DESI-RCA: A Residual-Covariance Advisory and Traceability Tool for DESI Spectral Products

The software was developed and tested on a standard Windows PC equipped with an Intel Core i7-14700 processor and 64 GB of RAM. No HPC, cloud computing, or specialized hardware is required.

Associated software, manuscript, and indexing

The software associated with the related manuscript is explicitly named DESI-RCA. DESI-RCA is a post-pipeline residual-covariance advisory and traceability tool for DESI spectral products.

DESI-RCA is archived on Zenodo under DOI:
https://doi.org/10.5281/zenodo.20740564

DESI-RCA is indexed in the Astrophysics Source Code Library as:
ASCL identifier: ascl:2606.002
ASCL code record: https://ascl.net/code/v/4759
Bibcode: 2026ascl.soft06002M

Preferred software citation:
F. Javier Martínez Sánchez, DESI-RCA: Residual-covariance advisory and traceability tool for DESI spectral products, Zenodo, DOI: https://doi.org/10.5281/zenodo.20740564

Methodological background

DESI-RCA is part of a broader residual-covariance diagnostic line developed from the methodological framework presented in the peer-reviewed article:

F. Javier Martínez Sánchez, Forensic Characterization of Correlated Residuals in Large Spectroscopic Pipelines: A Case Study of SDSS/BOSS Data, Astronomy and Computing.
DOI: https://doi.org/10.1016/j.ascom.2026.101155

Please use the DOI as the permanent citation link for the peer-reviewed methodological article.

Description

DESI-RCA is a local post-pipeline residual-covariance advisory and traceability tool for DESI spectral products. It provides an author- and reviewer-facing diagnostic workflow for inspecting residual independence, stacking, coaddition, simplified uncertainty propagation, and residual-covariance assumptions in selected DESI spectral subsets.

Large spectroscopic surveys such as DESI produce highly processed data products that are often stacked, coadded, filtered, compared, or reused in downstream analyses. These workflows may rely, explicitly or implicitly, on assumptions that residual fluctuations are effectively independent or sufficiently well characterized. DESI-RCA provides a practical way to inspect those assumptions for selected DESI spectral products.

The tool prepares controlled spectral matrices from DESI coadd products, constructs residual diagnostic products, computes spectrum-by-spectrum residual-correlation diagnostics, evaluates null controls, and issues conservative advisory decisions focused on stacking caution and residual-covariance risk.

DESI-RCA does not replace the DESI pipeline, certify DESI data products, validate cosmological conclusions, claim pipeline failure, or perform astrophysical interpretation. Its purpose is narrower and methodological: to make residual-covariance structure visible, reproducible, and auditable for selected DESI spectral subsets.

Release content

This v0.2.0 beta-candidate release includes the DESI-RCA source code, documentation, configuration files, tests, common RCA command-line workflow, DESI-specific ingestion layer, guided workflow utilities, diagnostic routines, report-generation tools, file manifest, SHA-256 release checksums, and citation metadata.

DESI-RCA provides:

- controlled DESI coadd selection and local ingestion;
- preparation of controlled spectral matrices from DESI products;
- residual construction for statistical inspection;
- residual-correlation diagnostics and eigenspectrum summaries;
- null-test controls for residual-covariance context;
- metadata-aware and fiber-aware group diagnostics;
- conservative advisory decisions focused on stacking caution;
- PDF reports, visual diagnostic summaries, machine-readable manifests, SHA-256 hashes, and report-package verification;
- a common RCA command-line workflow aligned with the broader RCA software family;
- a guided menu layer for scientist-facing operation without assembling long commands manually.

Large DESI FITS products and heavy local run outputs are not included in the archived release.

Scientific and operational boundaries

DESI-RCA is not a replacement for the DESI reduction pipeline, not an official DESI certification system, and not a cosmological or astrophysical inference tool. It does not validate or invalidate DESI physical results, does not claim DESI pipeline failure, and does not redistribute DESI FITS products.

The generated report packages may be useful as supplementary diagnostic material accompanying analyses that rely on stacked spectra, coadded products, or simplified independence assumptions. Such reports do not certify the scientific result; they provide a traceable record of one statistical diagnostic relevant to the analysis.

Planned roadmap — DESI-RCA v2

DESI-RCA v2 is planned as a focused extension of the current v1 core, incorporating technical recommendations received through the DESI Data Q&A.

Planned developments include:

• explicit use of IVAR and MASK in residual-covariance diagnostics;
• comparison of nominal and mask-aware analyses;
• use of OBJTYPE='sky' spectra as a DESI-specific control/null-test population;
• preferential analysis of tile cumulative, non-coadded spectra products when investigating residual structure;
• reduced reliance on coadded/healpix products where data combination may obscure the origin of correlated residuals;
• assessment of whether detected dependence is consistent with the reported uncertainty model.

The existing DESI-RCA framework for residual covariance, N_eff, VIF, null tests, provenance, reproducibility, and reporting will be preserved. Resolution information may be considered where useful, but is not expected to be a primary component of the v2 analysis.

The goal is to refine DESI-RCA using DESI-specific quality information and more appropriate spectroscopic products while maintaining its diagnostic-only and traceable approach.

Independence statement: The recommendations were provided in response to an independent technical inquiry and do not imply that DESI commissioned, endorsed, validated, approved, or co-developed DESI-RCA.

Status: planned functionality for a future v2; these capabilities are not part of the current v1 release unless explicitly stated.