Using saved inputs

cesm-hawc save-inputs writes each column as a small NetCDF file. These files can drive the simulator directly, with native sasktran2 and hawcsimulator calls and no cesm_hawc import.

File contents¶

Every file contains the extracted WACCM profile on altitude_m [m]:

Variable	Units	Description
`pressure_pa`	Pa	Pressure
`temperature_k`	K	Temperature
`specific_humidity`	kg/kg	Specific humidity
`vmr_o3`, `vmr_no2`, `vmr_so2`	mol/mol	Gas volume mixing ratios
`vmr_h2o`	mol/mol	Water vapour, blended from `Q` and chemistry `H2O`
`n_air_cm3`	cm⁻³	Air number density
`sulfate_a1_N_cm3`, `sulfate_a3_N_cm3`	cm⁻³	Accumulation / coarse mode number concentration
`sulfate_a1_r_um`, `sulfate_a3_r_um`	μm	Accumulation / coarse mode wet median radius
`sulfate_a1_dry_mass_kg_m3`, `sulfate_a3_dry_mass_kg_m3`	kg m⁻³	Accumulation / coarse mode dry sulfate mass concentration

with attributes latitude, longitude, time_index, sigma_a1, sigma_a3, time (the observation time), and radius_source_a1 and radius_source_a3: "wet_diameter" if the radius came from CESM’s wet diameter, "dry_mass" if it was derived from dry sulfate mass (see Methods). Files from orbit mode also record the satellite position as observer_latitude, observer_longitude [degrees] and observer_altitude [m].

When sasktran2 was installed at save time (and --profiles-only was not used), the file also holds, for each mode aerosol_accum and aerosol_coarse:

Variable	Dims	Description
`{mode}_extinction_per_m`	`wavelength_nm`, `altitude_m`	Truth extinction at the configured wavelengths [m⁻¹]
`{mode}_reference_extinction_per_m`	`altitude_m`	Extinction at the 745 nm reference wavelength [m⁻¹]
`{mode}_median_radius_nm`	`altitude_m`	Median radius, clipped to the Mie database range [nm]

and the attributes needed to rebuild the Mie databases: extinction_reference_wavelength_nm, mode_width_accum, mode_width_coarse, mie_refractive_index, mie_wavelength_grid_nm and mie_median_radius_grid_nm.

The includes_constituents attribute (0 or 1) says which kind of file it is.

Running the simulator from a saved file¶

The Mie databases themselves can’t be stored in a file, so they are rebuilt from the saved parameters; everything else is read directly.

import xarray as xr
import sasktran2 as sk
from hawcsimulator.ali.configurations.ideal_dolp_imager import IdealALISimulator
from hawcsimulator.noise import ALINoiseModel

ds = xr.open_dataset("my_case.cam.h0.2035-02.nc")   # a file written by save-inputs
assert ds.attrs["includes_constituents"], "file was saved with --profiles-only"
alt_m = ds["altitude_m"].values


def mode_constituent(mode: str) -> sk.constituent.ExtinctionScatterer:
    width_key = "mode_width_accum" if mode == "aerosol_accum" else "mode_width_coarse"
    mode_db = sk.database.MieDatabase(
        sk.mie.distribution.LogNormalDistribution().freeze(mode_width=ds.attrs[width_key]),
        sk.mie.refractive.H2SO4(),          # ds.attrs["mie_refractive_index"]
        ds.attrs["mie_wavelength_grid_nm"],
        median_radius=ds.attrs["mie_median_radius_grid_nm"],
    )
    return sk.constituent.ExtinctionScatterer(
        mode_db,
        altitudes_m=alt_m,
        extinction_per_m=ds[f"{mode}_reference_extinction_per_m"].values,
        extinction_wavelength_nm=ds.attrs["extinction_reference_wavelength_nm"],
        median_radius=ds[f"{mode}_median_radius_nm"].values,
    )


constituents = {
    "o3": sk.constituent.VMRAltitudeAbsorber(sk.optical.O3DBM(), altitudes_m=alt_m, vmr=ds["vmr_o3"].values),
    "no2": sk.constituent.VMRAltitudeAbsorber(sk.optical.NO2Vandaele(), altitudes_m=alt_m, vmr=ds["vmr_no2"].values),
    "aerosol_accum": mode_constituent("aerosol_accum"),
    "aerosol_coarse": mode_constituent("aerosol_coarse"),
}

sim_input = {
    "tangent_latitude": ds.attrs["latitude"],
    "tangent_longitude": ds.attrs["longitude"],
    # Fixed geometry. For a file from orbit mode, instead pass
    # observer_latitude, observer_longitude and observer_altitude from
    # ds.attrs, and the solar angles are computed from the time.
    "tangent_solar_zenith_angle": 60.0,
    "tangent_solar_azimuth_angle": 0.0,
    "altitude_grid": alt_m,
    "polarization_states": ["I", "dolp"],
    "sample_wavelengths": ds["wavelength_nm"].values,
    "time": ds.attrs["time"],
    "l1b_cfg": {"noise_model": ALINoiseModel(straylight_fraction=0.0)},
    "constituents": constituents,
}
data = IdealALISimulator().run(["l2", "front_end_radiance", "l1b"], sim_input)

This matches what cesm-hawc run does: the ideal_dolp_imager instrument model with the same noise model. The one difference is that cesm-hawc caps the Mie databases’ single-scattering albedo just below 1 (see Methods), which changes the simulated radiance by less than 0.1%.