ELM coupler bypass - unhandled condition

[rfiorella]

@jonschwenk and I have been troubleshooting an unhandled condition in the coupler bypass logic that results in only shortwave radiation and precipitation advancing through a time series if the resolution of the forcing dataset is exactly half the ELM timestep.

Looking at lnd_import_export.F90 L. 503-519, and assuming hourly forcing data (e.g., ERA5) and a half-hour ELM time step, atm2lnd_vars%npf(v)) evaluates to 2.0. For variables that are not v = 4 or 5 (8 to 13 only come from EAM output, L. 237-238), this puts us in the L. 510-514 block, where there is a conditional that tests: mod(tod/get_step_size()-1,nint(atm2lnd_vars%npf(v))) <= atm2lnd_vars%npf(v)/2._r8 .and. mod(tod/get_step_size(),nint(atm2lnd_vars%npf(v))) > atm2lnd_vars%npf(v)/2._r8. mod(tod/get_step_size()-1,nint(atm2lnd_vars%npf(v))) can only evaluate to 1 or 0 if npf = 2, and since the right hand side evaluates to 1, there is no way for this condition to be met.

If there remains a need to treat v = 4, 5, 8-13 differently than v = 1-3, 6, 7, one potential fix would be:

          do v=1,met_nvars
            if (atm2lnd_vars%npf(v) - 1._r8 .gt. 1e-3) then
              if (v .eq. 4 .or. v .eq. 5 .or. (v .ge. 8 .and. v .le. 13)) then    !rad/Precipitation
                if (mod(tod/get_step_size(),nint(atm2lnd_vars%npf(v))) == 1 .and. nstep .gt. 3) then
                  atm2lnd_vars%tindex(g,v,1) = atm2lnd_vars%tindex(g,v,1)+1
                  atm2lnd_vars%tindex(g,v,2) = atm2lnd_vars%tindex(g,v,2)+1
                end if
              else
                if (atm2lnd_vars%npf(v) .ne. 2._r8) then
                  if (mod(tod/get_step_size()-1,nint(atm2lnd_vars%npf(v))) <= atm2lnd_vars%npf(v)/2._r8 .and. &
                      mod(tod/get_step_size(),nint(atm2lnd_vars%npf(v))) > atm2lnd_vars%npf(v)/2._r8) then
                    atm2lnd_vars%tindex(g,v,1) = atm2lnd_vars%tindex(g,v,1)+1
                    atm2lnd_vars%tindex(g,v,2) = atm2lnd_vars%tindex(g,v,2)+1
                  end if
                else
                  if (mod(tod/get_step_size(),nint(atm2lnd_vars%npf(v))) == 1 .and. nstep .gt. 3) then
                     atm2lnd_vars%tindex(g,v,1) = atm2lnd_vars%tindex(g,v,1)+1
                     atm2lnd_vars%tindex(g,v,2) = atm2lnd_vars%tindex(g,v,2)+1
                  end if
                end if
              end if
            else
              atm2lnd_vars%tindex(g,v,1) = atm2lnd_vars%tindex(g,v,1)+nint(1/atm2lnd_vars%npf(v))
              atm2lnd_vars%tindex(g,v,2) = atm2lnd_vars%tindex(g,v,2)+nint(1/atm2lnd_vars%npf(v))
            end if

I could add this to #7480 if desired, or let me know if a different approach would be preferred here.