Add a seafloor age (isochron) layer

CLAUDE.md

@@ -30,6 +30,7 @@ Forked from [SceneryStackTemplate](https://github.com/OpenPhysics/SceneryStackTe
 | `src/common/data/hotspots.ts` | Hand-maintained hotspot list |
 | `src/common/data/generated/` | **Generated — do not edit.** `npm run build-data` owns it |
 | `src/common/data/generated/motionFrameData.ts` | Rotations belonging to boundaries rather than plates |
+| `src/common/data/generated/seafloorAgeData.ts` | Isochrons of the ocean floor, and the ages they are drawn at |
 | `src/plate-tectonics/model/PlateTectonicsModel.ts` | All AXON state |
 | `src/plate-tectonics/model/EarthquakeDepthFilter.ts` | Depth bands and the filter predicate |
 | `src/plate-tectonics/view/PlateTectonicsScreenView.ts` | Layout, view switching, `pdomOrder` |
@@ -42,7 +43,7 @@ Forked from [SceneryStackTemplate](https://github.com/OpenPhysics/SceneryStackTe
 | `src/plate-tectonics/view/CrossSectionNode.ts` | Section + localized annotations |
 | `src/plate-tectonics/view/LegendSwatches.ts` | Map symbols, shared by legend and checkboxes |
 | `scripts/build-data.ts` | Fetches and reshapes every dataset |
-| `scripts/data/` | Fetch cache, GeoTIFF reader, geodesy, emitters |
+| `scripts/data/` | Fetch cache, GeoTIFF and netCDF readers, geodesy, contouring, emitters |
 
 ## Working on this sim
 
@@ -55,15 +56,16 @@ keyed by URL hash, so re-runs are fast and a changed query parameter re-fetches.
 
 `tests/geophysicalData.test.ts` guards the regeneration: it checks structural
 integrity and a few facts about the Earth (deep earthquakes cluster around the
-Pacific; the Chile profile's deep events sit inland of its shallow ones). Run
-`npm test` after any regeneration.
+Pacific; the Chile profile's deep events sit inland of its shallow ones; the Atlantic
+isochrons step out symmetrically from the ridge as they get older). Run `npm test`
+after any regeneration.
 
 `npm run build-data` with no arguments rebuilds everything. Naming steps —
-`plate-model`, `land`, `earthquakes`, `volcanoes`, `relief`, `cross-sections` —
-rebuilds only those, which is how the PB2002 model can be regenerated without also
-pulling a newer earthquake catalogue and a fresh DEM into an unrelated diff.
-`plate-model` covers the plates, their boundaries and the motion frames together,
-because those three index into each other.
+`plate-model`, `land`, `earthquakes`, `volcanoes`, `seafloor-age`, `relief`,
+`cross-sections` — rebuilds only those, which is how the PB2002 model can be
+regenerated without also pulling a newer earthquake catalogue and a fresh DEM into an
+unrelated diff. `plate-model` covers the plates, their boundaries and the motion frames
+together, because those three index into each other.
 
 ### What moves when the clock runs
 
@@ -126,6 +128,12 @@ Boundary colors (red / cyan / violet) and earthquake-depth colors (yellow → or
 magenta) have to stay mutually distinguishable *and* readable over the relief raster.
 Check both profiles after changing any of them.
 
+`seafloorAgeRampColorProperties` is a five-stop red → blue ramp, interpolated by age
+rather than indexed, so the number of isochron ages and the number of color stops are
+independent. Its young end deliberately sits near the divergent-boundary red — the
+youngest crust *is* the crust at the ridge — and the two are kept apart by line width
+and draw order instead of by hue.
+
 ## Common components
 
 ### PlateTectonicsPanel

CREDITS.md

@@ -9,7 +9,7 @@ GNU Affero General Public License v3.0 or later — see [org LICENSE](https://gi
 
 ## Data sources
 
-The simulation renders published observational data. Two of these sources ask for
+The simulation renders published observational data. Three of these sources ask for
 attribution and all of them deserve it; they are credited on screen, in the About
 dialog, and here.
 
@@ -52,6 +52,21 @@ Program holdings. Public domain.
 Public domain. The shaded relief raster shipped with the sim is rendered from it by
 `npm run build-data`.
 
+### Seafloor age — EarthByte
+
+Seton, M., Müller, R. D., Zahirovic, S., Williams, S., Wright, N. M., Cannon, J.,
+Whittaker, J. M., Matthews, K. J., & McGirr, R. (2020). *A global data set of
+present-day oceanic crustal age and seafloor spreading parameters.* Geochemistry,
+Geophysics, Geosystems, 21, e2020GC009214.
+[doi:10.1029/2020GC009214](https://doi.org/10.1029/2020GC009214)
+
+The 6-arc-minute present-day age grid (version 2020.1, GTS2012 timescale) published by
+the [EarthByte group](https://www.earthbyte.org/category/resources/data-models/seafloor-age/)
+at the University of Sydney, under the
+[Creative Commons Attribution 4.0 licence](https://creativecommons.org/licenses/by/4.0/).
+The isochrons shipped with the sim are contoured from it by `npm run build-data`; the
+grid itself is not redistributed.
+
 ## Acknowledgments
 
 Built with [SceneryStack](https://scenerystack.org/), from the

README.md

@@ -21,6 +21,10 @@ fitted to the earthquakes that trace it.
 - **Volcanoes and hotspots**: ~1 600 Holocene volcanoes plus the major mantle plumes.
 - **Topography / bathymetry**: shaded relief rendered from the NOAA global DEM, so
   trenches, ridges and mountain belts are visible rather than implied.
+- **Seafloor age (isochrons)**: lines of equal crustal age at 10, 20, 40 … 180 Ma,
+  contoured from the EarthByte age grid. The youngest crust hugs every spreading
+  ridge and the same ages appear at the same distance on both flanks, getting older
+  out to the continental margins — seafloor spreading, drawn from the measurements.
 
 **Cross-sections** — the Chile trench, the Mid-Atlantic Ridge, and the San Andreas
 fault. The surface profile comes from the DEM, the earthquakes and volcanoes are real
@@ -82,6 +86,7 @@ and full Interactive Description support.
 | Earthquakes | [USGS ANSS ComCat](https://earthquake.usgs.gov/fdsnws/event/1/) | Public domain |
 | Volcanoes | [NOAA NCEI](https://www.ngdc.noaa.gov/hazel/) / Smithsonian GVP Holocene volcano list | Public domain |
 | Elevation and bathymetry | [NOAA NCEI global DEM mosaic](https://gis.ngdc.noaa.gov/arcgis/rest/services/DEM_mosaics/DEM_global_mosaic/ImageServer) | Public domain |
+| Age of the ocean floor | [EarthByte](https://www.earthbyte.org/category/resources/data-models/seafloor-age/) / Seton et al. (2020), doi:10.1029/2020GC009214 | CC BY 4.0 |
 
 ## License
 

doc/implementation-notes.md

@@ -40,7 +40,8 @@ src/
       LegendSwatches.ts            the symbols, shared by legend and checkboxes
 scripts/
   build-data.ts                  fetches and reshapes every dataset
-  data/                          fetch cache, GeoTIFF reader, geodesy, emitters
+  data/                          fetch cache, GeoTIFF + netCDF readers, geodesy,
+                                 marching-squares contouring, emitters
 ```
 
 ## Why the map is a canvas
@@ -174,6 +175,16 @@ Notable pieces:
 - `scripts/data/dem.ts` contains a ~120-line GeoTIFF reader. NOAA's image service
   returns an uncompressed, tiled, signed-16-bit TIFF, which is little enough format to
   parse directly and saves a dependency.
+- `scripts/data/netcdf.ts` reads the classic netCDF-3 format, for the same reason: the
+  EarthByte age grid is published as a header of big-endian counted fields followed by
+  the raw values, which is small enough to parse directly and saves the build a
+  dependency on the netCDF/HDF5 stack.
+- `scripts/data/contour.ts` is marching squares, used once, to turn that age grid into
+  isochron polylines. Cells with a missing corner are skipped whole, so a contour stops
+  at the edge of the ocean instead of being interpolated onto land, and it works in
+  grid coordinates so that two neighbouring cells produce bit-identical crossings on
+  their shared edge — which is what lets the segments be strung together by exact key
+  match rather than by proximity.
 - Boundary segments are rebuilt from PB2002's *step* file rather than its boundary
   file, because only the steps carry the class (`OSR`, `SUB`, `CTF`, …) that the
   divergent / convergent / transform colouring needs.

doc/model.md

@@ -42,6 +42,7 @@ the map show, and it is why the Nazca arrow is long and the Antarctic arrow is a
 | Volcanoes | NOAA NCEI Holocene volcano list (Smithsonian GVP holdings) | ~1 600 volcanoes |
 | Hotspots | Hand-maintained list of the plumes named in introductory texts | `src/common/data/hotspots.ts` |
 | Topography and bathymetry | NOAA NCEI global DEM mosaic | rendered to a shaded relief raster, and sampled along each cross-section |
+| Age of the ocean floor | EarthByte / Seton et al. (2020) present-day age grid, doi:10.1029/2020GC009214 | contoured into isochrons at 10, 20, 40 … 180 Ma |
 
 ### From PB2002's poles to absolute plate motion
 
@@ -102,6 +103,37 @@ The outlines are subdivided until this stops showing: an edge whose ends ride mo
 far enough apart to stretch it by more than 200 km over the slider's range is split
 and reconsidered. `tests/PlateEvolution.test.ts` holds the whole scheme in place.
 
+## Seafloor isochrons
+
+The **Seafloor age** layer draws the ocean floor's isochrons: the lines along which
+the crust is all one age, at 10, 20, 40 … 180 Ma. They come from the EarthByte
+present-day age grid, which is built by identifying marine magnetic anomalies —
+stripes of alternating remanent magnetisation frozen into basalt as it cools through
+the Curie point — and dating them against the geomagnetic polarity timescale, itself
+calibrated on radiometrically dated rock. `scripts/build-data.ts` contours the
+6-arc-minute grid with marching squares on a 0.3° mesh and simplifies what comes out.
+
+The picture is the argument for seafloor spreading, and it is worth reading in three
+steps:
+
+1. **The youngest crust is at the ridges.** The 10 Ma line runs a degree or two either
+   side of every spreading axis and nowhere else.
+2. **The pattern is symmetric.** The same age appears at about the same distance on
+   *both* flanks, because both plates take roughly half the new crust. Across the
+   Atlantic at 24° N the 40 Ma line sits about 5° west and 7° east of the axis, the
+   80 Ma line about 12° and 13°; `tests/geophysicalData.test.ts` asserts that ordering
+   and that symmetry.
+3. **It stops at about 180 Ma.** There is no older ocean floor to find, because it has
+   all been subducted, while the continents carry rock a *hundred* times older. The
+   sea floor is not old and permanent; it is a conveyor.
+
+An isochron is frozen into the crust, so unlike a plate boundary it rides its plate —
+vertex by vertex, because one isochron can cross several. Two consequences follow when
+the clock runs, and both are honest rather than cosmetic. The two flanks of a pair walk
+back towards the ridge that made them. And crust younger than the reconstruction has
+reached did not exist yet, so at 50 Myr ago the 10, 20 and 40 Ma isochrons are simply
+not drawn.
+
 ## Earthquake depth bands
 
 Hypocentres are grouped at **70 km** and **300 km**, the conventional shallow /
@@ -156,7 +188,10 @@ distorting the picture silently.
    and watch the events step down along the slab from the trench to 600 km.
 3. **Seafloor spreading.** Switch on boundaries and topography, and follow the ridges:
    a continuous mountain range down the middle of the Atlantic, marked by shallow
-   earthquakes and no deep ones at all.
+   earthquakes and no deep ones at all. Then add **Seafloor age** and turn the other
+   layers off: the isochrons fan out from that same mountain range, matched pair by
+   matched pair, red at the axis and blue at the margins. Run the clock back and watch
+   the young ones disappear into the ridge that had not yet made them.
 
 ## What this model is not
 
@@ -186,6 +221,12 @@ distorting the picture silently.
 - **The relief raster is present-day.** It is hidden as soon as the reconstruction
   moves off the present day, because sea floor that has not been created yet cannot be
   shown.
+- **Isochrons are carried rigidly, not un-made.** Hiding the ones younger than the
+  reconstruction is right, but the ones that remain are only *rotated* back with their
+  plates: the ocean between them should also be closing up, and here it is not, because
+  the model has no way to un-make crust. So running to 50 Myr ago narrows the Atlantic
+  isochron fan by rather less than it should. The 10 Ma pair walking together onto the
+  ridge is the honest part of that picture; the 160 Ma pair barely moving is not.
 - **Hotspots do not move.** That is deliberate, and it is the physics: a plume is
   anchored in the deep mantle while the plate slides over it, which is why the Hawaiian
   chain gets older to the north-west.