Error in GO amplitude calculation

[Ian Culverwell]

Michael Gorbunov (Russian Academy of Sciences, Russia) has reported an error in the calculation of refractive amplitude in ropp_pp/preprocess/ropp_pp_amplitude_go.f90:

     snr_R(i) = SQRT(ABS( (1.0_wp/(dgns*dleo))*                        &
              (1.0_wp - (ps(i)/(rxgns(i)*dgns))*rvgns(i)/vtheta(i)     &
                 - (ps(i)/(rxgns(i)*dgns))*rvgns(i)/vtheta(i))*        &
                 pv(i)/vtheta(i)))

should be replaced by

     snr_R(i) = SQRT(ABS( (1.0_wp/(dgns*dleo))*                        &
              (1.0_wp - (ps(i)/(rxgns(i)*dgns))*rvgns(i)/vtheta(i)     &
                 - (ps(i)/(rxleo(i)*dleo))*rvleo(i)/vtheta(i))*        &
                 pv(i)/vtheta(i)))

Kjartan Kinch (DMI) agrees.

Test this in ROPP6.0.

[Ian Culverwell]

ropp_pp_amplitude_go is called by ropp_pp_preprocess, which is called by ropp_pp_occ_tool, so we should see a difference if we pass the before and after versions of the code through tests 2 (COSMIC) & 3 (GRAS) of t_ropp_pp.sh.

We do - but it's very small. |delta BA| < 1e-8 rad and |delta N| < 1e-5 N-units. (See attached for diff plots.) So it seems safe to commit to ropp6.0 devt branch.

[Ian Culverwell]

No problems in beta testing, so close ticket.

[Ian Culverwell]

While merging ropp6.0 devts to trunk, I noticed that in fact I'd replaced

     snr_R(i) = SQRT(ABS( (1.0_wp/(dgns*dleo))*                        &
              (1.0_wp - (ps(i)/(rxgns(i)*dgns))*rvgns(i)/vtheta(i)     &
                 - (ps(i)/(rxgns(i)*dgns))*rvgns(i)/vtheta(i))*        &
                 pv(i)/vtheta(i)))

by

     snr_R(i) = SQRT(ABS( (1.0_wp/(dgns*dleo))*                        &
              (1.0_wp - (ps(i)/(rxgns(i)*dgns))*rvgns(i)/vtheta(i)     &
                 - (ps(i)/(rxleo(i)*dleo))*rxleo(i)/vtheta(i))*        &
                 pv(i)/vtheta(i)))

ie rxleo/vtheta, not rvleo/vtheta. (The former isn't even dimensionally consistent.)

It turns out that this error is a large component of the difference plotted above. When redone properly, we find differences resulting from the bugfix Michael Gorbunov originally wanted, to be:

COSMIC

GRAS

Even smaller differences than before, so safe to put in ROPP6.0 and trunk.

[Ian Culverwell]

amp_go_cosmic.png

[Ian Culverwell]

amp_go_gras.png