Skip to contents

Regional Spectra

Source: vignettes/flux-preserving-products.Rmd
flux-preserving-products.Rmd

The segmentation map is categorical. Summing the spaxels in each region gives the flux-preserving spectrum used for fitting.

Regional summaries 

products <- summarize_cluster_spectra(
  seg,
  var_cube = variance_cube,
  variance_inflation = 1
)

products$cluster_ids
products$n_spaxels
products$sum_spectra
products$sum_variance

For region (k), channel λ\lambda, and assigned spaxels pp, Capivara computes

Fk,λsum=∑p∈kFp,λ. F_{k,\lambda}^{\mathrm{sum}} = \sum_{p \in k} F_{p,\lambda}.

When a variance cube is supplied, the regional variance is the sum of valid input variances multiplied by variance_inflation. Capivara does not estimate spatial covariance from the flux cube.

Means and medians describe spectral shape; only the summed spectra preserve the total regional flux.

Representative cubes 

representative <- reconstruct_cluster_cube(
  seg,
  template = "median",
  preserve_mask = TRUE,
  return_residual = TRUE
)

flux_preserving <- reconstruct_flux_preserving_cube(
  seg,
  fill_mode = "na",
  return_residual = TRUE
)

reconstruct_cluster_cube() fills a region with a representative spectrum. reconstruct_flux_preserving_cube() distributes each summed regional spectrum over its assigned spaxels so summing the reconstruction over the region recovers the regional total.

Map–table correspondence

Region identifier k in seg$cluster_map corresponds to k in products$cluster_ids and to the matching row of the returned spectral matrices. Region identifiers remain categorical even when stored as integers.