NEOExchange -- An online portal for NEO and Solar System science
T. A. Lister, E. Gomez, J. Chatelain, S. Greenstreet, J. MacFarlane, A. Tedeschi, I. Kosic
NNEOExchange - An online portal for NEO and SolarSystem science (cid:63)
T. A. Lister a, ∗ , E. Gomez c , J. Chatelain a , S. Greenstreet a,f,d,e , J.MacFarlane b,g , A. Tedeschi b , I. Kosic b a Las Cumbres Observatory, 6740 Cortona Drive Suite 102, Goleta, CA 93117, USA b Intern at Las Cumbres Observatory, 6740 Cortona Drive Suite 102, Goleta, CA 93117,USA c Las Cumbres Observatory, School of Physics and Astronomy, Cardiff University, QueensBuildings, The Parade, Cardiff CF24 3AA, UK d Asteroid Institute, 20 Sunnyside Ave, Suite 427, Mill Valley, CA 94941, USA e Department of Astronomy and the DIRAC Institute, University of Washington, 3910 15thAve NE, Seattle, WA 98195, USA f University of California, Santa Barbara, Santa Barbara, CA 93106, USA g Freedom Photonics LLC, 41 Aero Camino, Santa Barbara, CA 93117, USA
Abstract
Las Cumbres Observatory (LCO) has deployed a homogeneous telescope net-work of ten 1-meter telescopes to four locations in the northern and southernhemispheres, with a planned network size of twelve 1-meter telescopes at 6 loca-tions. This network is very versatile and is designed to respond rapidly to targetof opportunity events and also to perform long term monitoring of slowly chang-ing astronomical phenomena. The global coverage, available telescope apertures,and flexibility of the LCO network make it ideal for discovery, follow-up, andcharacterization of Solar System objects such as asteroids, Kuiper Belt Objects,comets, and especially Near-Earth Objects (NEOs).We describe the development of the “LCO NEO Follow-up Network” whichmakes use of the LCO network of robotic telescopes and an online, cloud-basedweb portal, NEOexchange, to perform photometric characterization and spec-troscopic classification of NEOs and follow-up astrometry for both confirmedNEOs and unconfirmed NEO candidates.The follow-up astrometric, photometric, and spectroscopic characterizationefforts are focused on those NEO targets that are due to be observed by theplanetary radar facilities and those on the Near-Earth Object Human SpaceFlight Accessible Targets Study (NHATS) lists. Our astrometric observationsallow us to improve target orbits, making radar observations possible for ob-jects with a short arc or large orbital uncertainty, which could be greater than (cid:63)
Available on the web at https://lco.global/neoexchange . Code available from GitHubat https://github.com/LCOGT/neoexchange/ ∗ Corresponding author
Email addresses: [email protected] (T. A. Lister), [email protected] (E. Gomez), [email protected] (J. Chatelain), [email protected] (S. Greenstreet)
Preprint submitted to Icarus February 23, 2021 a r X i v : . [ a s t r o - ph . E P ] F e b he radar beam width. Astrometric measurements also allow for the detec-tion and measurement of the Yarkovsky effect on NEOs. The photometric andspectroscopic data allows us to determine the light curve shape and ampli-tude, measure rotation periods, determine the taxonomic classification, andimprove the overall characterization of these targets. We are also using a smallamount of the LCO NEO Follow-up Network time to confirm newly detectedNEO candidates produced by the major sky surveys such as ATLAS, CatalinaSky Survey (CSS) and PanSTARRS (PS1). We will describe the construction ofthe NEOexchange NEO follow-up portal and the development and deploymentmethodology adopted which allows the software to be packaged and deployedanywhere, including in off-site cloud services. This allows professionals, ama-teurs, and citizen scientists to plan, schedule and analyze NEO imaging andspectroscopy data using the LCO network and acts as a coordination hub forthe NEO follow-up efforts. We illustrate the capabilities of NEOexchange andthe LCO NEO Follow-up Network with examples of first period determinationsfor radar-targeted NEOs and its use to plan and execute multi-site photometricand spectroscopic observations of (66391) 1999 KW4, the subject of the mostrecent planetary defence exercise campaign. Keywords:
Asteroids – Near-Earth object – Experimental techniques –astrometry – photometry
1. Introduction
Near Earth Objects (NEOs) are our closest neighbors and research into themis important not only for understanding the Solar System’s origin and evolu-tion, but also to understand the consequences of, and to protect human societyfrom, potential impacts. NEOs consist of two subclasses, Near Earth Aster-oids (NEAs) and a smaller fraction of Near Earth Comets (NECs). NECs arethought to be extinct comets that originally came from the Kuiper Belt or OortCloud, whereas NEAs originate in collisions between bodies in the main aster-oid belt and have found their way into near-Earth space via complex dynamicalinteractions. Understanding these interactions, the populations of the source re-gions, and the resulting orbital element distributions requires accurate orbits forrobust samples of the NEO population. Substantial numbers of objects mustbe observed in order to properly debias the sample and correctly model theNEO population. The existing surveys such as the Asteroid Terrestrial-impactLast Alert System (ATLAS; Tonry et al. 2018), Catalina Sky Survey (CSS), thePanSTARRS1 (PS1) & PanSTARRS2 (PS2) surveys (Wainscoat et al., 2014)and NEOWISE (Mainzer et al., 2014) are not normally capable of following-uptheir own NEO candidate detections and require additional programs of NEOfollow-up on other telescopes to confirm and characterize the new NEOs.LCO has deployed a homogeneous telescope network of ten 1-meter tele-scopes and ten 0.4-meter telescopes to six locations in the northern and south-ern hemispheres. These have joined the two existing 2-meter Faulkes telescopes2 igure 1: Network map of LCO facilities (FTN & FTS). The global coverage of this network and the apertures of tele-scope available make the LCO network ideal for follow-up and characterizationof Solar System objects in general and for Near-Earth Objects (NEOs) in par-ticular.We describe the creation, including the testing and software developmentphilosophy, of the “LCO NEO Follow-up Network” and the central observingportal, NEOexchange (often abbreviated as ‘NEOx’), in Section 3. We illustratethe use of the LCO NEO Follow-up Network in Section 4 to perform the sciencecases of NEO candidate follow-up and NEO characterization described above.We summarize some of the results to date and outline plans for future work anddevelopment in Section 5.
2. Overview of the LCO Network
LCO completed the first phase of the deployment (see Figure 1) with theinstallation and commissioning of the ten 1-meter telescopes at McDonald Ob-servatory (Texas), Cerro Tololo (Chile), SAAO (South Africa) and Siding SpringObservatory (Australia). These 1-meter telescopes join the two existing 2-meterFaulkes Telescopes which LCO has operated since 2005. The whole telescopenetwork has been fully operational since 2014 May, and observations are exe-cuted remotely and robotically. Two additional 1-meter telescopes for the sitein the Canary Islands will be deployed during 2020–2021. Future expansion toa site at Ali Observatory, Tibet is also planned, but the timescale is uncertainand dependent on partner funding.The 2-meter FTN and FTS telescopes originally had identical 10 (cid:48) × (cid:48) fieldof view (FOV) Charge Coupled Device (CCD) imager with 18 filters and a low-3esolution ( R ∼ (cid:48) × (cid:48) FOV CCD with21 filters. Each site also has a single high-resolution ( R ∼ g (cid:48) r (cid:48) i (cid:48) z s filters, we have apowerful tool for simultaneous color and coarse taxonomy determination. Thisis particularly true for rapid response characterization of small diameter NEOswhich would be too faint for spectroscopy.
3. Overview of the LCO NEO Follow-up Portal: NEOExchange
We consider NEOexchange to be an example of what are now being calledTarget and Observation Management (TOM) systems (Street et al., 2018). Thegoal of these systems is to ingest a large number of targets of interest and carryout selection of a subset according to some merit function. This subset is thenevaluated for observing feasibility and follow-up observations are requested fromavailable telescopes. The resulting status and any data from these observationsare returned to the TOM system, which records these details. The results ofdata reduction on the data are also recorded and this is used as feedback into thenext round of observation planning. NEOexchange is an implementation of sucha TOM system, focusing on Solar System bodies, with a particular emphasis onNEOs. An outline of the NEOexchange system is shown graphically in Figure 2.The major sections of the NEOexchange TOM system are the sources oftargets, the database and associated web front-end and interaction tools, theplanning and scheduling functions, and the data reduction processes. Each ofthese will be treated in more detail in the following sections.
The major source of target Solar System bodies is the Minor Planet Center(MPC), specifically the MPC’s NEO Confirmation Page (NEOCP). This web-page is parsed to extract the NEO candidates from the surveys that are in needof confirmation. Metadata for each candidate such as the digest2 (Keys et al.,2019) score (ranging from 0 . . . igure 2: Schematic overview of the NEOexchange TOM system. Targets for follow-up areingested into the NEOx DB (center) from the data source (left side), planned and scheduledand sent to follow-up telescopes (right). The resulting data flows through the observatorydata reduction pipeline and into the archive. This is retrieved by NEOexchange and processedthrough science specific pipelines and the results of these are also stored in the NEOx DB. andArecibo . These are supplemented by potential mission destinations from theHuman Space Flight Accessible Targets Study (NHATS) database. The radartargets are obtained by fetching and parsing webpages, whereas the NHATStargets are obtained by parsing emails sent to a mailing list. Both of thesemethods are somewhat fragile and at risk of missing or misparsing targets if thewebpage or email format changes. This illustrates the need for a lightweightprotocol that can be used by data requestors to signal targets of interest in away that is easy and simple to implement as well as parseable and readable byboth machines and humans.Given a list of characterization targets from any of these methods, we thenmake additional queries in the MPC database in order to extract and storeor update the orbital elements of these targets. At present this is also doneby parsing webpages (using the BeautifulSoup4 library), but the MPC hasexpressed desires to implement a webservices application programming interface(API) that provides this information in a more robust manner in the future. Wewill implement and switch to this method of obtaining the information once itbecomes available.Our final source of requests for an increased priority of a particular NEOcandidate is from the JPL SCOUT system. This performs analysis of theshort-arc NEO candidates that are on the NEOCP and determines if there isany risk of a potential impact or close passage to the Earth. Alerts from theSCOUT system are sent out via email and these can result in the triggering offollow-up, including potentially disruptive ‘rapid response’ that can interruptalready running observations on the LCO network.For the NEO candidates from the NEOCP, tasks are periodically executedto update the target lists and cross identifications from the MPC. In the case ofthe radar and other characterization targets, we update the observations storedin the NEOexchange database (DB) and perform orbit fitting to update the https://echo.jpl.nasa.gov/asteroids/goldstone_asteroid_schedule.html https://cneos.jpl.nasa.gov/scout/ find orb code in a non-interactive mode with the orbit fit to the measurementsfor the object exported from the database. These tasks are performed with afrequency determined by the typical update frequency of the data source. Thisis balanced in order to avoid putting undue strain on the remote data source. Inthe case of the NEO candidates, updates occur every 30 minutes; for the muchsmaller and less dynamic list of characterization targets, the update tasks arerun twice a day. The targets for follow-up and their associated metadata as described in theprevious section, are persisted in a SQL database. This is supplemented byinformation on user and proposal details for follow-up resources, follow-up re-quests, data frames obtained, and the catalog products and source measure-ments derived from those data frames. An overall database schema is shown inFigure 3.We use the
Python web framework,
Django , to provide a vendor-independentdatabase abstraction and web front-end. Django provides a means to definea model of a particular object (such as an asteroid target) and the relevantconcepts and parameters that are associated with that object, and then havethe resulting database table created automatically without the user needing toworry about the low-level database details. It also provides a means to definefunctions on the model that can provide associated calculations such as a skyposition computed from the target model’s orbital elements.As shown in Figure 3, the database structure is divided into 3 main areas: • the Body table and related tables which store additional information (the
ColorValues, Designations, PhysicalParameters, PreviousSpectra and
SpectralInfo tables). These tables hold the details about the tar-gets (such as target type, the data’s origin, orbital elements) and any pastcolor or taxonomic determinations, • the SuperBlock and
Block tables that record the details of scheduled ob-servational follow-up, along with the ancillary
StaticSource which holdsdetails about static (sidereal) calibration targets, • the Frame table, which stores the metadata about our and others’ obser-vations and associated derived quantities resulting from the observed dataframes and their data processing. These include the
CatalogSources ta-ble (the largest in the DB) which holds the source catalog extracted fromevery frame and the
SourceMeasurement table which records definite de-tections and measurements of a particular
Body igure 3: Overview of the NEOexchange database schema. Django
Model classes becometables in the database, which are shown as rectangles in the figure with the class/table nameon the top. Primary and foreign keys are shown in bold within the tables and the relationshipsbetween the tables are shown as black lines between the tables.
8n addition to these major tables, there are a small number of ancillary tables tohold details about the registered users and proposals for the LCO network andan associated table to link users to proposals, which verifies if they are allowedto submit observations to the telescopes.The primary mode of user interaction with NEOexchange is through the webfront-end. This allows users to view prioritized lists of targets, example detailsof the targets and the data obtained for them so far. Users can also scheduleadditional follow-up observations as well as analyze and report data. Certainfunctions, such as submitting observations, require that the user is registeredand authenticated with NEOexchange and has been associated with a proposal(or proposals) that have time on the LCO network.In addition to the website, various other command-line tools interact withthe database. These include those that update the target lists described inthe previous section, data download and processing scripts (discussed in Sec-tion 3.4), and data analysis tools that can build light curves of targets that havebeen observed and processed.
For the purposes of ranking NEO candidates, we compute the following meritfunction for each candidate:
F OM = (cid:16) e t last t arc − (cid:17) + (cid:16) e V − (cid:17) + (cid:16) e H − (cid:17) + 0 . × e (cid:18) − . × ( score − (cid:19) + e (cid:18) − . × ( SPD − . )2180 (cid:19) (1)where t last is the time since the object was last seen in decimal days, t arc isthe arc length, also in decimal days, V is the current visual magnitude, score is the NEOCP digest2 score, H is the absolute magnitude (approx. diameter,assuming an albedo) and SP D is the South Polar Distance in degrees.This merit function prioritizes targets that have not been seen in a while,have a short arc, have a large diameter (small value of H ) and are bright,have a high likely NEO ‘score’, and will go directly overhead of our southernhemisphere sites (this last weighting factor is because the LCO network hasmany more telescopes in the southern hemisphere; see Figure 1). The firstthree terms for ‘last seen’/‘arc length’, V magnitude, and absolute magnitude( H ) in the FOM computation are exponential i.e., for brighter, larger, seen lessrecently, shorter arc targets, the FOM rises exponentially. The remaining termsinvolving the ‘score’ and south polar distance (SPD) are gaussian, where theexpected values are 100 and 60 deg, respectively. The ‘score’ term is weightedlower (multiplied by 0.5) than the others to avoid it dominating the priorityranking. The ‘not seen’ and ‘arc length’ parameters are linked together suchthat targets with high ‘not seen’ values and low ‘arc length’ values (those thathaven’t been seen in a while and have short arcs) are ranked higher than thosewith both values high (those that haven’t been seen in a while and have longerarcs) or both values low (those that were seen recently and have short arcs).9n order to calculate the values needed for the merit function above and ingeneral, the ephemeris calculations make use of the SLALIB library, which hasbeen wrapped in Python, to handle the position and time calculations. Thislibrary provides the basic functions for the Earth, Moon, and object Cartesianpositions and velocities, precession/nutation matrices and time conversions forUTC and UT1 to TT and TDB. We have used these routines as the basis for ourcode that can calculate the brightness, sky position, motion and angle, altitude,and moon separation for a particular site as a function of time. The sky motionis used to set a suitable maximum exposure time that will not produce trailingbeyond the size of the seeing disk. The computed ephemeris is then intersectedwith both the calculated darkness times at the site as well as the telescope class-specific altitude and hour angle limits. The expected brightness of the target isused to set the length of the requested observation block based on 1 magnitudebins. The number of exposures of the previously calculated maximum exposuretime that will fit in the block, given the block setup (e.g. slew and settle) andper-frame overheads is calculated. After review by the user, the observationrequest is sent to the LCO observing portal and scheduling system for possiblescheduling on the LCO network.For spectroscopy observations, the situation is similar but a little more com-plicated as additional lamp flat and arc calibration observations are also re-quested, bracketing the main science spectrum. For calculating the expectedsignal-to-noise ratio (SNR) on the asteroid target, we make use of a generalizedtelescope and instrument model to build an Exposure Time Calculator (ETC).This makes use of a parameterized taxonomic class to photometric passbandtransformation based on Veres et al. (2015) and the mean ( S + C ) result basedon the findings of Binzel et al. (2015) that S- and C-type asteroids account for ∼
50 % and ∼
15 % of all NEOs respectively. We then calculate an effectivecapture cross-section based on the unobstructed area of the primary mirror.This is then reduced by the throughput, expressed as t = t atm × t tel × t inst × t grating × t ccd (2)where t atm = 10 ( − k/ . X is the throughput of the atmosphere, with k beingthe atmospheric extinction coefficient in the band (in mag/airmass ) and X is the airmass, t tel = 0 . n tel mirr is the telescope throughput with 0.85 beingthe typical reflectivity of overcoated aluminium and n tel mirr is the number oftelescope mirrors, t inst is the instrumental throughput (excluding the grating)which tends to be more uniform with wavelength, t grating and t ccd are theefficiency of the grating and CCD in the observed band respectively. Slit lossesare calculated based on the width of the slit and the typical FWHM of theseeing disk.For the noise sources, we calculate the expected sky background based onthe contributions from airglow (we use the 10.7 cm radio flux data as a proxyfor the progression through the solar cycle which has been shown to correlatewith the airglow intensity e.g. Tapping 2013), zodiacial light (as a functionof ecliptic latitude), the stellar background (as a function of galactic latitude)10nd the Moon. The model for the brightening due to moonlight is based onKrisciunas and Schaefer (1991). The readout noise per pixel is also includedin the noise sources, though we neglect the dark current as CCD cameras inspectroscopic instruments are normally operated cold enough that the darkcurrent is negligible.We are developing a more sophisticated and generalized ETC which willmake use of an “ETC language” to describe all of the elements and surfaces(e.g. atmosphere, mirror, lens, grating, CCD etc) and the relationship betweenthem as encountered by a photon from the top of the atmosphere to the detector.For each surface, the ETC language will support use of a scalar (e.g. a target V magnitude), short vector/per-filter values (e.g. extinction per unit airmass asa function of filter/passband) or a “spectrum” file (e.g. a reflectance spectrum,measured mirror reflectivity as a function of wavelength). This can be combinedwith the selection of the best-matching atmospheric transmission spectrum froma library of pre-calculated versions (such as the ESO Advanced Sky Model; Nollet al. 2012) with selection based on the values of precipitable water vapor, ozone(O ) and aerosols determined from remote sensing. This will be described inmore detail in a future publication.For spectroscopic calibration, nightly flux standards are automatically ob-served using each FLOYDS instrument. These publicly available flux standardsare used to create an airmass curve and account for nightly perturbations inatmospheric transmission. Additionally, for Solar System observations, Solaranalogs are required to properly obtain a reflectance spectrum. NEOexchangeis capable of automatically selecting and scheduling a suitable star from a listof Sun-like options. Stars that are closer on the sky to the target Solar Systemobject are given preference so that observing conditions might be as similar aspossible between the analog and the primary target. The exposure time of theanalog spectrum is automatically determined based on the brightness of theanalog and the resulting spectra are then stored and uploaded to the NEOex-change website. Though this process has been fully automated and streamlined,it should be noted that several fundamental limitations are imposed by the LCOobserving portal and scheduler. The greatest of these limitations is the inabilityto create single observations that include multiple targets. Because of this, wemust schedule the target and the analog separately, giving no guarantee thatboth will be observed on a given night. As neither observation is completewithout the other, this adds some risk of lost time to Solar System spectro-scopic observations with LCO. Additionally, due to the constant updating ofthe LCO observing schedule, it can be difficult to schedule both target and ana-log at the same airmass, which is ideal for optimal reduction. We have found,however, that this latter limitation can be mitigated somewhat by use of theaforementioned flux standards so long as both observations were made on thesame night. During and after the window of validity of the observing requests, we checkwith the LCO observing portal whether the request has been executed on the11CO network. If it has, we check with the LCO Science Archive for the pres-ence of reduced frames. Data taken on the telescopes of the LCO networkare automatically transferred back to the headquarters in Santa Barbara, CAand pipeline processed. This occurs in near real-time, typically within ∼ BANZAI pipeline (Mc-Cully et al., 2018) which performs the standard steps of assembling mastercalibration frames from the individual biases, darks, and flat fields. These cal-ibrations are then applied to the science images to perform bad pixel masking,bias subtraction, dark current correction, and flat field division. Crosstalk cor-rection and gain normalization between the individual quadrants and amplifiersof the Sinistro cameras’ Fairchild CCDs are also performed. An astrometric so-lution is performed using the astrometry.net software (Lang et al., 2010) whichmakes use of the 2MASS catalog (Skrutskie et al., 2006) as input. A catalog ofsources detected in the frame (having a certain minimum number of pixels morethan 10 σ above the fitted sky background) is also produced using SExtractor (Bertin and Arnouts, 1996). Finally the reduced frames, source catalog, and as-sociated master calibration frames are uploaded to the LCO Science Archive for distribution to end-users. These data products are then retrieved by NEOex-change for further astrometric and photometric analysis.Although the BANZAI pipeline does a good job at removing the instrumen-tal signature from the data and providing a “first pass” astrometric solution,we cannot make use of this pipeline-produced astrometric solution and sourcecatalog as-is. This is because the camera distortions, although small, are toolarge for our astrometry goals to be met with a simple linear fit. In addition,because many of our NEO targets are very faint, they will not be includedin the relatively shallow pipeline-produced source catalog. To solve both ofthese problems, we instead re-determine a new astrometric solution, incorpo-rating spatially-varying distortion polynomials. This astrometric solution ini-tially used the PPMXL catalog (Roeser et al., 2010), progressed to using theGaia-DR1 catalog (Gaia Collaboration et al., 2016) and now makes use of theGaia-DR2 catalog (Gaia Collaboration et al., 2018) to derive the astrometricsolution and a per-frame photometric zeropoint using the SCAMP software .The zeropoint is determined by cross-matching the detected CCD sourceswith sources in the Gaia-DR2 catalog by position and then iterating, with out-lier rejection, to determine the difference in mean magnitude between the in-strumental CCD magnitudes and the Gaia catalog G magnitudes. This ig- https://archive.lco.global/ G passband and the equally broadPanSTARRS-w, which comprise over 85% of our frames, and does not takeinto account the source color (almost always unknown) or stellar color. Thecolor correction for reference star colors was not possible when this part of thepipeline was originally written with the absence of accurate colors in either thePPMXL or Gaia DR1 catalogs. Given the large field of view of the LCO tele-scopes and the corresponding large number (hundreds) of potential referencestars in a typical field, the outlier rejection procedure for the zeropoint is robustagainst including stars with large discrepancies in color and magnitude. Withthe availability of G BP and G RP magnitudes in Gaia DR2, and the prospect ofspectral types for large numbers of reference stars from low resolution spectrain Gaia DR3, there is the potential to revisit this in the future versions of thepipeline with an improved treatment.The result of the fitting process and a record of the resulting processed framedata product is stored in the database. After checking that the results of the fitsare satisfactory, we perform a source extraction again using the SExtractor (Bertin and Arnouts, 1996) software but to a lower threshold than in BANZAI(3.0 vs 10.0 σ above the sky background) to include many more faint sources.The resulting source catalog contents are ingested into the database (formingthe largest fraction of the database size).This additional processing allows us to extract light curves for any object inthe database through either user tools or via the web frontend (see Figure 2).The source catalogs can also be exported to specialized moving object detectionsoftware written by the Catalina Sky Survey team (Shelly, 2016). The resultsof any detections are also ingested into the database and can be overlaid in a“Candidate Analyzer” in the web frontend. This allows the user to blink throughthe acquired frames for that candidate object and the moving object detectionsare overlaid, along with details about the position (both on the CCD and on-sky), magnitude, and the rate and direction of motion, allowing comparison withthe predicted motion to assist confirmation of a NEOCP candidate’s recovery.The user can then reject or confirm the identification which will then show asummary of the observation in the MPC1992 80 column format which can thenbe sent to the MPC. Moving object detections of real but unknown objects canalso be confirmed, in which case a new local candidate object is created in thedatabase.As discussed above, we make use of the Gaia-DR1 (Gaia Collaboration et al.,2016) and Gaia-DR2 (Gaia Collaboration et al., 2018) catalogs to perform theastrometric reduction. Gaia-DR1 greatly improved the quality of astrometryobtained by substantially reducing the systematic error contribution by vastlyreducing the catalog zonal errors (Spoto et al., 2017). The overall astrometricuncertainty is a combination of centroiding error (which is unaffected by thechoice of reference catalog), systematics from the reference catalog and othernormally smaller second-order effects such as stellar proper motion and differ-ential chromatic refraction.Switching from PPMXL (Roeser et al., 2010), which we used prior to theavailiability of the Gaia catalogs, to Gaia-DR1/2 has reduced the systematic13atalog error from ∼
300 mas to ∼
30 mas and the overall uncertainty ( ∼ . (cid:48)(cid:48) –0 . (cid:48)(cid:48) ) is now dominated by the centroiding error. With the release of DR2 inApril 2018 and the availability of good reference star colors, as well as the releaseof parallaxes and proper motions in later data releases, it would be possible totake other more subtle effects into account in the astrometric reduction.These effects include the differential chromatic refraction (DCR), space mo-tions of the reference stars, and unmodelled optical distortions. DCR is causedby differences in the spectral energy distribution of the Solar System targets andreference stars being refracted differently in the atmosphere, which in terms de-pends on the variation of temperature, pressure and water vapor (Stone, 1996)both from site to site and with time. The amount of DCR also depends on thefilters used and the optical path length through the atmosphere, which dependson the (changing) zenith distance and hour angle of the target. Historically, thelack of even accurate colors for the majority of reference stars, has made correc-tion of DCR difficult without obtaining large amounts of additional multi-colorcalibrated photometry. Similarly correcting for the proper motions, and lesscommonly for the parallax, of the reference stars has been difficult due to lackof accurate available data for the fainter ( V ∼ ∼ . The FLOYDS Spectroscopy Pipeline is automatically run on all FLOYDSdata obtained by the LCO network in a manner similar to the BANZAI pipelinedescribed above. The pipeline converts the raw, folded, multi-order fits framesinto a 1-dimensional extracted trace of the entire merged spectrum. Duringthis processing, lamp flats are used to minimize fringing in the red branch, thenstored flux calibration frames and telluric lines are used for atmospheric cor-rection. The final trace, as well as all intermediate data products, are wrappedin a tarball along with the guider images and made available to users via theLCO archive. A full description of the data processing, wavelength calibra-tion, and extraction is given on the LCO website . The data resulting fromthis automated pipeline is typically of sufficient quality to determine a roughspectral slope for the target and serves as a good first look. However, whenhigher quality reduction is needed, a manual version of the pipeline can be runthat uses lines from an arc lamp observed before and after the object frames https://lco.global/documentation/data/floyds-pipeline/
14o perform wavelength calibration, as well as more recent flux standards thatcan improve atmospheric correction. NEOexchange retrieves these data fromthe LCO archive and automatically creates a reflectance spectrum with themost proximate Solar Analog spectrum observed with the same telescope as theobject of interest.
One of the priorities for creating TOM (Target and Observation Manage-ment) systems like NEOexchange is to allow them to be developed, adapted,and deployed by different groups for their own particular science interests andfollow-up assets. Furthermore, scientific reproducibility is enhanced by havingthe full chain of software that was used to produce a scientific result available,along with the platforms the software operates on, which can be replicated bythird parties via virtual machines (Morris et al., 2017).Mindful of the above, we have adopted the following philosophies: • the use of Python as the programming language, • the use of version control ( git and github.com ) to manage revisions to thesoftware, • the adoption of Test-Driven Development (TDD) methodologies to de-velop and maintain the software, • the use of container technology to package and deploy the running soft-ware,The NEOexchange platform is written in Python which has had a high takeup for astronomy software and allows access to a large variety of astronomy-specific packages such as astropy (Astropy Collaboration et al., 2013) and astroquery (Ginsburg et al., 2019). This is the most flexible choice at presentfor open access astronomy software.Although the use of Test-Driven Development (TDD) in scientific softwareis currently small (Nanthaamornphong and Carver, 2017), it can provide manybenefits. The use of unit tests for the individual low-level functions and func-tional tests for the overall website operation helps in building more reliable andmaintainable software over the long term. The combination of tests, along withthe use of packaging and deployment technologies (described later), improvesthe reproducibility of scientific results by allowing others to reproduce them ona wide variety of platforms. Given the use of the Python programming languagefor NEOexchange, we decided to adopt pytest and Selenium for developingand running the unit and functional tests respectively. This builds on our pre-vious efforts (e.g. Lister et al. 2016) but uses more modern web developmentand deployment frameworks such as Django and Docker to make a redeployable container platform to package and run the software.Docker has quickly emerged as the technology of choice for software contain-ers. Docker is an operating system level virtualization environment that usessoftware containers to provide isolation between applications (compared to vir-tual machines (VMs) which virtualize at the hardware level and require a guestoperating system per VM). Through the use of a standardized file format (the Dockerfile ) for describing and managing the setup of the containers, we canseparate the sub-components of NEOexchange (the database, web frontend, andtasks backend) into standardized containers for each component.This ability to be able to completely describe a software deployment envi-ronment with Docker has the potential to improve the reproducibility and thesharing of data analysis methods and techniques for the science and researchcommunity. This can allow other researchers the ability to reproduce and buildon the original work (e.g. Boettiger 2015) or to develop and distribute con-tainerized versions of tools (e.g. Nagler et al. 2015) which can otherwise behard to install and setup.It is often the case that the research team developing the software does nothave (or desire) direct control and management over the software environmentwhere their software will be deployed. The deployment platform may be con-figured with different versions of the operating system, Python libraries, andmodules than those that are required by their software, and yet are necessaryfor the stability and long-term support of other applications. This can presentproblems when attempting to update the version of these infrastructure compo-nents, as it is normally very difficult to isolate the different system componentsfrom each other, requiring all components to be updated at the same time.With Docker, each application can be wrapped in a container configuredwith the specific version of operating system, Python interpreter, and Pythonpackages and modules it needs to function properly. The common interface withthe system is set at the container level, not at the operating system, program-ming language, or application server level. In this container-based approach todeploying a service like NEOexchange, the development process includes a con-tainer specifically designed for the service, with only the dependencies needed bythe service. The same container that is used during the development and testingof the NEOexchange software on the developer’s local machine can be deployedon a local computer cluster or in a private part of a cloud service, and evenbecomes part of the final project deliverable. The final product is shipped anddeployed as the container, with all of its dependencies already installed, ratherthan as an individual software component which requires a set of libraries andcomponents that need to be installed along with it. This not only simplifies the find orb orbit fitting code (which needs more moderncompilers than our base CentOS SEx-tractor , scamp , and the CSS moving object code), the main NEOexchangeapplication code, and finally the crontab entries used for the object and datadownloads (as described in Section 3.1 and Section 3.4) and other maintenancetasks. The containers can be built and rebuilt using the Docker command-linetools and are also rebuilt automatically by the LCO-wide Jenkins automatedbuild servers whenever commits are made to the NEOexchange github reposi-tory.For deployment we make use of Amazon Elastic Kubernetes Service (EKS)to deploy our Docker containers to the Amazon Cloud and allow users to accessthe system. Kubernetes is a container orchestration system for automatingdeployment and management of containerized applications. EKS makes useof Amazon’s Elastic Compute Cloud (EC2) to provide the virtual servers forrunning the NEOexchange application and the S3 storage system to store thefiles needed. We use Kubernetes to deploy several containers that setup andthen run the NEOexchange application. Two initialization containers are run atstartup; one collects all of the static content (HTML, CSS, images, JavaScript)needed by the Django framework and webserver that runs the NEOexchangeapplication and the other is used to download the DE430 JPL ephemeris file(Folkner et al., 2014) to the shared S3 storage for use by find orb during orbitfitting. Following the completed execution of these initialization containers, wedeploy three other containers for the main NEOexchange application. Theseare the nginx webserver, the backend NEOexchange application itself, and thecontainer for running the background and update tasks through the crontab .The Kubernetes system and the Amazon EKS handles constructing the neededservices to direct web traffic from the internet to the NEOexchange applicationand its webserver. It also handles redeploying the NEOexchange containersshould the application or one of the underlying virtual servers go down, or ifan upgrade to a new version of our application is deployed. In the latter case,the new version is brought up in parallel and confirmed to be working beforerouting traffic over from the old version, at which point the older deployedversion is removed. Kubernetes also allows us to run a developmental version ofthe NEOexchange code in parallel, using a separated parallel database, withoutdisturbing the production system. This allows us to build additional confidencethat the system and any new features are working correctly, beyond the checks https://jenkins.io https://kubernetes.io
17f our unit and functional tests, before making the new version live.The use of a common, widely adopted programming language and web frame-work, coupled with modern software development methodologies, particularlyTest Driven Development, has enabled us to develop and add many new fea-tures to the NEOexchange codebase while ensuring that the code continues towork as expected. The adoption of containerizing technologies to package thesoftware and all needed dependencies has also, after an initially steep learn-ing curve, been greatly beneficial in that it has served to insulate the codeand ourselves from many of the demands and trials of system administration.The more recent move to a more virtualized deployment strategy using cloud-based services, has similarly produced great medium and long-term benefits byallowing NEOexchange to be kept continuously running during seamless andsafe upgrades alongside parallel deployments for testing of new versions withoutimpact to the production system.
4. Use of the LCO Network for NEO follow-up and characterization
One of the original motivations for building the NEOexchange system (andits predecessor) was the large number of NEO candidates produced by the NEOsky surveys (thousands/year) requiring follow-up confirmation through astrom-etry and photometry. As described in Sections 3.1 and 3.3, the NEOexchangesystem retrieves new NEO candidates from the MPC, computes ephemerides,and can plan observations and automatically schedule them for follow-up on therobotic telescopes of the LCO network.The NEOexchange homepage shows the candidates in need of follow-up,ranked according to the FOM equation (Equation (1) and Section 3.3) alongwith recently confirmed NEOs. Individual candidates can be scheduled fromhere by following their associated links. There is also a command line user toolwhich can look at the current state of the candidate follow-up needs by query-ing the database and the state of the telescope network and then distributingcandidates to available telescopes at a given time (planning for a future time issupported). By default, this tool retrieves all of the NEOCP candidates thatwere ingested into the database within the last 5 days. In addition to older ob-jects, we also filter out those that have not already been followed up and thosethat haven’t gone more than 2.5 days without being observed. These cuts arenecessary as the short ∼
30 minute discovery arc of typical NEOCP candidatesmeans that their positional uncertainty can grow to many times our field of viewafter this time. All of these cuts can be customized by the user using commandline options.For each of the selected candidates, we compute the ephemeris and they arethen filtered as follows: • V magnitude in the range 19 ≤ V ≤ • on-sky motion ≤ (cid:48)(cid:48) / min, 18 Moon-object separation ≥ ◦ , • not already scheduled to be observed by NEOexchangeFiltered objects are then split into four lists : first by declination (Objects withDec. < = +5 ◦ go into the southern hemisphere list, objects with Dec. > +5 ◦ gointo the northern hemisphere list), and then into a “telescope class” list with V ≤ . find orb orbit fitter, the measurements are sent to the MPC. For targets whichare not automatically identified or which are moving too fast to be identifiedin the individual frames and which require stacking on the object’s motion, thedata is downloaded and analyzed on a local workstation. Manually detectedobjects have their measurements checked for validity and then are reported tothe MPC in the same way as described above. The combination of a large number of telescopes distributed around theworld, coupled with the ability of the LCO scheduler and the site softwareto re-plan, schedule and start execution of a new observing request within ∼
15 minutes, allows the LCO network to respond rapidly to new targets ofopportunity. These objects are primarily those NEO candidates that are deter-mined to have a possibility of impact by the JPL SCOUT early alert system.Since the particular candidate will already be in the NEOexchange system froman earlier ingest of candidates (see Section 3.1), all that is necessary is to triggera disruptive rapid-response request. This involves setting a particular value ofthe observing mode in the observing request which tells the LCO system thatthis request is of high enough priority that it can potentially disrupt and cancel19n already running request at the telescope. This can be done through eitherthe web frontend when scheduling an object or by adding a flag to the commandline tool that was described in Section 4.1.These rapid-response observations are considered first by the LCO schedulerin a separate scheduling run and sent out to the sites, before the rest of theobserving requests are scheduled, which reduces the latency. Once executed atthe telescope, the resulting data is treated in the same way as that from othertargets.As briefly discussed for NHATS and SCOUT targets in Section 3.1, there isa need for a better machine-readable format for requesting transient follow-upgenerally. Although the original SCOUT email alerts (which were machine gen-erated and also semi-structured and machine readable to some extent) have nowbeen supplemented by an API endpoint, the subsequent discussion about the na-ture of the object and validity of the alert and recommendations for continuingor halting follow-up are all conducted through free-form email conversations.This is also the situation in a number of other of time-domain and transientastronomy fields and illustrates the need to take the “next step” to allow better,faster, and more automated responses to, and prioritization of, new transients.This extends the potential for automation and faster response further down thechain from the producers of alerts and the broker/triage stage to follow-up andcharacterization resources, as has already been done by many surveys e.g. ZTF(Patterson et al., 2019).
Radar observations are a very powerful tool that are used to spatially resolveNEO targets, determine binary fraction and improve orbits, the last of whichhelps us to avoid loosing the target and improves impact risk assessment (e.g.Ostro et al. 2007). The LCO NEO Follow-up Network allows rapid responseastrometry that makes pointing and imaging by radar assets (such as Gold-stone and Arecibo) possible for targets that have only recently been discovered.Potential mission destinations such as NEO Human Space Flight AccessibleTargets Study (NHATS) targets are often small in size with a limited visibilitywindow of days to a few weeks. The LCO NEO Follow-up Network can quicklyrespond for characterization efforts for these objects. The network can alsoobtain colors and photometric light curves which allows the determination ofrotation rates, pole directions, spectral classes, and shapes as well as performrobotic spectroscopy to determine taxonomic classes. The majority of our timeallocation for the project is now spent on characterization efforts for these typesof targets.In addition to the ingest of the characterization targets discussed in Sec-tion 3.1, we also ingest any known physical parameters such as a light curveperiod and amplitude, pole orientation, and taxonomic class if known. This isperformed via a query and parsing of the JPL Small Body Database and storageof the results. In the absence of a period, we perform an initial observation blockof 1 hr in the PanSTARRS-w filter for maximum throughput, with the expo-sure time automatically calculated based on the object’s speed as described in20 igure 4: Example NEO light curves obtained with LCO NEO Follow-up Network: Lightcurve from the 0.4-m telescope in Tenerife (MPC site/telescope code Z17) of NEO (3122)Florence (left) which was observed with the Goldstone radar in 2017 August–September andwas discovered to be a rare triple system and a light curve of NEO 2015 EG from the LCO 1-min Chile (MPC site/telescope code W86) on 2019 March 6 (right) which was also observed thatmonth by the Arecibo planetary radar. The right figure shows an example of the automaticallyproduced plot products, with the light curve plotted on top with the pipeline-determined per-frame zeropoint underneath.
Section 3.3. Following the results of these initial observations, further follow-upfalls into one of three categories:1. The block covered at least ∼ . × the period: period is determined andannounced to community. Move onto taxonomic spectroscopy (for V (cid:46)
17) or colors (for V (cid:46) . . . . ∼ . H V = 14 . ∼ .
047 au) approach to the Earth in 2017 September; this was the closest ap-proach until the 2057 apparition, making it an important opportunity to studythis potentially very damaging large NEO. It was also the target of a citizenscience Asteroid Tracker campaign for that year (see Section 4.4).Observations were carried out with the southern 1-meter telescopes andnorthern 0.4-meter telescopes of the LCO network prior to and following closeapproach respectively. One of the light curves from one of the northern 0.4-meter telescopes in Tenerife (which has a MPC site/telescope code of Z17) isshown in Figure 4.3. These data confirmed the 2.36 hr period found in the 1996–1997 apparition by Pravec et al. (1998). (3122) was subsequently shown to bea triple system with 2 small companions in radar images.21015 EG was discovered by Catalina Sky Survey’s Mount Lemmon Stationin 2015 March and its Earth-crossing short period ( P ∼
293 day) and eccentric( e ∼ .
36) Aten orbit means it makes frequent close approaches to both Earth andVenus. It was targeted by the Arecibo planetary radar on the next return in 2019March, five orbital periods later. A 1.7 hour observation block two days afterclosest approach with the LCO 1-meter at Cerro Tololo (MPC site/telescopecode W86) showed a complex light curve with two unequal maxima and minima(Figure 4.3). The existence of distinct maxima and minima in the light curvesuggests an elongated shape but the fact that the two maxima and two minimaare themselves unequal, shows that the situation is more complicated. Thiswould need a more thorough light curve inversion, ideally incorporating theradar data in a combined solution. We determine a period of ∼ .
717 hours basedon the two maxima, in contrast to the previous determination of 1.29 hoursof Thirouin et al. (2016), which was an estimate based on incomplete phasecoverage. The radar ranging also secured a provisional weak detection of theYarkovsky acceleration (see next section) on this object, making the Yarkovskyacceleration a more secure 2 . σ detection. The semi-major axis drift some asteroids undergo due to the non-gravitationalYarkovsky effect, caused by the unequal absorption and re-radiation of thermalenergy on a rotating asteroid, can have a large influence on determining impactprobabilities for asteroids that pass close to the Earth (Giorgini et al. 2002,2008; Farnocchia et al. 2013; Farnocchia et al. 2014; Chesley et al. 2014; Spotoet al. 2014; Vokrouhlick´y et al. 2015). Even a small drift in semimajor axis canmake the difference between a near-miss and an impact for close-passing aster-oids when calculating long-term impact probabilities, especially when planetaryclose encounters perturb the orbit.The various command-line tools available through NEOexchange (describedin Section 3.2) were used to calculate the observability, brightness, and rate ofmotion of 36 asteroids determined to yield a high likelihood of a detection ofthe non-gravitational Yarkovsky effect (Greenstreet et al., 2019). The targetswere ingested into the NEOexchange database, which allowed for ephemeriscalculation, planning, and scheduling of observations on the LCO network.
Through the robotic FLOYDS spectrographs on the two 2-m telescopeson Maui, HI and Siding Spring, Australia (see Figure 1), we can performrapid-response low resolution spectroscopy of NEOs and other asteroids. Af-ter scheduling of spectroscopic observations (described in Section 3.3) and datareduction (described in Section 3.4), the data products for the target and the so-lar analog are extracted and made available through the NEOexchange websitefrom the requested observation block. This also allows display and download ofthe acquisition and guide movies made from the autoguider frames taken duringthe observations. 22 igure 5: Examples of optical reflectance spectra of NEO and Main Belt asteroids obtainedwith the FLOYDS spectrographs. (left) Normalized reflectance of NEO (433) Eros (blue)along with its optical/NIR template spectrum from SMASS (black line). Also shown aresome other examples of S- and X-type asteroids. (right) Spectra of (2) Pallas, (7) Iris and(16) Psyche, along with the boundaries of the mean B-, S-, and X-type taxonomic classesfrom SMASS (dotted lines).
The NEOexchange site also allows a normalized reflectance spectrum to beautomatically generated with the nearest-in-time suitable solar analog auto-matically chosen, plotted, and compared to taxonomic class templates from theSMASS library DeMeo et al. (2009); DeMeo et al. (2009). The plotting tool inthe website permits users to interactively pan and zoom the spectra, and allowscustomized zooms of the plotted spectra to be saved to the user’s computer.We plan to add the ability to reports wavelength and reflectance values at eachdatapoint and display known important lines from line lists such as atmosphericO , O , & H O and mineral spectra. Some example NEO and Main Belt Aster-oid reflectance spectra taken during the early commissioning phase of FLOYDSfor moving objects are shown in Figure 5.
We participated in both the first NASA Planetary Defense Coordination Of-fice (PDCO) Planetary Defense exercise in 2017 October on 2012 TC4 (Reddyet al., 2019) and the second exercise in 2019 May on NEO (66391) 1999 KW4.Due to the approach geometry, with the NEO approaching from low in theSouthern sky, LCO’s Faulkes Telescope South (FTS) was one of the few tele-scopes available to observe it before close approach. In preparation for thisevent, we worked extensively with the LCO software team to improve the ac-quisition and guiding performance of the FLOYDS low resolution spectrographs,for fast moving objects ( > (cid:48)(cid:48) / min), with a particular view to enhancing thecapabilities to obtain spectra in support of the PDCO exercise on 1999 KW4.We were the only people able to get spectra before close approach and wereable to get the data reduced and analyzed within a few hours of acquisition andcirculated to the campaign participants approximately 17 hours before closeapproach. The FTS spectra showed that the taxonomic class of the NEO was23 M a g n i t u d e M a g n i t u d e M a g n i t u d e Figure 6: Results from the LCO NEO Follow-up Network on PDCO target (66391) 1999 KW4(left) Light curves obtained with the 1m telescopes at Siding Spring Observatory, Australiaon 2019 May 25, 26 and 30. The integer part of the MJD has been subtracted from the timeof each datapoint and the remaining decimal day is used for x /time axis. A small numberof outlier points have been clipped from each light curve where the target passed close to afield star or thin cloud intervened. (right) Normalized reflectance of NEO (66391) 1999 KW4(purple line) along with the boundaries of the mean C-, S-, X- and Q-type taxonomic classesfrom SMASS (dotted lines). also more like the rarer Q-type rather than the S-type previously measured (seeFigure 4.3.3). We were also able to obtain a few light curves from several siteson the 1-meter network but extensive bad weather prevented us from gatheringenough consecutive data from the southern sites of the LCO network to providefull coverage of the 17.4 hour orbital period that would enable the full modellingof the system.During the preparation for the 1999 KW4 campaign, we developed code toconstruct on-demand long-term status plots for any object in the NEOexchangedatabase to show sky position, distance of the object from the Earth/Sun, mag-nitude and positional uncertainty. These will allow easier visualization andtargeting of observing campaigns. We used the software developed for this program to create Asteroid Tracker ,with the aim of increasing public awareness of asteroid phenomena, timed withAsteroid Day 2016. This virtual event resulted in an audience of over 900 peoplesigning up to each request observations on the LCO network of two NEO radartargets we had selected for Asteroid Day 2016. This provided a greatly simpli-fied and streamlined version of the NEOexchange portal for the general publicallowing them to request a small number of observations on the network, usinga single button, which would be combined together into a full-length sequence. https://minorplanetcenter.net/iau/lists/ObsCodesF.html https://asteroidtracker.lco.global able 1: Table of observations for (66391) 1999 KW4 with LCO Block start and end LCO Telescope MPC Observation Num.(UTC) Code Class Site Code Type Exposures2019-05-25 08:20 → → → → → → → → → → → → , called AgentNEO , using the Project Builder interface. This was a citizen science project toidentify and report NEO candidates in LCO data as described in Section 4.1.We used the Moving Target Detection algorithm to identify possible new NEOsin each follow up frame, as well as the target being followed up. For each can-didate we created a cropped thumbnail image around the target, and uploadedthe sequence of images. The Zooniverse users would step through the each can-didate sequence and classified as either a genuine target or artifact. Each imagesequence was given a unique identifier and stored in NEOexchange, so the re- Agent NEO on 2017 June 1 to tie in with Asteroid Day 2017 (30 June)and operated it for two months during the summer of 2017. A rapid cycle ofredesign was necessary when creating
Agent NEO to match the NEOexchangeworkflow to that of the Zooniverse Project Builder (chiefly the handling of an-imated image streams). Supporting such a large number of volunteers on theZooniverse required considerably more person-effort than initially considered.We also encountered issues with the timeliness of the data flow through the LCOand NEOexchange pipelines and with large numbers of detector/flatfielding ar-tifacts in the LCO Sinistro cameras on the 1-meter telescopes. These artifactsmimicked the moving NEO candidates, which often have uncertain positions,making identification and recovery of the correct moving object NEO candi-date difficult. This problem also affects the main LCO NEO Follow-up Networkprogram for candidates (Section 4.1. We had hoped that the situation would im-prove in 2018–2019 with the deployment of new Archon CCD controllers for theSinistro cameras (Harbeck et al., 2020) but this proved overly optimistic. Thishas been part of the reason for the change in focus of the LCO NEO Follow-upNetwork from candidate confirmation towards NEO characterization.
5. Discussion and Future Work
We have described the design, implementation and operation of NEOex-change, an online portal and TOM (Target and Observation Management) sys-tem for NEO and other small body science. We have also described the use ofNEOexchange to operate the LCO NEO Follow-up Network conduct a programof follow-up to conduct a program of NEO candidate confirmation initially, witha transition into more detailed characterization of radar and other high valueNEO targets, as usage and competition for time on the LCO network increased.Follow-up of NEO candidates with the LCO telescopes over the July 2014–July 2018 time period for which complete statistics are available from theMPC shows that 39300 measurements were reported. Of these, almost 9000measurements were of nearly 1300 objects that were confirmed to be NEOsfrom Catalina, PS1, NEOWISE and other surveys. Although part of this period(before 2015 July) of follow-up of NEO candidates predates NEOexchange sowe cannot search our database for these statistics (and the MPC database isnot externally queryable in this way). Examining the subset of 468 reportedconfirmed NEOs in the NEOexchange database and observed and reported tothe MPC by LCO network in 2015 July to 2018 July, shows that 310 of 468( ∼ < R ∼
23 showing the potential of the 1-meter network forNEO follow-up.We will be continuing to develop NEOexchange to allow use by other users(professional and citizen scientists), observation campaign planning and onlinedata analysis and reporting. Some intended areas of improvement have alreadybeen discussed but they are also collected below: • additional interactive tools to allow users to perform period finding of lightcurves and taxonomic determinations from asteroid spectra, • automated stacking of frames for faster moving objects and detection ofthe asteroids in the resulting stacked images, • investigation of the effect of 2nd-order astrometry effects such as DCR,proper motions and residual optical distortions (see Section 3.4 • completion of the development of the generalized Exposure Time Calcu-lator (ETC) framework for representing the atmosphere above a site, tele-scope and instruments and allowing more accurate prediction of neededexposure time to reach a particular SNR (Section 3.3), • use of the above ETC framework to allow modelling of instruments onlarger telescopes such as the Goodman optical spectrograph on the SOAR4.1-meter and the SCORPIO instrument (optical-NIR imager/spectrograph)on the Gemini South 8-m telescope as part of the Astronomical EventsObservatory Network (AEON ). • additional tools for coordination of characterization efforts such as ag-gregation and display of planned radar targets, citizen-science campaignswith Asteroid Tracker, follow-up observations from LCO, AEON otherfacilities and status of these efforts.AEON aims to bring together a network of observatories to allow time-domain follow-up and science to be performed in a rapid, efficient and homo-geneous way. It adds the SOAR 4.1-m telescope and Goodman spectrograph(optical) and the Gemini South 8-m telescope and likely the SCORPIO in-strument (optical-NIR imager/spectrograph) once commissioned to the existingLCO telescope network of 0.4, 1, and 2-m telescopes, creating a more powerful“virtual facility”. This “virtual facility” becomes a powerful tool for the studyof all kinds of solar system objects and transient behavior and subsequent evo-lution with access to telescope resources from 0.4–8-m with optical and NIRimaging and low resolution spectroscopy instruments.The production of transient alerts from current and future sky surveys, theintegration of alert streams from alert brokers such as ANTARES (Saha et al.,2014; Narayan et al., 2018), with the target coordination and follow-up capabili-ties provided by TOMs such as NEOexchange, and rapid response telescopes and https://lco.global/aeon Acknowledgements
Funding from the NASA NEOO program through grants NNX14AM98Gand 80NSSC18K0848 to Las Cumbres Observatory (LCO) is acknowledged. Weare very grateful to all of the LCO Engineering, IT, and Software teams fordesigning and building the hardware and software that makes the LCO NEOFollow-up Network possible. In particular we would like to thank Steve Foalefor the improvements to the guider software to make the FLOYDS robotic spec-troscopy of moving objects possible and Ira Snyder for the development of, andvery valuable assistance with, Kubernetes and the LCO Amazon Cloud infras-tructure. This research has made use of data and/or services provided by theInternational Astronomical Union’s Minor Planet Center and the VizieR cat-alogue access tool, CDS, Strasbourg, France (DOI: 10.26093/cds/vizier). Thiswork makes use of observations from the Las Cumbres Observatory global tele-scope network.S. Greenstreet acknowledges support from the Asteroid Institute, a programof B612, 20 Sunnyside Ave, Suite 427, Mill Valley, CA 94941. Major funding forthe Asteroid Institute was generously provided by the W.K. Bowes Jr. Founda-tion and Steve Jurvetson. Research support is also provided from Founding andAsteroid Circle members K. Algeri-Wong, B. Anders, R. Armstrong, G. Baehr,The Barringer Crater Company, B. Burton, D. Carlson, S. Cerf, V. Cerf, Y.Chapman, J. Chervenak, D. Corrigan, E. Corrigan, A. Denton, E. Dyson, A.Eustace, S. Galitsky, L. & A. Fritz, E. Gillum, L. Girand, Glaser Progress Foun-dation, D. Glasgow, A. Gleckler, J. Grimm, S. Grimm, G. Gruener, V. K. Hsu& Sons Foundation Ltd., J. Huang, J. D. Jameson, J. Jameson, M. JonssonFamily Foundation, D. Kaiser, K. Kelley, S. Krausz, V. Laˇsas, J. Leszczenski,D. Liddle, S. Mak, G.McAdoo, S. McGregor, J. Mercer, M. Mullenweg, D. Mur-phy, P. Norvig, S. Pishevar, R. Quindlen, N. Ramsey, P. Rawls Family Fund, R.Rothrock, E. Sahakian, R. Schweickart, A. Slater, Tito’s Handmade Vodka, T.Trueman, F. B. Vaughn, R. C. Vaughn, B. Wheeler, Y. Wong, M. Wyndowe,and nine anonymous donors.S. Greenstreet acknowledges the support from the University of Washing-ton College of Arts and Sciences, Department of Astronomy, and the DIRACInstitute. The DIRAC Institute is supported through generous gifts from theCharles and Lisa Simonyi Fund for Arts and Sciences and the Washington Re-search Foundation. 28 eferencesReferences
Astropy Collaboration, Robitaille, T.P., Tollerud, E.J., Greenfield, P., Droet-tboom, M., Bray, E., Aldcroft, T., Davis, M., Ginsburg, A., Price-Whelan,A.M., Kerzendorf, W.E., Conley, A., Crighton, N., Barbary, K., Muna, D.,Ferguson, H., Grollier, F., Parikh, M.M., Nair, P.H., Unther, H.M., Deil,C., Woillez, J., Conseil, S., Kramer, R., Turner, J.E.H., Singer, L., Fox, R.,Weaver, B.A., Zabalza, V., Edwards, Z.I., Azalee Bostroem, K., Burke, D.J.,Casey, A.R., Crawford, S.M., Dencheva, N., Ely, J., Jenness, T., Labrie,K., Lim, P.L., Pierfederici, F., Pontzen, A., Ptak, A., Refsdal, B., Servillat,M., Streicher, O., 2013. Astropy: A community Python package for astron-omy. Astronom. Astrophys. 558, A33. doi: , arXiv:1307.6212 .Bertin, E., Arnouts, S., 1996. SExtractor: Software for source extraction. A&AS117, 393–404. doi: .Binzel, R.P., Reddy, V., Dunn, T.L., 2015. The Near-Earth Object Population:Connections to Comets, Main-Belt Asteroids, and Meteorites, in: Michel, P.,DeMeo, F.E., Bottke, W.F. (Eds.), Asteroids IV, pp. 243–256. doi: .Boettiger, C., 2015. An Introduction to Docker for Reproducible Research.SIGOPS Oper. Syst. Rev. 49, 71–79. URL: https://doi.org/10.1145/2723872.2723882 , doi: .Brown, T.M., Baliber, N., Bianco, F.B., Bowman, M., Burleson, B., Conway,P., Crellin, M., Depagne, ´E., De Vera, J., Dilday, B., Dragomir, D., Dub-berley, M., Eastman, J.D., Elphick, M., Falarski, M., Foale, S., Ford, M.,Fulton, B.J., Garza, J., Gomez, E.L., Graham, M., Greene, R., Haldeman,B., Hawkins, E., Haworth, B., Haynes, R., Hidas, M., Hjelstrom, A.E., Howell,D.A., Hygelund, J., Lister, T.A., Lobdill, R., Martinez, J., Mullins, D.S., Nor-bury, M., Parrent, J., Paulson, R., Petry, D.L., Pickles, A., Posner, V., Ros-ing, W.E., Ross, R., Sand, D.J., Saunders, E.S., Shobbrook, J., Shporer, A.,Street, R.A., Thomas, D., Tsapras, Y., Tufts, J.R., Valenti, S., Vander Horst,K., Walker, Z., White, G., Willis, M., 2013. Las Cumbres Observatory GlobalTelescope Network. Publ. Astron. Soc. Pac. 125, 1031. doi: , arXiv:1305.2437 .Chesley, S.R., Farnocchia, D., Nolan, M.C., Vokrouhlick´y, D., Chodas, P.W.,Milani, A., Spoto, F., Rozitis, B., Benner, L.A.M., Bottke, W.F., Busch,M.W., Emery, J.P., Howell, E.S., Lauretta, D.S., Margot, J.L., Taylor, P.A.,2014. Orbit and bulk density of the OSIRIS-REx target Asteroid (101955)Bennu. Icarus 235, 5–22. doi: .29eMeo, F., Binzel, R.P., Slivan, S.M., Bus, S.J., 2009. Bus-DeMeo Aster-oid Taxonomy V1.0. EAR-A-VARGBDET-5-BUSDEMEOTAX-V1.0. NASAPlanetary Data System.DeMeo, F.E., Binzel, R.P., Slivan, S.M., Bus, S.J., 2009. An extension of theBus asteroid taxonomy into the near-infrared. Icarus 202, 160–180. doi: .Eastman, J.D., Brown, T.M., Hygelund, J., van Eyken, J., Tufts, J.R., Barnes,S., 2014. NRES: the network of robotic Echelle spectrographs, in: Ground-based and Airborne Instrumentation for Astronomy V, SPIE. p. 914716.doi: .Farnocchia, D., Chesley, S.R., Chodas, P.W., Micheli, M., Tholen, D.J., Milani,A., Elliott, G.T., Bernardi, F., 2013. Yarkovsky-driven impact risk analysisfor asteroid (99942) Apophis. Icarus 224, 192–200. doi: , arXiv:1301.1607 .Farnocchia, D., Chesley, S.R., Tholen, D.J., Micheli, M., 2014. High preci-sion predictions for near-Earth asteroids: the strange case of (3908) Nyx.Celestial Mechanics and Dynamical Astronomy 119, 301–312. doi: .Folkner, W.M., Williams, J.G., Boggs, D.H., Park, R.S., Kuchynka, P., 2014.The Planetary and Lunar Ephemerides DE430 and DE431. InterplanetaryNetwork Progress Report 42-196, 1–81.Gaia Collaboration, Brown, A.G.A., Vallenari, A., Prusti, T., de Bruijne, J.H.J.,Babusiaux, C., Bailer-Jones, C.A.L., Biermann, M., Evans, D.W., Eyer, L.,Jansen, F., et al., 2018. Gaia Data Release 2. Summary of the contentsand survey properties. A&A 616, 616. doi: , arXiv:1804.09365 .Gaia Collaboration, Brown, A.G.A., Vallenari, A., Prusti, T., de Bruijne,J.H.J., Mignard, F., Drimmel, R., Babusiaux, C., Bailer-Jones, C.A.L., Bas-tian, U., et al., 2016. Gaia Data Release 1. Summary of the astromet-ric, photometric, and survey properties. Astronom. Astrophys. 595, A2.doi: , arXiv:1609.04172 .Ginsburg, A., Sip˝ocz, B.M., Brasseur, C.E., Cowperthwaite, P.S., Craig,M.W., Deil, C., Guillochon, J., Guzman, G., Liedtke, S., Lim, P.L., Lock-hart, K.E., Mommert, M., Morris, B.M., Norman, H., Parikh, M., Pers-son, M.V., Robitaille, T.P., Segovia, J.C., Singer, L.P., Tollerud, E.J.,de Val-Borro, M., Valtchanov, I., Woillez, J., the astroquery collaboration,2019. astroquery: An Astronomical Web-querying Package in Python. AJ157, 98. URL: https://doi.org/10.3847%2F1538-3881%2Faafc33 , doi: . 30iorgini, J.D., Benner, L.A., Ostro, S.J., Nolan, M.C., Busch, M.W., 2008.Predicting the Earth encounters of (99942) Apophis. Icarus 193, 1–19. doi: .Giorgini, J.D., Ostro, S.J., Benner, L.A., Chodas, P.W., Chesley, S.R., Hud-son, R.S., Nolan, M.C., Klemola, A.R., Standish, E.M., Jurgens, R.F., Rose,R., Chamberlin, A.B., Yeomans, D.K., Margot, J.L., 2002. Asteroid 1950DA’s Encounter with Earth in 2880: Physical Limits of Collision ProbabilityPrediction. Science 296, 132–136. doi: .Greenstreet, S., Farnocchia, D., Lister, T., 2019. Measuring the Yarkovskyeffect with Las Cumbres Observatory. Icarus 321, 564–571. doi: .Harbeck, D., Smith, C., Douglas, D., Rosing, W., McCully, C., 2020. UpgradingLas Cumbres Observatory’s Sinsitro imagers: New Archon CCD Controllersand Cryotiger Operations Improvement. In preparation.Keys, S., Vereˇs, P., Payne, M.J., Holman, M.J., Jedicke, R., Williams, G.V.,Spahr, T., Asher, D.J., Hergenrother, C., 2019. The digest2 NEO Classifica-tion Code. Publ. Astron. Soc. Pac. 131, 064501.Krisciunas, K., Schaefer, B.E., 1991. A model of the brightness of moonlight.PASP 103, 1033.Lang, D., Hogg, D.W., Mierle, K., Blanton, M., Roweis, S., 2010. Astrome-try.net: Blind Astrometric Calibration of Arbitrary Astronomical Images. AJ139, 1782–1800. doi: , arXiv:0910.2233 .Lister, T.A., Greenstreet, S., Gomez, E., Christensen, E., Larson, S., 2016. TheLCOGT NEO Follow-up Network, in: Chesley, S.R., Morbidelli, A., Jedicke,R., Farnocchia, D. (Eds.), Asteroids: New Observations, New Models, pp.321–323. doi: .Mainzer, A., Bauer, J., Cutri, R.M., Grav, T., Masiero, J., Beck, R., Clarkson,P., Conrow, T., Dailey, J., Eisenhardt, P., Fabinsky, B., Fajardo-Acosta, S.,Fowler, J., Gelino, C., Grillmair, C., Heinrichsen, I., Kendall, M., Kirkpatrick,J.D., Liu, F., Masci, F., McCallon, H., Nugent, C.R., Papin, M., Rice, E.,Royer, D., Ryan, T., Sevilla, P., Sonnett, S., Stevenson, R., Thompson, D.B.,Wheelock, S., Wiemer, D., Wittman, M., Wright, E., Yan, L., 2014. InitialPerformance of the NEOWISE Reactivation Mission. Astrophys. J. 792, 30.doi: , arXiv:1406.6025 .Mainzer, A., Grav, T., Bauer, J., Masiero, J., McMillan, R.S., Cutri, R.M.,Walker, R., Wright, E., Eisenhardt, P., Tholen, D.J., Spahr, T., Jedicke, R.,Denneau, L., DeBaun, E., Elsbury, D., Gautier, T., Gomillion, S., Hand,E., Mo, W., Watkins, J., Wilkins, A., Bryngelson, G.L., Del Pino Molina,A., Desai, S., G´omez Camus, M., Hidalgo, S.L., Konstantopoulos, I., Larsen,J.A., Maleszewski, C., Malkan, M.A., Mauduit, J.C., Mullan, B.L., Olszewski,31.W., Pforr, J., Saro, A., Scotti, J.V., Wasserman, L.H., 2011. NEOWISEObservations of Near-Earth Objects: Preliminary Results. Astrophys. J. 743,156. doi: , arXiv:1109.6400 .McCully, C., Volgenau, N.H., Harbeck, D.R., Lister, T.A., Saunders, E.S.,Turner, M.L., Siverd, R.J., Bowman, M., 2018. Real-time Processing of theImaging Data from the Network of Las Cumbres Observatory Telescopes us-ing BANZAI, in: Observatory Operations: Strategies, Processes, and SystemsVII, SPIE. p. 107070K.Morris, D., Voutsinas, S., Hambly, N., Mann, R., 2017. Use of Docker fordeployment and testing of astronomy software. Astronomy and Computing20, 105 – 119. URL: , doi: https://doi.org/10.1016/j.ascom.2017.07.004 .Nagler, R., Bruhwiler, D., Moeller, P., Webb, S., 2015. Sustainability andreproducibility via containerized computing. Synchrotron Radiation 4769,145–146. ArXiv preprint arXiv:1509.08789.Nanthaamornphong, A., Carver, J.C., 2017. Test-Driven Development in scien-tific software: a survey. Software Quality Journal 25, 343–372.Narayan, G., Zaidi, T., Soraisam, M.D., Wang, Z., Lochner, M., Matheson,T., Saha, A., Yang, S., Zhao, Z., Kececioglu, J., Scheidegger, C., Snodgrass,R.T., Axelrod, T., Jenness, T., Maier, R.S., Ridgway, S.T., Seaman, R.L.,Evans, E.M., Singh, N., Taylor, C., Toeniskoetter, J., Welch, E., Zhu, S.,ANTARES Collaboration, 2018. Machine-learning-based Brokers for Real-time Classification of the LSST Alert Stream. Astrophys. J. Suppl. S 236, 9.doi: , arXiv:1801.07323 .Narita, N., Fukui, A., Kusakabe, N., Watanabe, N., Palle, E., Parviainen, H.,Monta˜n´es-Rodr´ıguez, P., Murgas, F., Monelli, M., Aguiar, M., Perez Prieto,J.A., Oscoz, ´A., de Leon, J., Mori, M., Tamura, M., Yamamuro, T., B´ejar,V.J.S., Crouzet, N., Hidalgo, D., Klagyivik, P., Luque, R., Nishiumi, T., 2019.MuSCAT2: four-color simultaneous camera for the 1.52-m Telescopio CarlosS´anchez. Journal of Astronomical Telescopes, Instruments, and Systems 5,015001. doi: , arXiv:1807.01908 .Narita, N., Fukui, A., Kusakabe, N., Watanabe, N., Palle, E., Parviainen, H.,Monta˜n´es-Rodr´ıguez, P., Murgas, F., Monelli, M., Aguiar, M., Perez Prieto,J.A., Oscoz, ´A., de Leon, J., Mori, M., Tamura, M., Yamamuro, T., B´ejar,V.J.S., Crouzet, N., Hidalgo, D., Klagyivik, P., Luque, R., Nishiumi, T.,2020. MuSCAT3: A 4-color simultaneous camera for the 2m Faulkes TelescopeNorth. In preparation.Noll, S., Kausch, W., Barden, M., Jones, A.M., Szyszka, C., Kimeswenger, S.,Vinther, J., 2012. An atmospheric radiation model for Cerro Paranal. I. The32ptical spectral range. A&A 543, A92. doi: , arXiv:1205.2003 .Ostro, S.J., Giorgini, J.D., Benner, L.A.M., 2007. Radar reconnaissance of near-Earth asteroids, in: Valsecchi, G.B., Vokrouhlick´y, D., Milani, A. (Eds.), NearEarth Objects, our Celestial Neighbors: Opportunity and Risk, pp. 143–150.doi: .Patterson, M.T., Bellm, E.C., Rusholme, B., Masci, F.J., Juric, M., Krughoff,K.S., Golkhou, V.Z., Graham, M.J., Kulkarni, S.R., Helou, G., ZwickyTransient Facility Collaboration, 2019. The Zwicky Transient Facility AlertDistribution System. Publ. Astron. Soc. Pac. 131, 018001. doi: , arXiv:1902.02227 .Pravec, P., Wolf, M., ˇSarounov´a, L., 1998. Lightcurves of 26 Near-Earth Aster-oids. Icarus 136, 124 – 153. doi: .Reddy, V., Kelley, M.S., Farnocchia, D., Ryan, W.H., Thomas, C.A., Ben-ner, L.A.M., Dotson, J., Micheli, M., Brucker, M.J., Bus, S.J., Brozovic, M.,Wheeler, L., Abbasi, V., Bauer, J.M., Bonsall, A., Brown, Z.L., Busch, M.W.,Chodas, P., Choi, Y.J., Erasmus, N., Fast, K.E., Faucher, J.P., Fernandes,R.B., Ghigo, F.D., Gilbank, D.G., Giorgini, J.D., Gustafsson, A., Hainaut, O.,Harris, W.M., Jao, J.S., Johnson, L.S., Kareta, T., Kim, M.J., Koschny, D.,Kramer, E.A., Landis, R.R., Laurin, D.G., Larsen, J.A., Lee, C.G., Lejoly, C.,Lister, T., McMillan, R., Masiero, J.R., Mathias, D., Mommert, M., Moon,H.K., Moskovitz, N.A., Naidu, S.P., Nallapu, R.T., Niazi, H.K., Noonan,J.W., Polishook, D., Ryan, E.V., Schatz, L., Scotti, J.V., Sharkey, B., Shus-tov, B.M., Sickafoose, A.A., Silva, M.A., Slade, M.A., Slick, L.R., Snedeker,L.G., Springmann, A., Tholen, D., Trilling, D.E., Vodniza, A.Q., Wainscoat,R., Weryk, R., Yoshikawa, M., 2019. Near-Earth asteroid 2012 TC4 observ-ing campaign: Results from a global planetary defense exercise. Icarus 326,133–150. doi: .Roeser, S., Demleitner, M., Schilbach, E., 2010. The PPMXL Catalog of Po-sitions and Proper Motions on the ICRS. Combining USNO-B1.0 and theTwo Micron All Sky Survey (2MASS). AJ 139, 2440–2447. doi: , arXiv:1003.5852 .Saha, A., Matheson, T., Snodgrass, R., Kececioglu, J., Narayan, G., Sea-man, R., Jenness, T., Axelrod, T., 2014. ANTARES: a prototype tran-sient broker system, in: Peck, A.B., Benn, C.R., Seaman, R.L. (Eds.),Observatory Operations: Strategies, Processes, and Systems V, p. 914908.doi: , arXiv:1409.0056 .Shelly, F., 2016. private communication.Siverd, R.J., Brown, T.M., Barnes, S., Bowman, M.K., De Vera, J., Foale, S.,Harbeck, D.R., Henderson, T., Hygelund, J., Kirby, A., McCully, C., Nation,33.S., Smith, C., Taylor, B., Tufts, J.R., 2018. NRES: the network of roboticechelle spectrographs, in: Proc. SPIE, p. 107026C. doi: .Skrutskie, M.F., Cutri, R.M., Stiening, R., Weinberg, M.D., Schneider, S., Car-penter, J.M., Beichman, C., Capps, R., Chester, T., Elias, J., Huchra, J.,Liebert, J., Lonsdale, C., Monet, D.G., Price, S., Seitzer, P., Jarrett, T.,Kirkpatrick, J.D., Gizis, J.E., Howard, E., Evans, T., Fowler, J., Fullmer, L.,Hurt, R., Light, R., Kopan, E.L., Marsh, K.A., McCallon, H.L., Tam, R.,Van Dyk, S., Wheelock, S., 2006. The Two Micron All Sky Survey (2MASS).AJ 131, 1163–1183. doi: .Spoto, F., Milani, A., Farnocchia, D., Chesley, S.R., Micheli, M., Valsecchi,G.B., Perna, D., Hainaut, O., 2014. Nongravitational perturbations and vir-tual impactors: the case of asteroid (410777) 2009 FD. A&A 572, A100.doi: .Spoto, F., Tanga, P., Bouquillon, S., Desmars, J., Hestroffer, D., Mignard, F.,Altmann, M., Herald, D., Marchant, J., Barache, C., Carlucci, T., Lister,T., Taris, F., 2017. Ground-based astrometry calibrated by Gaia DR1: newperspectives in asteroid orbit determination. A&A 607, A21. doi: .Stone, R.C., 1996. An Accurate Method for Computing Atmospheric Refraction.Publ. Astron. Soc. Pac. 108, 1051–1058. doi: .Street, R.A., Bowman, M., Saunders, E.S., Boroson, T., 2018. General-purposesoftware for managing astronomical observing programs in the LSST era, in:Software and Cyberinfrastructure for Astronomy V, SPIE. p. 1070711.Tapping, K.F., 2013. The 10.7 cm solar radio flux (F10.7). Space Weather 11,394–406. doi: .Thirouin, A., Moskovitz, N., Binzel, R.P., Christensen, E., DeMeo, F.E., Person,M.J., Polishook, D., Thomas, C.A., Trilling, D., Willman, M., Hinkle, M.,Burt, B., Avner, D., Aceituno, F.J., 2016. The Mission Accessible Near-Earth Objects Survey (MANOS): First Photometric Results. Astron. J. 152,163. doi: , arXiv:1607.03517 .Tonry, J.L., Denneau, L., Heinze, A.N., Stalder, B., Smith, K.W., Smartt, S.J.,Stubbs, C.W., Weiland, H.J., Rest, A., 2018. ATLAS: A High-cadence All-skySurvey System. Publ. Astron. Soc. Pac. 130, 064505.Veres, P., Jedicke, R., A., F., Denneau, L., Granvik, M., Bolin, B., Chastel,S., Wainscoat, R.J., Burgett, W.S., Chambers, K.C., Flewelling, H., Kaiser,N., Magnier, E.A., Morgan, J.S., Price, P.A., Tonry, J.L., Waters, C., 2015.Absolute magnitudes and slope parameters for 250,000 asteroids observed byPan-STARRS PS1 - Preliminary results. Icarus 261, 34–47.34okrouhlick´y, D., Farnocchia, D., ˘Capek, D., Chesley, S.R., Pravec, P.,Scheirich, P., M¨uller, T.G., 2015. The Yarkovsky effect for 99942 Apophis.Icarus 252, 277–283. doi:10.1016/j.icarus.2015.01.011