Leila at al.,
Thanks. Thus the problem was that the provided WSA fits file is different from the file used by SWPC, i.e. the provided ENLIL log file directed us to using a different file. Now, I used the correct WSA file and, as you also showed, the results match very nicely now.
(1) There is no need do match the temporal profiles during the relaxation phase (values at Earth between -15 to about -3 days). This phase starts with ad hoc values in the heliosphere (at time=-15 days) and there might be slight differences in this assumption (i.e, temperature is scaled (from the boundary values) either by 1/r or by 1/r^(2*(gamma-1)) which would provide different temperature profile if, say, gamma=5/3 but not gamma=1.5. And, correspondingly, there might be different "corrective wave" that propagates from the boundary. But details of this process are not important, one should always reach exactly the same "physical" state.
(2) In the temporal profiles at Earth, there is small systematic shift in time. This can be caused by different ways how the values at Earth are calculated from the values in the heliospheric computational domain. The issue is that the Earth is fixed at 180 deg longitude and moves +/-7.25 degs in latitude. The code takes values at the nearest grid points and uses area-weighting to calculate values at the Earth position. The "problem" is that on 2-deg (4-deg) grid, the nearest longitudes are 179 and 181 (178 and 182), i.e., not 180, and thus there is always need to interpolate values in longitude. The recent ENLIL versions do that because it is the same approach to all other observing points and there is need for longitudinal interpolation in the sidereal frame (Earth moves both in longitude and latitude). However, in some old versions, and especially for runs on 4-deg grids I might avoid longitudinal interpolation by assuming that the Earth position is at 179 (or 178) degs. I suggest to ignore such details and use the new version.
(3) And, just for completeness, see the attached slides. They are as before; results are computed by 2.9 (solid lines) and compared with 2.6 from SWPC (dashed lines). All with correct WSA file (i.e, bnd.nc calculated by 2.9 is the same).
(4) The left and right sides show results for the "azimuthal" and "temporal" shift, respectively. The latter is new v2.9 feature (used if new parameter "nshift=2"; default is "nshift=1" and "azimuthal" shift is used). Note that the azimuthal shift is given by (1) the imposed values of "shift" (propagation delay from 2.5 to 21.5 Rs and by (2) differences in the Julian dates of the central meridian. This can provide more intuitive plots (like "bnd2" in the attached slides) because values corresponds to the beginning of the computation. These values are calculated from the WSA values by interpolation in azimuthal direction (longitude). This if for time=0. This values co-rotates and thus another interpolation is needed at each numerical time step.
(5) The "temporal" shift avoids that first interpolation by using a different time (that correspond to the 27.2753-day azimuthal rotation). This approach probably makes no sense with current WSA maps (they are on 2.5-deg, thus there is always need in the first interpolation). However, it might by used with new WSA maps (on the same 2-deg as ENLIL uses), since it avoids one interpolation (i.e., diffusion/erosion of sharp peaks/transitions).
(6) However, results look very similar. This is good, since the bnd* plots are more intuitive for the "azimuthal" shift.
Dusan