GitHub

Reading and Writing Radar Data

Daisho.jl works with radar data in CfRadial (CF/Radial) NetCDF format, a standard for weather radar data.

Reading CfRadial Files

The primary function for reading radar data is Daisho.read_cfradial:

# Create a moment dictionary mapping variable names to column indices
moment_dict = Dict("DBZ" => 1, "VEL" => 2, "WIDTH" => 3)

# Read the CfRadial file
volume = Daisho.read_cfradial("path/to/cfradial.nc", moment_dict)

The returned Daisho.radar struct contains all the data from the file, including:

  • Azimuth, elevation, and range arrays
  • Platform motion vectors (for moving platforms)
  • Nyquist velocity
  • Geographic coordinates (latitude, longitude, altitude)
  • Sweep information
  • All requested radar moments in a 2D array

The radar Struct

The Daisho.radar struct is the core data container. Moments are stored as a 2D array where rows correspond to range gates across all rays (flattened) and columns correspond to different radar variables.

Moment Dictionaries

Daisho uses dictionaries to map moment names to array column indices:

raw_names = ["DBZ", "ZDR", "KDP", "RHOHV", "VEL", "WIDTH", "PHIDP", "SQI", "SNR"]
qc_names = ["DBZ", "ZDR", "KDP", "RHOHV", "VEL", "WIDTH", "PHIDP", "SQI"]
grid_types = [:linear, :linear, :weighted, :weighted, :weighted, :weighted, :weighted, :weighted]

raw_dict, qc_dict, grid_dict = Daisho.initialize_moment_dictionaries(raw_names, qc_names, grid_types)

Grid types determine how moments are interpolated during gridding:

  • :linear - Convert to linear units before averaging (e.g., reflectivity)
  • :weighted - Weighted average in native units (e.g., velocity)
  • :nearest - Nearest-neighbor interpolation

Writing QC'd CfRadial Files

After quality control, write the QC'd data back to CfRadial format with Daisho.write_cfradial. The CfRadial 2.1 refactor unified the per-radar variants (SIGMET, single-/dual-pol, P-3) into this single Volume-based writer.

Splitting Sweeps

To work with individual sweeps from a volume scan:

sweeps = Daisho.split_sweeps(volume)
for sweep in sweeps
    # Process each sweep independently
end

Platform Orientation

For moving platforms (ships, aircraft), retrieve heading, pitch, and roll:

orientation = Daisho.get_radar_orientation("path/to/cfradial.nc")
# Returns [heading pitch roll] matrix with one row per ray