Problem in snow albedo computation

[maignan]

Dan Zhu and Pascale Braconnot have reported an inconsistency in the snow albedo computation, depending on whether you're using DGVM or not.
In condveg_snow, we find:

         ! For  vegetated surfaces
         fraction_veg(:) = un - totfrac_nobio(:)
         snowa_veg(:) = zero
         IF (ok_dgvm) THEN
            DO ji = 1, kjpindex
               IF ( fraction_veg(ji) .GT. min_sechiba ) THEN
                  snowa_veg(ji) = snowa_veg(ji) + &
                       tot_bare_soil(ji)/fraction_veg(ji) * ( snowa_aged(1)+snowa_dec(1)*agefunc_veg(ji) )
               END IF
            END DO
            
            DO jv = 2, nvm
               DO ji = 1, kjpindex
                  IF ( fraction_veg(ji) .GT. min_sechiba ) THEN
                     snowa_veg(ji) = snowa_veg(ji) + &
                          veget(ji,jv)/fraction_veg(ji) * ( snowa_aged(jv)+snowa_dec(jv)*agefunc_veg(ji) )
                  ENDIF
               ENDDO
            ENDDO
         ELSE
            DO jv = 1, nvm
               DO ji = 1, kjpindex
                  IF ( fraction_veg(ji) .GT. min_sechiba ) THEN
                     snowa_veg(ji) = snowa_veg(ji) + &
                          veget_max(ji,jv)/fraction_veg(ji) * ( snowa_aged(jv)+snowa_dec(jv)*agefunc_veg(ji) )
                  ENDIF
               ENDDO
            ENDDO
         ENDIF

Catherine Ottlé says the right formulation is the one using tot_bare_soil and veget, meaning the ok_dgvm case.
The code should then be replaced with:

         ! For  vegetated surfaces
         fraction_veg(:) = un - totfrac_nobio(:)
         snowa_veg(:) = zero
         DO ji = 1, kjpindex
            IF ( fraction_veg(ji) .GT. min_sechiba ) THEN
               snowa_veg(ji) = snowa_veg(ji) + &
                    tot_bare_soil(ji)/fraction_veg(ji) * ( snowa_aged(1)+snowa_dec(1)*agefunc_veg(ji) )
            END IF
         END DO
            
         DO jv = 2, nvm
            DO ji = 1, kjpindex
               IF ( fraction_veg(ji) .GT. min_sechiba ) THEN
                  snowa_veg(ji) = snowa_veg(ji) + &
                          veget(ji,jv)/fraction_veg(ji) * ( snowa_aged(jv)+snowa_dec(jv)*agefunc_veg(ji) )
               ENDIF
            ENDDO
         ENDDO

[jgipsl]

Due to decision on Tuesday meeeting : The method for the case ok_dgvm=false is now used in both cases, see commit [4691].

The ticket stays open because as Catherine says above, the other method might be better but needs more tests.

[jgipsl]

Said during ORCH meeting today:

PP has discussed with CO, they will prepare a version with the modifications. PP will then ask AD for check before commit. PP will meet with Josefine to commit.

[peylin]

Philippe P: we have proposed together with Catherine to use two different approaches for high vegetation covers (Trees) versus low vegetation cover (Grass, crop).
For high vegetation cover, we should keep the formulation with vegetmax as the bare soil part when there is no/few leaves (difference between veget and vegetmax) we should still consider that the "aging function" for the albedo of Tree with branches/trunk is similar to the one when there is leaves.
On the other hand, for short grass/crop vegetation, the bare soil part when there is not/few leaves should be treated as the standard bare soil fraction (veget(1)).
The code has been modified by PP for Josefine to commit.
Philippe

[jgipsl]

Done in the trunk [6523]

[peylin]

The Routine CONDVEG_ALBEDO has been modified to account for the proposed modifications.
The comments are:

! The new rational for the snow albedo and its dependency to snow age is (see ticket 223; PP Fevrier 2020) :
! i) Short vegetation (grass, crop): we assume that when there is no leaves (LAI=0), the snow aging is not
! impacted by the vegetation. The difference between veget and vegetmax (bare soil fraction within the PFT
! should be considered in the same way as the bare soil fraction (PFT1).
! ii) High vegetation (Tree): In this case we consider that the vegetation impacts the snow aging even when
! LAI=0 because of the trunks, branches. So these PFTs should be treated differently without any
! consideration of the bare soil fraction within the PFT.

[nvuilsce]

A bug was introduced in the last commit [6523] for the treatment of bare soil (PFT1). For PFT1, veget equals veget_max and not zero. It resulted in an albedo of zero for Greenland.
A correction has been proposed by P. Peylin and commited in branch ORCHIDEE_3 [6907]

[peylin]

Following the correction on the snow albedo (for bare soil PFT), several actions are needed:

• Test the comit in branch ORCHIDEE_3 in the current Trunk
• Evaluate if the "non intuitive" fact that veget equals veget_max for PFT1 and not zero may cause some problem elsewhere in the code ?

Make these test before closing the ticket

[maignan]

The same concerns was already expressed in #404.

[xnwang]

In trunk:
tot_bare_soil_notree(: ) = 0
DO ipts = 1, kjpindex

DO ivm=1,nvm

IF ( fraction_veg(ipts) .GT. min_sechiba .AND. .NOT. is_tree(ivm) ) THEN

tot_bare_soil_notree(ipts) = tot_bare_soil_notree(ipts) + veget_maipts,ivm) - veget(ipts,ivm)

In orchidee tag 3:
tot_bare_soil_notree( : ) = veget_max(:,1)
The iterration over pfts starts from 2, not from 1.

Without the modification, we compute tot_bare_soil_notree by summing over ivm from 1 to nvm, so the veget_max from baresoil is taken into account.

With the modification (to be the same as Tag3), the tot_bare_soil_notree is initialized with veg_max at bare soil. Then we compute tot_bare_notree by summing over imv from 2 to nvm. So finally tot_bare_soil_notree is computed in the same way, with or without the modification.

Some test simulations were done by applying the modifications to the trunk. Then comparisons were made on albedo results with and without modification. There is no difference in the albedo results, which confirms the above analysis of code.

[xnwang]

The question mentioned in this ticket is clarified. The ticket can be closed. Some new issues were found while studying this ticket, which would be put into a new ticket.