Echo Products

Echo products are quantities derived from gridded polarimetric fields: fuzzy-logic hydrometeor identification (HID / FHC) and polarimetric rain-rate estimates. Daisho currently computes them on the gridded radar variables — the beam-power-weighted averages produced by the gridding step — rather than on raw gates. Gridding the smooth polarimetric moments first and classifying the result avoids the ill-posed problem of regridding integer hydrometeor categories. The methods are ported from the CSU Radar Tools Python package.

All echo products are configured through the [echo] block and surface as EchoProductsParameters on the loaded DaishoParameters. They are written after gridding, so they are not part of [fields]; the gridded readers know to surface them via echo_output_names.

Three ways to run it

1. Inline, as part of gridding (recommended). Add an [echo] block with enabled = true and grid as usual — the products are appended to the output NetCDF automatically for every geometry (volume, lat/lon, PPI, RHI) and for the accumulator path:

p = DaishoParameters("mygrid.toml")          # [echo] enabled = true
volume = read_cfradial("radar_file.nc")
grid_radar_volume(volume, "out.nc", volume.time_coverage_start, p)
# out.nc now also contains HID_CSU and RATE_CSU_BLENDED

2. Standalone, reprocessing existing grids. Append products in place to one or more already-written Daisho grids (single- or multi-time; volume / PPI / RHI) with add_echo_products! — no regridding required:

p = DaishoParameters("mygrid.toml")
written = add_echo_products!("leg_gridded.nc", p)   # modifies the file in place

A ready-to-use script for this path ships at docs/examples/add_echo_products_demo.jl:

julia --project=. docs/examples/add_echo_products_demo.jl config.toml grid.nc [more.nc ...]

3. In-memory, on a field dictionary. For custom pipelines, apply_echo_products is pure: it takes a Dict of gridded fields and returns a Dict of new product fields, leaving I/O to the caller.

Configuration

The full annotated [echo] block lives in the bundled template (write it with print_config("mygrid.toml")). The essentials:

[echo]
enabled              = true         # master switch (off by default)
band                 = "S"          # radar band: "S", "C", or "X"
compute_fhc          = true         # write hydrometeor classification
compute_blended_rain = true         # write the blended rain rate
# Individual rain components to also write (any subset):
#   RATE_Z, RATE_Z_CONV, RATE_Z_STRAT, RATE_KDP, RATE_Z_ZDR, RATE_KDP_ZDR
rain_components      = []

# Input field names read from the grid (ldr_field = "" disables LDR):
dbz_field            = "DBZ"
zdr_field            = "ZDR"
kdp_field            = "KDP"
rhohv_field          = "RHOHV"
ldr_field            = ""

# FHC options:
fhc_method           = "hybrid"     # "hybrid" or "linear"
use_temp             = true         # include temperature in the FHC

# Output variable names:
fhc_output           = "HID_CSU"
rain_output          = "RATE_CSU_BLENDED"
rain_method_output   = ""           # "" suppresses the per-cell method field

The input names default to DBZ/ZDR/KDP/RHOHV — set them to match the fields actually present in your grid (e.g. the QC'd *_QC copies).

Temperature

The fuzzy classifier uses temperature to separate ice and liquid categories. Three sources are selectable with temp_source:

temp_source          = "profile"    # "profile", "field", or "reference_state"
temp_field           = "TEMP_FOR_PID"  # gridded field, used when temp_source = "field"
temp_field_units     = "C"          # "C" or "K" (K is converted to °C)
height_field         = ""           # see below
temp_factor          = 1.0          # > 1 broadens the temperature memberships

# Piecewise-linear T(z): height (m) -> temperature (°C), linearly interpolated.
[echo.temperature]
heights      = [0.0,  1000.0, 3000.0, 5000.0, 10000.0, 15000.0]
temperatures = [25.0, 18.0,    5.0,   -8.0,   -40.0,   -70.0]

"profile" samples the [echo.temperature] T(z) profile at each cell's height. On a 3-D grid the grid z-axis is used directly. On a 2-D PPI/RHI grid (no Z axis), set height_field to a gridded beam-height field (e.g. "HEIGHT") so the profile can be sampled per cell. See TemperatureProfile and read_temperature_profile.
"field" uses a gridded temperature field directly (per cell, may be 3-D), converting K → °C when temp_field_units = "K".
"reference_state" is reserved for a future Springsteel reference state (not yet implemented).

Set use_temp = false to classify without temperature.

Outputs and the sentinel policy

The variables written are exactly those reported by echo_output_names: the HID field (fhc_output, default HID_CSU), the blended rain field (rain_output, default RATE_CSU_BLENDED), the optional method field, and any requested rain_components. Re-running overwrites these rather than duplicating them.

The integer HID labels index into FHC_SUMMER_CLASSES (class i == FHC_SUMMER_CLASSES[i]); a label of 0 marks an unclassified cell.

Echo products preserve Daisho's true-missing vs clear-air distinction (see the [io] sentinels):

Input reflectivity cell	HID	Rain
true-missing (`io.fill_value`)	`fill_value`	`fill_value`
undetect / clear air (`io.undetect`)	`undetect`	`undetect` (not `0`)
valid, but a required input invalid	`fill_value`	`fill_value`

Writing the undetect sentinel through to the rain fields (rather than 0) keeps downstream masking able to tell "scanned, no echo" from "rain rate of zero".

Echo Products

Three ways to run it

Configuration

Temperature

Outputs and the sentinel policy

See also