Palaeoclimate cluster (T40–T45): tidy pass, rim/coastline fixes, PLASIM-GENIE naming

Notebooks/paleoclimate_helpers.py

@@ -0,0 +1,73 @@
+"""Shared display helpers for the paleoclimate-cluster notebooks (T40–T45).
+
+These functions clean up how a coarse climate-model field on a lat/lon grid is
+drawn with ``pygmt.grdimage`` in a Mollweide projection. They are display-only:
+they wrap/clamp/resample a field for *plotting* and never change the underlying
+data used for analysis.
+
+Used by T40 (deep-time SAT snapshots) and T41 (cGENIE SST vs proxies); import
+with, e.g.::
+
+    from paleoclimate_helpers import close_lon_seam, clamp_lat_poles, refine_for_plot
+
+Requires ``numpy`` and ``xarray``.
+"""
+
+import numpy as np
+import xarray as xr
+
+
+def close_lon_seam(da):
+    """Pad a −180→180 grid cyclically so grdimage reaches the ±180 seam.
+
+    Climate-model SAT/SST grids typically stop short of the +180 meridian (a
+    3.75° grid's last column is at +176.25°), so grdimage leaves an unpainted
+    wedge at the antimeridian where the Mollweide boundary curves past the last
+    data column (GMT fills it grey). Wrapping one column round each seam lets the
+    raster reach the edge; the extra half-cells fall outside ``region="d"`` and
+    are clipped by the projection boundary.
+    """
+    lon = da.lon.values
+    dlon = float(lon[1] - lon[0])
+    west = da.isel(lon=-1).assign_coords(lon=lon[0] - dlon)   # +180 side → west edge
+    east = da.isel(lon=0).assign_coords(lon=lon[-1] + dlon)   # −180 side → east edge
+    return xr.concat([west, da, east], dim="lon").sortby("lon")
+
+
+def clamp_lat_poles(da):
+    """Extend the outermost lat rows out to ±90 (latitude can't wrap, so clamp).
+
+    A grid that stops short of the poles would leave NaN slivers there when
+    interpolated onto a finer grid. Repeating the edge rows out to ±90 keeps the
+    interpolation in-range. Only adds a pole row when the grid actually stops
+    short — a grid that already reaches −90 would get a duplicated row, making
+    the latitude index non-unique (which breaks ``interp``).
+    """
+    lat = da.lat.values
+    parts = [da]
+    if float(lat[0]) > -90.0:
+        parts.insert(0, da.isel(lat=0).assign_coords(lat=-90.0))
+    if float(lat[-1]) < 90.0:
+        parts.append(da.isel(lat=-1).assign_coords(lat=90.0))
+    return xr.concat(parts, dim="lat").sortby("lat") if len(parts) > 1 else da
+
+
+def refine_for_plot(da, step=0.5):
+    """Resample a coarse field onto a fine regular grid for display.
+
+    grdimage draws each native model cell as a quadrilateral, which staircases
+    along the curved Mollweide rim at high latitudes (and aliases into stray
+    cells right at the ±180 seam). Wrapping the seam, clamping the poles, and
+    bilinearly resampling onto a fine (default 0.5°) grid before plotting gives a
+    smooth rim. Display-only: it linearly interpolates a continuous field and
+    changes no analysis.
+
+    The final ``.transpose("lat", "lon")`` is essential: the concat/sortby steps
+    can silently reorder the dims to (lon, lat), and pygmt infers the plot region
+    from dim order — a transposed grid renders with lon/lat swapped (the data
+    collapses to a central blob). Forcing (lat, lon) keeps grdimage correct.
+    """
+    src = clamp_lat_poles(close_lon_seam(da))
+    lon = np.arange(-180.0, 180.0 + step / 2, step)
+    lat = np.arange(-90.0, 90.0 + step / 2, step)
+    return src.interp(lon=lon, lat=lat, method="linear").transpose("lat", "lon")

claude_palaeoclimate.md

@@ -0,0 +1,95 @@
+# Palaeoclimate cluster (T40–T45) — change summary
+
+Notes on a tidy/fix pass over the paleoclimate-cluster notebooks (**T40–T45**),
+done with Claude. T42 was deliberately left untouched (see *Known issues*).
+
+> **Submitted as a pull request:** [EarthByte/GPlately-pyGMT-tutorials#1](https://github.com/EarthByte/GPlately-pyGMT-tutorials/pull/1),
+> from branch `palaeoclimate-cluster-fixes` on the `amer7632` fork.
+
+## Shared infrastructure
+
+- **`Notebooks/paleoclimate_helpers.py`** — new module with three display-only
+  helpers used across the raster-map notebooks:
+  - `close_lon_seam` — cyclically pads a −180→180 grid so `grdimage` reaches the
+    ±180° seam (kills the grey wedge at the antimeridian).
+  - `clamp_lat_poles` — extends the outermost latitude rows to ±90° before
+    interpolation (no NaN slivers at the poles).
+  - `refine_for_plot` — wraps the seam, clamps the poles, and bilinearly
+    resamples a coarse GCM field onto a fine (0.5°) grid so the Mollweide rim is
+    smooth instead of a staircase. **Display-only** — analysis still uses the
+    native grids. It ends with `.transpose("lat","lon")` because pyGMT infers the
+    plot region from dim order, and the intermediate steps can silently transpose.
+  T40, T41, T43, T44 import from it (via a `sys.path` insert of `Notebooks/`).
+
+- **Data rename** — `data/scion_permian/` → `data/plasim_genie_permian/`, and the
+  three bundled files `scion_252Ma_*` → `plasim_genie_252Ma_*`. The bundle README
+  was rewritten (correct model name, Holden 2016 citation, T43/T44 cross-refs, and
+  a leaked local path removed).
+
+- **`environment.yml`** — added `cmcrameri` + `openpyxl` (conda-forge) and
+  `cgeniepy` (pip). `bayfox` (T42) omitted; pySCION (T45) is a separate clone, not
+  a package.
+
+## Model-naming correction (T40, T43, T44)
+
+The bundled end-Permian climate fields were variously labelled "cGENIE-PLASIM"
+and "SCION". They are **PLASIM-GENIE** runs (PLASIM atmosphere + GOLDSTEIN ocean;
+the cGENIE biogeochemistry module was off), cited as Holden et al. (2016). All
+three notebooks were renamed/relabelled accordingly, and their SCION references
+were replaced. (T45 is genuinely pySCION-based, so SCION naming stays there.)
+
+## Per-notebook changes
+
+- **T40 — Deep-time snapshots.** Fixed the Mollweide rim/gridding artefacts
+  (seam + high-latitude staircase); a transposed Gaussian-lat grid that had
+  collapsed the deep-time map to a central blob; prose/code mismatches (palette
+  clamps); `cGENIE`→`GENIE`→`PLASIM-GENIE`; `jet`→`vik` anomaly palette; removed
+  SCION framing; reference fixes. Data caveat below.
+
+- **T41 — cGENIE SST vs proxies.** Removed the unused/advertised `cgeniepy`
+  dependency (loads the bundled netCDF with xarray); fixed the rim artefact via
+  `refine_for_plot`; switched both panels to reconstructed 50 Ma coastlines
+  (were modern); corrected a `viridis`→`thermal` mismatch; fixed a wrong bundled
+  filename; stripped inline-URL clutter.
+
+- **T43 — PLASIM-GENIE Permian SAT** *(renamed from `..._SCION_...`)*. SCION →
+  PLASIM-GENIE throughout; fixed wrong plate-model labels (prose said
+  Müller2022 / Zahirovic2022 for coastlines that are actually Merdith2021);
+  `refine_for_plot` on the SAT + land-sea-mask panels.
+
+- **T44 — CO₂-sensitivity ensemble** *(renamed from `..._SCION_...`)*. Fixed the
+  SAT-map coastlines (were **modern** GSHHG on a 252 Ma map → now the model's own
+  land-sea-mask contour); fixed a mangled "CO₂" title (Unicode subscript doesn't
+  render in GMT); `refine_for_plot` rim fix; moved the spin-up / equilibrium
+  check up to §2 (validate before analysing); **fixed a 64× normalization bug** —
+  the bundled spin-up scalars were area-weighted but never divided by the number
+  of longitudes, so GAST and |TOA thermal| were 64× too large. SCION →
+  PLASIM-GENIE; broken version probe fixed.
+
+- **T45 — pySCION Phanerozoic biogeochemistry.** Fixed a **years→Ma bug**:
+  `age` came from `state.time` (in years), so every time-series x-axis read
+  0–6×10⁸ "Ma" and, worse, SET 5's inflection lookup compared Ma against years —
+  all three paleo-Earth maps stamped the same ~349 ppmv. After `/1e6` the run
+  window reads "0–600 Ma" and the maps show age-correct pCO₂ (250 Ma → 8690,
+  30 Ma → 336 ppmv). Also adapted to the current `amer7632/pySCION` API
+  (`pySCION_initialise` now returns a `(run, interpstack, lipstack)` tuple), and
+  set the reference to Merdith et al. (2025, *Science Advances*).
+
+## Known issues / follow-ups
+
+- **T42 (bayfox SST)** — left untouched. It won't run with `bayfox` installed:
+  the code calls a non-existent `bayfox.predict_sst` (real API is
+  `predict_seatemp`, returning a `Prediction`); `SPECIES="G_ruber"` is invalid
+  and unused (valid code is `"G. ruber"` via `foram=`). The bundled CSV is now a
+  planktonic WPWP (ODP 806) record, but the prose/Data-availability still describe
+  the old LR04 benthic stack and `CORE_LON_LAT` points at the North Atlantic.
+
+- **`data/paleoclimate/hadcm3_modern_mat.nc`** (T40 modern panel) is nearly
+  zonal — no land-sea contrast and far too warm at the poles — i.e. not a
+  faithful HadCM3 field. Left as-is; worth replacing at source.
+
+- **`plasim_genie_252Ma_spinup.nc`** still stores the 64×-inflated GAST/thermal
+  values at source; T44 divides by `n_lon` as a workaround.
+
+- **pySCION** is not bundled: T45 needs a clone at `external/pySCION`
+  (`https://github.com/amer7632/pySCION`, gitignored) plus `openpyxl`.

data/plasim_genie_permian/README.md

@@ -0,0 +1,45 @@
+# PLASIM-GENIE end-Permian (252 Ma) CO2-sensitivity ensemble — compact bundle
+
+Source: Andrew Merdith's PLASIM-GENIE coupled-model output, version `MER25rev6`
+(2025-08-01). PLASIM-GENIE = PLASIM atmosphere + GOLDSTEIN ocean (the cGENIE
+biogeochemistry module was off for these runs). PLASIM atmosphere is T21
+(64 lon x 32 lat, ~5.625 deg). GOLDSTEIN ocean is on a 64 lon x 32 lat C-grid
+with 16 depth levels (only the surface and 2D bathymetry are bundled here).
+
+## Files
+
+- `plasim_genie_252Ma_equilibrium.nc` — last-year (year 3000) snapshot per CO2
+  level. Dims (co2_ppm: 10, latitude: 32, longitude: 64). Variables: SAT,
+  surface_T, sea_ice_cover, sea_ice_thickness, snow_depth, surface_albedo,
+  landsea_mask, orography, top_thermal, top_solar.
+- `plasim_genie_252Ma_bathymetry.nc` — GOLDSTEIN bathymetry per CO2. Identical
+  across CO2 levels (same input geography); bundled per-CO2 for symmetry.
+- `plasim_genie_252Ma_spinup.nc` — (co2_ppm, year) energy-balance time series
+  for the model-stability / spin-up curves (T44 §5). 30 yearly steps per CO2
+  level (years 100..3000 in steps of 100).
+
+## CO2 levels
+
+140, 280, 560, 1120, 2240, 3360, 4480, 8960, 17920, 35840 ppm.
+End-Permian best-estimate ~1120 ppm; runaway scenarios up to 35840 ppm.
+
+## Size
+
+Full bundle ~50 KB compressed (vs 1.5 GB for the raw model output).
+That covers everything T43 + T44 need.
+
+## Provenance
+
+For the full model output (every yearly snapshot, every variable), contact
+Andrew Merdith.
+
+## Citation
+
+Cite the PLASIM-GENIE model description:
+
+  Holden, P.B., Edwards, N.R., Fraedrich, K., Kirk, E., Lunkeit, F. & Zhu, X.
+  (2016). PLASIM-GENIE v1.0: a new intermediate complexity AOGCM.
+  *Geoscientific Model Development* 9, 3347-3361.
+
+And acknowledge Andrew Merdith's 252 Ma ensemble as personal communication
+pending publication.

environment.yml

@@ -17,4 +17,8 @@ dependencies:
   - pyarrow         # for parquet caches (T10)
   - imageio         # for T03 animations
   - ffmpeg          # for T03 MP4 encoding
+  - cmcrameri       # Crameri scientific colour maps (T44, T45)
+  - openpyxl        # Excel forcings for pySCION (T45)
   - pip
+  - pip:
+    - cgeniepy      # cGENIE output reader (T41); pip-only, not on conda-forge