Issue
Currently, the NT_Cor flag toggles NT corrections at the quadrature nodes (Section 3.7.2 in the Comprehensive Documentation) and subroutines.interpolate separately interpolates those nodes to output the intensity/radiance at arbitrary viewing angles. In operations, it may be more beneficial to apply NT corrections at the interpolated viewing angles than at the quadrature nodes. This would be more accurate, and if there are fewer viewing angles than quadrature nodes then this would also save compute.
Potential Solution
Add a subroutine to perform NT corrections at arbitrary angles. The workflow would then be pydisort (with NT_cor=False) -> subroutines.interpolate -> NT_corrections. Note that if there are many more viewing angles than quadrature nodes, then it may be cheaper (and perhaps accurate enough) to do the current workflow of applying NT corrections at the quadrature nodes then interpolating; applying NT corrections at every viewing angle is likely why Stamnes' DISORT takes so long to compute the solution at many points (Section 6.5).
Issue
Currently, the
NT_Corflag toggles NT corrections at the quadrature nodes (Section 3.7.2 in the Comprehensive Documentation) andsubroutines.interpolateseparately interpolates those nodes to output the intensity/radiance at arbitrary viewing angles. In operations, it may be more beneficial to apply NT corrections at the interpolated viewing angles than at the quadrature nodes. This would be more accurate, and if there are fewer viewing angles than quadrature nodes then this would also save compute.Potential Solution
Add a subroutine to perform NT corrections at arbitrary angles. The workflow would then be
pydisort (with NT_cor=False) -> subroutines.interpolate -> NT_corrections. Note that if there are many more viewing angles than quadrature nodes, then it may be cheaper (and perhaps accurate enough) to do the current workflow of applying NT corrections at the quadrature nodes then interpolating; applying NT corrections at every viewing angle is likely why Stamnes' DISORT takes so long to compute the solution at many points (Section 6.5).