[patnr]

Adds rows for other analysis flavours to the MiniRes benchmark, at the settings of the rows they vary:

• ES-MDA (4 steps, α = 4) with the full and the enif analysis;
• GN-EnRML (4 iterations) with the subspace2 and the margis analysis.

ES-MDA's subspace and subspace2 flavours are left out. Their weight-space step is applied to the prior ensemble, not the current one (propose_state), so under ES-MDA each step is a fresh update of the prior with that step's inflated data error, not an assimilation in sequence.

RESULTS.md is not rerun here.

Tentative numbers, from run.py --quick (coarse case, 5 truths; means):

	misfit	CRPS forecast WWCT	logK RMSE	logK CRPS	sor CRPS	simulations
prior	97.4	0.103	0.772	0.437	0.0178
GN-EnRML (4 iterations)	5.53	0.025	0.779	0.440	0.0160	250
GN-EnRML, subspace2 analysis	6.02	0.027	0.768	0.431	0.0146	250
GN-EnRML, margis analysis	6.21	0.027	0.763	0.428	0.0133	250
ES-MDA (4 steps, α = 4)	1.56	0.039	0.795	0.464	0.0269	250
ES-MDA, full analysis	1.68	0.039	0.793	0.456	0.0231	250
ES-MDA, enif analysis	13.0	0.036	0.732	0.412	0.0119	250

Checked locally: the benchmark's smoke test, and ruff.

🤖 Generated with Claude Code