SKYSURF Sky-SB Measurements
We give sky-SB measurements described in O'Brien et al. (2023) for WFC3/UVIS, WFC3/IR, ACS/WFC and WFCPC2 (in prep). If you use these measurements, please reference this paper.
Documentation for the csv files is given here.
We present sky-SB measurements using two algorithms: The Percentile-clip method and the ProFound Median method. These are described in detail in O'Brien et al. (2023). The most important difference between these methods is that the Percentile-clip method provides a lower limit to the sky-SB, while the ProFound Median method provides an upper limit. Please refer to the bottom of this page for a description of the methods.
WFC3/UVIS
Download all WFC3/UVIS for the Percentile-clip Method
Download all WFC3/UVIS for the ProFound Median Method
WFC3/IR
Download all WFC3/IR for the Percentile-clip Method
Download all WFC3/IR for the ProFound Median Method
ACS/WFC
Download all ASC/WFC for the Percentile-clip Method
Download all ASC/WFC for the ProFound Median Method
WFPC2 (in prep)
Description of Sky-SB Measurement Algorithms
We describe our sky-SB measurement algorithms here. Please refer to O'Brien et al. (2022) for more details.
The Percentile-clip (Per-clip) method follows three steps: 1) Divide each HST chip into sub-regions and calculate sky levels for each sub-region, 2) Flag sub-regions with unreliable sky measurements, then 3) Choose the sub-regions with the lowest sky-SB levels to determine the true sky-SB of the image. We begin by masking all pixels flagged (flag >0) in the Data Quality (DQ) array associated with the science image (SCI) in each flt or flc file. Next, we divide each image into sub-regions. WFC3/IR is the simplest case, where all images in our sample are 1014x1014 pixels. We split each WFC3/IR image into 39x39 pixel regions, for a total of 26x26=676 sub-regions per image. For WFC3/UVIS, a full-frame detector image has dimensions of 2051x4096 pixels. For this case, we make each sub-region 64x64 pixels, where remaining pixels will be added to the sub-region on the top-most and right-most sub-regions. For ACS/WFC, the detector has dimensions 2048x4096, and we follow the same binning technique as for WFC3/UVIS. Both WFC3/UVIS and ACS/WFC images include two CCD detectors (stored as separate science extensions), and we perform sky-SB measurements on each detector independently. Next, we calculate a sky level and sky rms level for each sub-region by iteratively clipping pixels. The median (standard deviation) value of the last iteration is saved as the sky-SB for that sub-region. Sub-regions are masked if they have too many masked DQ pixels or are identified as contianing a discrete object. The final sky-SB level of an image is the 5th-percentile of unmasked sub-regions.
The ProFound Median (Pro-med) method is similar to the Per-clip method with two main differences: 1) it calculates sky-SB values using ProFound (Roobotham et al. 2018) SKY maps, and 2) it assumes the true sky to be the median of reliable sub-regions. ProFound is unique in its ability to dilate objects, remove objects, and interpolate behind them to create a robust sky map. It is able to estimate the sky-SB by using a sliding box car filter on a grid, then interpolating between grid points to construct a sky-SB (SKY) map. We create ProFound SKY maps for all images in the SKYSURF database. The Pro-med method splits the ProFound SKY Map into sub-regions, and estimates the sky-SB using the same methods as for the Per-clip method. Because it uses a median, the Pro-med method tends to approach the true sky-SB value from above.
The Pro-med method gives sky-SB values that are on average 2% higher for ACS/WFC, 3% higher for WFC3/UVIS, and 0.8% higher for WFC3/IR than the Per-clip method. The difference in methods is due to the fact that the Per-clip method assumes the true sky-SB to be the darkest part of the sky (5th-percentile) while the Pro-med method assumes it to be the median.
We emphasize that both methods are robust and reliable depending on the science goal. Both methods can recover the input sky-SB to within 0.3% uncertainty. The Per-clip method provides the darkest possible sky-SB for every HST image. This method is useful for subtracting the sky-SB from images where large objects take up a majority of the field of view, or where the extended profiles of discrete objects is of concern. For SKYSURF purposes, the Per-clip method measures a lower limit to EBL and Diffuse Light signals. Although ProFound masks light from objects and their surroundings, excess light from large objects remains in SKYSURF ProFound SKY maps. We consider the Pro-med method to be less reliable for images where a larger portion of the field of view is contaminated by bright discrete objects. Based on tests on the simulated images, the Pro-med method is generally a better estimator of the sky-SB for flat, empty fields where the Per-clip method is negatively biased. For SKYSURF purposes, the Pro-med method thus provides an upper limit to EBL and Diffuse Light signals.