ice velocity issues in AGRIF sponge layer with LIM2 VP

[torgemartin]

Context

We found unusually high (order 10¹ to 10² greater than expected) sea ice drift velocities near coastal grid cells within the sponge layer of the AGRIF nested model. The false velocities are advected into the nest domain. As a consequence, thick ice formed (convergence downstream of ice flow) and ocean received too much momentum at these grid cells. The high velocities are not found in the base model. We use NEMO 3.6 with LIM2 and VP rheology.

Analysis

Issue with interpolation of ice drift velocity from base model to nest domain. Problem does not occur when using EVP (C-grid), only in case of VP (B-grid).

Recommendation

Problem seems to be solved when applying the following change to agrif_lim2_interp.F90 (note, this is for key_lim2_vp):

In SUBROUTINE agrif_rhg_lim2_load the Agrif_specialValue must be set to -9999. before calling interp_u/v_ice.

67,68c67,72
<- Agrif_SpecialValue=0.
<- Agrif_UseSpecialValue = .FALSE.
---
-> ! Agrif_SpecialValue=0.
-> ! Agrif_UseSpecialValue = .FALSE.
-> Agrif_SpecialValue=-9999.
-> Agrif_UseSpecialValue = .TRUE.

B.t.w. this correction compares well to the EVP case.