API Reference¶
Everything listed in radtrax.__all__ is importable from the top level. The
capability index lists every public name on one page, with
a one-line summary each.
Modules¶
| Module | Contents |
|---|---|
| Forward model | ForwardModel, unit_absorption_target |
| Atmosphere | AtmosphericState, ViewingGeometry, layers_from_levels, afgl_us_standard, ideal-gas constants, number_density, air_mass_factor |
| Spectroscopy | OpacityProvider, TableOpacity, v0's table cache (load_opacity), the solar spectrum (solar_irradiance), tables (single-slab σ interpolation, Beer–Lambert), hitran (table building, [hitran]) |
| Optics | OpticalProperties |
| Radiometry | planck_radiance, log_planck_radiance, the radiation constants, the per-µm and per-GHz views; brightness_temperature, band_planck_radiance, band_brightness_temperature |
| Solvers | RTSolver, BeerLambert |
| Surface | Surface, Lambertian |
| Slab models | single-slab forward models, the nB table, observation config |
| Instrument | Instrument, BandInstrument, SpectralResponseFunction, PointSpreadFunction, GroundSamplingDistance, NoiseModel (DiagonalNoise, SNRNoise, NEdTNoise) |
| Retrieval | the matched filter; 3D-Var; the MBMP signal; band tables; the absorption-shape test (AbsorptionWindow, transmittance_table, pair_pixels, ratio_spectrum, fit_transmittance, d_norm) |
| Geo | sentinel2_srf; sentinel2_instrument, sentinel2_geometry, sentinel2_radiance (the [geo] extra); emit_instrument, emit_geometry; landsat_srf, landsat_instrument, landsat_radiance, landsat_geometry, landsat_noise, LANDSAT_NETD_K, LANDSAT_CALIBRATION_K; ecostress_bands, ecostress_srf, ecostress_instrument, ecostress_noise, ecostress_radiance, ecostress_geometry |
| Exceptions | RadtraxError, MissingTableError |
Package overview¶
Differentiable radiative transfer in JAX and Equinox.
radtrax (formerly rtmx) is the GeoStack's radiative transfer package: a
forward model y = F(x) from an atmospheric state to what a sensor
measures, written so that every Jacobian a retrieval needs comes from
automatic differentiation. Its first releases port plumax's band-integrated
Beer–Lambert models, HITRAN look-up tables, matched filter and 3D-Var
retrieval, then rebuild them as one layered forward model.
The design and roadmap live in the GeoStack docs (radtrax · 3).