David V. Stark

As a member of the Hubble Advanced Camera for Survey (ACS) team, much of my work is devoted to studying the performance of the instrument, and helping observers make the best of their data. Below are some major projects I've been involved in.

Satellite Trail Identification and Masking

ACS has had its data occasionally photobombed by satellites since its installation in 2022. As satellites pass through the field of view during an exposure, they leave narrow streaks in the data due to the light they reflect (the figure on the right shows such a trail in a single exposure). To create the best final data products, satellite trails should be identified and masked.

One of my first tasks as a staff scientist at STScI was to work on a new and improved algorithm to automatically identify satellite trails. The workhorse method for our approach is technique called the "Median Radon Transform" (MRT) which is a slight modification to the standard Radon Transform. The MRT calculates the median of the data along every possible straight path through an image. Those paths mostly covering empty sky return a median of zero (or the sky level), even if there is the occasional star/galaxy along the path. Those paths that align with a persistent linear feature (like a satellite trail!) will return a median value greater than zero. We found that our method was 5-10 times more sensitive than the previously developed satellite identification software developed in our team, so definitely represented a substantial improvement, and it is able to find things that are not easily seen by eye. The code was added into the acstools Python package, and can be applied to most types of astronomical imaging data.

We used this new algorithm to create a database of satellite trails in ACS/WFC data spanning over 20 years, which we used to examine how the rate of satellite trails and their brightness evolved over time. The typical trail brightness has stayed pretty much constant, but the rate of trails has roughly doubled (see the figure below), from a trail every 3-4 hours in 2002, to a trail every 1-2 hours in 2022. About 5% of images were affected in 2002, while 10% were affected in 2022. So satellite trails are (unsurprisingly) getting more common, but it's worth remembering that it is highly unlikely they will completely ruin a science program. They are narrow, likely affecting less than half a percent of the pixels in an image (meanwhile cosmic rays affect 3-6 times that in a 1000s exposure). Furthermore, we can identify them, mask them, and since we take multiple images per field, the trail can be effectively removed from our final data products.

For more info, check out ACS ISR 2022-08. To use the new Python module, be sure to update to acstools v3.6.0 or greater. This work was also the subject of a press release!

Glint in ACS/WFC Imaging Data

In 2023-2024, I supervised a fantastic INS intern, Genevieve Mehra, who led a project to understand when glint (a narrow streaky artifact) occurs in ACS/WFC data (see example to the right). The assumption has been that glint is caused by light reflecting off material the gap between the WFC chips, but this had never been tested. By looking at where bright stars fall in exposures with and without glint, Genevieve found two regions on the ACS detectors where the presence of a bright star seemd to coincide with glint, one at a specific location on the chip gap, but the other suprisingly right in the middle of one of the WFC chips. Whatever the cause, her work has enabled observers to avoid placing bright stars in those regions, ensuring the highest quality data can be obtained. For more on Genevieve's work, see ISR 2024-05.

The impact of degraded charge transfer efficiency on image quality

Charge Tranfer Efficiency (CTE) refers to how well the charge stored in individual CCD pixels transfers to neighboring pixels during the read-out process. The harsh environment progressively damages the ACS Wide Field Channel (WFC) detectors, creating charge traps, and reducing their charge transfer efficiency. This leads to measured flux loss, especially for sources that are dim, on top of low backgrounds, and/or far from the CCD serial registers. If we want to continue making precise measurements using ACS/WFC, we need to thoroughly characterize how degraded CTE impacts measured brightnesses.

One approach to this study is to create mock data sets, allowing us to use very large data sets in highly controlled experiments. Making use of the CTE modeling routine () I created large mock data sets to study how point-source signal-to-noise ratios (SNR) and detectability are impacted by degraded CTE at different sky background levels. We found that an exposure sky background level of around 30 electrons per pixels is a "sweet spot" where SNR is maximized (increased sky backgrounds reduce CTE-related losses further, but at the expense of added sky Poisson noise). This is shown in the figure below, which displays SNR vs background for sources with different brightnesses. The thin lines are the true values while the thick lines show what we would get if CTE were perfect. We also studied if CTE leads to false detections that contaminate data sets. They can, but if you're using something that performs PSF fitting, you should be in good shape. If you're using something simpler, there are still basic cuts that can dramatically clean up data sets. For more info about this study, check out ISR 2024-02-

I am now pursuing a project using actual data to understand how degraded CTE is impacting extended sources, like galaxies. Stay tuned for results!