Rubin Science Platform Overview

• Rubin Science Platform is an integrated interactive environment that enables reproducible, containerized analysis of multi-wavelength, joint survey data.
• It provides direct access to pixel-level images and catalogs, facilitating advanced tasks like deblending, calibration, and multi-epoch profile fitting.
• Its scalable architecture and unified API support high-impact cosmological research by ensuring consistent, reproducible computational setups.

The Rubin Science Platform (RSP) is the Vera C. Rubin Observatory’s integrated, web- and container-based interactive analysis environment, developed to facilitate advanced data exploration, joint survey science, and pipelined processing for astronomical studies at petascale. Within multi-survey frameworks—most notably the Joint Survey Processing (JSP) effort for Euclid, Rubin/LSST, and Roman (WFIRST)—the RSP is architected as the Tier 3 “Science Platform,” providing a scalable, reproducible interface for users to manipulate concordance multi-wavelength images and catalogs, perform novel analysis, and deploy custom pipelines directly on the flagship datasets (Chary et al., 2020).

1. Architectural Integration and Roles

The RSP serves as the user-facing analytical tier within JSP’s tiered architecture. It is constructed atop containerized environments (using Singularity or analogous technologies), ensuring that a consistent and reproducible computational setup is available regardless of execution context—data center, localized HPC node, or distributed cloud platform. Its core roles include:

• Providing direct access to jointly processed (concordance-frame) pixel-level images and catalogs from Euclid, Rubin/LSST, and Roman,
• Enabling both standard and user-defined calibration algorithms (e.g., astrometric/photometric recalibration; deblending; source extraction) tested across early JSP tiers,
• Supporting interactive exploration, data visualization, and pipeline development within a unified API and notebook environment.

The RSP thus becomes the prime online portal for scientific analysis, ensuring seamless integration between backend joint processing and front-end community engagement (Chary et al., 2020).

2. Multi-Wavelength Image and Catalog Handling

The RSP enables coherent access to pixel-aligned, astrometrically and photometrically recalibrated datasets produced by the JSP pipeline. Specific capabilities include:

• Application of source detection, star–galaxy separation, and simultaneous multi-band flux fitting,
• Implementation of prior-based deblending: leveraging the high spatial resolution of Roman and Euclid as priors to extract deconfused photometry from LSST’s high-S/N but seeing-limited images,
• Multi-epoch, multi-wavelength profile fitting: for variable/transient object analysis, proper motion studies, and moving object (e.g., solar system) detection,
• Out-of-the-box support for tasks such as coadd generation, non-sidereal stack production (central to solar system science), and fake source injection to quantify completeness.

Within this context, the RSP enables transformation of heterogeneous, instrument-specific data into unified, science-ready catalogs suitable for precision cosmology and time-domain astronomy (Chary et al., 2020).

3. Algorithmic Implementation and Optimization

Concordance image and catalog analysis on the RSP employs rigorous statistical model-fitting. At the pixel level, this is formalized as minimization of the global chi-square:

$$\chi^2 = \sum_i \left( \frac{I_i - M_i}{\sigma_i} \right)^2,$$

where $I_i$ denotes the observed intensity in pixel $i$, $M_i$ is the model value from a joint fit across surveys, and $\sigma_i$ is the per-pixel uncertainty. For deblending,

$M_i = \sum_j f_j P_{ij},$

where $f_j$ is the flux of source $j$ and $P_{ij}$ encodes the PSF-weighted contribution of each source to pixel $i$.

These models are iteratively optimized using containerized algorithms accessible via the RSP’s interactive and batch environments. Researchers can modify algorithm parameters, substitute custom models, or adapt new joint photometry and deblending schemes, leveraging the reproducible software stack at all participating facilities (Chary et al., 2020).

4. Reproducibility and Containerization

RSP’s operational reliability hinges on strong reproducibility guarantees:

• All user-facing computation occurs within containers, guaranteeing consistent software states irrespective of the underlying hardware or facility,
• Core tasks—algorithm development, data ingest, visualization—utilize the same API and data formats,
• The RSP can be accessed remotely, supporting flexible deployment from the Rubin Observatory Data Center to external HPCs with identical analysis semantics.

This reproducibility not only solidifies scientific conclusions but also democratizes joint survey science by removing computational environment disparities (Chary et al., 2020).

5. Scientific Impact: Cosmological and Astrophysical Discoveries

Use of RSP within JSP directly enhances the accuracy of cosmological parameter inference by reducing photometric and astrometric systematics and improving deblended photometry. Specific impacts include:

• Lower scatter in photometric redshifts and lensing shear calibrations, leading to narrower constraints on the dark energy equation of state ($w$, $w'$) and superior systematics control in weak gravitational lensing analyses,
• Enablement of innovative investigations such as high-cadence solar system object tracking (exploiting non-sidereal stacking), time-domain cosmology (supernova delay measurements, microlensing, reionization mapping), and dark matter substructure studies.

The RSP thus acts as a "virtual observatory" unifying ground-based and space-based data for advanced user-led scientific projects, elevating fidelity and breadth of astrophysical results (Chary et al., 2020).

6. Practical Deployment and Community Access

Deployment of the RSP within the JSP context ensures that the community has access to:

• Tiered access rights for raw, processed, and concordance data products,
• Interactive, browser-accessible environments supporting rich visualization, exploratory computing, and custom pipeline deployment,
• Documentation and APIs optimized for large-scale, collaborative projects and rapid algorithm development.

The structure inherently fosters collaborative, high-throughput science and allows rapid hypothesis testing on petabyte-scale datasets, centralizing next-to-data research for Euclid, Rubin/LSST, and Roman (Chary et al., 2020).

---

The Rubin Science Platform, as the user-facing scientific and computational environment embedded in Joint Survey Processing, is fundamental to enabling cutting-edge, cross-survey scientific analysis, reproducible research, and high-impact discoveries by integrating precision-calibrated multi-wavelength datasets with state-of-the-art algorithmic infrastructure.