The Pan-STARRS1 Surveys
Pan-STARRS1 has carried out a set of distinct synoptic imaging sky surveys including the Steradian Survey and the Medium Deep Survey in 5 bands (). The mean 5 point source limiting sensitivities in the stacked 3 Steradian Survey in are (23.3, 23.2, 23.1, 22.3, 21.4) respectively. The upper bound on the systematic uncertainty in the photometric calibration across the sky is 7-12 millimag depending on the bandpass. The systematic uncertainty of the astrometric calibration using the Gaia frame comes from a comparison of the results with Gaia: the standard deviation of the mean and median residuals ( ) are (2.3, 1.7) milliarcsec, and (3.1, 4.8) milliarcsec respectively. The Pan-STARRS system and the design of the PS1 surveys is described and an overview of the resulting image and catalog data products and their basic characteristics are described together with a summary of important results. The images, reduced data products, and derived data products from the Pan-STARRS1 surveys are available to the community from the Mikulski Archive for Space Telescopes (MAST) at STScI.
Subject headings:astronomical databases, catalogs, standards, surveys
The Panoramic Survey Telescope and Rapid Response System (Pan-STARRS) is an innovative wide-field astronomical imaging and data processing facility developed at the University of Hawaii’s Institute for Astronomy Kaiser et al. (2002, 2010). The Pan-STARRS1 Science Consortium (PS1SC) was formed to use and extend the Pan-STARRS System for a series of surveys to address a set of science goals and in the process the PS1SC continued the development of the Pan-STARRS System. An original goal the PS1SC set for itself was to insure the data would eventually become public.
This is the first in a series of seven papers that describe the Pan-STARRS1 Surveys, the data reduction techniques, the photometric and astrometric calibration of the data set, and the resulting data products. These papers are concurrent with and are intended to support the public release of the Pan-STARRS1 data products 111http://panstarrs.stsci.edu/ from the Barbara A. Mikulski Archive for Space Telescopes (MAST) at the Space Telescope Science Institute.
There are two Data Releases funded: Data Release 1, (DR1) containing the stacked images and the supporting database of the Steradian Survey, and Data Release 2 (DR2) containing all of the individual epoch data of the Survey. Further Data Releases will depend on the availability of resources to support them.
This Paper (Paper I) provides an overview of the fully implemented Pan-STARRS System, the design and execution of the Pan-STARRS1 Surveys, the image and catalog data products, a discussion of the overall data quality and basic characteristics, and a summary of scientific results from the Surveys.
Magnier et al. (2016b, Paper II) describes how the various data processing stages are organised and implemented in the Imaging Processing Pipeline (IPP), including details of the the processing database which is a critical element in the IPP infrastructure .
Waters et al. (2016, Paper III) describes the details of the pixel processing algorithms, including detrending, warping, and adding (to create stacked images) and subtracting (to create difference images) and resulting image products and their properties.
Magnier et al. (2016a, Paper IV) describes the details of the source detection and photometry, including point-spread-function and extended source fitting models, and the techniques for “forced” photometry measurements.
Magnier et al. (2016c, Paper V) describes the final calibration process, and the resulting photometric and astrometric quality.
Flewelling et al. (2016, Paper VI) describes the details of the resulting catalog data and its organization in the Pan-STARRS database. Huber et al. 2017 (in preparation - Paper VII) describes the Medium Deep Survey in detail, including the unique issues and data products specific to that survey. The Medium Deep Survey is not part of DR1.
Note: These papers are being placed on arXiv.org to provide crucial support information at the time of the public release of Data Release 1 (DR1). We expect the arXiv versions to be updated prior to submission to the Astrophysical Journal in January 2017. Feedback and suggestions for additional information from early users of the data products are welcome during the submission and refereeing process.
The paper is laid out as follows. In Section 2 of this paper we begin with an overview of the completed Pan-STARRS1 System, and a brief description of its associated subsystems: the Pan-STARRS Telescope #1, (PS1), the Gigapixel Camera #1 (GPC1), the Image Processing Pipeline (IPP), hierarchical database or Pan-STARRS Products System (PSPS), and the Science Servers: the Moving Object Pipeline (MOPS), Transient Science Server (TSS), Photo-Classification Server (PCS). Section 3 describes the various Pan-STARRS1 Surveys and their characteristics; the details of the observing strategy and the resulting impact on the time sampling and survey depth as a function of position on the sky. Section 4 provides a summary of the Pan-STARRS1 data products. Section 5 summarizes the overall astrometric and photometric calibration of the surveys. Section 6 provides an overview of the features and characteristics of the 3 Survey. Finally, a summary of the legacy science of the PS1 Science Consortium and a brief discussion of the future of Pan-STARRS is provided in Section 7.
2. The Pan-STARRS System
2.1.1 The Pan-STARRS Project
The Panoramic Survey Telescope and Rapid Response System (Pan-STARRS) is an innovative wide-field astronomical imaging and data processing facility developed at the University of Hawaii’s Institute for Astronomy Kaiser et al. (2002, 2010). Approximately 80 percent of the construction and development funds came from the US Air Force Research Labs (AFRL) in response to a Broad Agency Announcement “to develop the technology to survey the sky”. The remainder of the development funds came from NASA, the PS1 Science Consortium (PS1SC), the State of Hawaii, and some private funds. The project’s goal was originally to construct 4 separate 1.8-meter telescope units each equipped with a 1.4 gigapixel camera, and operate them in union. The ambitious nature and full scale cost of the project led to a decision to build a prototype system of a single 1.8-meter telescope unit. This provided an opportunity not only to test the hardware, software and design but also to carry out a unique science mission. This system, located on the island of Maui, was named Pan-STARRS1 (PS1).
2.1.2 The PS1 Science Consortium
In order to execute and deliver a competitive and scientifically interesting set of sky surveys, the Institute for Astronomy (IfA) of the University of Hawaii (UH) assembled the PS1 Science Consortium (PS1SC). This group of interested academic institutions established a set of science goals Chambers (2007), and a Mission Concept Statement Chambers (2006b) and funded the operations of PS1 for the purpose of executing the PS1 Science Mission (Chambers, 2006a; Chambers & Denneau, 2008). The founding institutions of the PS1SC defined 12 Key Projects to ensure that the definition of the surveys and their implementation were shaped by science drivers covering a range of topics from solar system objects to the highest redshift QSOs. The Memorandum of Agreement of the PS1SC established that the funding for operations was provided in return for the proprietary use of the Pan-STARRS1 data for scientific purposes. As the PS1 Mission went on, additional members were added to bring in additional resources. The member institutions of the PS1 Science Consortium are provided in Table 1
|University of Hawaii Institute for Astronomy|
|Max Planck Institute for Astronomy|
|Max Planck Institute for Extraterrestrial Physics|
|The Johns Hopkins University|
|University of Edinburgh|
|Queen’s University Belfast|
|Harvard-Smithsonian Center for Astrophysics|
|Las Cumbres Observatory Global Telescope Network|
|National Central University of Taiwan|
|Space Telescope Science Institute|
|National Aeronautics and Space Administration|
|National Science Foundation|
|University of Maryland|
|Eötvös Loránd University|
|Los Alamos National Laboratory|
2.1.3 The PS1 Science Mission
The PS1 Telescope began formal operations on 2010 May 13, with the start of the PS1 Science Mission, funded by the PS1SC and with K. Chambers as PI and Director of PS1. At the beginning of the PS1 Mission, the Image Processing Pipeline (IPP) - the software and hardware for managing and processing the data - was not at an advanced stage of development, nor were the characteristics of the unusual OTA devices well understood. Furthermore, because of the AFRL funding, the imaging data was initially required to be censored. The AFRL “Magic” software was devised so that the pixels surrounding any feature in an individual image that could be interpreted as a potential satellite streak were masked. This meant removal of pixels in a broad streak, or elongated box, which was large enough to prevent the determination of any orbital element of the artificial satellite before the images left the IfA servers to the consortium scientists. This requirement hindered analysis of the very features that were triggering the censor, nearly all of which were not satellite streaks, but were inherent detector characteristics. This effectively delayed the full and rapid analysis of the pixel data by consortium scientists until the ARFL finally dropped the requirement on 2011 Dec 12. From that date, all Pan-STARRS1 images, including prior data taken during commissioning and from the start of survey operations, were no longer subject to any such masking software. Earlier data was re-processed from the untouched original raw data without the streak removal. There is no real time nor archival censorship of any Pan-STARRS data. None of the data now being released in DR1 and DR2 suffer from any application of the “Magic” streak removal software either in the individual or in the stack images.
At the start of the PS1 Mission the development of the IPP (software and hardware), and eventually the development of the PSPS, shifted from the Pan-STARRS Project Office (2003-2014) to the PS1 Science Consortium funded PS1 Operations team. The Project Office went on to develop the second Pan-STARRS facility, Pan-STARRS2. In August of 2014 the Pan-STARRS Office closed, and the Operations team also took over responsibility for the completion and commissioning of PS2 with the support of the NASA NEO Program, the State of Hawaii, and private funding. No further involvement with the AFRL is expected.
2.1.4 The STScI Mast Archive and Data Releases
To fully exploit the scientific potential of the PS1 survey data the PS1SC committed to make all PS1SC data public and accessible as soon as possible, but not before one year after the end of PS1SC survey operations. The science consortium made this commitment in principle in order that the data reach as wide a usage as possible, however the original founding members of the PS1SC did not contain the resources or expertise to provide a public interface server. To enable such a public release, the PS1SC joined forces with the Space Telescope Science Institute (STScI) and the Barbara A. Mikulski Archive for Space Telescopes (MAST). The STScI joined the PS1 Science Consortium through a Memorandum of Agreement to contributing resources to create an archive of the PS1 Data Products that will serve the entire astronomical community.
2.2. Flow of information in the Pan-STARRS System
An overview of the flow of information through the Pan-STARRS System is shown in Figure 1. In brief: photons from astronomical objects are brought to a focus by the Telescope onto the focal plane of the Gigapixel Camera #1 (GPC1). As discussed below, a feedback signal is generated from selected areas of GPC1 and fed back to the telescope through the Observatory, Telescope, and Instrument Software system or OTIS, see Section 2.4. During the night, as new images are downloaded, they are processed by the IPP, see Section 2.7. The results are passed to the Moving Object System (Section 2.9.1) and the Transient Science Server (Section 2.9.2) Near Earth Object (NEOs) candidates from MOPS are sent to the Minor Planet Center, and stationary transient objects are now posted on the IAU Transient Name Server222https://wis-tns.weizmann.ac.il/ for use by the community. Offline from nightly processing, the IPP uses a variety of tools for calibration (Section 2.7.9). The catalog data products produced by IPP are passed on the PSPS database (Section 2.8). Both the PSPS database and all the image products from the IPP are then available to the community from the Barbara Mikulski Archive for Space Telescopes (MAST) at STScI.
The Pan-STARRS telescopes (both PS1 and PS2) are located at Haleakala Observatories (HO) on the island of Maui on the site of the Lunar Ranging Experiment (LURE) (Carter & Williams, 1973). Measurements by the HO Differential Image Motion Monitor (DIMM) show the site has a median image quality of 0.83 arc-seconds (the mode is 0.66 arc-seconds). On average 35% of the nights on Haleakala are photometric, with an additional 30% usable with very low extinction or more than 60% of the sky clear of clouds. The wind pattern is predominately trade winds from the east-northeast, with occasional “Kona” winds from west-southwest. PS2 is due north of PS1, the center of the two telescope piers is separated by 20.05 meters. The domes are situated in the wake of the flow from trade winds into the crater wall. Detailed metrics of the site characteristics will be published elsewhere (Chambers, 2017 in prep). More recently the Daniel K. Inouye Solar Telescope (DKIST)333http://dkist.nso.edu/ has been erected to the south-south west of the Pan-STARRS facility. The ultimate impact of DKIST operations on the Pan-STARRS environment is not yet fully known, their operational plan is to manufacture ice at night for use in the daytime cooling of DKIST, and subsequently dissipation of heat into the atmosphere at the summit.
The International Astronomical Union has determined that the acceptable level of Radio Frequency Interference outside an observatory doing optical and infrared observations should be less than integrated over the radio spectrum. This is exceeded at Haleakala and at the start of the PS1 Mission, radio frequency interference from various Federal and commercial transmission sites near the summit was an issue. However with the relocation of TV broadcasters to the Ulukalapua site, this problem has been mitigated and we see no evidence of RFI in GPC1. However cellphone transmission, wifi transmission, and microwave ovens have a noticeable effect and are not allowed at the Observatory.
2.4. Telescope, optics, and control system
The Pan-STARRS1 Telescope (PS1) is an alt-az telescope with an instrument rotator built by Electro Optic Systems Technologies Inc., Tucson, (EOST) with an enclosure by Electro Optic Systems Ltd. (EOS), Australia. The PS1 Dome motion closely follows the telescope through a featherweight direct coupling. The dome has four independently controllable vents for air flow through the dome. The dome slit is covered by two independently controllable shutters that can be deployed over the top on to the back side of the dome. When the moon is up the dome slit shutters are used to mitigate scattered light from the moon.
The Observatory, Instrument, Telescope, Software (OTIS) system controls all these aspects of the Observatory and collects and stores a wide variety of auxiliary and metadata on the conditions and all the functions of the Observatory.
The Pan-STARRS1 optical design (Hodapp et al., 2004a, b; Morgan & Kaiser, 2008) has a wide field Richey-Chretien configuration with a 1.8 meter diameter /4.44 primary mirror, and 0.9 m secondary. The resulting converging beam then passes through two refractive correctors, one of six possible interference filters with a clear aperture diameter of 496 mm, and a final refractive corrector that is the cryostat window. Note that the Pan-STARRS1 as-built optics are described by the Zemax model NOADC-3.0.
The optical design has 4 aspheric surfaces; one each on the primary and secondary mirrors, one on the first corrector lens, L1, and a final aspheric on L3, the last corrector lens in the optical path and which also serves as the cryostat window. The secondary mirror has a conic constant of and a order aspheric term of , which made it a challenge to fabricate (Morgan & Kaiser, 2008).
The Secondary Mirror is mounted on a hexapod and can be moved in five axes: x,y,z,tip, tilt. The Primary Mirror is on a pneumatic support system and can be commanded in y,z, tip, and tilt. The Primary Mirror can be moved in the x direction as well, but this is not on a powered actuator and must be done manually. Furthermore the Primary Mirror has a 12 point astigmatic correction system. Thus there are 22 independent mirror actuators that can be used to bring the optics into proper collimation and alignment with the optical axis as defined by the axis of the instrument rotator. These actuators allow for modest amounts of Primary mirror deformation to remove trefoil, coma, and astigmatism. The procedure for establishing the proper collimation and alignment is described in Morgan & Kaiser (2008). Given the system matrix, only minor adjustments are required to maintain collimation and alignment. PS1 does have significant flexure, so empirical models have been determined to correct for that. In practice the largest corrections are in the M2 tip and y (tangent to altitude) de-center. The M1 figure correction also has an altitude dependent term. The OTIS software applies these corrections for the destination of any commanded slew, corrections are disabled during exposures and the system tracks quiescently during the short exposures - generally not more than 2 minutes. A focus offset is determined from each exposure based on the measured astigmatism, and this offset is applied to the empirically derived focus model. The offset is calculated from an analysis of the ellipticity of the PSF across the focal plane calculated by the GPC1 software. The calculation of the correction takes approximately one minute, and then can not be applied until the next pause between exposures while the camera is reading out. Thus the telescope focus is maintained by the local focus model with an observationally based offset determined within a few minutes of a new exposure. After large slews or starting a new chunk, a short exposure (10 seconds) is made to obtain a current focus correction. This system maintains the correct M2 focus position to within microns of true focus. The collimation and alignment do drift occasionally, especially if there is maintenance performed on the telescope. These drifts are corrected by a procedure of using above and below focus images of stars (donuts) Morgan & Kaiser (2008) to make a correction. The system to maintain the image quality is imperfect, and the results can be seen in some images. Typically the impact is some combination of higher order aberrations that result in a asymmetric PSF. The IPP fits only an elliptical PSF, so there is no systematic measure of this asymmetry or its effect on photometry, albeit it must be small. The telescope illuminates a diameter of 3.3 degrees, with low distortion, and mild vignetting at the edge of this illuminated region. The field of view is approximately 7 square degrees. The 8 meter focal length at gives an approximate 10 micron pixel scale of 0.258 arcsec/pixel.
|Focal Length||8000 mm|
|Nominal Field of view||3.0 degree diameter circle|
|Primary mirror||1800 mm diameter|
|M1 coating||protected aluminum|
|Secondary mirror||947 mm diameter|
|M2 coating||protected silver|
|including diffraction and obscuration|
|Rotator range||179 degrees|
|Telescope/Dome wrap||420 degrees|
2.5. GPC1 - the Gigapixel Camera #1
The Gigapixel Camera #1 (GPC1) uses Orthogonal Transfer Arrays devices, a concept developed by Tonry et al. (1997) and their development was key to the Pan-STARRS concept (Kaiser et al., 2000). The detectors in GPC1 are CCID58 back-illuminated Orthogonal Transfer Arrays (OTAs), manufactured by Lincoln Laboratory Tonry et al. (2006, 2008). They have a novel pixel structure with 4 parallel phases per pixel (Tonry et al., 2008) and required the development of a new type of controller (Onaka et al. (2008)). GPC1 is actually populated with two different kinds of CCID58s, the CCID58a with a three phase serial register, and the CCID58b which has a two phase serial register (Onaka et al., 2012). Table 3 has summary of GPC1 characteristics. The intent of the OTA design was to allow charge to be moved in orthogonal directions providing an on-CCD tip-tilt image correction given a guide signal from a nearby cell being read at video rates, and Tonry’s OPTIC camera did this successfully, (e.g. Stalder et al., 2009). However, with GPC1 when the Orthogonal Transfer mode of the detectors was turned on ,it produced an unacceptable level of non-uniform background noise (Onaka et al., 2012). The Pan-STARRS1 Surveys did not use the detectors in Orthogonal Transfer mode. All Pan-STARRS1 Survey data was taken with the GPC1 devices operating as ”normal” CCDs.
The detectors are read out using a StarGrasp CCD controller, with a total overhead of 10.3 seconds for a full unbinned image, see Table 3 for a breakdown of the overhead. Other performance characteristics of GPC1 are presented in Tonry et al. (2008); Onaka et al. (2012).
The focal plane of Pan-STARRS1 comprises a total of 60 CCID58 OTA devices Tonry et al. (2008). Each of these devices consists of an array of individual addressable CCDs called “cells”. The overall format of a single OTA is a pixel array with a pixel size of 10 m which subtends 0.258 arcsec. Each OTA device is made up of 64 cells where each cell is pixels. The cells are separated by a gap between columns, that is 18 “inactive” pixels in size, and a gap between rows that is 12 inactive pixels in size. Thus a single OTA device contains a single piece of silicon with 64 cells in an array separated by a grid of internal streets. We will often refer to the OTA devices as ”chips” in the data processing discussions. Further more, there is a physical gap between the devices as mounted in GPC1. The placement of the devices in the focalplane is shown in Figure 3. The relative positions of each device, including rotation, were determined from a vast number of astrometric measurements on sky.
The separation between the OTA devices is 1400 microns (approximately 36 arcsec) in the direction and 2800 microns ( approximately 70 arcsec) in the direction. In practice the devices are not perfectly spaced and can have some small rotation with respect to one another. The astrometric solution for each device is solved independently without reference to one another, the only place where the determined relative position is used is telescope pointing and guiding. Note there is a slight optical pin-cushion distortion of the sky on the focal plane, all of this is removed in the process of the astrometric registration (warping) by the IPP (see Magnier et al., 2016a).
|Device||CCID58a; three phase|
|Device||CCID58b; two phase|
|Charge transfer||bad regions masked|
|Pixel size||0.258 arcseconds|
|Camera fill factor||90%:|
|Shutter||opening time 1 sec|
|Shutter||precision 10 msec|
|Total Overhead||10.3 sec|
|Exposure start||2.0 sec|
|Exposure readout||7.0 sec|
|Exposure save/clean||1.0 sec|
|Pixel Mask fractions|
|Poor Charge Transfer Efficiency||2.2%|
|Other defect flags||1%|
The telescope, detector devices, and control electronics each contribute a variety of artifacts to the GPC1 images. Where possible these artifacts are identified and the pixels are masked or modified during processing and flags are set in the database. These include optical ghosts from reflections in the optics, glints from scattered moon light, glints from structure in the camera, regions of poor charge transfer in the devices, persistence or “sticky charge” from saturation leaving “burn-trails” that persist for all successive images for tens of minutes, electronic ghosts from cross-talk in the electronics, and correlated read noise from the fiberflex that transmit the signal through the cryostat wall. These are identified and masked where possible in the detrending procedure as part of the chip processing stage in the IPP. There is a detailed discussion of the defects and how they are masked in Waters et al. (2016). These defects are visible on the focal plane in the single exposure frames, but the stacked images made from the multiple images taken over the survey duration are composed of dithered frames.
If we take the sky area covered by the GPC1 footprint to be the area of the inner blue circle in Figure 3 (7 sq degrees) then the dead cells, pixel gaps and masking of defective pixels account for an overall loss of 20% of the focal plane in any one exposure. There is an additional dynamic masking of around 2-3% per exposure, which mostly covers the “burn-trails”. Therefore the overall fill factor of the camera is % per exposure and this is mitigated by the dither and stack techniques that were employed in the and Medium Deep Surveys. The first data release (DR1) from the STScI MAST archive is the stack images only and hence the images will mostly look continuous, although there are areas where a combination of poor devices and fewer than 12 exposures mean masked regions creep through to the final product.
A subset of bright stars (mag) which fall on the focal plane are selected to be used as guide stars (suitably located across the camera), and a pixel box is defined, centered on the position where these stars are predicted to land based on the commanded telescope position. This set of sub-arrays on different devices are read at video rates. The centroid from these video frames are used to send a guide signal to the telescope control system.
Typically there are 4 to 10 stars chosen, which means these cells are then masked in the science exposures. These additional masked cells are included in the “dynamic” mask developed for each exposure that includes the masking due to the artifacts of that particular exposure and is added to the ”static” mask as seen in Figure 3.
The shutter, built by the team at Bonn University, is a dual blade design. The shutter aperture is approximately 40 cm across, and in closed position one blade covers the aperture and one is stored to the side. When the shutter is opened, one side of the focal plane is exposed first. At the conclusion of the exposure, the second blade traverses the aperture in the same direction, hence the total exposure time seen by each pixel is the same to the precision of the movement, or 10 milliseconds. For the subsequent exposure the motion is in the opposite direction. Short exposures are possible, where the blades follow each other trailing closely. This does mean that the center time of the exposure is different by up to 0.5 seconds depending on placement in the focal plane. The metadata exists to give every object a time of exposure depending on its position in the focal plane. In principle a correction to the exact time of the center of the exposure can be calculated for every detection, such a correction is not currently made, and thus the UT of the exposure can be in error for any object by up to 0.5 seconds. Even for moving asteroids this is not a serious limitation.
In practice the total overhead time between adjacent exposures, including the 1 second shutter movement time, is 10 seconds, see Table 3 for details.
2.6. Filter bandpasses and PS1 sensitivity
The Pan-STARRS1 observations are obtained through a set of five broadband filters, designated as , , , , and . Under certain circumstances Pan-STARRS1 observations are obtained with a sixth, “wide” filter designated as that essentially spans , , and . There is full sky sky coverage in each of but not in , which was mostly used for near-earth object surveys. Although the filter system for Pan-STARRS1 has much in common with that used in previous surveys, such as the Sloan Digital Sky Survey (SDSS, (York et al., 2000)), there are important differences, which is why the filters are labelled specifically with the P1 subscript. The filter extends 20 nm redward of with the intention of providing greater sensitivity and lower systematics for photometric redshifts. The strong [O I] 5577Å sky emission is on the filter edge but only at 1% transmission. The filter has a sharply defined cut-off at 922 nm, which is contrast to the SDSS band which has no red cut off and the response is defined by the detector response. The and filters are very similar to SDSS and colour differences between the two magnitude systems are small. SDSS has no corresponding filter. The transmission of the Pan-STARRS1 filters, optics and total throughout were precisely measured with a calibrated photodiode and a tuneable laser, without use of celestial standards by Stubbs et al. (2010) and this procedure was repeated in November 2016 (Stubbs et al. in prep). The definition of the photometric system has already been discussed in detail and published in Tonry et al. (2012b). Tabular data of the overall throughput of the PS1 system is available in the online data of Tonry et al. (2012b) and the individual filter throughputs are in Stubbs et al. (2010). The PS1 total filter throughputs from Tonry et al. (2012b) are reproduced here in Figure 4.
Pan-STARRS1 magnitudes are interpreted as being at the top of the atmosphere, with 1.2 airmasses of atmospheric attenuation being included in the system response function. No correction for Galactic extinction is applied to the Pan-STARRS1 magnitudes. We stress that, like SDSS, Pan-STARRS1 uses the AB photometric system and there is no arbitrariness in the definition. Flux representations are limited only by how accurately we know the system response function vs. wavelength.
2.7. IPP - Image Processing Pipeline
All images obtained by the Pan-STARRS1 system are processed through the Image Processing Pipeline (IPP) on a computer cluster at the Maui High Performance Computer Center. The pipeline runs the images through a succession of stages, including de-trending or removing the instrumental signature, a flux-conserving warping to a sky-based image plane, masking and artifact removal, and object detection and photometry. The IPP also performs image subtraction to allow for the prompt detection of moving objects, variables and transient phenomena. Mask and variance arrays are carried forward at each stage of the IPP processing. Photometric and astrometric measurements performed by the IPP system are published in a mysql relational database. Below we give a brief summary of the Pan-STARRS image processing, full details are provided in the companion papers of Magnier et al. (2016a, c, b); Flewelling et al. (2016); Waters et al. (2016). Figure 5 gives a simplified schematic of the processing stages.
2.7.1 Chip Stage
In the “Chip Stage” raw exposures are detrended (dark subtracted, flattened, masked, etc, Waters et al., 2016) and sources in the images are detected and basic instrumental characterization is performed. A PSF model is generated and all sources fitted with that model. For sources above a minimal signal-to-noise limit (nominally 20), a simple galaxy model is fitted if the source appears to be extended. The best model (PSF or galaxy) is subtracted and an additional source detection pass is made (down to S/N = 5). This provides for some de-blending. Reported values include instrumental positions, fluxes (PSF, seeing-matched aperture, Kron aperture), moments, and various quality flags are recorded for each source in the image. The output from this stage consists of fits tables of detections and their properties called CMF files and detrended images and their associated variance, and mask pixel images.
2.7.2 Camera Stage
In the “Camera Stage” the instrumental measurements from all the chips in one exposure are gathered together for astrometric and photometric calibration by comparison with a reference catalog. Initially a synthetic reference catalog was created based on 2MASS, USNO-B, and Tycho. This was used for a photometric calibration as the survey proceeded. In the re-processing and re-calibration that produced the data in DR1, the reference catalog uses Pan-STARRS itself, to create a precise and consistent internal calibration (Magnier et al., 2016c, b) based on the “ubercal” methods described in Schlafly et al. (2012) and Finkbeiner et al. (2015). The primary data product from the Camera Stage is the collection of calibrated detection tables.
2.7.3 Warp Stage
In the “Warp Stage” the detrended pixel images generated by the chip stage are geometrically transformed to a predefined set of images which tessellate the relevant portion of the sky. Specific examples are discussed in Section 3 below. A set of virtual rectilinear images with square pixels of 0.25 arcseconds size, on a local tangential projection center no bigger than about 4 degrees across are defined. These virtual images are called “projection cells” and one or more projection centers can be defined for specific areas of interest or arranged in some defined tessellation of the entire celestical sphere. The total output from the warp stage is the collection of images that describe the signal, variance, and masking for each skycell.
2.7.4 Stack Stage
Individual epoch skycell images (from the Warp Stage) are combined together to form deeper stack images of the sky, the details of the algorithms are in Waters et al. (2016). In the IPP analysis, stacks of different depths/quality may be made depending on the individual survey goals. This is of particular application to the “Medium Deep” survey. The output from the stack stage consists of the signal, variance, and mask stack images.
2.7.5 Difference Image Stage
The primary means for detecting a transient, moving, or variable object is through the process of subtracting a template image of a source from a single image to create a “difference” or “diff” image. The IPP generates Alard-Lupton convolved difference images for skycells in various combinations depending on the survey goals. The output from the diff stage is a collection of detections from the difference images, including both positive and negative difference detections.
2.7.6 Static Sky Stage
The “Static Sky” refers to a final stacked image. The stack images from all filters are processed in a single analysis step to perform the deep source detection and characterization of objects detectable in the stacks. This analysis step is similar to the source detection and characterization performed at the chip stage, with some important additions: First, 3 PSF-convolved galaxy models (Sersic, DeVaucouleurs, Exponential) are fitted to all objects with sufficient signal-to-noise and in regions outside the densest portions of the Galactic plane. In addition, sources which are detected in only two of the 5 filters (or just in the band, to allow for the presence of astrophysical objects which are dropouts in the bluer bands) are then used to force PSF photometry (and aperture and Kron flux measurements) at that same location in the other 3 (or 4) filters. Finally, flux is measured for 7 radial aperture annuli, using apertures of the same radii in arc-seconds on the sky as used by SDSS. These radial aperture fluxes are measured for the raw stack with its natural seeing as well as on a version of the stack convolved to match 1.5 and 2.0 arc-second seeing.
2.7.7 Skycal Stage
The ”Skycal” stage is similar to the chip stage, the staticsky stage analysis generates measurements in instrumental coordinates (X,Y,DN). The skycal stage performs the calibration of the staticsky outputs relative to the reference catalog in an analogous fashion to the camera stage.
2.7.8 Full Force Stage
Image quality variations between different exposures (and even within a single exposure) result in a stack PSF which can vary discontinuously on small scales. PSF photometry and PSF-convolved galaxy model fitting on the stack cannot follow these variations. The result is degraded performance in the stack photometry and morphology analysis. To avoid this problem, we use the outputs from the “Static Sky” stage analysis as the input to a “forced” photometry analysis on each of the input warp images.
In this analysis, the positions of all objects detected in the stacks are used to measure the PSF photometry of those objects on each of the input warps images, using the appropriate PSF model determined for that position on that warp image. The individual warp measurements are then combined in catalog space (in our photometry databasing system) to determine the mean photometry for each object.
In this step, input measurements with excessive masking are also excluded from this mean photometry calculation. The result is a reliable photometry measurement for all objects down to the detection limits of the stack, as well as the data to study the variability and transient nature of the faintest sources.
In this stage, we also perform an analysis of the galaxy morphology using the “static sky” galaxy model measurements as the seed (Magnier et al. (2016c)).
2.7.9 Post-Processing and DVO
After the pixel-level processing is performed, the catalogs of measurements extracted from the images are ingested into an instance of the Desktop Virtual Observatory or DVO (for more details see Magnier et al., 2016a). DVO is a set of stand alone tools within the IPP system created to perform calibrations and provide further analysis of systematic effects.
In addition to the ingest into DVO at the IPP, the team of Eddie Schlafly (MPIA, LBL), Doug Finkbeiner (Harvard), and Greg Green (Stanford) also ingest the camera-stage data into a separate databasing system called LSD (2012ascl.soft09003). This system is similar in scope to DVO and allows similar calibration operations. This team runs the “ubercal” analysis on the detections from the chip and camera stage to measure zero points for photometric data. In this analysis, relative photometry of overlapping images is used to constrain the zero points and airmass terms. A rigid solution is determined by requiring a single zero point and airmass term for each night. The resulting photometric system is shown to have a precision of 8, 7, 9, 11, 12 millimags for each of respectively (as described in Schlafly et al., 2012).
Given the way the Pan-STARRS System evolved, it has been necessary to implement a translation layer to collate the catalog products produced by the IPP (so called “CMF” and “SMF” fits tables containing measured attributes Magnier et al., 2016a) in an optimal manner for ingest into the PSPS. The IPP-to-PSPS produces batches of binary fits files containing catalog data. There is a different kind of batch for each type of database table (e.g. objects, stacks, detections, difference detections). Each batch contains data from a localized region of the sky. Some units are rationalized in the IPP-to-PSPS, so there is some manipulation of data values in this subsystem. See Flewelling et al. (2016) for a detailed discussion.
2.8. PSPS - Published Science Products System
The Pan-STARRS Project teamed with the database development group at Johns Hopkins University to undertake the task of providing a hierarchical database for Pan-STARRS (Heasley, 2008). Since the JHU team was the major developer of the SDSS database (Thakar et al., 2003), our goal was to reuse as much of the software developed for the SDSS as possible. The Pan-STARRS database is commonly refered to as the “PSPS”.
The key to moving from the SDSS database to a system capable of dealing with Pan-STARRS data is the design of the Data Storage layers. It was immediately clear that a single monolithic database design (like SDSS) would not work for the PS1 problem. Our approach has been to use several features available within the Microsoft SQL Server product line to implement a system that would meet our requirements. While SQL Server does not have (at present) a cluster implementation, this can be implemented by hand using a combination of distributed partition views and slices (Heasley, 2008). This allows us to partition data into smaller databases spread over multiple server machines and still treat the information as a unified table (from the users’ perspective). Further, by staying with SQL Server we are able to retain a wealth of software tools developed for SDSS, including the use of Hierarchical Triangular Mesh indexing for efficient spatial searches.
An overview of the PSPS system is shown in Figure 6.
2.8.1 Object Data Manager
The Object Data Manager is a collection of systems that are responsible for publishing data attributes measured by the IPP or other Pan-STARRS Science Servers to the end user (scientist). The ODM manages the ingest of data products from the IPP (or other sources), integrates the new products with existing information in its data stores, and then makes the information available to the users in relational databases.
Catalog data from the IPP as prepared by the IPP-to-PSPS layer is contained in batches of binary fits tables. These fits tables are read by a Data Transformation (or DX) Layer where data are grouped by declination zone and throttled in Right Ascension by the IPP-to-PSPS layer. Then the data is loaded into ‘cold’ or load slice machines by the DLP or data loading pipeline. The slices are variable bands in declination, established to have nearly constant data density. Once data are loaded on all declination slices through a given RA range, the data are merged, wherein they are stored and indexed on the slice machines so that data that are nearby in the sky are similarly nearby on disks and grouped by machine. Once the data are successfully merged across the whole sky, the database is copied from the load/merge machines to the data storage machines where the user can access the database through the Query Manager (QM) and web-based Pan-STARRS Science Interface (PSI).
2.8.2 The Data Retrieval Layer
The Data Retrieval Layer or DRL is the unseen hub of the PSPS system. It sits between software that provides user access and the underlying data stores themselves. The DRL provides the access to users and the databases through web browsers. Only those users who want to write their own access clients will interact with the DRL directly. A simple application programming interface (API) has been developed to allow one to develop such applications.
The DRL also provides the internal mechanisms for routing result sets from the PSPS databases back to the user.
The DRL API allows the PSPS to expand to incorporate the addition of new databases that can make science products created by PS1SC science servers available to the user community. The API has been demonstrated to work with Microsoft SQL Server, MySQL, and PostgreSQL databases.
The DRL Layer is accessible through the CasJobs interface at the Pan-STARRS1 Archive at MAST.
2.8.3 PSI Interface
The Pan-STARRS Science Interface (PSI) is a web application that has been developed by the PSPS development team. It is designed provide users with easy access to the PSPS through a web browser. PSI has tools to simplify the construction of querys and flags and a variety of useful features. PSI is built on a improved version of CASJOBS, but it is not immediately backwards compatible with the version of CASJOBS at STScI. Access to the Pan-STARRS1 archive at MAST at STScI is through the standard CASJOBS (O’Mullane et al. (2005); Thakar & Li (2008)) interface.
2.9. Science Servers
The PS1 Science Servers were a project concept to add science value to the basic data products of object, position and flux. The three projects that evolved to provide working code and data products are breifly described here.
2.9.1 MOPS - Moving Object Processing System
The Pan-STARRS Moving Object Processing System (MOPS; Denneau et al. (2013)) is a modern software package that produces automatic asteroid discoveries and identifications from catalogs of transient detections from Pan-STARRS or any next-generation astronomical survey telescope.
As implemented as a subsytem in the Pan-STARRS System, it obtains difference detections from the IPP, performs linkages between detections, and makes initial orbit determinations. Potential moving objects are evaluated by a human inspection system, and candidates are passed to the Minor Planet Center of the IAU.
Funded by the Pan-STARRS Project prior to the formation of the PS1SC, MOPS was the first integrated asteroid detector system able capable of automatically producing high-quality orbits from individual per-exposure transient catalogs. MOPS is also able to search its own historical data for orphaned one-night detections after an orbit is generated.
As implemented as a subsytem in the Pan-STARRS System, it obtains difference detections from the IPP, performs linkages between detections, and makes initial orbit determinations. Potential moving objects are evaluated by a human inspection system, and candiates are passed to the Minor Planet Center of the IAU.
MOPS has additional value as a research tool in survey design, able to simulate years of observations and detections given a catalog of synthetic asteroids and a hypothetical observation schedule. The synthetic solar system model (S3M; Grav et al., 2011), containing objects representing populations of all major solar system bodies, remains the standard synthetic population for evaluating survey performance.
2.9.2 TSS - Transient Science Server
The vast majority detections in difference images requires a system for classifying real vs artifacts to manually select the most promising candidates. The Queen’s University Belfast group developed the Transient Science Server to systematically process difference detections from stationary transients from the IPP stream and apply machine learning techniques to classify them. (Wright et al. (2015)). This system continues to process transient events from Pan-STARRS and post discoveries on the IAU Transient Name Server. In parallel the team at CfA, Harvard developed a custom version of the photpipe image subtraction and analysis pipeline and analyse the MDS data in real time (Berger et al., 2012; Rest et al., 2014) The two teams cross-correlated transient discoveries and photometric measurements from both streams to improve efficiency and measurement precision of the IPP products. Both were successful in different ways, and the QUB based TSS was the only one currently in operation for the 3 based searches and the ongoing Pan-STARRS Survey for Transients(Huber et al., 2015; Smartt et al., 2016).
2.9.3 PCS - Photometric Classification Server
The Photometric Classification Server (Saglia et al. (2012)) is a set of software tools and hardware set up to compute photometric, color-based star/QSO/galaxy classification and best-fitting spectral energy distribution (SED) and photometric redshifts (photo-z) with errors for (reddish) galaxies. The system can establish an interface to the PSPS database and results can be ingested back into the PSPS. Results from the Photometric Classification Server will not be available in DR1.
2.10. Pan-STARRS Operations
The observatories are operated remotely from the Pan-STARRS Remote Operations Center in the Institute for Astronomy (IfA) Advanced Technology Research Center (ATRC) in Pukalani, Maui. There is no one at the summit at night or on weekends except in urgent or emergency situations. The Observatory is approximately 45 minute drive from the ATRC. A Pan-STARRS observer on a swing shift schedules the night’s operations based on the overall science goals, state of the survey, and expected conditions. The night Observer executes the plan prepared by the swing shift observer and modifies it in real time as circumstances demand. The observing staff rotate through the swing and night shift and support the day crew at the summit. The Staff at the ATRC also provides support for the telescope, scheduling software, and system administration for the IPP cluster in Kihei.
The IPP is a linux cluster that currently has 3100 cores and 5.5 Petabytes of storage. The IPP cluster is currently located at the Maui Research and Technology Center in Kihei, Maui, about 20 minutes drive from the ATRC. The computing facility (power, cooling, network connectivity to the outside world) is administered by the Maui High Performance Computing Center. Additional computing resources were required for the PS1 Surveys including the Mustang Cluster (30,000 cores) at Los Alamos National Laboratory and the Cray cluster at the University of Hawaii (3600 cores).
Operationally the IPP and the PSPS are run remotely by the IPP team from IfA Manoa. During night time operations, the raw exposures are immediately downloaded to the IPP cluster in Kihei, Maui. Nighty data processing occurs automatically for exposures as they are obtained, with the analysis emphasis on the discovery of transient events, as well as data characterization for future re-processing. The reprocessing versions and status are discussed in detail below and in the companion papers. These data products have been loaded and merged in the PSPS database and transferred to STScI for exposure through the MAST archive.
3. The PS1 Surveys
3.1. The PS1 Science Goals
The primary science design drivers for PS1 were originally put forth in the PS1 Science Goals Statement ((Chambers, 2007)). The top level goals were:
Precision photometric and astrometric survey of stars in the Milky Way and the Local Group;
Surveying our Solar System, including searching for Potentially Hazardous Objects amongst Near Earth Asteroids;
New constraints on Dark Energy and Dark Matter;
Exploration and categorization of the astrophysical time domain, including, but not limited to, explosive transients, microlensing events in M31, and a transit search for exo-planets.
Providing a development platform for prototyping PS4 components, subsystems, and survey strategy.
These goals drove the initial design and engineering requirements, and shaped real time development decisions. On the last point, while the PS4 system has not yet been funded, PS1 did serve in this capacity for the development of PS2 (Morgan et al., 2012). The above outline goals do not begin to cover the vast array of solar system, galactic, extragalactic, and cosmological studies that can be done with the PS1 data products. To refine this, the project and the PS1 Science Consortium Science Council generated the PS1 Mission Concept Statement Chambers (2006b) with a set of surveys as follows: (1) A 3 Steradian Survey; of 60 epochs in five passbands () of the entire sky north of declination degrees, (2) A Medium Deep Survey with data in all of of ten PS1 footprints on well studied fields totaling 70 square degrees at high galactic latitudes spaced around the sky, (3) A solar system ecliptic plane survey in the wide passband with cadencing optimized for the discovery of Near Earth Objects and Kuiper Belt Objects, (4) a Stellar Transit Survey of 50 square degrees in the galactic bulge; and (5) a Deep Survey of M31 with an observing cadence designed to detect micro-lensing events and other transients. In addition a special series of observations of spectro-photometric standards was carried out for calibration, and the Celestial North Pole was observed nightly for the last two years of the survey to track performance and measure atmospheric properties. Table 4 summarises these surveys and the approximate percentage time spent on each of the total operational science time.
The operational plan for execution of these surveys was articulated in the PS1 Design Reference Mission (Chambers & Denneau (2008)) or DRM, that served as a benchmark as the system transitioned from commissioning to operations. This survey strategy evolved into a Modified Design Reference Mission as lessons learned were incorporated as the surveys progressed.
|3 Steradian Survey||56||2009-14|
Medium Deep Survey
|Solar System Survey||5 11||2012-14|
|Pan-Planets Transit Survey||4||2010-12|
|PAndromeda Survey of M31||,||2||2010-12|
|Small Area Survey 2||1||2010|
|Celestial North Pole||1||2012-14|
3.2. The 3 Steradian Survey
The Steradian Survey covers the sky north of Dec degrees in five filters () and includes data taken between 2009-06-02 and 2014-03-31. This means that for a given sky tessellation, a field center was included in the survey only if it was above declination degrees. For pointings with field centers that are close to , close to half the field (up to 1.5 degrees) extended below the limit. This means there is a ragged edge and an uneven declination limit to the survey between .
The survey pattern and scheduling followed two different strategies over the course of the 3 survey: the initial pattern layed out in the Design Reference Mission (DRM) Chambers & Denneau (2008) followed by the Modified Design Reference Mission (MDRM). We switched to MDRM on 2012-01-14. All exposures in the DRM were taken in pairs, with each exposure separated by a Transient Time Interval or TTI of 12 to 24 minutes, for the purpose of detecting moving objects within the Solar System. These were referred to as “TTI pairs”. The original plan was then to take 2 TTI pairs over an observing season with , and taken within the same lunation and separated by days to weeks. The and were to be taken approximately 6 months apart to optimise stellar parallax and proper motion measurements (for low mass stars). Over 3.5 years this would give (allowing for weather interruptions) 12 exposures in each band or 60 in total over all 5 filters.
In the MDRM, a series of 4 exposures, “quads”, all separated by approximately 15 minutes (therefore completed within about 1hr), were implemented for about half of the exposures with the express purpose of increasing the recovery of Near Earth Objects (NEOs). The relative exposure times in each were also chosen to make an asteroid of mean solar color (taken to be ) to have approximately the same signal-to-noise.
3.2.1 The boresight sky tessellation
3.2.2 The reduced image tessellation
During data processing, the “Warp Stage” takes the detrended pixel images generated by the “Chip stage” and geometrically transforms (warps) and re-samples them onto a predefined set of images which tessellate the relevant portion of the sky (these processing stages are discussed in Section 2.7). For the survey, PS1 uses a modification (RINGS.V3) of the Budavari rings tessellation with tangential projection centers spaced degrees apart. A set of virtual images called “projection cells” are defined to cover the sky about these projection centers without gaps. These virtual projection cells are subdivided along cartesian pixel boundaries into “skycells”, the image regions onto which the native device pixels are warped. All skycells have a pixel scale of 0.25 arcsec per pixel and are roughly 20 arcminutes on a side, which is comparable in size to the native device images (these chip images are the pixel arrays which are 0.258 arscsec per pixel). The main output from this stage is the collection of three separate pixel images each representing the signal, variance, and masking for the skycells. The MD and similar surveys use special local projection cells centered on the fields of interest.
3.2.3 Primary object resolution on the sky
The skycells and projection cells are defined to have an overlap of 60 arcseconds (120 pixels) on each edge in order to avoid objects being split between adjacent skycells. Note that it is the same data which goes into the overlap regions - there is no new data involved here. The problem of identifying a unique area, and thus assigning an object to a particular skycell, is called the primary resolution problem. This is important, as data analysis is performed on each skycell independently, so an object near a boundary will have duplicate measurements. IPP produces a tessellation tree file which contains RA and DEC limits for each projection cell, which can be used to define unique areas. Objects landing within these limits are classed as primary objects and have the primary flag set in PSPS. This flag should always be used to define a unique sample of objects on the sky.
3.2.4 Scheduling of PS1 Surveys
The primary reason for a discussion of the scheduling of the PS1 Surveys is to explain why the time domain of the 3 Survey has the detailed structure that it has. Prior to the formal start of the PS1 Mission on May 13, 2010, we used a contemporaneous version of the LSST scheduler to model the PS1 Mission as defined by the Design Reference Mission Chambers & Denneau (2008) and smaller in summer in accordance with the length of night. We further tweaked the size of the slices to accommodate time for the smaller PAndromedra and Pan-Planets surveys, and assumed that the MD surveys, which are fairly evenly distributed in RA, could be fit into a constant nightly time allocation.
The observing pattern from the DRM (applicable from 2010-05-10 - 2012-01-14) is schematically shown in Figure 8. An Observing Cycle (OC) is defined as one lunation. The sky areas and filter coverage observed in an example OC are illustrated in this figure. Clearly one needs to observe, on average, 1/12.37 of the sky per Observing Cycle per filter (12.37 is the number of lunations in one year). This corresponds to a slice of sky from the pole to which is roughly 4 hrs in right ascension. This mean value was expanded or compressed a-priori for the length of night and to adjust for the non-uniform impact of the smaller surveys in their RA distribution. One aspect of the PS1 3 Survey is that the bands are observed out of phase with by months, whereas might be taken in the same night or be out of phase by days. We defined two distinct kinds of slices, the Opposition slices, where the sky within about 2 hours RA of opposition was observed in and bands, and “Wing” slices which were near the meridian at twilight. There were several reasons for this “strategic” approach: (i) because twilight (defined as the moment when the night sky reaches a constant sky brightness) occurs at increasing solar elongation as one proceeds through the filter set from red to blue , there is a period of time when the sky is as dark as it is going to get in band, but it is still in twilight in bands. Thus it is most advantageous to use this time in band. Once the sky becomes dark in band, the same is true. The time differences in the other bands are more modest. To illustrate this quantitatively Table 5 gives details of the solar elongation angle at which the sky reaches its constant dark level (i.e. end of twilight) in each filter. (ii) We desire to measure as many stars with measurable parallaxes as possible. These are the closest stars and are thus most likely to be brown dwarfs. PS1 is a red sensitive instrument, and is already delivering on its goal of finding new populations of L and T dwarfs (Deacon et al., 2011a; Liu et al., 2013a). It is therefore desirable to observe in bands at maximum parallax, i.e. with a cadence of nearly six months. As illustrated in Figure 8, the Wing slices here in the bands are separated by nearly six months: as the pattern marches to the left from “Month A” to “Month B”. This shows that in about six months time the same region of the sky will be observed again in the same filters. (iii). This approach also ensures that the sky areas surveyed in the bands are observed near the meridian, or close to optimum airmass. During operations there typically was not quite enough time to get all of the and band observations in the twilight time of . However near full moon, the sky is bright even in band, and some band fields were observed closer to the middle of the night. This required that the polar regions, which were beyond the 30 degree moon avoidance region, be shifted slightly closer to opposition and yet could still be observed at reasonable airmass.
During a night’s observing the pattern from the above strategy was to observe “chunks” where a chunk is simply a contiguous region of sky of approximately GPC1 footprints, with a pair of visits separated by a TTI in band, and then if possible the same chunk in band as the sky got darker. Then the available Medium Deep Survey fields were inserted between the Wing Slices and the Opposition Slices. Depending on the sky brightness (lunar illumination and distance) a chunk in or band would be observed near opposition. Once opposition passed it would be back to the other programs and the morning Wing slice at twilight. The prioritization of the chunks in declination was by image quality and transparency, so generally by airmass, followed by wind direction or partial clouds.
It was eventually realized that a modification to the DRM was necessary. The DRM was done entirely in pairs with the assumption that observations of NEOs in pairs separated by one or two nights could be linked. As a matter of experience that turned out not to be the case. However simply switching to quads, or four exposures per night separated by TTIs would put the years worth of exposures for a given field in a given filter all into one night. This would have endangered the photometric survey, reducing the number of opportunities to have a photometric night, critical for ubercalibration (see Magnier et al., 2016c). The solution, called the Modified Design Reference Mission or MDRM, which was settled upon is shown in Figure 9. The Wings pairs in remained the same in the MDRM as the DRM. The sensitivity to discovering asteroids was low in although a few have been found by a pair in and a pair in . In this compromise solution, one third of the data in a given filter is taken in a quad, while the remainder is taken in pairs on different nights. The total number of exposures per field per year is still 4, but the cadence will be different in different parts of the sky. This is roughly smoothed out with data over different years where the pattern is shifted by the position of opposition at new moon. Furthermore the pattern is altered by the scheduling around weather and wind. One way of visualizing the time domain is shown in Figure 10 where the date (in Unix seconds 444Unix time is defined as the number of seconds that have elapsed since 00:00:00 Coordinated Universal Time (UTC), Thursday, 1 January 1970) of every exposure over the 3 Survey is plotted vs the right ascension. The slanting bands are the yearly revisiting of the RA in opposition in bounded by two visits in per year. For a given day, one can look along a row of constant time and see the range in RA ascension covered in a night. This banding is by design in the “strategic” approach, but because the constraints of airmass and sky brightness are generic to an all sky survey, one imagines that the results of the LSST scheduler should show the same pattern if the tension between the parallax cadence (six months) and the sky brightness (twilight) is balanced. Compare Figure 10 with Figure 1 of (Chambers, 2006b) in advance of the survey.
The effectiveness of the strategic scheduling approach and the patient efforts of the PS1 Observers who used the scheduling tools to solve the travelling salesman problem in multiple dimensions and responded interactively to the nightly conditions (clouds, wind speed, cadence, sky brightness, survey completeness) is demonstrated in Figure 11. This shows the actual distribution of pointings in Dec vs Hour Angle for the entire Survey. The hole in the middle is the keyhole characteristic of an Alt-Az telescope. The hour angle distribution shows that 65% of the data is taken within 1.5 hours of the meridian.
3.3. The Medium Deep Survey
The Medium Deep Survey consisted of 10 single PS1 footprints on well studied fields spaced approximately uniformly around the sky in Right Ascension. The pointing centers of these 10 fields are listed in Table 7. The table includes two additional fields of M31 which can be considered an MDS like field (see Section 3.6) and a field at the north ecliptic pole (NEP). The latter was not observed as extensively as the 10 main fields and was only observed over the period 2010-09-20 to 2011-06-17. The individual exposure times per filter were considerably longer than those for the 3 survey (see Table 6).
The Medium Deep Survey (MDS) component of the program regularly visited these 10 fields ( 7 sq. deg. each). Each field was picked to have significant multi-wavelength overlap from previous and concurrent surveys by other teams and facilities (e.g. DEEP2, ELIAS-N, CDFS, COSMOS, GALEX). In total, 25% of the PS1 time was allocated to the MDS. The cadence was generally composed of and together on one night, followed by on the second and on the third. This pattern was repeated continually on a 3 day cycle over the 6-8 month observing season for the field, interrupted only by weather and the moon. Around full moon, the filter was primarily used and hence it does not have the same time cadence as the other 4. Figure 12 illustrates the cadence and observing seasons while Table 6 (originally presented in Rest et al., 2014) summarises the exposure times. On each night the 8 separate exposures were dithered and the field was rotated. The images were then combined into nightly stacks of 904 sec (and ) and 1902 sec (, and ). Roughly one a year these stacks are further combined to produce so-called reference stacks, which are then used as templates for difference imaging. Finally, all the data are combined to produce very deep stacks, which contain several tens of hours worth of exposure.
The description of the MDS was initially presented in Tonry et al. (2012a) and many papers on transients have already given an overview of the data products and survey (e.g. Chomiuk et al., 2011; Gezari et al., 2012; McCrum et al., 2015; Sanders et al., 2015; Lunnan et al., 2016). Estimates of the typical 5 depths of the MDS nightly stacks were given in Rest et al. (2014) and are also listed in Table 6 here. Development work continued to improve the single exposure processing though to deep stacks during the transient event discovery and other science consortium programs over the course of the survey, the culmination of those improvements being applied in a more uniformly reprocessed dataset used for the public data release. A full discussion of the Medium Deep Fields, including improved estimates of depths and their special processing will be presented inHuber et al. 2017 (in preparation - Paper VII). No Medium Deep data will be released in DR1.
|1||,||8113 each||23.1, 23.3|
|Field||RA J2000||Dec J2000||Overlaps|
|MD08||242.787||Elias N1- DXS|
|MD11||270.000||North Ecliptic Pole|
3.4. Solar System Survey
3.5. Pan-Planets stellar transit survey
For Pan-Planets, seven slightly overlapping fields with overall 40 sq. deg. were observed with PS1, making up about 4% of the total survey time (see Table 8). Data were collected between 2009 and 2012 in the -band. Depending on seeing, exposure times were either 30 sec or 15 sec. In the first two years of the survey, three fields were observed. From 2011 on, four additional fields were added to the survey area, meaning that the previous three fields have a higher number of visits. On each survey night, the exposures were cycled through the seven fields to minimize saturation effects. We obtained at least 2000 exposures for each point in our FOV and up to 6000 in the overlapping areas between the fields. The main goal of Pan-Planets is the search for transits from extrasolar planets, mainly hot gas giants close to their star with a special focus on M-dwarfs (Afonso & Henning, 2007) There are up to 60000 M-dwarfs in the FOV with magnitudes between 13mag and 18mag in the i-band, which makes the survey one of the most comprehensive transit searches for M-dwarf exoplanets. A description of the scientific results and analysis can be found in Obermeier et al. (2016). The Pan-Planets stellar transit data is not included in DR1.
|Field||RA J2000||Dec J2000||Overlaps|
3.6. PAndromeda, the M31 transient survey
PS1 had a special monitoring survey for M31 for 2% of the original PS1 survey time. Data were taken from 2010 to 2012 (3 seasons), during the second half of each year when M31 was easily visible. M31 was also covered in the regular 3 Steradian Survey. As part of the separate survey, M31 was visited up to two times per night in the and filters. Depending on the weather conditions, we obtained up to 14 times 60-second exposure in and 10 times 60-second exposures in . These exposures were spread across the two visits per night to give some intra-night time resolution. The survey strategy was optimized to detect short-term M31 microlensing events, but to also allow one to identify and analyze the variable star content in M31. Observations were taken much more sparsely in the remaining filters (, and ) in order to give multicolour maps of M31 in the full PS1 filter complement. The first results and demonstration of data quality from the first 90 nights in 2010 were presented in Lee et al. (2012). M31 data will not be released in DR1.
3.7. Calibration observations, CNP, SAS2
3.7.1 Spectro-photometric and Calspec Standard Stars
The AB magnitude system calibration of the Pan-STARRS1 photometric system by Bohlin et al. (2001) used data from a single photometric night. and special observations of the HST Calspec sample Bohlin et al. (2001). All standard stars were placed on OTA 34 and cell 33, so their integration was on the same silicon and used the same amplifier for read-out. However, this position was very close to the center of the focal plane, where it has been noted that there is a strong gradient in the behavior of the chip (Rest et al. 2014), and thus these observations were not included in the subsequent study by Scolnic et al. (2014, 2015). They analyzed a sample of faint Calspec standards observed over the course of the survey and re-determined the AB offsets for the , , , bands of the PS1 system. The super-cal Scolnic et al. (2015) AB offsets were used in the calibration of all the DR1 and DR2 data Magnier et al. (2016b). However Scolnic et al. (2015) note that primary difference in the update arises from changes in the Calspec standards.
3.7.2 The Celestial North Pole
The 3 Steradian Survey extends to the North Pole. It was soon realized that a dedicated nightly pointing near the Celestial North Pole would provide continuous time coverage that could monitor the performance of the system as well as be of scientific interest for the unique cadence. So a set of exposures of 30 seconds was obtained each night on the meridian and at a declination of 89.5 degrees. Observations were obtained every clear night between 2010-10-13 and 2014-02-13. The net result is about a 4 square degree area with regular observations for 3.3 years. This data is not included in DR1.
3.7.3 Small Area Survey 2
In July 2011 a test area of the 3 survey, consisting of about 70 deg centered on (J2000), was observed to the expected final depth of the survey in . These data are described in depth in Metcalfe et al. (2013) where general issues of the data and the PS1 reduction software is subject to a rigorous investigation, with emphasis on the depth of the stacked survey. A further paper (Farrow et al., 2014) demonstrates how galaxy number counts and the angular two-point galaxy correlation function, w(),can be reliably measured. This data is not included in DR1.
|processing||image||file||ID||Avg No.||No.||No. IDs||location||Release|
|convolved 6&8 signal||fits||skycell||1||1||1050k|
|convolved 6&8 variance||fits||skycell||1||1||1050k|
|convolved 6&8 mask||fits||skycell||1||1||1050k|
|convolved 6&8 number||fits||skycell||1||1||1050k|
|convolved 6&8 exp||fits||skycell||1||1||1050k|
|convolved 6&8 expwt||fits||skycell||1||1||1050k|
|Table Class||PSPS Table Name||Release|
4. Overview of PS1 Data Products
The PS1 Data Products consist of images of various kinds, catalogs of attributes measured from the images organized in a hierarchical relational database, derived data products such as proper motions and photometric redshifts, and metadata for linking and tracking all of the above. Here we provide a brief overview, see Flewelling et al. (2016) for details. We refer below to types of images and data files listed in Tables 9 and 10.
The proper convention when reporting Pan-STARRS1 magnitudes is to use the nomenclature (see Section 2.6) and the convention for IAU names is
where the PSO identifier stands for Pan-STARRS Object, and the coordinates are in decimal degrees. Another point of interest is the use of magnitudes and fluxes. There are advantages and disadvantages to the use of these, and we employ both magnitude and fluxes where useful. Luptitudes (Lupton et al., 1999) also have some advantages, but we have made the decision not to use them in the PS1 data products. One noteable advantage of fluxes, for example is that the magnitude of an aperture flux measurement can correctly be negative when measured on a sky subtracted image and, when the mean of a series of such measurements is computed, the result is well behaved. All Pan-STARRS magnitudes are in AB magnitudes (Tonry et al., 2012b) and the fluxes are reported in the corresponding Janskys, where the absolute calibration is discussed in Scolnic et al. (2014). See also Section 2.6.
4.1. Image Data Products
In this section we discuss the specific kinds of images, their properties and location as produced by the IPP processing stages discussed above in Section 2.7. A summary of the kinds of data files that exist including images is provided in Table 9.
The raw pixel data are archived in two geographically separate locations; one archival copy is retained on storage machines at the IfA’s ATRC on Maui, and another is stored in the IPP cluster, which has moved from its inital location at the Maui High Performance Computing Center (MHPCC) to the Maui Research and Technology Center (MRTC-B) to its permanent location in the UH Information Technology Center on the Manoa campus. Within the IPP all other files have at least two instances on separate Raid 10 machines.
Each Pan-STARRS image (an “exposure” or “frame”) creates 60 fits image files, one for each device in the camera, and each fits file has 64 extensions, where each extension is the pixel data from one OTA cell (see Section 2.5) Table 9 lists the various image and binary fits table files produced by the IPP by each of its stages. Some images are intermediate products and are not saved permanently, although they can be reproduced from the raw data.
The “chip” images are the detrended images. The signal image is now a float, and a matching mask and variance image are also produced. The detection of objects and measurements of their positions and attributes (in detrended pixels) are stored binary fits tables (internally called CMF files). These measurements are therefore in pixel coordinates and can have any orientation on the sky. Together with the astrometric calibration from the “camera stage” these measurements and their position in are the basis of the “Detection Table” Flewelling et al. (2016), These are also binary fits tables, (internally called SMF files).
The “warp” images are astrometrically registered by a geometric transformation onto rectilinear North-South pixels in a tangential projection using the nearest projection center as defined by the RINGS3 tessellation The warp stage also produces warped mask and variance images, see Waters et al. (2016) for complete discussion.
The “stack” images are additions of accumulated warp images which should be precisely registered. Variance and mask images for the stack are also created as well as a number image that shows the number of warps that contributed to the stack at any pixel. Note, the different warps likely have different PSFs having been taken at different times and at different places in the focal plane, leading to what are essentially intractable problems in PSF measurements performed on the stacks. This is the motivation for the “Forced Photometry” stage (Magnier et al., 2016a). The results of the analysis of the stacked images are stored in binary fits tables (again, internally labelled as CMF files) and are available in the Stack tables, (Flewelling et al., 2016).
Two convolved versions of the stack images are created by convolving with gaussians of width 6 or 8 pixels (precisely 1.5 and 2.0 arcseconds in 0.25 arcsecond sky cell pixels). These are then used for aperture measurements (Magnier et al., 2016a). The convolved images are intermediate products and are not saved. The aperture measurments are stored in binary fits CMF files and are available in the Stack Aperture tables.
Difference images and their associated variance and mask images are created by the “Diff Stage” and measured, with the results going in the Difference Tables. The Difference images are not retained, but could in principle be regenerated from the stacks and warps.
4.2. Fundamental Data Products
The Pan-STARRS1 database schema ((Flewelling et al., 2016)) is organized into four sections:
Fundamental Data Products. These are attributes that are calculated from either detrended but untransformed pixels or warped pixels. The instrumental fluxes or magnitudes have been re-calibrated, as have their positions. Because of these calibrations, the catalog values are to be preferred to making a new measurement from the images. See Table 10.
Derived Data Products. These are higher order science products that have been calculated from the Fundamental data products, such as proper motions, photometric redshifts, associations of detections of moving objects by MOPS.
Observational Metadata. This is metadata that provides detailed information about the individual exposures (e.g. PSF model fit) or which exposures went into an image combination (stacks and diffs) of exposures, as well as information such as detection efficiencies.
System Metadata. These tables have fixed information about the system and the database itself, including descriptions of various flags.
Various database ”Views” or logical combinations of Tables are also constructed as an aide for common kinds of queries. Note in the PSPS architecture, large tables (almost all except the ”head node” ObjectThin, MeanObject, and GaiaFrameCoordinate) are actually ”views” joining subsections (slices) of the data across different file partitions, but this structure is hidden from the casual user.
The classes of tables in the Fundamental Data Products include Detection, Object, Stack, Difference, and Forced (Table 10). We now consider each of these in turn.
4.2.1 Detection Table
At the most basic level, an individual “Detection” is a feature, likely a star or galaxy or artifact, detected above the noise in an individual exposure. There are likely, but not always, multiple “detections” of the same astronomical object from subsequent exposures. The majority of single detections at the faint end are not real, but arise from systematic noise, primarily correlated read noise, in the GPC1 (Waters et al., 2016). A wealth of flags are provided to help distinguish between real detections and artifacts (Flewelling et al., 2016). Nonetheless there are detections that arise from systematic noise that are, by themselves, indistinguishable from real features
For each single epoch detection, the Detection Table contains PSF magnitudes, total aperture based magnitudes, Kron magnitudes (Kron (1980)), assorted radial moments and combinations of moments, and circular radial aperture magnitudes in SDSS radii R3 through R11 Stoughton et al. (2002). See Magnier et al. (2016c) for details.
4.2.2 Object Tables
Individual “detections” are associated into ”Objects” by virtue of being approximately at the same location in . The IPP makes this association of detections into objects in the “static sky stage” if they are within 1.0 arcseconds and there are various complications for blended objects or single objects that become resolved in a subsequent, higher quality image (Magnier et al., 2016c). It is possible that an astronomical object is only measured once even with multiple exposures with good pixels at the same location - for example a moving object, or a transient object, or an object that only rises above the noise in one image. Systematic noise, especially the correlated read noise in the GPC1 detector (Waters et al., 2016), can also contribute a faint artifact that is interpreted as a single detection and becomes an Object. Such single instances must also be elevated into ”Objects” because at the time there is no independent way of knowing. Thus the association of detections into objects is one-or-more to one. Thus one-time-only false detections from artifacts are also promoted to Objects. In the Pan-STARRS1 dataset, as a consequence of these features produced by the GPC1, such artifacts dominate the Detection and Object Tables. One simple way to exclude them is to require an Object to have two or more detections; an occurrence which is decreasingly likely to happen if the feature isn’t real. This could also obviously exclude real moving and transient objects. Sample queries to produce robust catalogs from the Pan-STARRS1 database are provided in Flewelling et al. (2016), but one should always be aware of this aspect of Pan-STARRS1 data.
The Object Tables described in Flewelling et al. (2016) include the ObjectThin Table, which contains the most minimal information set about an object, primarily its position and various indexes linking it to other tables. There are two (ra,dec) positions provided, a “Mean” position and a “Stack” position. Mean positions are the most accurate if available, as they come from a mean of all the individual epoch measurements, each of which have been calibrated on the Gaia (Lindegren et al., 2016) reference frame. Objects that are only detected in the stack are fainter and their positions in DR1 have not been re-calibrated on the Gaia frame. This is because their uncertainties are intrinsically larger and hence this is only an issue for the most demanding astrometry.
The MeanObject Table contains the mean photometric information for objects based on the single epoch data, calculated as described in Magnier et al. (2016c). To be included in this table, an object must be bright enough to have been detected at least once in an individual exposure. PSF, Kron Kron (1980) and aperture magnitudes and statistics are provided for all filters.
The GaiaFrameCoordinate Table contains the re-calibration of the astrometric positions of all MeanObjects on the Gaia reference frame (Magnier et al., 2016b).
4.2.3 Stack Tables
Attributes measured on the stacked images are reported in the Stack Tables. The StackObjectThin table contains the most minimal positional and photometric information for point-source photometry of stack detections. The information for all filters are joined into a single row, with metadata indicating if this stack object represents the primary detection. See Section 3.2.3.
The StackObjectAttributes table is analogous to the Detection table for single epoch images and it contains the PSF, KronKron (1980), and aperture fluxes for all filters in a single row, along with assorted radial moments and combinations of moments.
The StackApFlx Table contains the unconvolved fluxes within the SDSS R5, R6, R7 apertures (Stoughton et al., 2002) for all Stack Objects. The StackApFlxExGal Unc, Conv6, Conv8, Tables contain the unconvolved fluxes within the SDSS R3 through R11 apertures (Stoughton et al., 2002) for objects in the extragalactic sky, i.e., they are not provided for objects in the Galactic plane because they are not useful in crowded areas. For each aperture we report: flux (janskys), flux error, flux standard deviation (from the individual measurements), and the fill factor of the aperture (masked pixels could reduce this from 1.0).
The StackPetrosian and StackModelFit Exp, DeV, and Ser tables report the results of fits of extended sources to model PSF convolved surface brightness profiles. The measurements include Petrosian magnitudes and radii, Exponential, de Vaucouleurs, and Sersic magnitudes and radii and elliptical aperture magnitudes and errors for a signal-to-noise ratio and galactic latitude limited sample. See Magnier et al. (2016c) for details.
4.2.4 Difference Tables
The IPP generates Alard-Lupton convolved difference images for skycells in various combinations depending on the survey goals. For the 3 Survey a difference image: is created for each epoch.
This difference image is then analyzed in the same fashion as an individual warp. The DiffDetection table is analogous to the Detection table and has the same measurements. If possible, DiffDetections are associated into DiffObjects, e.g. different points on a light curve are associated into a Difference Object.
No attempt is made to associate Diff Objects with Objects. While this might make sense for a variable object, a transient source, e.g. a supernovae in a galaxy, could be undetectable in either an individual warp or stack, and yet be clear in the difference image. In this case the closest Object would be the host galaxy, but that association would be incorrect. Hence Difference Objects are a unique class and not contained in the Object Table. On the other hand, a “good” match between Objects and DiffObjects would provide a candidate for a variable object.
4.2.5 Forced Photometry Tables
Forced photometry is carried out at the positions of all significant objects found in at least two bands. This requirement keeps the number of forced objects to a practical number. Single band detections, especially z-band dropouts or objects found only in y-band are a non-trivial subject of active research.
The forced detection measurements made on individual warp images are reported in the ForcedWarpMeasurement table. Where the field-of-view of the exposure contains the position of the object, but its properties can not measured because the data happens to be masked at the position, the object’s identify is stored in the ForcedWarpMasked Table. The ForcedWarpExtended table contains the single epoch forced photometry fluxes within the SDSS R5, R6, and R7 apertures (Stoughton et al., 2002). ForcedWarpLensing contains the contains the mean lensing parameters (Kaiser et al., 1995) of objects detected in stacked images measured on the individual single epoch data. The individual epoch measurements are not reported, only their mean. ForcedMeanObject has the mean properties of the individual forced measurements, including PSF, Kron, and aperture magnitudes, and R5, R6, and R7 apertures. See Magnier et al. (2016c).
4.3. Derived Data Products
Derived data products are results that can not be traced directly back to the pixels but are the result of systematic analysis of the Fundamental Data Products discussed above. These include (i) measurements of proper motion and parallax (Magnier et al in prepation) made from an analyis of the minute changes in the positions of objects; (ii) Photometric redshifts deduced from aperture magnitudes using a variety of machine learning techniques ,(iii) the extinction and stellar parameters deduced from stellar photometry ; (iv) associating detections in the database with known or discovered moving objects in the MOPS database. We have the tools to ingest these derived data products back into the PSPS and make them widely available to the community. Our intention is to include these derived data products in future Data Releases.
5. Accuracy, and precision of the PS1 data set
The Pan-STARRS1 astrometry has been re-calibratedMagnier et al. (2016b) using Gaia (Gaia Collaboration et al., 2016b) The Gaia DR1 (Gaia Collaboration et al., 2016a) catalog (Lindegren et al., 2016) was used as the input reference catalog. After recalibrating all individual epoch measurements to the Gaia Frame, and then re-constructing the mean PS1 positions we can estimate the astrometric error of the resulting PS1 postions. The systematic uncertainty of the astrometric calibration using the Gaia frame comes from a comparison of the results with Gaia: the standard deviation of the mean and median residuals ( ) are (2.3, 1.7) milliarcsec, and (3.1, 4.8) milliarcsec. The latter is a measurement of the bright end errors for average positions while the former is a measurement of the consistency of the PS1 and Gaia systems(Magnier et al. (2016b)).
The photometric accuracy of the PS1 data products has been demonstrated in the ubercal analysis (Schlafly et al., 2012) and relative photometric analysis (Magnier et al., 2013). Zero points for photometric data are determined with a reliability of 7-12 millimags. Individual detections in the 3 survey have photometric accuracy limited at the bright end to 12 millimags per epoch. The current limits on the photometric precision are driven by our ability to model the 2D variations in the shape of the PSF. The PSF shape in a given exposure changes on a variety of spatial scales due to 3 major effects: the atmosphere, the optics, and the detector. To the extent that the PSF model is unable to follow the PSF variations, the PSF photometry is biased either high or low as the model PSF under or over predicts the size of the PSF. The optics introduce image quality variations due to ripples in the focal surface. These variations occur on spatial scales of 10 arcminutes and are relatively stable between exposures, introducing photometry errors of a few millimags. The atmosphere introduces stochastic variations due to uncorrelated seeing across the focal plane, with a similar level of impact. The detectors introduce PSF changes due to variable diffusion resulting from variations in the doping characteristics with spatial scales down to 10s of arcseconds. With the density of PSF stars available in a typical PS1 exposure at high Galactic latitude, we are able to model the PSF variations on spatial scales of 3 arcmin, placing a limit on the accuracy of the PSF model on small scales. See Magnier et al. (2016b) for details.
|validated||exposure||single||star||of PSF||of PSF||of sky||of sky||stack|
6. 3 Survey Characteristics
The Steradian Survey is comprised of 374,446 validated images taken between 2009-06-03 and 2015-02-25. This includes some commissioning data taken before the start of the Mission and a modest number of images taken after the formal end of the survey, primarily in during twilight to smooth out the spatial distribution.
6.1. Summary of performance metrics
The image quality in the Pan-STARRS1 Surveys varies significantly, this is one reason for the forced photometry measurements. The site, Haleakala Observatories (HO), is a well characterized site, and the lower limit to the seeing distribution is equivalent to Mauna Kea, but the distribution is broader and the median seeing as recorded by the HO Differential Image Motion Monitor (DIMM) is 0.84 arcsec, with a mode of 0.66 arcsec. However, PS1 has a floor to its image quality, arising primarily from the wide field optics, so even the best images do not have a FWHM . The image quality also depends on the filter, with the reddest bands displaying the best. Figure 15 shows the cumulative distribution of the image quality as characterized by a FWHM for each filter for the PS1 Surveys. Haleakala is known for very low atmospheric scattering, that is why it is preferred to Mauna Kea for the Solar Telescope, the sky is even darker than Mauna Kea. Solar astronomers assert this is due to the fact that the summit of Haleakala is primarily rock, whereas the summit of Mauna Kea is primarily cinder, and that summit of Mauna Kea is constantly surrounded by a halo of microscopic volcanic cinders. Figure 15 also shows the cumulative distribution of the sky brightness in each filter for the PS1 Surveys.
Table 11 provides a summary of the characteristics of the 3 Survey.
6.2. Simple Star/galaxy separation
For the DR1 and DR2 releases we recommend using a simple cut in (PSF - Kron) magnitude space to separate stars from galaxies. Figure 16 shows v - for -band data around the galactic pole region. Unresolved objects form the tight sequence around PSF-Kron. A cut of (PSF-Kron) does a reasonable job of selecting stars down to . Figure 17 shows the star and galaxy counts resulting from such a cut. Faintward of the number of stars is over-predicted by a simple linear cut like this. Also, at the brightest magnitudes, saturated stars tend to get classified as extended by this technique, resulting in a peak in the galaxy counts at . The use of the IPP flags or a more sophisticated non-linear cut can relieve this problem to some extent. The distribution of stars and galaxies on the sky is visualised in Figure 18.
A more detailed discussion of this technique applied to PS1 data, including the behaviour of synthetic stars and galaxies, can be found in Farrow et al. (2014).
6.3. Variation of Steradian Survey Depth
Although by the design of the survey each pixel on the sky notionally has 12 visits, in practice the coverage can be much more variable than this. Figure 19 shows the distribution of the number of exposures which contribute to each 16x16 binned pixel (4x4″) over the whole of the 3 stacked survey. The result of this is that the depth of the stacked survey varies significantly on quite small scales. To estimate the depth, in the reduction of a skycell, artificial point-sources are added in magnitude bins and run through the process of being detected. The numbers of these fake sources recovered and inserted as a function of magnitude is stored, for the stacked data, in the StackDetEffMeta table. Maps of depth can be produced by finding at what magnitude a particular percentage of fake point-sources is recovered for each skycell, using linear interpolation between different magnitude bins when necessary. To visualise the results across the whole survey, it is convenient to take the mean of these magnitudes for each skycell landing in a particular HEALpix555http://healpix.sourceforge.net (Górski et al., 2005) pixel. Figures 20 and 21 show the results of this procedure for recovery rates of 50% and 98%. Not all the variation in limiting magnitude seen is due to the coverage. For instance, in the galactic plane crowding can significantly reduce the number of recovered fakes. It should also be noted that these limits are for point sources - Metcalfe et al. (2013) showed that the limits for extended sources are roughly mag brighter, although this is, of course, depends on the profile of the source.
Tests of how well these fake sources reproduce the true point-source recovery fractions, as well as a method of producing even higher resolution maps of depth will be presented in Farrow et al (in preparation).
6.4. Examples of stellar and galactic photometry
In Figure 22 we show examples of stellar color-magnitude diagrams for a variety of well-known galactic globular star clusters, as well as the Local Group dwarf galaxy Leo I. These data were taken from the MeanObject table, and hence represent the mean of the measurements on individual exposures. Despite the crowded nature of these fields, the stellar sequences are still quite tightly defined.
Figure 23 demonstrates the use of Kron magnitudes for galaxies, taken from the stacked data. Here we show the Coma galaxy cluster, with its prominent sequence at -.
Finally, in Figure 24 we display the -v -, -v -and -v -color-color locii for stars and galaxies around the galactic pole region.
7. The PS1 Science Consortium Science Legacy
The Pan-STARRS1 Surveys of the PS1 Science Consortium have enabled science on topics ranging from Near Earth Objects to the most distant quasars. New discoveries will be enabled by providing access to the community to the Pan-STARRS1 Archive at the Barbara A. Mikulski Archive for Space Telescopes. While Pan-STARRS1 is not a space telescope, among other kinds of science these surveys will advance is a means to reprocess the astrometry of the Hubble Space Telescope Archive based on the Pan-STARRS1 extention of the Gaia Reference Frame Magnier et al. (2016b). Below we provide a brief summary of the legacy science from the PS1 Science Consortium as examples of the kind of science that can be done with the Pan-STARRS1 Surveys.
A primary goal of the PS1 mission was the Solar System Survey, designed to discover previously unknown Near-Earth Objects (NEOs) and provide additional orbital information on known bodies. So far, PS1 has been responsible for discovering over 2900 NEOs, including potential targets for both robotic and manned space missions. PS1 survey data has also led to the discovery of 129 comets and tens of thousands of new main-belt asteroids. A major legacy of both the PS1SC Solar System Survey and the continuing NEO survey (PI Wainscoat) has been the reporting to date of astrometric and photometric measurements of moving objects to the IAU Minor Planet Center. A key science result from this treasure trove has been a determination of the luminosity distribution of NEOs down to diameters of just a few metres by Schunová-Lilly et al. (2017) Additionally, the photometric properties of approximately a quarter of a million main-belt asteroids has been published by Vereš et al. (2015) Looking at rarer objects, PS1 has allowed characterisation of the Main-Belt Comet (Hsieh et al. (2015)), and constrained the number of observable asteroid breakups (Denneau et al. (2015)). Pre-discovery imaging of comet ISON out to Saturns’ orbit demonstrated the value of deep solar system surveys in constraining the cometary activity of inbound long-period comets (Meech et al. (2013)).
The rapid nightly processing of data to search for NEOs yielded objects with motions as slow as 0.05 deg/day, depending on seeing, and yielded numerous Centaurs, making Pan-STARRS1 one of the most prolific discovery telescopes for Centaurs. The cadence of the observations was tuned for discovery of faster moving objects such as NEOs, making it not ideal for discovery of outer solar system objects. Nevertheless, the Pan-STARRS1 dataset is rich in observations of the outer solar system. Weryk et al. (2016) describe a search for distant solar system objects using the archival PS1 data for the period 2010 Feb 24 to 2015 July 31. A total of 607 distant solar system objects were identified, 332 being new first observation discoveries, with an additional 24 significantly improving the astrometry of previously designated objects. While a large number of new objects were found, no new extreme TNOs showing a clustering in their argument of perihelia were found, which, if present, could support the presence of a distant planetary sized perturber in the outer solar system. Lin et al. (2016) describe the discovery of five new Neptune Trojans in the PS1 data. Four of these may be primordial, but the fifth is likely a recent capture. Chen et al. (2016) describe the discovery of a retrograde TNO in the PS1 dataset, and show that this object has similar orbital characteristics to other low semimajor axis high-inclination TNOs and Centaurs, hinting at a common orbital plane.
The scheduling and filter set of Pan-STARRS1 was designed from the beginning to enable the complete census and study of the ultracool dwarfs ( K) of the Solar neighborhood, as well as white dwarfs. The combination of proper motions and parallaxes allows for relaxed color selections to obtain complete, volume-limited samples (Deacon et al., 2011b). Proper motions and colors also allow for the detection of substellar companions to main sequence stars, putting constraints on their age and mass (Deacon et al., 2014, 2012a, 2012b); peculiar brown dwarfs such as low-gravity, young brown dwarfs (Liu et al., 2013b); members of nearby stellar structures (Aller et al., 2016; Best et al., 2015; Goldman et al., 2013). The depth of the stacked images provides the first colors in the visual and red part of the spectrum for a large number of ultra-cool stars and brown dwarfs (Best et al. in prep). The accuracy of both Pan-STARRS 1 astrometry and photometry also allowed the characterisation of Kepler target stars and the discovery of wide binary companions to planet hosts Deacon et al. (2016).
The panoptic Pan-STARRS1 survey provides a unique opportunity to map the distribution of stars (e.g. Morganson et al. 2016) in our own Milky Way and its outskirts and place it in a cosmological context. In particular, it revealed the presence of multiple very faint Milky Way satellites (e.g. Laevens et al. 2015) and stellar streams likely stemming from the tidal disruption of globular clusters by our Galaxy (Bernard et al., 2016). Exploiting PS1 as a time-domain survey, the largest and deepest sample of RRLyrae candidate stars was identified (Hernitschek et al., 2016), which provides for unprecedented 3-D mapping of the Milky Way’s stellar halo out to 120 kpc. A major goal of Pan-STARRS1 Milky Way science was to map the interstellar dust in 3 dimensions using star colors. For this purpose, PS1 has three main advantages over SDSS: it is deeper, goes one band redder, and covers more of the low-latitude sky. The collaboration developed a method to infer the posterior on distance and reddening of each star, and then group stars into angular pixels and estimate the reddening as a function of distance in each (Green et al., 2014). They applied this method to PS1 photometry of 800 million stars (some with 2MASS photometry as well) and created a map with 2.4 million angular pixels and 31 distance bins, covering 3/4 of the sky (Green et al., 2015). They used a variant of the same technique to produce the largest catalog of molecular cloud distances (Schlafly et al., 2014). More recently, PS1 photometry has served as the basis of a new parameterization of the reddening law, and study of its variation in 2 and 3 dimensions (Schlafly et al., 2016b, a).
The Pan-Planets survey was a dedicated exo-planet transit survey within the PS 1 project. The survey covered an area of 42 sq degrees in the galactic disk for about 165 hr with the goal to constrain the occurrence rate of hot Jupiters around M dwarfs. A combination of SED fiting, dust maps, and proper motion information allowed to identify more than 60 000 M dwarfs in the field. This is the largest sample of low-mass stars observed in a transit survey. With this large sample size, the Pan-Planets survey resulted in an occurrence rate of hot Jupiters of 0.11 (+0.37-0.02) % in case one of our candidates turns out to be a real detection. If, however, none of our candidates turn out to be true planets, we are able to put an upper limit of 0.34% with a 95% confidence on the hot Jupiter occurrence rate of M dwarfs. This is the best limit for the occurrence rate of hot Jupiters around M stars so far.
The major science goal for the Andromeda monitoring with Pan-STARRS1 (PAndromeda) originally was to identify a large number of gravitational microlensing events towards M31. The final depths and image quality of the survey meant that early expectations of event rates were not met. Nevertheless we found 6 events in a subfield of the first year of the survey (Lee et al., 2012) and the data are suitable for other PAndromeda stellar science. Using only a subset of the PAndromeda data we identified and classified Cepheids and analysed their period luminosity/Wesenheit relations (Kodric et al., 2013) and we found eclipsing binaries in M31 (Lee et al., 2014b) from which a handful of the brightest ones are suitable to derive an independent M31 distance. We furthermore searched for rare variable stars which help to understand stellar structure (17 Beat Cepheids; Lee et al. (2013)) and we also identified four new LBVs, i.e. potential supernova progenitors (Lee et al., 2014a). The combination of the PAndromeda variability analysis and the HST-PHAT data (Dalcanton et al., 2012) turned out to be very powerful. In Kodric et al. (2015) we presented the largest M31 (HST) near infrared band and band sample at this time (371 Cepheids), studied their near infrared period luminosity relations and showed that the bright part of our sample is well suited for determination using M31 (having no metallicity issues compared to LMC/SMC) for the Cepheid distance ladder. In summer 2016 we finished to completely (re)do the PAndromeda difference imaging for the full survey (time and area) by optimzing our pipeline for the data characteristics and by increasing the masking fraction. We hence now have the final PAndromeda data products, i.e. light curves for “all” variable sources in M31. The analysis of these light curves is ongoing, e.g., the final catalogue and analysis for PAndromeda Cepheids (750 with HST NIR photometry) will be made available to the community in 2017 (Kodric et al. 2017, in preparation).
The Medium Deep Survey (MDS, Section 3.3) was designed for both deep fields and transient science with multi-colour temporal coverage of 70 square degrees in total. In addition the multi-epoch aspect of the 3 survey also provided a new opportunity for time domain science. The data for the MDS are not in DR1, but they will be released at a future date. The cadence and filter coverage of the MDS fields was designed to both discover type Ia supernovae (SN Ia) before maximum light and to sample their lightcurves sufficiently for distance measurements. We discovered 3000 SNe Ia within , and obtained spectroscopic confirmation for 500 SNe Ia. Rest et al. (2014); Scolnic et al. (2014) published data from the first 1.5 years of the MDS, in which 146 SNe Ia were used to constrain the dark energy equation-of-state parameter, , to 7% An analysis of the full sample is in preparation, and we have undertaken a first step in analyzing the photometric sample (Jones et al., 2016). The MDS was also a rich source of exotic transients, with the discovery of high redshift superluminous supernovae. These supernovae are 100 times brighter than normal core-collapse evens and peak at mag. They are UV bright and PS1 has discovered some of the most distant SLSNe, including one at (Berger et al., 2012). In a series of papers we studied their physical parameters (Chomiuk et al., 2011; Lunnan et al., 2013; McCrum et al., 2014; Lunnan et al., 2016), and their host-galaxy environments (Lunnan et al., 2014, 2015) and rates (McCrum et al., 2015). The MDS discovered two tidal disruption events, PS1-10jh (Gezari et al., 2012, , the best studied TDE to date) and PS1-11af (Chornock et al., 2014), defined a class of fast-declining transients Drout et al. (2014) and provided an extensive study of type II explosions(Sanders et al., 2015). In addition, the combination of the MDS and GALEX led to complementary UV data on some transients Botticella et al. (2010); Gezari et al. (2010). The 3 survey provided discovery and critical lightcurve points for a number of interesting objects. Low redshift super-luminous supernovae (Pastorello et al., 2010; Inserra et al., 2013; Nicholl et al., 2013) were either discovered with the 3 survey or had lightcurve data at critical points. The survey also provided the detection of a pre-supernova outburst of a type IIn explosion (Fraser et al., 2013). Initially we ran a transient search by catalogue matching with SDSS, which provided the discovery of slowly evolving blue transients at the centres of galaxies (Lawrence et al., 2016). After the creation of an all sky stack, we progressed to routine difference imaging, leading to discovery of some super-luminous supernovae at lower redshift than in MDS (Inserra et al., 2016; Nicholl et al., 2016). Pan-STARRS1 is now the world leading disoverer of low-redshift supernovae, according to the IAU statistics666https://wis-tns.weizmann.ac.il/stats-maps.
The Steradian Survey is being used for a citizen-scientist enabled nearby galaxy survey based on the optical imaging from Pan-STARRS1, but also incorporating multi-wavelength data from the ultraviolet and infrared regimes (GALEX, WISE). This project is called the PS1 Optical Galaxy Survey or POGS777http://pogs.theskynet.org (Vinsen & Thilker, 2013, Thilker et al. in prep). Distributed computing resources contributed by tens of thousands of volunteers allow comprehensive pixel-by-pixel spectral energy distribution (SED) fitting for galaxies, which in turn provides key physical parameters such as the local stellar mass surface density, star formation rate (SFR), and dust attenuation. Sufficiently nearby galaxies are being processed using complete UV-optical-IR SED coverage, whereas distant (but more numerous) galaxies are analyzed with optical only data due to the resolution of ancillary observations. With pixel SED fitting output, the POGS pipeline constrains parametric models of galaxy structure and measures non-parametric morphology indicators in a more meaningful way than ordinarily achieved, by operating on images of estimated physical parameters. The depth, sky-coverage and time-domain capabilities of PS1 have also been leveraged to conduct various focused studies of galaxy properties, including host galaxy properties of variability selected AGN (Heinis et al., 2016) the structure of outer galactic disks (Zheng et al., 2015) and the influence of group environment on the SFR-stellar mass relation Lin et al. (2014).
Early data from the 3 survey have been used to quantify galaxy angular clustering (Farrow et al., 2014), confirm and determine redshifts for Planck cluster candidates (Liu et al., 2015) and detect a large void in front of the CMB Cold Spot (Szapudi et al., 2015). Preliminary data from the MDS has been used to find galaxy groups and clusters and investigate the dependence of star formation on environment (Jian et al., 2014; Lin et al., 2014).
One of the key science goals from the inception of PS1 was the discovery of quasars at the highest redshifts (z6), and to push the redshift barrier of z=6.4 (imposed at the time by the choice of the SDSS filters). These highâredshift quasars are thought to be one of the most massive structures that exist in the first Gyr of the Universe. The quasar host galaxies harbour accreting supermassive black holes, which allow detailed studies of key quasar properties, as well as the impact of the quasars on the surrounding intergalactic medium. PS1 has now become the survey in which most of the z6 quasars have been discovered, and a number of papers have resulted from this effort (e.g., first papers by Morganson et al. (2012) and Bañados et al. (2014)). The highest-redshift quasars found by PS1 are discussed in Venemans et al. (2015), and the most complete catalog of the high-z quasar population is published in Bañados et al. (2016). The latter discusses the properties of 77 newly detected PS1 quasars (out of a total of 124 known) at z5.6.
We have presented an introduction to and overview of the Pan-STARRS1 surveys in preparation for the first data release (DR1) on 2016 December 19. Under the auspices of the Pan-STARRS1 Science Consortium, PS1 observed the entire sky north of Dec deg (the 3 survey) in (to ), additional ecliptic fields in for a Solar System survey (to per visit CHECK!), several extragalactic deep fields in (to per visit), and some specialized fields (M31, transit survey fields). These data have been calibrated to mmag internal photometric precision and mas internal astrometric precision. DR1 consists of the image stacks and associated catalogs for the 3 survey, distributed through the MAST system at STScI. The second data release (DR2, expected May 2017) will distribute the 3 time-domain data. Future data releases will release the Medium Deep data, difference image data, and photometric redshifts from the Photo-Classification Server.
These data have already produced a variety of science, on subjects as varied as asteroids, Milky Way structure, galaxy formation, supernovae and cosmology, but we hope they will find further utility in achieving science goals beyond the scope of the Pan-STARRS1 Science Consortium. The PS1 data are currently being used to provide targeting information for the SDSS-IV Time Domain Spectroscopic Survey (Morganson et al., 2015), and to provide high resolution, deep, multi-colour reference images for transients in many transient surveys (e.g. PESSTO, ATLAS, ASASSN, GAIA Smartt et al., 2015; Tonry et al., 2016; Holoien et al., 2017; Blagorodnova et al., 2016). Besides additional science investigations based directly on these data, the data have great legacy value in providing a high-quality network of calibration sources across the sky. The data are already being used to calibrate the Hyper Suprime-Cam Survey (Toba et al., 2015; Chan et al., 2016) and will also be useful in cross-checking the calibration of the northern areas imaged by the Dark Energy Survey and the Large Synoptic Survey Telescope. But in addition to large-scale surveys, individual programmes with relatively small observing fields north of Dec deg will all have good PS1 calibration sources in the field, observed through exactly the same column as the target sources, allowing simple relative calibration for astrometry and photometry.
Finally, the legacy of the PS1 surveys extends beyond just the data. The experience and lessons learned from designing and executing the PS1 surveys are assisting in the development of future survey projects. Some algorithms and code from PS1 are being used in the development of LSST, and students and postdocs who built careers starting with PS1 are applying their experience to new and larger surveys. We hope the PS1 surveys will be useful to the astronomical community for many years to come.
- Afonso & Henning (2007) Afonso, C., & Henning, T. 2007, in Astronomical Society of the Pacific Conference Series, Vol. 366, Transiting Extrapolar Planets Workshop, ed. C. Afonso, D. Weldrake, & T. Henning, 326
- Aller et al. (2016) Aller, K. M., Liu, M. C., Magnier, E. A., et al. 2016, ApJ, 821, 120
- Bañados et al. (2014) Bañados, E., Venemans, B. P., Morganson, E., et al. 2014, AJ, 148, 14
- Bañados et al. (2016) Bañados, E., Venemans, B. P., Decarli, R., et al. 2016, ApJS, 227, 11
- Berger et al. (2012) Berger, E., Chornock, R., Lunnan, R., et al. 2012, ApJ, 755, L29
- Bernard et al. (2016) Bernard, E. J., Ferguson, A. M. N., Schlafly, E. F., et al. 2016, MNRAS, 463, 1759
- Bertin (2012) Bertin, E. 2012, in Astronomical Society of the Pacific Conference Series, Vol. 461, Astronomical Data Analysis Software and Systems XXI, ed. P. Ballester, D. Egret, & N. P. F. Lorente, 263
- Best et al. (2015) Best, W. M. J., Liu, M. C., Magnier, E. A., et al. 2015, ApJ, 814, 118
- Blagorodnova et al. (2016) Blagorodnova, N., Van Velzen, S., Harrison, D. L., et al. 2016, MNRAS, 455, 603
- Bohlin et al. (2001) Bohlin, R. C., Dickinson, M. E., & Calzetti, D. 2001, AJ, 122, 2118
- Botticella et al. (2010) Botticella, M. T., Trundle, C., Pastorello, A., et al. 2010, ApJ, 717, L52
- Carter & Williams (1973) Carter, W. E., & Williams, J. D. 1973, in The Earth’s Gravitational Field and Secular Variations in Position, ed. P. V. Angus-Leppan, A. G. Bomford, J. C. Dooley, & R. S. Mather, 433
- Chambers (2006a) Chambers, K. 2006a, in The Advanced Maui Optical and Space Surveillance Technologies Conference, E39
- Chambers (2006b) Chambers, K. C. 2006b, Mission Concept Statement for PS1, doi:10.5281/zenodo.199843
- Chambers (2007) —. 2007, PS1 Science Goals Statement, doi:10.5281/zenodo.199832
- Chambers & Denneau (2008) Chambers, K. C., & Denneau, L. J. 2008, PS1 Design Reference Mission, doi:10.5281/zenodo.199860
- Chan et al. (2016) Chan, J. H. H., Suyu, S. H., More, A., et al. 2016, ApJ, 832, 135
- Chen et al. (2016) Chen, Y.-T., Lin, H. W., Holman, M. J., et al. 2016, ApJ, 827, L24
- Chomiuk et al. (2011) Chomiuk, L., Chornock, R., Soderberg, A. M., et al. 2011, ApJ, 743, 114
- Chornock et al. (2014) Chornock, R., Berger, E., Gezari, S., et al. 2014, ApJ, 780, 44
- Dalcanton et al. (2012) Dalcanton, J. J., Williams, B. F., Lang, D., et al. 2012, ApJS, 200, 18
- Deacon et al. (2011a) Deacon, N. R., Liu, M. C., Magnier, E. A., et al. 2011a, AJ, 142, 77
- Deacon et al. (2011b) —. 2011b, AJ, 142, 77
- Deacon et al. (2012a) Deacon, N. R., Liu, M. C., Magnier, E. A., et al. 2012a, Astrophys. J., 755, 94
- Deacon et al. (2012b) —. 2012b, Astrophys. J., 757, 100
- Deacon et al. (2014) Deacon, N. R., Liu, M. C., Magnier, E. A., et al. 2014, ApJ, 792, 119
- Deacon et al. (2016) Deacon, N. R., Kraus, A. L., Mann, A. W., et al. 2016, Mon. Not. R. Astron. Soc., 455, 4212
- Denneau et al. (2013) Denneau, L., Jedicke, R., Grav, T., et al. 2013, PASP, 125, 357
- Denneau et al. (2015) Denneau, L., Jedicke, R., Fitzsimmons, A., et al. 2015, Icarus, 245, 1
- Drout et al. (2014) Drout, M. R., Chornock, R., Soderberg, A. M., et al. 2014, ApJ, 794, 23
- Farrow et al. (2014) Farrow, D. J., Cole, S., Metcalfe, N., et al. 2014, MNRAS, 437, 748
- Finkbeiner et al. (2015) Finkbeiner, D. P., Schlafly, E. F., Schlegel, D. J., et al. 2015, ArXiv e-prints, arXiv:1512.01214
- Flewelling et al. (2016) Flewelling, H. A., Magnier, E. A., Chambers, K. C., et al. 2016, ArXiv e-prints, arXiv:1612.05243
- Fraser et al. (2013) Fraser, M., Magee, M., Kotak, R., et al. 2013, ApJ, 779, L8
- Gaia Collaboration et al. (2016a) Gaia Collaboration, Brown, A. G. A., Vallenari, A., et al. 2016a, A&A, 595, A2
- Gaia Collaboration et al. (2016b) Gaia Collaboration, Prusti, T., de Bruijne, J. H. J., et al. 2016b, A&A, 595, A1
- Gezari et al. (2010) Gezari, S., Rest, A., Huber, M. E., et al. 2010, ApJ, 720, L77
- Gezari et al. (2012) Gezari, S., Chornock, R., Rest, A., et al. 2012, Nature, 485, 217
- Goldman et al. (2013) Goldman, B., Röser, S., Schilbach, E., et al. 2013, A&A, 559, A43
- Górski et al. (2005) Górski, K. M., Hivon, E., Banday, A. J., et al. 2005, ApJ, 622, 759
- Grav et al. (2011) Grav, T., Jedicke, R., Denneau, L., et al. 2011, PASP, 123, 423
- Green et al. (2014) Green, G. M., Schlafly, E. F., Finkbeiner, D. P., et al. 2014, ApJ, 783, 114
- Green et al. (2015) —. 2015, ApJ, 810, 25
- Heasley (2008) Heasley, J. N. 2008, in American Institute of Physics Conference Series, Vol. 1082, American Institute of Physics Conference Series, ed. C. A. L. Bailer-Jones, 352–358
- Heinis et al. (2016) Heinis, S., Gezari, S., Kumar, S., et al. 2016, ApJ, 826, 62
- Hernitschek et al. (2016) Hernitschek, N., Schlafly, E. F., Sesar, B., et al. 2016, ApJ, 817, 73
- Hodapp et al. (2004a) Hodapp, K. W., Siegmund, W. A., Kaiser, N., et al. 2004a, in Proc. SPIE, Vol. 5489, Ground-based Telescopes, ed. J. M. Oschmann, Jr., 667–678
- Hodapp et al. (2004b) Hodapp, K. W., Kaiser, N., Aussel, H., et al. 2004b, Astronomische Nachrichten, 325, 636
- Holoien et al. (2017) Holoien, T. W.-S., Stanek, K. Z., Kochanek, C. S., et al. 2017, MNRAS, 464, arXiv:1604.00396
- Hsieh et al. (2015) Hsieh, H. H., Denneau, L., Wainscoat, R. J., et al. 2015, Icarus, 248, 289
- Huber et al. (2015) Huber, M., Carter Chambers, K., Flewelling, H., et al. 2015, IAU General Assembly, 22, 58303
- Inserra et al. (2013) Inserra, C., Smartt, S. J., Jerkstrand, A., et al. 2013, ApJ, 770, 128
- Inserra et al. (2016) Inserra, C., Smartt, S. J., Gall, E. E. E., et al. 2016, ArXiv e-prints, arXiv:1604.01226
- Jian et al. (2014) Jian, H.-Y., Lin, L., Chiueh, T., et al. 2014, ApJ, 788, 109
- Jones et al. (2016) Jones, D. O., Scolnic, D. M., Riess, A. G., et al. 2016, ArXiv e-prints, arXiv:1611.07042
- Kaiser et al. (1995) Kaiser, N., Squires, G., & Broadhurst, T. 1995, ApJ, 449, 460
- Kaiser et al. (2000) Kaiser, N., Tonry, J. L., & Luppino, G. A. 2000, PASP, 112, 768
- Kaiser et al. (2002) Kaiser, N., Aussel, H., Burke, B. E., et al. 2002, in Proc. SPIE, Vol. 4836, Survey and Other Telescope Technologies and Discoveries, ed. J. A. Tyson & S. Wolff, 154–164
- Kaiser et al. (2010) Kaiser, N., Burgett, W., Chambers, K., et al. 2010, in Proc. SPIE, Vol. 7733, Ground-based and Airborne Telescopes III, 77330E
- Kodric et al. (2013) Kodric, M., Riffeser, A., Hopp, U., et al. 2013, AJ, 145, 106
- Kodric et al. (2015) Kodric, M., Riffeser, A., Seitz, S., et al. 2015, ApJ, 799, 144
- Kron (1980) Kron, R. G. 1980, ApJS, 43, 305
- Laevens et al. (2015) Laevens, B. P. M., Martin, N. F., Bernard, E. J., et al. 2015, ApJ, 813, 44
- Lawrence et al. (2016) Lawrence, A., Bruce, A. G., MacLeod, C., et al. 2016, MNRAS, 463, 296
- Lee et al. (2012) Lee, C.-H., Riffeser, A., Koppenhoefer, J., et al. 2012, AJ, 143, 89
- Lee et al. (2013) Lee, C.-H., Kodric, M., Seitz, S., et al. 2013, ApJ, 777, 35
- Lee et al. (2014a) Lee, C.-H., Seitz, S., Kodric, M., et al. 2014a, ApJ, 785, 11
- Lee et al. (2014b) Lee, C.-H., Koppenhoefer, J., Seitz, S., et al. 2014b, ApJ, 797, 22
- Lin et al. (2016) Lin, H. W., Chen, Y.-T., Holman, M. J., et al. 2016, AJ, 152, 147
- Lin et al. (2014) Lin, L., Jian, H.-Y., Foucaud, S., et al. 2014, ApJ, 782, 33
- Lindegren et al. (2016) Lindegren, L., Lammers, U., Bastian, U., et al. 2016, A&A, 595, A4
- Liu et al. (2015) Liu, J., Hennig, C., Desai, S., et al. 2015, MNRAS, 449, 3370
- Liu et al. (2013a) Liu, M. C., Magnier, E. A., Deacon, N. R., et al. 2013a, ApJ, 777, L20
- Liu et al. (2013b) —. 2013b, ApJ, 777, L20
- Lunnan et al. (2013) Lunnan, R., Chornock, R., Berger, E., et al. 2013, ApJ, 771, 97
- Lunnan et al. (2014) —. 2014, ApJ, 787, 138
- Lunnan et al. (2015) —. 2015, ApJ, 804, 90
- Lunnan et al. (2016) —. 2016, ApJ, 831, 144
- Lupton et al. (1999) Lupton, R. H., Gunn, J. E., & Szalay, A. S. 1999, AJ, 118, 1406
- Magnier et al. (2013) Magnier, E. A., Schlafly, E., Finkbeiner, D., et al. 2013, ApJS, 205, 20
- Magnier et al. (2016a) Magnier, E. A., Chambers, K. C., Flewelling, H. A., et al. 2016a, ArXiv e-prints, arXiv:1612.05240
- Magnier et al. (2016b) Magnier, E. A., Schlafly, E. F., Finkbeiner, D. P., et al. 2016b, ArXiv e-prints, arXiv:1612.05242
- Magnier et al. (2016c) Magnier, E. A., Sweeney, W. E., Chambers, K. C., et al. 2016c, ArXiv e-prints, arXiv:1612.05244
- McCrum et al. (2014) McCrum, M., Smartt, S. J., Kotak, R., et al. 2014, MNRAS, 437, 656
- McCrum et al. (2015) McCrum, M., Smartt, S. J., Rest, A., et al. 2015, MNRAS, 448, 1206
- Meech et al. (2013) Meech, K. J., Yang, B., Kleyna, J., et al. 2013, ApJ, 776, L20
- Metcalfe et al. (2013) Metcalfe, N., Farrow, D. J., Cole, S., et al. 2013, MNRAS, 435, 1825
- Morgan & Kaiser (2008) Morgan, J. S., & Kaiser, N. 2008, in Proc. SPIE, Vol. 7012, Ground-based and Airborne Telescopes II, 70121K
- Morgan et al. (2012) Morgan, J. S., Kaiser, N., Moreau, V., Anderson, D., & Burgett, W. 2012, in Proc. SPIE, Vol. 8444, Ground-based and Airborne Telescopes IV, 84440H
- Morganson et al. (2012) Morganson, E., De Rosa, G., Decarli, R., et al. 2012, AJ, 143, 142
- Morganson et al. (2015) Morganson, E., Green, P. J., Anderson, S. F., et al. 2015, ApJ, 806, 244
- Morganson et al. (2016) Morganson, E., Conn, B., Rix, H.-W., et al. 2016, ApJ, 825, 140
- Nicholl et al. (2013) Nicholl, M., Smartt, S. J., Jerkstrand, A., et al. 2013, Nature, 502, 346
- Nicholl et al. (2016) Nicholl, M., Berger, E., Smartt, S. J., et al. 2016, ApJ, 826, 39
- Obermeier et al. (2016) Obermeier, C., Koppenhoefer, J., Saglia, R. P., et al. 2016, A&A, 587, A49
- Oke & Gunn (1983) Oke, J. B., & Gunn, J. E. 1983, ApJ, 266, 713
- O’Mullane et al. (2005) O’Mullane, W., Li, N., Nieto-Santisteban, M., et al. 2005, Batch is back: CasJobs, serving multi-TB data on the Web, doi:10.1109/ICWS.2005.29
- Onaka et al. (2012) Onaka, P., Rae, C., Isani, S., et al. 2012, in Proc. SPIE, Vol. 8453, High Energy, Optical, and Infrared Detectors for Astronomy V, 84530K
- Onaka et al. (2008) Onaka, P., Tonry, J. L., Isani, S., et al. 2008, in Proc. SPIE, Vol. 7014, Ground-based and Airborne Instrumentation for Astronomy II, 70140D
- Pastorello et al. (2010) Pastorello, A., Smartt, S. J., Botticella, M. T., et al. 2010, ApJ, 724, L16
- Rest et al. (2014) Rest, A., Scolnic, D., Foley, R. J., et al. 2014, ApJ, 795, 44
- Saglia et al. (2012) Saglia, R. P., Tonry, J. L., Bender, R., et al. 2012, ApJ, 746, 128
- Sanders et al. (2015) Sanders, N. E., Soderberg, A. M., Gezari, S., et al. 2015, ApJ, 799, 208
- Schlafly et al. (2016a) Schlafly, E. F., Peek, J. E. G., Finkbeiner, D. P., & Green, G. M. 2016a, ArXiv e-prints, arXiv:1612.02818
- Schlafly et al. (2012) Schlafly, E. F., Finkbeiner, D. P., Jurić, M., et al. 2012, ApJ, 756, 158
- Schlafly et al. (2014) Schlafly, E. F., Green, G., Finkbeiner, D. P., et al. 2014, ApJ, 786, 29
- Schlafly et al. (2016b) Schlafly, E. F., Meisner, A. M., Stutz, A. M., et al. 2016b, ApJ, 821, 78
- Schunová-Lilly et al. (2017) Schunová-Lilly, E., Jedicke, R., Vereš, P., Denneau, L., & Wainscoat, R. J. 2017, Icarus, 284, 114
- Scolnic et al. (2014) Scolnic, D., Rest, A., Riess, A., et al. 2014, ApJ, 795, 45
- Scolnic et al. (2015) Scolnic, D., Casertano, S., Riess, A., et al. 2015, ApJ, 815, 117
- Smartt et al. (2015) Smartt, S. J., Valenti, S., Fraser, M., et al. 2015, A&A, 579, A40
- Smartt et al. (2016) Smartt, S. J., Chambers, K. C., Smith, K. W., et al. 2016, MNRAS, 462, 4094
- Stalder et al. (2009) Stalder, B., Chambers, K. C., & Vacca, W. D. 2009, ApJS, 185, 124
- Stoughton et al. (2002) Stoughton, C., Lupton, R. H., Bernardi, M., et al. 2002, AJ, 123, 485
- Stubbs et al. (2010) Stubbs, C. W., Doherty, P., Cramer, C., et al. 2010, ApJS, 191, 376
- Szapudi et al. (2015) Szapudi, I., Kovács, A., Granett, B. R., et al. 2015, MNRAS, 450, 288
- Thakar & Li (2008) Thakar, A. R., & Li, N. 2008, Computing in Science & Engineering, 10, 18
- Thakar et al. (2003) Thakar, A. R., Szalay, A. S., Vandenberg, J. V., Gray, J., & Stoughton, A. S. 2003, in Astronomical Society of the Pacific Conference Series, Vol. 295, Astronomical Data Analysis Software and Systems XII, ed. H. E. Payne, R. I. Jedrzejewski, & R. N. Hook, 217
- Toba et al. (2015) Toba, Y., Nagao, T., Strauss, M. A., et al. 2015, PASJ, 67, 86
- Tonry et al. (1997) Tonry, J., Burke, B. E., & Schechter, P. L. 1997, PASP, 109, 1154
- Tonry et al. (2016) Tonry, J., Denneau, L., Stalder, B., et al. 2016, The Astronomer’s Telegram, 8680
- Tonry et al. (2006) Tonry, J., Onaka, P., Luppino, G., & Isani, S. 2006, in The Advanced Maui Optical and Space Surveillance Technologies Conference, E47
- Tonry et al. (2008) Tonry, J. L., Burke, B. E., Isani, S., Onaka, P. M., & Cooper, M. J. 2008, in Proc. SPIE, Vol. 7021, High Energy, Optical, and Infrared Detectors for Astronomy III, 702105
- Tonry et al. (2012a) Tonry, J. L., Stubbs, C. W., Kilic, M., et al. 2012a, ApJ, 745, 42
- Tonry et al. (2012b) Tonry, J. L., Stubbs, C. W., Lykke, K. R., et al. 2012b, ApJ, 750, 99
- Venemans et al. (2015) Venemans, B. P., Bañados, E., Decarli, R., et al. 2015, ApJ, 801, L11
- Vereš et al. (2015) Vereš, P., Jedicke, R., Fitzsimmons, A., et al. 2015, Icarus, 261, 34
- Vinsen & Thilker (2013) Vinsen, K., & Thilker, D. 2013, Astronomy and Computing, 3, 1
- Waters et al. (2016) Waters, C. Z., Magnier, E. A., Price, P. A., et al. 2016, ArXiv e-prints, arXiv:1612.05245
- Weryk et al. (2016) Weryk, R. J., Lilly, E., Chastel, S., et al. 2016, ArXiv e-prints, arXiv:1607.04895
- Wright et al. (2015) Wright, D. E., Smartt, S. J., Smith, K. W., et al. 2015, MNRAS, 449, 451
- York et al. (2000) York, D. G., Adelman, J., Anderson, Jr., J. E., et al. 2000, AJ, 120, 1579
- Zheng et al. (2015) Zheng, Z., Thilker, D. A., Heckman, T. M., et al. 2015, ApJ, 800, 120