StarDICE IV: correcting visible photometry from atmospheric gray extinction using thermal infrared observations

The StarDICE IV paper presents a method that utilizes simultaneous thermal infrared observations to correct visible photometry for atmospheric gray extinction, enabling ground-based surveys to achieve sub-percent calibration precision even under non-photometric conditions.

The Gist

Here is an explanation of the paper "StarDICE IV," translated into simple language with creative analogies.

The Big Problem: The "Dirty Window" Effect

Imagine you are trying to take a perfect photograph of a star through a window. Usually, the window is clean, and the photo looks great. But sometimes, a thin layer of fog or a smudge of grease appears on the glass.

In astronomy, this "fog" is the Earth's atmosphere. Sometimes, clouds (even invisible ones) drift in front of the telescope. These clouds dim the starlight, making the stars look fainter than they really are. This is called extinction.

For most astronomers, if the sky isn't perfectly clear, they just cancel the observation and wait for a better night. But for massive projects like the Vera C. Rubin Observatory (which will map the entire sky), waiting for perfect weather is too slow. They need to take pictures every night, even when the sky is a bit cloudy.

The challenge? How do you know exactly how much the clouds dimmed the light so you can fix the photo later?

The Old Way vs. The New Way

• The Old Way: Astronomers used to look at the stars themselves to guess how cloudy it was. They would say, "Hmm, that star looks 10% dimmer than usual, so the clouds must be 10% thick."
  • The Flaw: This is like trying to guess how dirty a window is by looking at a single speck of dust on it. If the clouds are patchy (thick here, thin there), looking at just a few stars doesn't tell you the whole story.
• The New Way (StarDICE IV): Instead of guessing based on the stars, the team built a special "cloud detector" that looks at the sky in a different color of light.

The Solution: The Thermal "Thermometer"

The team attached a thermal infrared camera (a camera that sees heat) right next to their main telescope.

Here is the magic trick:

1. Visible Light: Clouds are almost invisible to our eyes and standard cameras. They look like clear air.
2. Infrared Heat: But those same clouds are made of water droplets and ice. They are cold, but they still radiate heat (thermal energy). To the thermal camera, the clouds glow brightly, like a warm blanket against the cold night sky.

The Analogy: Imagine you are in a dark room. You can't see the smoke from a candle with your eyes (visible light). But if you hold your hand near it, you feel the heat (infrared). The thermal camera is like a super-sensitive hand that can "feel" the clouds even when they are invisible to the eye.

How They Fixed the Photos

The team created a clever recipe to fix the photos:

1. The Setup: They pointed a standard telescope at a patch of stars and, at the exact same time, pointed the thermal camera at the same patch of sky.
2. The Measurement: The thermal camera measured how much "heat" (radiance) the clouds were emitting.
3. The Math: They used a computer model to calculate: "If the clouds are glowing this much with heat, they must be blocking this much visible starlight."
4. The Correction: They applied this math to the star photos. If a star looked dim, they used the thermal data to figure out exactly how much to "brighten" it back up to its true value.

The Results: Turning "Bad" Data into Gold

The paper tested this method over three months. Here is what they found:

• Before: On cloudy nights, the star measurements were all over the place. Some stars looked 0.64 magnitudes dimmer than they should have. It was like trying to read a book through a foggy window; the text was blurry and inconsistent.
• After: Using the thermal camera correction, the measurements became incredibly sharp. The error dropped from 0.64 down to 0.11.
• The Analogy: It's like taking a blurry, foggy photo and running it through a filter that instantly clears the fog. The stars that looked fuzzy and dim suddenly popped into focus, looking just as clear as they would have on a perfect, cloudless night.

Why This Matters

This technique is a game-changer for the future of astronomy.

• No More Waiting: Telescopes don't have to sit idle waiting for perfect weather. They can work through thin clouds, saving time and money.
• Better Maps: By correcting the "fog" in real-time, we can build more accurate maps of the universe.
• Finding the Unexpected: In time-domain astronomy (studying things that change quickly, like exploding stars), you can't wait for a clear night. This method ensures we don't miss a cosmic event just because a cloud drifted by.

In a Nutshell

The StarDICE team realized that while clouds are invisible to our eyes, they are very visible to heat sensors. By using a thermal camera as a "cloud ruler," they can measure exactly how much the clouds dim the starlight and mathematically "erase" the clouds from the data. This turns bad weather nights into productive observation nights, allowing us to see the universe more clearly than ever before.

Technical Summary

Here is a detailed technical summary of the paper "StarDICE IV: correcting visible photometry from atmospheric grey extinction using thermal infrared observations."

1. Problem Statement

Ground-based astronomical surveys, such as the Vera C. Rubin Observatory's Legacy Survey of Space and Time (LSST), require photometric calibration stable to the sub-percent level. A major obstacle to achieving this precision is grey extinction caused by thin, undetected clouds (cirrus).

• The Challenge: These clouds are often transparent in the visible spectrum but opaque in the thermal infrared (8–13 µm). They induce extinction that varies spatially and temporally on scales smaller than typical telescope fields of view and exposure times.
• Limitations of Current Methods: Standard mitigation involves discarding non-photometric nights or using dense stellar fields to model extinction residuals (e.g., Forward Global Calibration Method). However, these methods struggle when stellar density is low, when extinction is high (>1.5 mag), or when cloud structures are complex and non-uniform.
• Goal: The StarDICE experiment aims to calibrate standard stars to the millimagnitude (mmag) level. To do this, it must correct for variable atmospheric effects without discarding data, even under non-photometric conditions.

2. Methodology

The authors propose a novel method that corrects optical photometry by simultaneously measuring atmospheric down-welling radiance using a thermal infrared (IR) camera. The approach relies on the physical correlation between visible cloud extinction and IR thermal emission.

A. Physical Basis

• Grey Extinction vs. IR Radiance: Thin cirrus clouds emit thermal radiation in the LWIR (8–14 µm) due to their temperature (200–300 K). The paper establishes a relationship where the visible magnitude excess ($\Delta m$) due to cloud extinction is proportional to the IR radiance excess ($\Delta L$) above the clear-sky background.
• Forward Modeling: To isolate the cloud signal, the system must subtract the clear-sky atmospheric radiance. This is achieved using:
  • Radiative Transfer Simulations: Using libradtran to simulate clear-sky down-welling radiance based on real-time atmospheric parameters.
  • Ancillary Data: Inputs include barometric pressure, precipitable water vapor (PWV) from GNSS, ozone columns, and aerosol optical depth.
  • Synthetic Photometry: Using Gaia DR3 stellar parameters (effective temperature, gravity, metallicity) and synthetic spectra (AMBRE/MARCS models) to calculate the expected chromatic extinction and instrumental throughput for each star.

B. Instrumental Setup

The prototype system at the Observatoire de Haute-Provence (OHP) consists of:

1. Optical System: A 72 mm apochromatic refractor with a ZWO ASI183MM CMOS camera (r-band filter), providing a 2.18° × 1.46° field of view.
2. Thermal IR System: A FLIR Tau2 3 radiometric camera (8–14 µm) with a 60 mm lens, providing a 10.37° × 8.30° field of view. It captures images at ~8.33 Hz.
3. Alignment: The optical and IR fields are aligned via coordinate transformation using the Moon as a common reference.
4. Environmental Monitoring: A weather station and a GNSS receiver provide real-time PWV and pressure data for the radiative transfer models.

C. Analysis Pipeline

1. Data Collection: Simultaneous optical (20s exposures) and high-cadence IR imaging.
2. Photometry: Forced aperture photometry is performed on Gaia DR3 stars. Chromatic extinction is corrected using the forward model.
3. Radiance Extraction: The IR radiance at the position of each star is extracted from the thermal images. The clear-sky radiance (simulated) is subtracted to find the radiance excess ($\Delta L$).
4. Model Fitting: A grey extinction model is fitted for each image using a training subset of stars. The model relates the measured grey extinction ($\Delta m$) to the radiance excess ($\Delta L$).
   • Physical Model: $\Delta m = -\frac{1}{B} \ln(\frac{A + C - \Delta L}{A})$ (non-linear).
   • Linear Model: $\Delta m = A \times \Delta L + B$ (used when extinction is low or data is sparse).
5. Correction: The fitted model is applied to the test set to correct the instrumental magnitudes.

3. Key Contributions

• First Quantitative Application: This is the first study to quantitatively apply thermal infrared imaging to correct grey extinction in optical photometry on a star-by-star basis.
• Independent Measurement: Unlike methods relying solely on stellar density (which fail in sparse fields), this method uses an independent physical measurement of the atmosphere (IR radiance).
• No Spatial Assumptions: The method does not assume a smooth spatial structure of extinction (unlike 2D polynomial fits), making it robust against complex, patchy cloud structures.
• Hybrid Approach: It successfully combines radiative transfer physics, Gaia stellar catalogs, and machine learning-style fitting to separate chromatic and grey extinction components.

4. Results

The method was tested on two fields (BD + 28 4211 and HD 210955) over three months, covering diverse atmospheric conditions.

• Correction Accuracy:
  • For exposures with significant grey extinction (up to ~3 mag), the mean absolute error (MAE) improved from 0.64 mag to 0.11 mag.
  • Temporal extinction variations were reduced to 0.025 mag per source.
  • The dispersion of corrected magnitudes for test stars was reduced by a factor of 2 to 5 compared to uncorrected data.
• Model Performance:
  • Group 1 (~5% of images): High extinction; corrected using the non-linear physical model.
  • Group 2 (~48% of images): Low-to-moderate extinction; corrected using a linear model.
  • Group 3 (~47% of images): Negligible extinction; no correction applied (noise dominated).
• Resolution: The method produces extinction maps with a resolution of 2 arcminutes (limited by the IR camera pixel scale).
• Robustness: The technique successfully recovered data under non-photometric conditions, achieving precision comparable to data taken under photometric conditions. Even under severe extinction (up to 4 mag), the RMS scatter remained around 0.134 mag.

5. Significance and Future Outlook

• Survey Efficiency: This technique allows large-scale surveys (like LSST) to utilize data from "non-photometric" nights, significantly increasing telescope duty cycles and data volume without sacrificing calibration precision.
• Transient Astronomy: It is particularly valuable for time-domain astronomy where transient objects may only be observed once under variable conditions.
• Limitations & Improvements:
  • Current resolution is limited by the IR camera pixel size (~2 arcmin). A larger lens could improve this.
  • The study used a single optical filter (r-band); future work will verify the "grey" nature of extinction across multiple bands (ugrizy).
  • Thermal noise from the environment remains a challenge; future setups should use roll-off-roof buildings to minimize thermal fluctuations.
• Conclusion: The StarDICE IV experiment demonstrates that combining thermal infrared radiometry with forward atmospheric modeling is a viable, high-precision solution for correcting grey extinction, paving the way for more robust ground-based photometric surveys.