BACKGROUNDS

OVERVIEW

To derive calibrated photometry, robust background subtraction is a necessary and familiar step. Source photometry is affected by a number of backgrounds contributors, both celestial and instrumental backgrounds. During nominal science data collection, pixels designated as background are measured in addition to target pixels. These data are used to correct the photometry within the Photometric Analysis pipeline module.

In addtion to correction of source photometry, measures of the spatial and temporal background flux are by themselves an important science topic, with a long and rich pedigree (e.g., the CMBR). Backgrounds at many wavelengths are the sum of a number of processes; peeling back each layer has provided valuable insight onto cosmic evolution [ultraviolet: Bowyer, AnnRev 29, 59 (1991); Henry, AnnRev 29, 89 (1991); infrared: Hauser & Dwek, AnnRev 39, 249 (2001); soft x-rays: McCammon & Sanders, AnnRev 657, (1990)]. Kepler provides a continuous measure of the celestial background in a wide optical/near-IR band over a 100 square degree FOV for 3.5 years nominal mission length.

Due to Kepler's unusual operational strategy, the method applied for background removal differs from the most common approach used in astronomical applications. Backgrounds are usually estimated from a annulus of pixels immediately adjacent to the source, scaled to a common area. Kepler does not use this approach, rather a separate set of background apertures are collected across the focal plane, and a background measure derived from these pixels. GOs should note that the pipeline does not use pixels within the source aperture to measure the local background.
 

COMPONENTS

Celestial backgrounds arise from a number of sources, both from spatially smooth, diffuse light which affects all pixels in the aperture, and from transient events, which affect a few pixels.

• Zodical light, produced by sunlight scatered from dust in the ecliptic plane.

• Diffuse scattered starlight, produced by dust in the Galaxy. Due to Kepler's large FOV, the Galactic component shows a spatial gradient, increasing at lower galactic latitudes.

• Unresolved starlight. Given the 4x4 arcsecond dimensions of the pixels, some light in the aperture arises from faint field stars. As with the diffuse Galactic emission, the contribution from unresolved starlight increases with decreasing Galactic latitude.

• Cosmic ray impacts which corrupt individual pixels. The pipeline software flags and removes cosmic ray events from the pixel counts, within the Photometric Analysis module. Each cosmic ray event is replaced with a temporally local average of the pixel's time series without the cosmic ray pixel events.

• Surrounding sources. A type of background which will affect source photometry are the residual wings of the PSFs produced by nearby stars, which may overlap the PSF of the target. Observers may visualize this effect by considering an image taken with poor seeing (equivalently poor focus). Faint PSF wings of surrounding sources will fall across the target, requring either PSF-fitting photometry or some form of aperture correction. A particular challenge for Kepler observers is that the user does not receive an entire image of the surrounding field, just the target pixel image. Stars outside of the this aperture may affect the target photometry, even though those stars were not actually measured. Also remember that nearby sources may be variable, producing a time-dependent background.

  A correction for contaminating flux in the source aperture produced by surrouding sourcs is applied within the PDC pipeline module - a single valued subtraction, termed the crowding metric. In ideal situations the PSFs of neighboring stellar sources would not change over time, expect for possible intrinsic source variabilty. Observers should be aware that spacecraft operations may induce changes in the source PSFs, through focus changes, and spacecraft motions (jitter and drift). Motion of the source center during an observing season, even at the millipixel level, will induce variations in the contaminating flux, introducing small levels of noise into thelight curve. In preparing proposals, GO should choose sources as isolated as possible, and be aware of the constraints for achieving excellent time series photometry under strong crowding situations.

Instrumental backgrounds include the detector bias level (also termed black level), which is removed in the CAL pipeline module, some anticipated issues, such as scattered light, unexpected electronic issues discovered during pre-flight characterization of the detectors, and some features seen during early flight operations, e.g.,v"argabrightening", an anomalous full-field illumination, whose origin is under investigation by the instrument team (Jenkins etal 2010). Spatially varying backgrounds produced by the detector electronis are full described in the Kepler Instrument Handbook (KIH). The Kepler project is investing considerable effort to identify, assess, and develop corrections for instrumental signatures discovered during the early phase of data collection. Additional details on the spatially-varying image artifacts, and their corrections can be found on the GO Data Artifacts page.
 

MEASUREMENT & REMOVAL

Kepler constructs a background flux map using a set of target pixels specifically assigned for this purpose. Background "targets" are small (nominally 2 x 2 pixels) postage stamp images which measure the background signal in the long cadence observations. Since backgrounds must be estimated and removed from all observed sources in the FOV, a method was adopted to interpolate background values from all targets pixels in a channel using the background apertures in that channel. A maximum of 1125 background apertures and 4,500 pixels (~4 pixels per target) are allocated for each of the 84 output channels. These background targets are selected to optimize a 2D polynomial representation of the background flux distribution, derived separately for each channel.

For robust fitting, background apertures should be uniformly distributed on each CCD array. To mitigate edge effects, more background apertures are positioned at the frame edges. During target management prior to each quarter, background pixels are selected, avoiding stars and locations affected by charge bleeds.

Distribution of background apertures in a typical Kepler CCD frame. Spacing of background pixels is increased along the frame edge to improve the fit to the background.

Background pixels are calibrated in the same manner as source pixels in the CAL pipeline module. Background channel maps are generated in PA; interpolated background values are then subtracted from the pixel values prior to aperture photometry.

A measure of the celestial background in the Kepler FOV is provided by VanCleve in Data Release Notes 5. The figure below shows the background flux time series for the Q0+Q1 observing season, derived from the reprocessed data released on 15-June-2010. Time series are shown for the full focal plane, and for channels closest to and furthest from the Galactic plane. From this plot, a few items are evident:

• The temporal and FOV-averaged value background value is ~2.7 × 10⁻⁵ electrons per cadence. This value corresponds to the expected signal from a star with Kp ~ 18.3.

• The difference in background counts produced by viewing perpendicular to the Galactic plane is about a factor of 1.8.

Background time series for observing quarters Q0 + Q1. The curves show average over all mod.outs (CCD channels), the modules furthest from (mod.out 2.4 = channel 4) and nearest to (mod.out 24.4 = channel 84) the Galactic plane. The vertical offset is produced by the spatially-dependent Galactic emission; the horizontal trend is caused by variation in the zodical light. The narrow spikes visible in all 3 curves are Argabrightening events. Observations affected by these diffuse brightenings are tabulated in the release notes.

GUEST OBSERVER ISSUES

What should GOs be concerned about regarding backgrounds ?? Background subtraction is normally a user-driver reduction step, not a pipeline function. The current assessment is that the pipeline background corrections do not produce significant noise or errors in the extracted light curves. A few steps which GOs can take to examine their specific situations:

1. Inspect the FOV around sources using a full-frame image. Look for nearby bright sources which may affect your target, especially the potential for unknown variable stars. Isolated sources should present no issues, whereas the treatment of the background by the pipeline for crowded sources may not be optimal. Also look for saturated stars in the same (+/-1) column to check for unfortunately placed charge bleeds.

2. Compare the Kepler view of your sources with a DSS image. The later have a resolution of about 1.5 arcseconds, and may show faint sources in the source apertures.

3. When the target pixels images become available for GO use, users may wish to contruct their own background estimates using the ancillary background aperture photometry. Background values are not presently supplied with the light curves, and can only be roughly estimated on a per pixel basis using a full-frame image from the same observing season.