, sweep_number, sweep_mode, fixed_angle, sweep_start_ray_index, sweep_end_ray_index, azimuth, elevation, npulses_max, velocity_bins)
def _get_spectra(xrobj, time): """ Retrieves the spectra values using time and locator mask. """ the_spectra_locs = xrobj.locator_mask.sel(time=time).values the_spectra_loc = np.zeros((len(the_spectra_locs), len(xrobj.speclength.values))) the_spectra_locs[np.isnan(the_spectra_locs)] = -9999.0 for i, locs in enumerate(the_spectra_locs): if locs != -9999.0: the_spectra_loc[i, :] = xrobj.spectra.values[int(locs), :] else: the_spectra_loc[i, :] = np.nan*np.ones(len( xrobj.speclength.values)) return the_spectra_loc def get_kazr_location(arm_site): """ Return the latitude, longitude and altitude of a ARM KAZR Radar. Parameters ---------- arm_site : str Three letter ARM site acroynm. Returns ------- lat, lon, alt : float Latitude (in degrees), longitude (in degrees), and altitude (in meters above mean sea level) of the KAZR radar. """ loc = ARM_SITES[arm_site.upper()] return loc['lat'], loc['lon'], loc['alt'] ARM_SITES = { "SGP": {'lat': 36.605, 'lon': -97.485, 'alt': 318}, "ENA": {'lat': 39.0916, 'lon': -28.0257, 'alt': 30}, "NSA": {'lat': 71.323, 'lon': -156.609, 'alt': 8}, }