
Resolve a writable, Docker-bind-mountable run root for calibration runs
growth_calibration_work_root.RdMirrors the fire-calibration scratch root, with one extra constraint: the value is
baked into the {targets} command at pipeline-definition time, so it must be
STABLE across sessions. tempdir() is therefore not usable as the fallback –
it would change every session and invalidate the run every time.
Details
Resolution order:
LANDIS_GROWTH_SCRATCH, if set/mnt/scratch/<user>/<project>/landis-growth-calibration, when/mnt/scratch/<user>exists and is writable (the cluster convention)~/.cache/landis-growth-calibration
The default of running in place is deliberately NOT used: in many deployments
./LANDIS-II/ is commonly a root-squashed NFS symlink, which Docker cannot
bind-mount (the run dies with exit code 126). Staging on a local filesystem
and letting tar_landis() archive the finished replicate back to the tracked
scenario directory works on the workstation, the cluster, and a laptop alike.
See also
Other growth calibration helpers:
extract_landscape_cohort_structures(),
growth_add_objective(),
growth_auto_window(),
growth_best_candidates(),
growth_bin_observations(),
growth_calibration_design(),
growth_calibration_partition(),
growth_climatic_distance(),
growth_climatic_weight(),
growth_expand_over_pseudo_species(),
growth_factorial_ratio_grid(),
growth_fitting_windows(),
growth_identifiability(),
growth_inflation_factor(),
growth_mortality_onset_frac(),
growth_plot_palette(),
growth_pseudo_species_name(),
growth_reference_curves(),
growth_reference_inflection(),
growth_score_fit(),
growth_scoring_for(),
growth_smooth_observations(),
growth_structure_cell_curves(),
growth_structure_cohort_table(),
growth_structure_design(),
growth_structure_summary(),
growth_window_for(),
plot_growth_calibration(),
plot_growth_candidate(),
plot_growth_factorial_sensitivity(),
plot_growth_structures(),
read_growth_scoring(),
read_landscape_cohort_structures(),
scale_linetype_growth_reference(),
write_growth_review_bundle()