OrbScore2_README_20181127.txt
by Peter Bird, UCLA

In November 2018, I completed the project which I had started 1~2 years earlier, 
   by updating all parts of the source code for OrbScore2.

This comprehensive edit included:
*Change from Fortran77 to Fortran90 syntax (e.g., free-form, with longer lines, and less use of line numbers).
*Modern capitalization (ALLCAPS for Fortran reserved words; Initial_cap for subprograms; camelCaps for variables).
*REAL ---> REAL*8 throughout, including values of inline constants (e.g., "2" --> "2.0D0" in REAL*8 calculations).
*Replaced IMSL routine with MKL(LAPACK) routine for solving the 3x3 symmetric REAL*8 system in subprogram Adjust.
*Removed "input", "modify", "output" in CALLs and SUBROUTINEs, and replaced them with INTENT(IN/INOUT/OUT).
*IMPLICIT NONE throughout, declaring all variable names.

All the changes above are intended to produce clearer code, more trustworthy executables,
and slightly better numerical precision.  However, they should not substantially change any output scores.

In addition, I made 2 substantive changes that affect ONLY the geodetic-velocity scoring section:

(1) I detected that SUBROUTINE Adjust (which is supposed to minimize RMS geodetic velocity misfit
by adjusting the velocity reference frame for the model and/or scoring dataset) was not always
succeeding.  I re-derived its equations and recoded it completely.  Now it does not reduce
the global RMS geodetic velocity misfit very much, but at least it does not increase it!

(2) I noticed that the brittle/ductile transition depths computed by SUBROUTINE Mohr
(and used in forming the rectangular dislocation patches for computing long-term-average
coseismic velocities at geodetic benchmarks) were not always reasonable.  For subduction
zone fault elements ONLY, I overrode these estimates with a fixed depth-range of 26 km,
based on Oleskevitch et al. [1999, J.Geophys.Res.], as cited in Bird & Kagan [2004, BSSA].
Taking the preferred model (Earth5-049) of Bird et al. [2008, J. Geophys. Res.] 
as an example, the RMS geodetic-velocity misfit is now reduced to 11.47 mm/a.

Note that neither of these improvements imply or require any changes to -Shells-.