GitHub

Daisho.jl

Daisho.jl is a Julia package for radar meteorology data processing, quality control, and gridding. It provides tools for working with weather radar data in CfRadial format, performing quality control operations, and gridding radar observations onto regular Cartesian or latitude-longitude grids.

Features

  • CfRadial I/O: Read and write CfRadial NetCDF radar data files
  • Quality Control: Threshold-based QC, despeckling, platform motion removal, terrain masking
  • Gridding: Beam-aware interpolation onto 3D Cartesian, lat/lon, RHI, PPI, composite, and column grids
  • Moving Platform Support: Full support for airborne and ship-based radars with platform motion correction
  • SRTM Integration: Digital elevation model integration for terrain-aware quality control
  • Coordinate Transforms: Transverse Mercator projections and approximate inverse projections

Installation

using Pkg
Pkg.add("Daisho")

For development work, clone the repository and pkg> dev /path/to/Daisho.jl to track local changes.

Quick Start

Daisho is driven by a single TOML parameter file loaded into a DaishoParameters struct. The struct bundles the canonical field list, QC thresholds, gridding knobs, grid geometry (Cartesian / lat-lon / RHI / spectral), CF-1.12 output metadata, and I/O fill values. To start a configuration, write the bundled template to a path you control and edit it for your deployment:

using Daisho

print_config("mygrid.toml")            # write the template
# edit mygrid.toml for your radar, grid, and CF metadata
p = DaishoParameters("mygrid.toml")    # strict load

The [fields] block in the TOML replaces the old initialize_moment_dictionaries helper. Each field maps to a flat array of tags drawn from a documented vocabulary. No tag is required; field order and tag order are insignificant.

[fields]
DBZ   = ["linear_interp",   "define_detection"]
ZDR   = ["linear_interp"]
KDP   = ["weighted_interp"]
RHOHV = ["weighted_interp"]
VEL   = ["weighted_interp",  "velocity"]
WIDTH = ["weighted_interp"]
PHIDP = ["weighted_interp"]
SQI   = ["weighted_interp",  "define_scanned"]

Interpolation tags (at most one per field; weighted_interp is the default when none is given): linear_interp (convert to linear units before averaging, e.g. reflectivity), weighted_interp (weighted average in native units), nearest_interp (winner-take-all). Role tags name the field whose presence proves a detectable echo (define_detection) or that the gate was scanned (define_scanned); velocity is reserved for forthcoming multi-Doppler synthesis.

[fields] and [io] are mandatory; [qc], [gridding], and the [grid.*] sub-tables are optional and default-construct when omitted (an operation that needs a missing block raises a clear point-of-use error). Any block that is present is validated strictly — a missing, unknown, or mis-typed key raises an ArgumentError at load time naming the offending section. Configs in the pre-tag schema (the old names/[fields.grid_type], [gridding] missing_key/valid_key, or [io] fill_value_missing/fill_value_clear) raise a targeted migration error pointing at the replacement.

With p in hand, the read → QC → grid pipeline is:

# Read a CfRadial file (auto-detects CfRadial 1.4 vs 2.1)
volume = read_cfradial("radar_file.nc")

# Apply per-sweep threshold QC using p.qc thresholds.
# Each input field FOO gains a QC'd copy FOO_QC in sweep.fields.
for sweep in volume.sweeps
    threshold_qc!(sweep, p)
end

# Grid the data using p.grid.cartesian, p.gridding, p.moments, p.grid.metadata
grid_radar_volume(volume, "output.nc", volume.time_coverage_start, p)

See the Gridding guide for the lat-lon, RHI, PPI, composite, column, and spectral variants, and the multi-sweep accumulator workflow for building rolling grids across flight legs.

Package Structure