THE PAN-STARRS SOLAR SYSTEM SIMULATION Larry Denneau, Jr. Pan-STARRS Team Pan-STARRS, Institute for Astronomy, University of Hawaii ABSTRACT The Institute for Astronomy at the University of Hawaii is developing a large optical astronomical surveying system–the Panoramic Survey Telescope and Rapid Response System (Pan-STARRS). The Moving Object Processing System (MOPS) client of the Pan-STARRS image processing pipeline is developing software to automatically discover and identify >90% of near-Earth objects (NEOs) 300m in diameter and >80% of other classes of asteroids and comets. In developing its software, MOPS has created a synthetic solar system model (SSM) with over 10 million objects whose distributions of orbital characteristics match those expected for objects that Pan- STARRS will observe. MOPS verifies its correct operation by simulating the survey and subsequent discovery of synthetically generated objects. MOPS also employs novel techniques in handling the computationally difficult problem of linking large numbers of unknown asteroids in a field of detections. We will describe the creation and verification of the Pan-STARRS MOPS SSM, demonstrate synthetic detections and observations by MOPS, describe MOPS asteroid-linking techniques, describe accuracy and throughput of the entire MOPS system, and provide predictions regarding the numbers and kinds of objects, including as yet undiscovered "extreme objects", that MOPS expects to find over its 10-year lifetime. INTRODUCTION The Institute for Astronomy at the University of Hawaii is developing a large optical astronomical surveying system–the Panoramic Survey Telescope and Rapid Response System (Pan-STARRS). Pan-STARRS will provide state-of-the-art capability in several areas: four wide-field (7 deg 2) identical telescopes operating in parallel, a gigapixel orthogonal transfer array (OTA) detector for each telescope, 0.01 arcsecond astrometric precision, and a limiting magnitude of R=24. In a single night Pan-STARRS will observe approximately 6000 deg 2 of sky. All combined, this capability will allow Pan-STARRS to perform automated asteroid searching on an unprecedented scale. The Pan-STARRS single prototype telescope, called PS1, is expected to see first light in 2006, followed by the complete four-telescope system, PS4, in 2009. MOVING OBJECT PROCESSING SYSTEM The Moving Object Processing System (MOPS) client of the Pan-STARRS Image Processing Pipeline (IPP) is developing software to automatically discover and identify >90% of near-Earth objects (NEOs) 300m in diameter and >80% of other classes of asteroids and comets. The Pan-STARRS Solar System Simulation (SSS) exists so that the MOPS can be developed and verified and so that PS4 operating efficiencies can be monitored. MOPS will not have any live data until the Pan-STARRS prototype telescope and subsequent astrometric/photometric survey are completed in 2007. So MOPS must construct its own population of Solar System objects and simulate their observation. Once synthetic telescope fields are created, MOPS can develop and verify algorithms that perform automatic discovery of asteroids. During the operation of PS4, Pan-STARRS and MOPS will require a way to measure how efficiently MOPS is processing its input data stream. MOPS will estimate its operating efficiency on real sky data by injecting its full synthetic model into the real data stream and then extract efficiency parameters from the synthetic stream after processing. As MOPS discovers and verifies new objects, in particular near-Earth objects (NEOs) and potentially hazardous objects (PHOs), orbital parameters and observations will be reported to external sources for further evaluation of impact hazard. MOPS SOLAR SYSTEM SIMULATION Synthetic Asteroid Population In order for MOPS survey strategies and software design to be tested, there must exist a source of fictitious asteroid- like bodies from which MOPS can synthesize observations. The MOPS team has synthesized over 10 million Solar System objects for its simulation, shown in Table 1. Table 1. MOPS Synthetic Asteroid Population NEOs 250,000 Main Belt 10,000,000 Trojans 420,000 Centaurs 60,000 Trans-Neptunian 72,000 Scattered Disc 20,000 Comets 20,000 Total 10,842,000 The NEO population is based on the Bottke et al. NEA population model [1], and the Main Belt population is based on current known orbital distributions of Main Belt asteroids, scaled to reproduce a sky-plane density of approximately 200 objects/deg 2 in the ecliptic [2]. Other populations were developed similarly, with some selection criteria for observability in the case of distant populations (Trans-Neptunian, Scattered Disc objects). Survey Strategy Considerable study has already been dedicated to optimal observing strategies for the detection of asteroids and in particular NEOs by Pan-STARRS [3]. During its nightly 1000-field scan, the preliminary Pan-STARRS asteroid survey will include two 2,200-deg 2 opposition regions (~660 fields each) flanked by two 550-deg 2 —sweet-spot“ regions (~160 fields each). The sweet-spots represent locations on the sky more likely to contain PHOs of highest impact risk, so the greatest impact reduction can be achieved by searching for asteroids in these regions. These regions of sky coverage will move along the ecliptic at approximately one degree per day to remain opposite the Sun; after one year the sky coverage pattern returns to its original location and the survey is repeated. Fig. 1 shows Pan-STARRS sky coverage at four different times during a single year of the MOPS simulated survey. 1ig. 1. MOPS simulated survey sky coverage during one year In addition to appropriate sky coverage, Pan-STARRS must consider the optimal numbers of visits per night and number of days between visits for a particular location on the sky. MOPS has developed several survey simulations using various intra-night and inter-night intervals, and for its testing and development has selected a survey that generates 30-minute intervals between visits to the same location per night, and on average 3-4 days between visits, for a 10-12 day observational arc. Previous Pan-STARRS studies have shown that 10-12 day arcs provide sufficient orbital definition to find objects the following month [4], at which point their orbits can be more precisely calculated. MOPS has programmed its survey scheduler to simulate random coarse weather losses. Currently MOPS can only simulate entire nights lost to weather, but future MOPS survey simulations may be able to simulate finer-grained losses. Synthetic Telescope After creating a population of synthetic asteroids and suitable surveys, MOPS simulates the observation of its synthetic Solar System essentially by calculating the position and magnitude on the sky for all objects and selecting the objects that appear within a single field. The naïve or brute-force approach to this problem is computationally prohibitive, and MOPS employs multiple-hypothesis testing and spatial indexing routines among other optimizations to substantially reduce computation required for this process. After calculating exact position and magnitude within a field, MOPS —fuzzes“ the observation by applying astrometric (position) and photometric (magnitude) errors consistent with expected total astrometric and photometric error for the PS4 system. The astrometric error can be modeled as a Gaussian distribution that is a function of photometric signal-to-noise, and for bright objects the expected Pan-STARRS astrometric precision will be about 0.01 arcsec. False detections are also added to the detection stream, at a rate of about 200 false detections per deg 2 at a 5 σ confidence level. Fig. 2 shows a complete synthetic Pan-STARRS field, as seen by MOPS. 1ig 2. Simulated Pan-STARRS field MOPS Software Upon generation of synthetic telescope fields of moving objects, MOPS can get to the business of finding asteroids. Algorithms for automatic discovery of asteroids by wide-field surveys is a complicated and rapidly changing area of research [5], and while MOPS has its own methods for finding asteroids, one consideration in the MOPS design is to allow other algorithms to be —plugged into“ the MOPS pipeline. Fig. 3 shows the general flow of data through the MOPS pipeline. Generally, MOPS asteroid discovery (—linking“) consists of the identification of candidate detection pairs (using the current survey strategy; triplets are under consideration) of moving objects within a single night, called tracklets . Conceptually, all detections consistent with quasi-linear motion within an acceptable velocity range become candidate tracklets. After three nights of observations have been performed at a given region on the sky, tracklets from these three nights are assembled into proposed tracks , or linkages, of detections. In the MOPS design, many of these tracks are unlikely to represent correct linkages of detections–that is, they may contain mixed detections from two or more different objects. These mixed tracks are not discarded until orbit computation is performed on the track. Computed orbits that satisfy a maximum-residual requirement are accepted as provisional discovered objects and are preserved in the MOPS system. QuickTime™ and a TIFF (Uncompressed) decompressor are needed to see this picture. 1ig 3. MOPS software pipeline MOPS does not introduce physics into its linking process due to computational cost. Instead the linking software assumes an approximately quadratic motion on the sky over the time interval under question. Only