Yelmo coupling

elsa as the layer backend of Yelmo’s passive-tracer subsystem

Yelmo drives elsa through its passive-tracer subsystem ytrc (src/yelmo_tracers.f90). ytrc can run three backends side by side: an Eulerian age tracer, the Lagrangian particle model tracer, and elsa. Each backend fills its own deposition-time field on Yelmo’s sigma grid, and one of them is designated as the authoritative source. This page describes the elsa side of that coupling as it stands in Yelmo. The Yelmo documentation remains the reference for ytrc itself.

Build

Yelmo expects an elsa checkout (or a link to one) at yelmo/elsa. Its Makefile builds libelsa.a in that checkout through the elsa-static target, passing on the openmp setting, and links it ahead of fesm-utils. Yelmox reaches the same checkout at yelmo/elsa. No separate installation step is needed once the checkout is in place.

Yelmo also carries a copy of the defaults file at yelmo/input/elsa_defaults.nml, since elsa reads this file by a path relative to the run directory (see the user guide). The copy must be updated whenever a parameter is added to or removed from elsa. Otherwise elsa_init stops on the first parameter that the copy does not declare.

Parameters

The coupling is controlled by the ytrc group of the Yelmo namelist.

&ytrc
    use_elsa        = True
    t_dep_source    = "elsa"         ! "euler", "trc" or "elsa"
    time_end        = 0.0            ! [yr] end of the experiment
    elsa_nml        = "par/elsa.nml"
    elsa_group      = "elsa"
/

Parameter	Content
`use_elsa`	runs the elsa backend
`t_dep_source`	the backend that fills the authoritative `t_dep` field
`time_end`	passed to `elsa_init` to size the layer stack
`elsa_nml`, `elsa_group`	the file and the group from which elsa reads its own parameters

Two settings need attention. First, time_end defaults to 0 and must be set to the end time of the run, on Yelmo’s time axis in years. elsa stops at init if it is not later than the start time. Second, the elsa group must keep grid_factor = 1. Yelmo checks this at init and stops otherwise, since the mapping described below is done column by column on a shared grid.

What Yelmo passes

At init, Yelmo passes its cell-centre axes, its zeta_aa levels and the ice thickness, with stagger = "acx_acy". At every Yelmo time step it calls

call elsa_update(trc%elsa,time,tpo%now%H_ice,dyn%now%ux,dyn%now%uy, &
                 tpo%now%smb,tpo%now%bmb)

with the arguments converted to double precision. The velocities are Yelmo’s native three-dimensional ux and uy on the acx and acy nodes, so nothing is destaggered. elsa decides internally whether an update is due, according to dt_coupling.

elsa integrates the mass balance and the velocities over every Yelmo time step, and applies their mean over the coupling period (see the user guide). The fields passed after a step are taken to hold over that step, which is the case for tpo%now%smb and tpo%now%bmb.

The `t_dep_elsa` field

After each update, Yelmo maps the layer stack onto its sigma grid as a deposition-time field t_dep_elsa(xc,yc,zeta). For each column, the layer interfaces from the top of the initialization layers to the surface define a set of knots in height, and the corresponding deposition times define the knots in time. The deposition time at each sigma level is then interpolated linearly between the knots.

Two conventions of this mapping should be known. The surface knot carries the time of the last elsa update. All ice below the oldest isochrone, i.e., within the initialization layers, is assigned the initial time of elsa, since its true age is unknown. The field is thus a lower bound on age in that part of the column.

If t_dep_source = "elsa", t_dep_elsa is copied to the authoritative t_dep field, from which Yelmo diagnoses the depth of the isochrones listed in time_iso (depth_iso). Both t_dep_elsa and depth_iso are available as Yelmo output variables.

Restart

Yelmo writes elsa’s restart as a sidecar file next to its own restart file, with _elsa inserted before the extension (e.g., yelmo_restart.nc is accompanied by yelmo_restart_elsa.nc). On a restarted run, Yelmo derives the same sidecar name from its restart path and passes it to elsa_init.

The rules given in the user guide apply. In particular, elsa extends its schedule and its stack if time_end of the restarted run lies beyond that of the original run. Each segment of an experiment thus only needs its own end time.

Native elsa output

The t_dep_elsa field is a diagnostic on Yelmo’s grid. The layer stack itself lives in ylmo%trc%elsa, and can be written in elsa’s own format with elsa_write_init and elsa_write_step from the driver program (see Output).