🔭 astrophysics

The Hubble Advanced Spectral Product (HASP) Program

The Hubble Advanced Spectral Products (HASP) program is an automated initiative within the MAST archive that generates high-quality, dynamically updated coadded spectra from COS and STIS instruments to significantly enhance signal-to-noise ratios and wavelength coverage for astronomers.

Original authors: John Debes, Ravi Sankrit, Travis Fischer, Elaine Frazer, Alec Hirschauer, Kate Rowlands, Matthew Burger, Robert Swaters, Robert Jedrzejewski, Sierra Gomez, Leonardo Dos Santos, Svea Hernandez, Lauren

Published 2026-05-11

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: John Debes, Ravi Sankrit, Travis Fischer, Elaine Frazer, Alec Hirschauer, Kate Rowlands, Matthew Burger, Robert Swaters, Robert Jedrzejewski, Sierra Gomez, Leonardo Dos Santos, Svea Hernandez, Lauren Miller, Anna Payne, Marc Rafelski, Thomas Wevers, Sara Anderson, Tom Bair, Kathryn Bello, Joleen Carlberg, Brian Charlow, Andrew Cortese, Nadia Dencheva, Tracy Ellis, Ben Falk, Scott Fleming, Peter Forshay, Syed Gilani, Patty Hall, Tim Kimball, Talya Kelley, Richard Kidwell, Jenn Kotler, Aiden Kovacs, Bethan James, Christopher Rahmani, David Rodriguez, Julia Roman-Duval, David Soderblom, Lisa Sherbert, Dan Welty, David Wolfe

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the Hubble Space Telescope as a massive, decades-long library of cosmic stories. For years, astronomers have been able to read individual pages (spectra) from this library, but if they wanted to read a whole chapter or a complete book, they had to do the heavy lifting themselves: finding the pages, checking if they were torn or smudged, and manually taping them together.

The Hubble Advanced Spectral Products (HASP) program, described in this report, is like a new, automated librarian who does all that work for you. It takes thousands of scattered, individual observations of the same star or galaxy and stitches them together into one perfect, high-quality "master copy."

Here is how HASP works, broken down into simple concepts:

1. The Problem: A Library of Scattered Pages

Hubble has two main "cameras" for reading light: COS and STIS. Over the years, they have taken millions of pictures of the universe. However, the standard files Hubble sends down are like individual snapshots.

Sometimes, a star was observed with different camera settings (called "gratings" or "central wavelengths").
Sometimes, the same star was observed multiple times over different visits.
The standard files don't automatically combine these. If an astronomer wanted a complete picture, they had to manually download every single file, check if they were good, and combine them. It was slow, tedious, and prone to human error.

2. The Solution: The Automated "Master Chef"

The HASP program is an automated system that acts like a master chef preparing a complex dish. It doesn't just throw ingredients together; it follows a strict recipe to ensure the final meal is delicious.

Step 1: The Shopping List (Collation)
The system automatically finds every relevant "ingredient" (spectra) in the Hubble archive for a specific target. It looks for the freshest data, including any new observations or re-calibrated old data.

Step 2: The Quality Control (Filtering)
This is the most critical step. The system acts like a strict food inspector. It throws out any "bad ingredients" that could ruin the dish.

The "Guide Star" Check: If the telescope lost its focus or the target drifted off-center (like a chef dropping a pan), that data is discarded.
The "Flux" Check: The system compares the brightness of each ingredient against the others. If one piece of data is significantly dimmer than the rest (perhaps because the telescope pointed at the wrong spot or the star flickered unexpectedly), it gets rejected.
The "Mode" Check: It automatically removes data from unsupported camera modes or settings that don't play well together.

Step 3: The Cooking (Coaddition)
Once the good data is selected, HASP combines them.

Same Settings: If you have 50 photos of a star taken with the same camera lens, HASP stacks them. This is like taking 50 blurry photos of a moving car and stacking them to create one crystal-clear image. The result is a much sharper picture with a higher Signal-to-Noise Ratio (SNR). In simple terms, the "signal" (the star's light) gets louder, and the "noise" (static) gets quieter.
Different Settings: Hubble often uses different lenses (gratings) to see different colors of light. HASP can stitch these together. Imagine taking a photo of a landscape with a wide-angle lens and another with a telephoto lens, then seamlessly joining them into one panoramic view. HASP does this for light, creating a continuous spectrum from ultraviolet to visible light.

3. The Result: A "Quick-Look" and a "Deep Dive"

HASP produces two main types of products for astronomers:

The "Visit" and "Program" Coadds: These are the high-quality, stitched-together spectra for specific observation sessions or entire research projects. They are ready for serious scientific analysis.
The "Abutted" Spectrum: This is a "quick-look" version. Since different lenses have different resolutions (some are sharp, some are blurry), HASP doesn't try to force them to blend perfectly. Instead, it "abutts" them—meaning it joins them end-to-end like puzzle pieces. This gives the astronomer a quick glance at the entire range of light the telescope captured, showing the full story of the object's energy.

4. Why This Matters

It's Automatic: As Hubble gets new data or better calibration rules, HASP automatically re-runs the process. You don't have to wait for a human to update the files; the "master copy" is always the best version available.
It's Accessible: The team didn't just build the machine; they gave the keys to the public. They released the computer code (scripts) and "cookbooks" (Jupyter notebooks) so that any astronomer can run their own custom combinations if they need to tweak the recipe for a specific scientific question.
It's Reliable: The team tested this system against thousands of known stars (like standard light bulbs in the sky). They found that in over 90% of cases, the combined data matched the original data with incredible accuracy (within 5%), proving the "stitching" didn't introduce errors.

5. The Caveats (When the Chef Says "No")

The report is honest about what HASP cannot do. It is designed for compact, steady objects (like a single, non-flickering star).

Variable Stars: If a star changes brightness wildly from day to day, HASP might reject the data because the "ingredients" don't match.
Extended Objects: If the target is a giant, fuzzy cloud of gas (like a nebula) and the telescope looked at it from different angles, the light might not line up perfectly. HASP might filter this out to avoid a "messy" result.
Moving Targets: Planets and asteroids move too fast for this automated stitching process at the program level.

Summary

In essence, the HASP program is a massive, automated quality-control and assembly line for Hubble's light data. It takes the messy, scattered raw data of the past 30 years, filters out the mistakes, and serves up clean, high-definition, combined spectra. It saves astronomers time, improves the quality of the data, and provides a foundation for future discoveries, all while giving the scientific community the tools to customize the process if they need to.

Technical Summary: The Hubble Advanced Spectral Products (HASP) Program

Problem Statement
The Hubble Space Telescope (HST) archives contain vast amounts of spectroscopic data from the Cosmic Origins Spectrograph (COS) and the Space Telescope Imaging Spectrograph (STIS). Historically, the default pipeline products for these instruments do not automatically combine spectra from different central wavelengths (CENWAVES), gratings, or apertures within a single program or visit. Consequently, users have been responsible for manually coadding these spectra to achieve higher signal-to-noise ratios (SNR) or broader wavelength coverage. This manual process is labor-intensive and prone to inconsistency. Furthermore, previous efforts to create high-level spectral products, such as the Hubble Spectroscopic Legacy Archive (HSLA) and the ULLYSES program, utilized specific algorithms that were not designed to be applied automatically across the entire archive or to support custom user-driven coaddition.

Methodology
The HASP program implements an automated, multi-stage workflow within the Mikulski Archive for Space Telescopes (MAST) to generate robust, combined 1-D spectra. The process consists of three primary phases:

Collation and Filtering: An automated workflow collects relevant x1d and sx1 FITS files from the archive. A rigorous multi-stage filtering process removes data that may compromise coaddition quality. This includes:
- Excluding exposures associated with AlertObs reports or failed target acquisitions.
- Removing special calibration programs (e.g., focus changes) and unsupported modes (e.g., STIS PRISM, COS BOA).
- Checking header keywords (EXPFLAG, EXPTIME, PLANTIME, FGSLOCK, POSTARG1/2, PATTERN1, P1 PURPS) to identify guiding failures, shutter issues, or non-standard observing patterns.
- Filtering out moving targets (Solar System objects) at the program level due to expected variability.
Flux Validation: A critical step involves comparing input spectra against a preliminary coadd to detect subtle failures, such as target acquisition errors or extraction issues. The algorithm calculates the median deviation of flux weighted by uncertainty. If an input spectrum's flux is significantly lower than the coadd (based on a threshold $C_{thresh}$ set to -50 for default products), it is rejected. This iterative process continues until no further spectra are rejected, ensuring that only consistent flux levels are combined.
Coaddition and Abutment:
- Coaddition: For spectra from the same grating, HASP performs a weighted average of flux measurements within output wavelength bins. Weights are determined by throughput (net count rate). This handles the combination of different CENWAVES within a grating and the merging of adjacent spectral orders in echelle data.
- Abutment: To create a single spectrum covering multiple gratings or instruments, HASP "abutts" (joins) spectra at specific transition wavelengths. A priority list (Table 1) governs which grating data is selected in overlapping regions, generally prioritizing higher spectral resolution. This results in a discontinuous spectral sampling at the transition but provides a seamless overview of the spectral energy distribution.

The code is written in Python and is publicly available via the Space Telescope Science Institute's GitHub repositories, allowing users to perform custom coadditions with adjustable parameters.

Key Contributions

Automated Archive Processing: HASP is the first system to automatically generate visit-level and program-level coadded spectra for nearly every COS and STIS spectrum in the archive.
Multi-Stage Filtering: The implementation of a robust filtering and flux-checking pipeline ensures that the resulting products meet strict quality standards, removing data from failed observations or inconsistent flux levels.
Customizability: Unlike previous legacy archives, HASP provides the source code and Jupyter notebooks to the user community. This allows astronomers to disable filters, adjust flux thresholds, or apply flux scaling for variable or extended sources, facilitating tailored scientific analysis.
Data Product Structure: The output products include a science extension with coadded spectra, noise, and SNR estimates, alongside a provenance table (metadata extension) detailing the input files, observation times, and aperture usage.

Results and Validation
The HASP team conducted extensive testing against the full COS and STIS archives to validate the accuracy of the coadded products:

Flux Accuracy: For datasets with sufficient SNR (at least 5 bins with SNR > 20), approximately 91% of comparisons between input x1d spectra and HASP coadds showed a median residual and standard deviation within the required 5% limit.
Wavelength Accuracy: Cross-correlation analysis demonstrated that 97% of comparisons showed wavelength offsets of less than ±1 pixel.
Standard Star Verification: Comparisons with CALSPEC models for standard white dwarfs (e.g., GD71, G191B2B, WD0308-565) revealed absolute flux consistency better than 0.032% across the spectrum.
SNR Enhancement: The program successfully demonstrated significant SNR improvements. For example, in Program 16196, individual spectra with an SNR of ~3 at 1200 Å were coadded to produce a program-level product with an SNR of 56.8.
Coverage: The abutment capability successfully combined data from multiple gratings (e.g., STIS G230L, G430L, G750L) to create continuous spectral energy distributions from 1800 Å to 10,000 Å.

Significance
The paper claims that HASP serves as a vital resource for astronomers by providing immediate, high-quality access to combined spectroscopic data without the need for users to reinvent coaddition techniques. By automating the combination of different CENWAVES and instruments, HASP significantly enhances wavelength coverage and SNR. The project emphasizes transparency and user empowerment; by making the coaddition code and diagnostic tools (such as log files and Jupyter notebooks) publicly available, HASP enables the community to perform custom analyses for specific science cases, including those involving variable or extended sources that may not fit the default automated criteria. The authors position HASP as the foundational step toward a future, updated Hubble Spectroscopic Legacy Archive (HSLA) that will further expand cross-program target-level coaddition.