Why are dtran_a and dtran_b used for both solar and satellite transmittance (tsol, tsen) in l2gen/aerosol.c?

[fengqiaoed]

In l2gen/aerosol.c, the calculation of solar and satellite transmittance (tsol, tsen) uses the same LUT coefficients (dtran_a and dtran_b) for both angles:

```c
read_lut_variable(file, nc_id, "dtran_a", aertab->model[im]->dtran_a[0]);
read_lut_variable(file, nc_id, "dtran_b", aertab->model[im]->dtran_b[0]);
```

These coefficients are applied in model_transmittance() for both solar and satellite angles:

```c
model_transmittance(modmin, wave, nwave, geom->solz, gmult, tauamin, tsolmin, dtmin); // Solar
model_transmittance(modmax, wave, nwave, geom->senz, gmult, tauamax, tsenmax, dtmax); // Satellite
```

However, the solar transmittance (tsol) should logically use dtran_a0 and dtran_b0 (sun-to-ground coefficients), while satellite transmittance (tsen) should use dtran_a and dtran_b (ground-to-sensor coefficients). When I modified the code to use dtran_a0/dtran_b0 for tsol, the results differed by ~0.6% compared to the current implementation.

Questions:

1. What is the rationale behind using dtran_a/dtran_b for both solar and satellite transmittance?
2. Is this a deliberate simplification, or is there an underlying assumption (e.g., symmetry in diffuse transmittance)?
Any insights would be appreciated!

[bryanfranz]

The choice to use the same coefficients for both paths (Sun to surface, and surface to sensor) is intensional. Reciprocity (symmetry) is assumed for the transmittance. See:

https://www.oceanopticsbook.info/view/atmospheric-correction/atmospheric-transmittances

Some RT studies were done by Z. Ahmad many years ago, and solar-specific path coefficients were put into the tables for testing, but it was ultimately decided that the original Gordon and Yang assumption of reciprocity was more accurate. Unfortunately, the analysis my Ahmad is unpublished.