Give the resource a weight-length relationship

[gustavdelius]

Stacked on #538 — review that one first; this PR's diff is against array-type-attribute.

Summary

plotSpectra(params, size_axis = "l") used to drop the resource silently — 79 rows on the weight axis, 0 on the length axis, no message — because the length conversion needs an allometric weight-length relationship and only the species had one.

resource_params() now carries a and b, defaulting to the equivalent spherical diameter of an organism with the density of water, $w = \frac{\pi}{6}l^3$, i.e. a = pi/6 ≈ 0.524, b = 3. ESD is the standard convention in plankton ecology precisely because it is geometric rather than taxonomic, so it degrades gracefully across a composite of many taxa. On a mizer grid it gives sensible sizes throughout:

w (g)	ESD	plausible organism
1e-12	1.2 µm	bacterium
1e-6	124 µm	ciliate / small copepod
1e-3	1.2 mm	copepod
9.8 (w_pp_cutoff)	2.7 cm	macrozooplankton

The resource therefore appears on length-based spectra, and the resource array classes gain size_axis and llim, animate() included — it used to raise "a length axis is not available for resource arrays".

The convention caveat

The resource sits on the length axis at its own convention: a fish of a given weight is about 3.7× longer than a sphere of that weight. Two things about that:

• It is real biology, not an artefact. A 1 mg copepod really is shorter than a 1 mg fish larva, so the step at the resource–consumer boundary is information.
• Because both conventions have b = 3, the change is a pure translation on a log-length axis: the resource curve keeps its shape and slope.

It is also not a new kind of compromise — a length axis is already per-species, since each species converts with its own a, b.

Design

The parameters feed none of the rates, so they are deliberately not arguments of setResource(). Instead:

• setResource() fills them in as defaults, so every newly built model shows them in resource_params() and users can change them there;
• they are also defaulted at the point of use (resource_length_params()), so a model saved before they existed needs no upgrade — and repair_params() and validation_key() stay untouched, which would otherwise have meant adding resource_params to the fingerprint of every object.
• resource_params<-() validates them when present.

The one shared change is plot_length_params(), which replaces the inline species_params lookup in convert_plot_size_axis() and the Jacobian in convert_plot_density_axis(). It returns a/b per row — from the species, from the resource, or NA for anything else — so both the coordinate conversion and the density Jacobian handle the resource through one place.

Testing

Full suite: 4288 passing, 0 failures. New tests cover the default and the fallback for old objects, validation, the ESD sanity check (µm at 1e-12 g, mm at 1e-3 g, the 3.741 fish/ESD ratio), the resource on a length axis with the Jacobian and llim, and the resource joining a length-axis plotSpectra(). One existing test asserted that both Resource and Total are hidden on a length axis; it now asserts that Resource appears and Total still does not — the latter is the next PR.

🤖 Generated with Claude Code