Dear Users,

This is a fix to make the model more stable when running the
multi-layer soil model (ISOIL=1).  It is a repair to a fix that was
introduced earlier for the same problem.  Modify the following line of
physics/pbl_sfc/util/slab.F:


          NSOILI=1+IFIX(4*DIFSL/DZS(1)*DELTSM/DZS(1))                            25JAN00.682

so that it reads:

          NSOILI=1+NINT(4*DIFSL/DZS(1)*DELTSM/DZS(1))                            25JAN00.682

instead. The change involves using NINT (round to nearest integer)
instead of IFIX (truncate). This has the effect of making the soil time
step more conservative. Instability in the soil model can manifest
itself as floating point exceptions or NaNs in explicit moisture or the
cumulus scheme.  If MPHYSTBL is set to 1 in the configure.user file,
the problem may manifest itself as a segmentation fault in the lexmoisr
routine (if IMPHYS=5).  The risk is particularly acute on coarse
domains with large time steps.

This fix also applies to non-MPP builds of the model.

Rotang,

May 15, 2000