Changeset 41197
- Timestamp:
- Dec 8, 2019, 12:14:51 PM (7 years ago)
- Location:
- trunk/doc/release.2015/ps1.calibration
- Files:
-
- 3 edited
-
calibration.tex (modified) (38 diffs)
-
pics/A4.pdf (modified) ( previous)
-
pics/A4.ps (modified) ( previous)
Legend:
- Unmodified
- Added
- Removed
-
trunk/doc/release.2015/ps1.calibration/calibration.tex
r41191 r41197 1102 1102 from the recalculated mean. 1103 1103 1104 Suspicious stars are also excluded from the analysis. We exclude stars 1105 with reduced $\chi^2$ values more than 20.0, or more than 2$\times$ 1106 the median, whichever is larger. We also exclude stars with standard 1104 Suspicious \textadd{(e.g., variable or otherwise poorly measured)} 1105 stars are also excluded from the analysis. We exclude stars with 1106 reduced $\chi^2$ values more than 20.0, or more than 2$\times$ the 1107 median, whichever is larger. We also exclude stars with standard 1107 1108 deviation (of the measurements used for the mean) greater than 0.005 1108 1109 mags or 2$\times$ the median standard deviation, whichever is greater. … … 1152 1153 \sigma_i^{-2})^2} 1153 1154 \end{equation} 1154 1155 These rejections and the over-weighting of the Ubercal measurements1156 are admittedly ad hoc. Since the goal at this stage is to tie the1157 non-Ubercal data to the Ubercal system, we1158 1155 1159 1156 The calculation of the relative photometry zero points is performed … … 1273 1270 For PV3, the relphot analysis was performed two times. The first 1274 1271 analysis used only the flat-field corrections determined by the 1275 ubercal analysis, with a resolution of 2x2flat-field values for each1272 ubercal analysis, with a resolution of $2 \times 2$ flat-field values for each 1276 1273 GPC1 chip (corresponding to \approx 2400 pixels), and 5 separate 1277 1274 flat-field 'seasons'. However, we knew from prior studies that there … … 1513 1510 \subsubsection{Iteratively Reweighted Least Squares Fitting} 1514 1511 1515 With an automatic process applied to hundreds of millions of stars, it1512 With an automatic process applied to hundreds of millions of objects, it 1516 1513 is important for the analysis to provide a measurement of the 1517 photometry of each object which is robust against failures. The 1518 Pan-STARRS\,1 detections have a relatively high rate of non-Gaussian 1519 outliers, partly because of the wide range of instrumental features 1520 affecting the data (see Paper III). \textmod{We have used Iteratively 1521 Reweighted Least Squares (IRLS) fitting to reduce the sensitivity of 1522 the fits to outlier measurements.} 1514 photometry of each object which is robust against failures or other 1515 outliers. \textadd{We would like to calculate an average magnitude 1516 for each filter in the assumption that the flux of the star is 1517 constant and all measurements are drawn from that population. 1518 However, even after rejecting bad measurements based on the quality 1519 information above, individual measurements may still be deviant.} 1520 The Pan-STARRS\,1 detections have a relatively high rate of 1521 non-Gaussian outliers, partly because of the wide range of 1522 instrumental features affecting the data (see Paper III). \textmod{We 1523 have used Iteratively Reweighted Least Squares (IRLS) fitting to 1524 reduce the sensitivity of the fits to outlier measurements.} 1523 1525 1524 1526 We have also used bootstrap resampling to determine confidence limits … … 1787 1789 for either DR1 or DR2. An update to the database will define fields 1788 1790 for each object which encapsulate the information about the ``primary'' 1789 and ``best'' detections. 1791 and ``best'' detections. Users should consult the help pages at MAST 1792 for further information. 1790 1793 1791 1794 \subsubsection{Warp Photometry} … … 1826 1829 than PSF-like, the object bit flag \code{ID_OBJ_EXT} is raised. If 1827 1830 more than half of the PS1 \ippstage{chip}-stage measurements within a 1828 single filter are extended, then the per-filter bit flag 1831 single filter are extended, then the per-filter bit flags 1829 1832 \code{ID_SEC_OBJ_EXT} and \code{ID_SEC_OBJ_EXT_PSPS} are set. The 1830 1833 latter bit is a duplicate bit defined because the high bit in a 32-bit … … 1832 1835 object which has any \ippstage{chip}-stage measurements for one of the 1833 1836 five filters has the per-filter bit flag \code{ID_SECF_HAS_PS1} set. 1837 \textadd{Since stack images are more sensitive than the individual exposures, 1838 faint sources which are detected in only the stacks will have the bit 1839 flag {\tt ID\_SECF\_HAS\_PS1\_STACK} set but not {\tt ID\_SECF\_HAS\_PS1} 1840 as the latter only refers to individual chip detections.} 1834 1841 1835 1842 In addition, if the object has measurements from the 2MASS point … … 1872 1879 1873 1880 \subsection{Photometry Calibration Quality} 1881 \label{sec:photcal} 1874 1882 1875 1883 % /data/kukui.1/eugene/cal.paper.images.20190217/scatter.sh : allsky.scatter.photom … … 1892 1900 position across the sky. For each pixel in these images, we selected 1893 1901 all objects with (14.5, 14.5, 14.5, 14.0, 13.0) $<$ ($g,r,i,z,y$) $<$ 1894 (17, 17, 17, 16.5, 15.5) magnitudes, with at least 3 measurements in $i$-band (to1902 (17, 17, 17, 16.5, 15.5) \textadd{magnitudes}, with at least 3 measurements in $i$-band (to 1895 1903 reject artifacts detected in a pair of exposures from the same night), 1896 1904 with \code{PSF_QF} $> 0.85$ (to reject excessively-masked objects), … … 1922 1930 18)$ millimagnitudes. 1923 1931 1932 % /data/ipp070.0/eugene/dr2.figs.20190205/ 1924 1933 % /data/kukui.1/eugene/cal.paper.images.20190217/kronrepair.sh : full.figure 1925 1934 \begin{figure*}[htbp] … … 1929 1938 PV3.4 photometry illustrating the impact of the issues identified 1930 1939 in the PV3.3 stack and warp photometry. All figures use \ips-band 1931 photometry. The left panels use data from PV3.3 while the right 1940 photometry, \textadd{restricted to objects brighter than 17 magnitudes with 1941 at least 10 chip measurements}. The left panels use data from PV3.3 while the right 1932 1942 use PV3.4. The top row shows the mean difference between the 1933 1943 average photometry from individual exposures (``chip'') and the … … 2025 2035 2026 2036 First, the astrometric calibration has a larger number of systematic 2027 effects which must be performed. These consist of: 1) the2037 effects which must be corrected. These consist of: 1) the 2028 2038 Koppenh\"ofer Effect, 2) Differential Chromatic Refraction, 3) Static 2029 2039 deviations in the camera. We discuss each of these in turn below. … … 2046 2056 shift of about one pixel. This effect was only observed in 2-phase 2047 2057 OTA devices, with 22 / 30 of these suffering from this effect. By 2048 adjusting the summing well high voltage down from a default +7 2058 adjusting the summing well high voltage down from a default +7V to 2049 2059 +5.5V on the 2-phase devices, the effect was prevented in exposures 2050 2060 after 2011-05-03. However, this left 101,550 exposures (27\%) already … … 2078 2088 Differential Chromatic Refraction (DCR) affects astrometry because the 2079 2089 reference stars used to the calibrate the images are not the same 2080 color (SED)as the rest of the stars in the image. For a given star2090 color as the rest of the stars in the image. For a given star 2081 2091 of a color different from the reference stars, as exposures are taken 2082 at higher airmass, the apparent position of the star will be biased2092 at higher airmass, the apparent position of the star will be \textadd{shifted} 2083 2093 along the parallactic angle. While it is possible to build a model 2084 2094 for the DCR impact based on the filter response functions and … … 2112 2122 stars used the calibrate a specific blue- or red-filter image, 2113 2123 respecitively, while $\zeta$ is the zenith distance. 2114 Figure~\ref{fig:DCRexample} shows the DCR trend for the 5 filters2115 \grizy, as well as the measured displacement in the direction2124 Figure~\ref{fig:DCRexample} shows the DCR trend for the \gps\ filter 2125 as an example, as well as the measured displacement in the direction 2116 2126 perpendicular to the parallactic angle. We represent the trend with a 2117 2127 spline fitted to this dataset. … … 2131 2141 \includegraphics[width=\hsize,clip]{{\picdir/DCR.example}.\plotext} 2132 2142 \caption{\label{fig:DCRexample} Example of the DCR trend in the 2133 g-band . {\bf top:} DCR trend in the parallactic direction {\bf2143 g-band, in which it is strongest. {\bf top:} DCR trend in the parallactic direction {\bf 2134 2144 bottom:} DCR trend perpendicular to the parallactic angle.} 2135 2145 \end{center} … … 2139 2149 $(g,r,i,z,y) = (0.010, 0.001, -0.003, -0.017, -0.021)$ arcsec 2140 2150 airmass$^{-1}$ magnitude$^{-1}$. We saturate the DCR correction if 2141 the term $\left[ gi_{\rm ref} - (g - i)\right] \tan \zeta$ or2142 $\left[ zy_{\rm ref} - (z - y)\right] \tan \zeta$ for a given2151 the term $\left[(g - i)_{\rm ref} - (g - i)\right] \tan \zeta$ or 2152 $\left[(z - y)_{\rm ref} - (z - y)\right] \tan \zeta$ for a given 2143 2153 measurement is outside of the range where the DCR correction is 2144 2154 measured. The maximum DCR correction applied to the five filters is … … 2340 2350 In order to perform this analysis, we need estimated distances for 2341 2351 every reference star used in the analysis. \cite{2014ApJ...783..114G} 2342 performed SEDfitting for 800M stars in the 3$\pi$ region using PV22352 performed spectral energy distribution (SED) fitting for 800M stars in the 3$\pi$ region using PV2 2343 2353 data. The goal of this work was to determine the 3D structure of the 2344 2354 dust in the galaxy. By fitting model SEDs to stars meeting a basic … … 2354 2364 and Solar motion parameters ($U_{\rm sol}, V_{\rm sol}, W_{\rm sol}$) 2355 2365 = (9.32, 11.18, 7.61) km sec$^{-1}$ as determined by 2356 \cite{1997MNRAS.291..683F} using Hipparc hus data. Proper motions are2366 \cite{1997MNRAS.291..683F} using Hipparcos data. Proper motions are 2357 2367 determined from the following: 2358 2368 \begin{eqnarray} … … 2366 2376 is independent of distance while the reflex motion induced by the 2367 2377 solar motion decreases with increasing distance. Also note that this 2368 model assumes a flat rotation curve for objects in the thin disk. Any2378 model assumes a flat rotation curve for objects in the thin disk. \textmod{Any 2369 2379 reference stars which are part of the halo population will have proper 2370 2380 motions which are not described by this model; the mostly random 2371 2381 nature of the halo motions should act to increase the noise in the 2372 measurement, but should not introduce detectable motion biases. Also, 2373 if the distance modulus is not well determined, we can assume the 2374 object is simply following the Galactic rotation curve and set a fixed 2375 proper motion. If we do not have a distance modulus from the Green et 2376 al analysis, we assume a value of 500pc. 2377 2378 \note{find the improvement by using 2MASS -- in the PS1 DRAVG pages} 2382 measurement. We do not attempt to compensate for asymmetric drift in 2383 the populations with higher radial velocity dispersion. This effect 2384 will introduce some bias in the azimuthal direction which our simple 2385 model cannot address. For stars for which the distance modulus is not 2386 well determined, we assume the object is simply following the Galactic 2387 rotation curve and set a fixed proper motion.} If we do not have a 2388 distance modulus from the Green et al analysis, we assume a value of 2389 500pc. \textadd{We find that applying our Galactic rotatation model improves 2390 the systematic proper motion errors to some extent. The standard 2391 deviation of the quasar proper motions (averaged on 12 arcminute 2392 superpixels across the sky) is reduced from $(\sigma_{\mu,\alpha}, 2393 \sigma_{\mu,\delta}) = (4.6, 2.4)$ mas yr$^{-1}$ for the uncorrected 2394 analysis to $(\sigma_{\mu,\alpha}, \sigma_{\mu,\delta}) = (2.9, 2.0)$ 2395 mas yr$^{-1}$ after correction for the Galactic rotation model. The 2396 remaining quasar motions continue to show some systematics which may 2397 suggest the need to include a correction for the asymmetric drift.} 2379 2398 2380 2399 For the initial PV3 analysis, we again used the 2MASS coordinates as … … 2388 2407 the Gaia DR1 coordinates. The Gaia DR1 coordinates used a fixed 2015 2389 2408 epoch. Coordinates were propagated from that epoch to the epoch for 2390 each PS1 image as described above. 2409 each PS1 image as described above. \textadd{In a future analysis, we will use 2410 the Gaia DR2 proper motions to tie the astrometric analysis to Gaia 2411 both in terms of the mean positions as well as the dynamical system.} 2391 2412 2392 2413 \subsection{Object Astrometry} … … 2406 2427 \code{ID_MEAS_USED_OBJ}. Some detections were identified as extreme 2407 2428 outliers if their position deviated from the mean object coordinate by 2408 more than 2 arcseconds. These detections were ignored and marked with 2429 more than 2 arcseconds. \textadd{Such a large deviation can only occur when 2430 the in-database calibration is poor, for example near the edges of a 2431 chip.} These detections were ignored and marked with 2409 2432 the bit flag \code{ID_MEAS_POOR_ASTROM}. 2410 2433 … … 2418 2441 \subsubsection{Iteratively Reweighted Least Squares Fitting} 2419 2442 2420 With an automatic process applied to hundreds of millions of stars, it 2421 is important for the analysis to provide a measurement of the 2422 astrometry of each object which is robust against failures. The 2423 Pan-STARRS\,1 detections have a relatively high rate of non-Gaussian 2424 outliers, partly because of the high degree of structure in the 2425 astrometric transformations introduced by the camera optics and the 2426 atmosphere, and partly due to the high masked fraction and other 2427 detector effects. We have used a techinique called Iteratively 2428 Reweighted Least Squares (IRLS) fitting to reduce the sensitivity of 2429 the fits to outlier measurements. We have also used bootstrap 2430 resampling to determine confidence limits on our fits given the 2431 observed collection of position measurements. 2443 \textmod{Just as with the photometric analysis, it is also important for the 2444 astrometric analysis to provide a measurement which is robust against 2445 failures. In addition to the detector effects artifacts which affect 2446 astrometry, the astrometric measurments may have non-Gaussian outliers 2447 due to the high degree of structure in the astrometric transformations 2448 introduced by the camera optics and the atmosphere. We have again 2449 used the IRLS technique to reduce the sensitivity of the fits to 2450 outlier measurements.} We have also used bootstrap resampling to 2451 determine confidence limits on our fits given the observed collection 2452 of position measurements. 2432 2453 2433 2454 We begin the astrometric analysis for each object by projecting the … … 2474 2495 weights. New values for $\omega_\eta,\omega_\zeta$ are calculated, 2475 2496 and the fit is tried again. On each iteration, the fitted parameters 2476 are compared to the values from the previous iteration. If the y2497 are compared to the values from the previous iteration. If the 2477 2498 parameters have not changed significantly ($< 10^{-6}$) or if the 2478 2499 fractional change is less than some tolerance ($10^{-4}$), then … … 2494 2515 2495 2516 Bootstrap-resampling analysis is used to assess the errors on the fit 2496 parameters: A number of measurements equal to the number of unclipped 2497 data points are randomly selected from the set of unclipped data 2498 points, with replacement after each selection. These data points are 2499 then used to fit for the astrometric parameters, using ordinary least 2500 squares fitting. The parameters are recorded and the process re-run 2501 300 times. For each astrometric parameter, the error is determined as 2502 half of the 68\% confidence range for the distribution of fitted 2503 parameter values. 2517 parameters in a fashion similar to the photometry analysis: A number 2518 of measurements equal to the number of the remaining unclipped data 2519 points are randomly selected from the set of the remaining unclipped 2520 data points, with replacement after each selection. These data points 2521 are then used to fit for the astrometric parameters, using ordinary 2522 least squares fitting. The parameters are recorded and the process 2523 re-run 300 times. For each astrometric parameter, the error is 2524 determined as half of the 68\% confidence range for the distribution 2525 of fitted parameter values. 2504 2526 2505 2527 \subsubsection{Object Astrometry Flags} … … 2511 2533 fitted to parallax without proper motion as well. If an object is 2512 2534 fitted for parallax, it is also fitted with a model including only 2513 proper motion and only a mean position. The chi sqfor all three fits2535 proper motion and only a mean position. The chi-square for all three fits 2514 2536 is saved. Currently, the highest order fit allowed is saved in the 2515 2537 database, regardless of the significance of the improvement in adding … … 2631 2653 \sigma_\delta)$ is 16 milliarcseconds. 2632 2654 2633 The Galactic plane is clearly apparent lyin these images. Like2655 The Galactic plane is clearly apparent in these images. Like 2634 2656 photometry, we attribute this to failure of the PSF fitting due to 2635 2657 crowding. The celestial North pole regions have somewhat elevated … … 2721 2743 Solar motion to correct the absolute proper motion (see 2722 2744 Section~\ref{sec:galactic.rotation}). We identify the resulting 2723 database as PV3. 1. This database was used to generate the positions2745 database as PV3.2. This database was used to generate the positions 2724 2746 in the \ippdbtable{gaiaObject} table, which are exposed in the DR1 2725 2747 release. … … 2786 2808 \begin{center} 2787 2809 \includegraphics[width=\hsize,clip]{{\picdir/gaia.photom.v1}.\plotext} 2788 \caption{\label{fig:gaia.photom} Comparison with Gaia DR1 2789 photometry. {\bf Left} Mean of PS1 - Gaia DR1, {\bf Right} Standard 2790 deviation of PS1 - Gaia DR1. For pixels with $|b| > 30$ and $\delta > 2791 -30$, the standard deviation of the PS1 - Gaia DR1 mean values is 6.9 2792 millimagnitudes, while the median of the standard deviations is 12.4 2793 millimagnitudes. The former is a statement about the consistency 2794 of the Gaia DR1 and Pan-STARRS\,1 photometry, while the latter 2795 reflects the combined bright-end errors for both systems. } 2810 \caption{\label{fig:gaia.photom} Comparison with Gaia DR1 photometry 2811 (see Section~\ref{sec:gaia.tie} for sample selection). {\bf Left} 2812 Mean of PS1 - Gaia DR1, {\bf Right} Standard deviation of PS1 - 2813 Gaia DR1. For pixels with $|b| > 30$ and $\delta > -30$, the 2814 standard deviation of the PS1 - Gaia DR1 mean values is 6.9 2815 millimagnitudes, while the median of the standard deviations is 2816 12.4 millimagnitudes. The former is a statement about the 2817 consistency of the Gaia DR1 and Pan-STARRS\,1 photometry, while 2818 the latter reflects the combined bright-end errors for both 2819 systems. } 2796 2820 \end{center} 2797 2821 \end{figure*} … … 2869 2893 signal to noise in Gaia; they were also apparent in the plots of the 2870 2894 statisics of the per-exposure measurement residuals 2871 (Figure~\ref{fig:allsky.astrom.sigma} . The standard deviations of the2895 (Figure~\ref{fig:allsky.astrom.sigma}). The standard deviations of the 2872 2896 median differences are ($\sigma_\alpha, \sigma_\delta) = (4.8, 3.1)$ 2873 2897 milliarcseconds. … … 2932 2956 \begin{figure*}[htbp] 2933 2957 \begin{center} 2934 \includegraphics[width= \hsize,clip]{{\picdir/A4}.pdf}2958 \includegraphics[width=0.95\hsize,clip]{{\picdir/A4}.pdf} 2935 2959 \caption{\label{fig:pole.bad.histogram} Histogram of the fraction of bad groups for each skycell (red line).} 2936 2960 \end{center} … … 2949 2973 based on a comparison between stack and mean object photometry. In the 2950 2974 presence of modest registration errors, mean object photometry would 2951 not be affected, as individual detection would shave the correct2975 not be affected, as individual detection would have the correct 2952 2976 signal, and averaging their flux in catalog space would yield the 2953 2977 correct total magnitude. On the other hand, imperfect stacking would … … 2960 2984 in poor stack photometry for the affected skycells. 2961 2985 2962 Further investiga ion revealed that the cause of these failures was an2986 Further investigation revealed that the cause of these failures was an 2963 2987 error in the internal reference catalog used for the PV3 analysis (see 2964 2988 Section~\ref{sec:synthdb}). This reference catalog used PS1 … … 2994 3018 We first used the PV3 mean astrometry and photometry to define a new 2995 3019 reference catalog in the assumption that the bulk of the failures 2996 would be eliminated by the astrometric recalibration. We reproces ed a3020 would be eliminated by the astrometric recalibration. We reprocessed a 2997 3021 section of the polar cap data using this PV3-based reference catalog 2998 and re-ran the astrometric registration test was repeatedon the3022 and re-ran the astrometric registration test on the 2999 3023 reprocessed exposures. The reprocessing greatly ameliorated the 3000 3024 registration issue, as shown in Figure~\ref{fig:pole.bad.histogram}. … … 3016 3040 3017 3041 We consider skycells with more than 10\% bad groups to have been 3018 adversely affected by this problem. Use s of DR2 should be aware that3019 the affected s kycells have poor astrometry and effective image3042 adversely affected by this problem. Users of DR2 should be aware that 3043 the affected stack skycells have poor astrometry and effective image 3020 3044 quality. However, as these images may be useful to the community, 3021 3045 they are available from the MAST cutout server. Users who attempt to 3022 3046 download these problem skycells will see a warning message and should 3023 avoid using the skycell images for quantitative measurements without 3047 only use the skycell images for quantitative measurements with 3024 3048 extreme caution. Since stack measurements from these skycells are 3025 3049 significantly damaged, the DR2 release has set the measured stack … … 3029 3053 \section{Conclusion} 3030 3054 3031 The Pan-STARRS Data Release 2 provides astrom try and photometry of3055 The Pan-STARRS Data Release 2 provides astrometry and photometry of 3032 3056 roughly 3 billion astronomical objects across the $3\pi$ survey 3033 3057 region. The photometry system has been shown to be reliable across … … 3047 3071 community. 3048 3072 3049 \note{need to add discussion of SDSS, DES, LSST, Gaia} 3073 \textadd{The past three decades have seen the digital release of a series of 3074 large-scale optical and near-IR astronomical surveys with generally 3075 steady improvements in quality. The trend begins in the mid 1990s 3076 with the digitized photographic plate surveys such as USNO-B 3077 \cite{2003AJ....125..984M} and SuperCOSMOS \cite{2001MNRAS.326.1295H} 3078 which have photometric errors of roughly 300 millimags and astrometric 3079 errors of roughly 200 milliarcseconds. The Hipparcos \& Tycho 3080 catalogs released in the mid 1990s have much smaller astrometric 3081 errors (roughly 0.6 milliarcseconds) but substantially limited depth 3082 ($V < 11.5$) compared to the ground-based work 3083 \citep{1997AA...323L..57H}.} 3084 3085 \textadd{The first generation of sky surveys using digital detectors, including 3086 SDSS \citep{2001ASPC..238..269L} and 2MASS 3087 \citep{2006AJ....131.1163S}, brought a substantial leap in the quality 3088 of both photometry and astrometry along with improvements in the depth 3089 and wavelength coverage. Glossing over the details of how exactly to 3090 determine the accuracy of the SDSS and 2MASS photometry, it is clear 3091 that the photometric accuracy of those surveys are in the vicinity of 3092 10 - 20 millimagnitudes for all filters, more than an order of 3093 magnitude improvement over the photographic plate surveys. The 3094 astrometric accuracy of these two surveys (roughly 50 - 80 3095 milliarcseconds) is also a large improvement.} 3096 3097 \textadd{The Pan-STARRS $3\pi$ Survey public release represents an important 3098 step in the ongoing progress towards covering the sky with 3099 well-characterized measurements. The nearly coincident data releases 3100 from Gaia \citep{2016AA...595A...4L,2018AA...616A...1G} complement the 3101 PS1 releases greatly. In the south, the Dark Energy Survey has 3102 produced its first public data release covering roughly 5000 square 3103 degrees of the sky \citep{2018ApJS..239...18A} with reported 3104 photometric precision of better than 10 millimagnitudes.} 3105 3106 \textadd{The next decade will see further advances in survey breadth and depth 3107 along with further improvements in calibration quality. Over the next 3108 2-3 years, the Ultraviolet Near-Infrared Optical Northern Sky (UNIONS) 3109 Survey collaboration (a meta-collaboration of the Pan-STARRS and 3110 Canada-France Imaging Survey, or CFIS, collaborations) is expected to 3111 release deep photometry in the {\it ugriz} bands for roughly 5000 3112 square degrees of the northern hemisphere with agressive photometric 3113 precision goals. This collaboration is in part motivated to support 3114 the Euclid satellite mission, which requires deep 8-band photometry to 3115 measure photometric redshifts, but only provides the {\it JHK} bands. 3116 The Large Synoptic Survey Telescope is also expected to produce 3117 high-precision photometry and astrometry to great depths over a very 3118 large portion of the sky available from the southern hemisphere.} 3119 3120 \textadd{From our experience with the Pan-STARRS survey, and the results of the 3121 comparisons between surveys, a few lessons stand out.} 3122 3123 \textadd{First, systematic errors come in many forms and dominate the 3124 calibration precision. Internal or relative examination of the data 3125 can reveal important and unexpected effects such as the Koppenh\"ofer 3126 and vertical diffusion effects we identified in the Pan-STARRS 3127 devices.} 3128 3129 \textadd{Second, cross-comparisons between independent datasets are critical to 3130 reveal the limitations. This lesson has appeared several times in our 3131 intestigations, in the comparison between Pan-STARRS and Gaia above, 3132 between Pan-STARRS and SDSS \citep{2016ApJ...822...66F}, and in the 3133 comparison between Pan-STARRS and 2MASS \citep{2013ApJS..205...20M}. 3134 The cross-comparison can be used to explicitly constrain the 3135 calibration on one survey based on another, as was done by 3136 \cite{2016ApJ...822...66F} for the SDSS Hypercalibration solution. 3137 Alternatively, the cross-comparison can be used to identified issues 3138 which may be solved by improved internal analysis. } 3139 3140 \textadd{The third lesson we have learned is that there is no substitute for 3141 photometric conditions. The cross-comparison of photometry between 3142 Pan-STARRS and Gaia suggests that the current Pan-STARRS calibration 3143 is limited in part by the excessive contribution of non-photometric 3144 observations. This can be seen in the elevated scatter in patches 3145 which correspond to single observing blocks (see 3146 Figure~\ref{fig:allsky.photom.sigma} and discussion in 3147 Section~\ref{sec:photcal}). A future re-analysis of the Pan-STARRS 3148 dataset will attempt to further limit the impact of the 3149 non-photometric data on the photometric calibration. The other 3150 critical improvement will be to include more data from the continuing 3151 observations to ensure every patch of the sky is covered with 3152 photometric observations.} 3153 3154 \textadd{Finally, while the systematics are still probably the limiting factor 3155 for the average calibration, for individual measurements of objects, 3156 we believe our current limitations come from a few specific factors. 3157 First, the quality of the aperture corrections, especially in the 3158 ability of the software to avoid extremely deviant results on occasion 3159 appears to be one of the main drivers of bad photometry measurements 3160 for brighter stars. Second, the quality of the background sky model 3161 currently appears to be the limitation for the faint sources. 3162 Finally, improvements to the PSF model, especially including 3163 color-dependent and non-linear effects such as the brighter-fatter effect 3164 \citep{2014JInst...9C3048A,2015JInst..10C5032G} will probably be 3165 necessary to push the limits of photometric and astrometric accuracy. } 3166 3167 \textadd{While there is clearly still room for improvement, the Pan-STARRS 3168 $3\pi$ Survey DR1 and DR2 photometry will be a critical resource for 3169 many years. We are confident that, in addition to the many science 3170 discoveries enabled by the large and accurate photometry, the 3171 high-quality photometry provided here will save observers countless 3172 hours of telescope time by obviating, or at least greatly reducing, 3173 the need to observe standard stars on a regular basis.} 3174 3175 %% USNO-A,B : 0.2 arcsec, 0.3 mag, R ~ 21 3176 %% https://arxiv.org/pdf/astro-ph/0210694.pdf 3177 %% https://ui.adsabs.harvard.edu/abs/2003AJ....125..984M/abstract 3178 %% SuperCOSMOS : 0.2 arcsec, 0.3 mag 3179 %% http://www-wfau.roe.ac.uk/sss/intro.html 3180 %% https://ui.adsabs.harvard.edu/abs/2001MNRAS.326.1295H/abstract 3181 %% https://ui.adsabs.harvard.edu/abs/2001MNRAS.326.1315H/abstract 3182 %% Hipparcos : XX mas, 0.XX mag 3183 %% https://ui.adsabs.harvard.edu/abs/1997A%26A...323..620K/abstract 3184 %% https://ui.adsabs.harvard.edu/abs/1997A%26A...323L..57H/abstract 3185 3186 % \note{need to add discussion of SDSS, DES, LSST, Gaia} 3050 3187 3051 3188 \acknowledgments … … 3134 3271 kukui:/data/kukui.3/eugene/pv3.stats.20161202/maps.measure 3135 3272 3273 % /data/ipp070.0/eugene/dr2.figs.20190205/ 3274 % /data/kukui.1/eugene/cal.paper.images.20190217/kronrepair.sh : full.figure
Note:
See TracChangeset
for help on using the changeset viewer.
