Missing iceberg restart in default configurations

[smueller]

Context

Default configurations with enabled icebergs (ln_icebergs = .true. in namelist &namberg) initially generate "test icebergs" as nn_test_icebergs=10 in file namelist_ref. Such configurations, however, do not seem to restart icebergs. In order to enable the restart mechanism of the ICB component, nn_test_icebergs can be set to a value of less than '1' in restart configurations. This behaviour is not intuitive and can lead to the incorrect assumption of active ICB restarting in long model runs.

Analysis

When "test icebergs" are enabled (ln_test_icebergs > 0), a reset of the iceberg distribution is triggered during the model restart (source:/NEMO/trunk/src/OCE/ICB/icbini.F90@13295#L282​).

Recommendation

The conditional at source:NEMO/trunk/src/OCE/ICB/icbini.F90@13295#L282​ appears to be superfluous and could be removed to enable ICB restarts in default configurations (this would also permit the removal of the ln_test_icebergs modification in the SHORT configuration for the SETTE ORCA2_ICE_PISCES restartability test at source:/utils/CI/sette/sette_reference-configurations.sh@13790#L406​).

Alternatively, the default value of nn_test_icebergs could be changed to 0 in source:/NEMO/trunk/cfgs/SHARED/namelist_ref​ (in which case, the ORCA2_ICE_PISCES SETTE tests would have to be adjusted to enable an ICB initialisation that is appropriate for the tests).

[mathiot]

This ticket should probably be fixed in the same time as this one: https://forge.ipsl.jussieu.fr/nemo/ticket/2467 or at least by keeping this one in mind during the development. This is roughly the same issue with extra suggestion for clarification of the iceberg namelist.

[acc]

The conditional in icbini.F90 is not wholly superfluous because it does allow testing of icebergs starting with a restart from an ocean that was spun-up without icebergs. In this respect option 2 is probably the better of the 'quick' fixes. With a longer term tidy up as suggested to follow.