and results

  pism_terra/raster.py: introduce local_scale_factor(crs, x, y) that returns
  the dimensionless local scale factor k such that 1 m of true ground
  displacement at (x, y) equals k m of crs map units.

  - Computes sqrt(|det J|) of the true→map Jacobian via a numerical
    finite difference of pyproj's geographic→crs transform.
  - Uses the source ellipsoid's prime-vertical N(φ) and meridional M(φ)
    radii of curvature to convert the geographic step to true metres,
    instead of a hardcoded 6378137 m (correct on non-WGS-84 datums too).
  - Vectorised on numpy arrays; scalar in → float out.

  Motivation: glacier_velocities_from_grid does a finite-difference
  advect-and-roundtrip to rotate vx/vy into the destination CRS axes,
  which silently assumes the source CRS has unit local scale. That holds
  for the ITS_LIVE v2.1 per-region polar-stereographic COGs but fails for
  EPSG:3857 (Web Mercator), where k = 1/cos(φ) ≈ 2 at 60° N and the
  returned vx/vy come out halved. Pre-multiplying vx_pts/vy_pts by k
  ahead of the FD makes the round-trip CRS-agnostic.

  The docstring documents the math (with the M, N, det J formulae),
  explains the conformal-vs-non-conformal behaviour, and ships three
  doctest examples covering Mercator at 60° N, polar-stereo near its
  true-scale latitude, and the array call signature.

  Single-line subject if you prefer the terse style:

  raster: add local_scale_factor for CRS-agnostic vector transforms

…a into aaschwanden/summer-school

Aaschwanden/summer school

…ra into aaschwanden/summer-school

…ra into aaschwanden/summer-school

impi renamed to iimpi. Uff.

…ra into aaschwanden/summer-school

…ra into aaschwanden/summer-school

get resolution from config.

…a into aaschwanden/summer-school

Split each ISMIP7 product scalar file into per-variable diagnostics
(lim, limnsw, iareagr/fl, tend*) via a packaged NCO script run at the end
of the generated forward job script.

- data/postprocess_ismip7_scalar.sh: NCO script (fix global attrs + time
  units; extract each diagnostic to float with _FillValue; correct flux
  units/time). Uses a per-input scratch file so concurrent sweep jobs
  don't collide on a shared tmp.nc.
- run.py: post_scalar_str runs the script on the product-leg scalar --
  scalar_hist for historical experiments (C001/C002), scalar_proj for the
  projection experiments, both for legacy non-counter runs.
- templates: add {{ post_scalar_str }} to the ISMIP7 forward templates
  (and the missing compliance-checker block to chinook-ismip7.j2).
- pyproject: ship data/*.sh in package-data.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014DGQGDhsstkb1faoVUwh4M

Aaschwanden/ismip7 gris