NELIOTA: The wide-field, high-cadence lunar monitoring system at the prime focus of the Kryoneri telescope
E.M. Xilouris, A.Z. Bonanos, I. Bellas-Velidis, P. Boumis, A. Dapergolas, A. Maroussis, A. Liakos, I. Alikakos, V. Charmandaris, G. Dimou, A. Fytsilis, M. Kelley, D. Koschny, V. Navarro, K. Tsiganis, K. Tsinganos
AAstronomy & Astrophysics manuscript no. paperv10 c (cid:13)
ESO 2018September 5, 2018
NELIOTA: The wide-field, high-cadence lunar monitoring systemat the prime focus of the Kryoneri telescope
E.M. Xilouris , A.Z. Bonanos , I. Bellas-Velidis , P. Boumis , A. Dapergolas , A. Maroussis , A. Liakos ,I. Alikakos , V. Charmandaris , , G. Dimou , A. Fytsilis , M. Kelley , D. Koschny , , V. Navarro ,K. Tsiganis , and K. Tsinganos Institute for Astronomy, Astrophysics, Space Applications & Remote Sensing, National Observatory of Athens, P.Penteli, GR-15236 Athens, Greece Department of Physics, University of Crete, GR-71003 Heraklion, Greece DFM Engineering, Inc., 1035 Delaware Avenue, Longmont, CO 80501, USA Scientific Support Office, Directorate of Science, European Space Research and Technology Centre (ESA/ESTEC),2201 AZ Noordwijk, The Netherlands Chair of Astronautics, Technical University of Munich, 85748 Garching, Germany European Space Astronomy Centre (ESA/ESAC), Camino bajo del Castillo, s/n, Urbanizacion Villafranca delCastillo, Villanueva de la Ca˜nada, E-28692 Madrid, Spain Department of Physics, Aristotle University of Thessaloniki, GR-54124 Thessaloniki, Greece Section of Astrophysics, Astronomy and Mechanics, Department of Physics, University of Athens, Zografos, GR-15783Athens, GreeceReceived / Accepted
ABSTRACT
We present the technical specifications and first results of the ESA-funded, lunar monitoring project “NELIOTA”(NEO Lunar Impacts and Optical TrAnsients) at the National Observatory of Athens, which aims to determine thesize-frequency distribution of small Near-Earth Objects (NEOs) via detection of impact flashes on the surface ofthe Moon. For the purposes of this project a twin camera instrument was specially designed and installed at the1.2 m Kryoneri telescope utilizing the fast-frame capabilities of scientific Complementary Metal-Oxide Semiconductordetectors (sCMOS). The system provides a wide field-of-view (17.0 (cid:48) × (cid:48) ) and simultaneous observations in twophotometric bands (R and I), reaching limiting magnitudes of 18.7 mag in 10 sec in both bands at a 2.5 signal-to-noiselevel. This makes it a unique instrument that can be used for the detection of NEO impacts on the Moon, as wellas for any astronomy projects that demand high-cadence multicolor observations. The wide field-of-view ensures thata large portion of the Moon is observed, while the simultaneous, high-cadence, monitoring in two photometric bandsmakes possible, for the first time, the determination of the temperatures of the impacts on the Moon’s surface andthe validation of the impact flashes from a single site. Considering the varying background level on the Moon’s surfacewe demonstrate that the NELIOTA system can detect NEO impact flashes at a 2.5 signal-to-noise level of ∼ . ∼ . × − events km − h − . Key words.
Instrumentation: detectors - Techniques: miscellaneous - Telescopes - Moon - Surveys
1. Introduction
Lunar monitoring provides a promising method for deter-mining the size-frequency distribution (SFD; Ivanov et al.2002; Werner et al. 2002; Harris & D’Abramo 2015) of smallnear-earth objects (NEOs) via the detection of NEO im-pact flashes on the Moon (see Bouley et al. 2012; Suggset al. 2014, and references therein). Determining the SFDfor objects in the decimeter to meter range is important forevaluating the danger of small NEOs impacting Earth, inlight of the recent Chelyabinsk event (Brown et al. 2013), aswell as the risk to artificial satellites and future man-madestations on the Moon. The Lunar Reconnaissance Orbiter
Send offprint requests to : E. M. Xilouris, e-mail: [email protected] (LRO) provides an indirect measurement of the SFD, bydetermining the cratering rate on the Moon via temporalimaging (Speyerer et al. 2016). So far, one new crater hasbeen associated with an observed impact flash (Suggs et al.2014; Robinson et al. 2015), demonstrating the potential ofthe synergy between lunar monitoring from the ground andspace.The estimation of the NEO size from lunar monitoringobservations depends on the assumed value of the luminousefficiency η λ , which is defined as the ratio of the measuredluminous energy emitted at a particular wavelength λ tothe total kinetic energy of the meteoroid. It therefore im-plies the knowledge of the impactor velocity, which cannotbe accurately known for sporadic meteoroids. The lumi-nous efficiency is best constrained for meteoroids originat- a r X i v : . [ a s t r o - ph . I M ] S e p ilouris et al.: The NELIOTA lunar monitoring system Fig. 1.
Section view of the Kryoneri Prime Focus Instrument (KPFI) optical layout. The position of the field correctorlenses η ∼ × − from lunar Leonids, while Moser et al. (2011)report values between 1 . × − and 1 . × − from lunarimpacts involving Geminid, Lyrid and Taurid meteoroids.Results of numerical simulations (Nemtchinov et al. 1998)predict η to range from a lower limit of 10 − − − to anupper limit of 10 − , whereas impact simulations modelingthe high velocity of the Leonids yield values of η on theorder of 10 − to 2 × − (Artem’eva et al. 2001). In anycase, it is evident that lunar impact observations duringmeteor showers are valuable for constraining the value of η .Suggs et al. (2014) report estimated sizes of their observedimpactors - originating from non-sporadic meteoroids - tobe on the order of centimeters.Ground-based lunar monitoring typically employsmodest-sized aperture ( ∼
30 cm) telescopes, which surveyon the order of ∼ km of the lunar surface at a time. Theadvantage this method offers, as compared to monitoringmeteors in the Earth’s atmosphere using all-sky cameras,is that the monitored surface area is larger by two orders of magnitude. Several groups have therefore undertaken lu-nar monitoring surveys for NEO flashes (e.g. Madiedo et al.2014; Ortiz et al. 2015; Rembold & Ryan 2015; Ait MoulayLarbi et al. 2015). Ground-based surveys routinely use twoor more telescopes often located at different sites in order todistinguish noise, seeing variations, cosmic rays and satel-lite glints from real impact events. A lunar impact eventis confirmed when a simultaneous detection has been madeindependently by two telescopes.The “NEO Lunar Impacts and Optical TrAnsients”(NELIOTA) Project is an activity launched by ESA at theNational Observatory of Athens (NOA) in 2015. It aimsto contribute to the SFD of small NEOs, by developinga system that will perform a 22-month lunar monitoringcampaign. NELIOTA brings three innovative aspects to thefield of lunar monitoring: (1) a single telescope with a twincamera system, which detects and confirms events, (2) thetwin cameras simultaneously monitor the Moon in two pho-tometric bands, allowing for a temperature determinationof each flash and the evolution of the flash, in case of flashes https://neliota.astro.noa.gr2ilouris et al.: The NELIOTA lunar monitoring system Fig. 2.
KPFI filters (red solid line for the R-band and greensolid line for the I-band) and dichroic beam splitter trans-mission curves (dashed and dashed-dotted lines) combinedwith the QE response of the Lunar imaging detectors (bluesolid line). Filter curves reflect the Johnson R- and I-band.The dichroic is centered at 730 nm with a >
90% through-put.detected on multiple-frames (Bonanos et al. 2018), and (3)a large aperture telescope, which, in principle, can detectfainter impact flashes than the modest-sized aperture tele-scopes typically used. The limitation to the faintness of ob-served lunar impacts arises from the variable brightness ofthe lunar background (i.e. earthshine; Goode et al. 2001),in combination with the wavelength observed and the framerate of the observations.The selection of telescope and cameras were made so asto satisfy the main project requirements: (1) to monitor theMoon at ≥
20 frames per second, (2) detect lunar impactsdown to 12 th magnitude, (3) have the ability to distinguishfalse positives (e.g. cosmic rays, foreground glints) from realflashes, and (4) to guarantee availability of the telescope forcontinuous usage that is better than 95%.To address the specific needs of the NELIOTA projectthe 1.2 m Kryoneri telescope of NOA was upgradedin 2016, commissioning a prime focus, high-speed, twin-camera Lunar imager. The project has deployed a hard-ware system for recording and processing images. We havealso developed a software system, which controls both thetelescope and the cameras, processes the images and auto-matically detects candidate NEO lunar impact flashes. Theimpact events are verified, characterized and made availableto the scientific community and the general public via theNELIOTA website within 24 hours of discovery. NELIOTAcompleted its commissioning phase in early 2017 and begana 22 month observing campaign in February 2017 in searchof NEO impact flashes on the Moon. The 1.2 m Kryoneritelescope is capable of detecting flashes much fainter thanall current, smaller-aperture, lunar monitoring telescopes.NELIOTA is therefore expected to characterize the size-frequency distribution of NEOs weighing as little as a fewgrams.Bonanos et al. (2018) present the first scientific resultsof the project, in particular, the first temperature mea- http://kryoneri.astro.noa.gr https://neliota.astro.noa.gr/DataAccess Fig. 3.
I-band image of the Moon observed with theNELIOTA system illustrating the available FOV, which is17.0 (cid:48) × (cid:48) .surements of impact flashes. This paper presents a detaileddescription of the NELIOTA system and its components.Sect. 2 presents the upgraded 1.2 m Kryoneri telescope andits new optical design, while the Lunar imager camera sys-tem is detailed in Sect. 3. The NELIOTA system is pre-sented in Sect. 4, while a brief description of the system’ssoftware is given in Sect. 5. The instrument’s performanceis evaluated in Sect. 6 with the results of the first year ofthe NELIOTA campaign presented in Sect. 7. Finally, asummary of the NELIOTA system and its first results aregiven in Sect. 8.
2. The 1.2 m Kryoneri telescope
The 1.2 m Kryoneri telescope was selected (after a trade-off analysis) as the optimal facility of NOA for theNELIOTA project. It was built by Grubb-Parsons and com-missioned in 1975. The telescope is situated at KryoneriObservatory (37 ◦ (cid:48) (cid:48)(cid:48) North, 22 ◦ (cid:48) (cid:48)(cid:48) East), in the dis-trict of Corinth in the northern Peloponnese, Greece, atthe top of mount Kyllini, at an altitude of 930 m, close toKryoneri village.During its first ∼
40 years of operation, the telescope’soptical system consisted of a paraboloidal primary mirror of1.2 m diameter and f/3 focal ratio and a hyperboloidal sec-ondary mirror (31 cm). Both mirrors are made of Zerodur.This configuration produced a final focal ratio of f/13 atCassegrain focus. The main scientific instrument in the lastyears has been a 2 . (cid:48) × . (cid:48) CCD camera Apogee Ap47p witha set of UBVRI filters.
The NELIOTA science objectives imposed strict require-ments on the optical design in order to provide: (i) thecapacity to image simultaneously in two bands, (ii) a largeFOV ( ∼ (cid:48) ) and (iii) seeing-limited image quality (with ∼ . (cid:48)(cid:48) being a typical site seeing at Kryoneri Observatory Table 1.
Basic characteristics of the optical elements.
Element Material Diameter Thickness(cm) (cm)Primary mirror Zerodur 120.0 -Lens (typical dome seeing ∼ . (cid:48)(cid:48) ), yielding a scale of ∼ . (cid:48)(cid:48) pixel − ).In 2016 the telescope underwent an extensive upgradeby DFM Engineering Inc. , within the NELIOTA project.The electro-mechanical upgrade included replacement ofthe telescope servo-motors and associated hardware (in-cluding new encoder systems), and deployment of a newsystem for dome opening and rotation as well as motor-ized primary mirror doors. A computerized telescope con-trol system (TCS), developed by DFM, replaced the legacyconsole and allowed moving the observation control out ofthe telescope housing area. A new, GPS-based time server(Meinberg, LanTime M200/GPS) has been installed and isused by the TCS and other systems in the Observatory.Furthermore, and according to the requirement for lu-nar observations, the optics of the telescope were modi-fied to operate with instruments at the prime focus, bring-ing the telescope back to its primary mirror f/3 focal ra-tio and providing an unvignetted field-of-view (FOV) of ∼ . (cid:48) × (cid:48) ofthe total corrected FOV at the prime focus of the tele-scope, providing simultaneous high-cadence observations intwo bands. With such a FOV, a significant fraction of thenon-sunlit part of the Moon, the only ‘usuable’, for our pur-poses, can be monitored to detect faint flashes coming fromNEO impacts. A direct imaging optical configuration, usinga separate CCD detector, was also added to the design toallow use of the full FOV. This was made possible through acomputer controlled camera slider plate mechanism allow-ing for two operating modes, either using the twin imagingsystem (the “Lunar imager” hereafter) or the direct imag-ing configuration (the “Direct imager” hereafter). Several optical elements were needed to overcome the chal-lenge of placing an instrument at the prime focus of the1.2 m Kryoneri telescope. The existing parabolic primarymirror (f/3) of the telescope was used along with a set offield corrector lenses in order to focus the beam in the primefocus. In total, three corrector lenses were used (see lay-out in Fig. 1 and characteristics in Table 1). The first andsecond field corrector elements (labeled as field correctorlens ∼ . Table 2.
Specifications of the Andor Zyla 5.5 sCMOS cam-era for the NELIOTA project with a 2 × Parameter SpecificationSensor type Front illuminated Scientific CMOSPixel size 6.48 µ mActive pixels 1280 × × × Shutter GlobalGain settings Low & High well capacity (16-bit)Gain 0.4 e − per A/D countRead noise 5.1 e − rmsRead-out rate 560 MHz (280 MHz × − (or ∼ − on the lunar surface)Frame rate 30 fpsField of view 17.0 × Exposure time 23 msecLinearity < ◦ C)Connection USB 3.0 ager producing a focal ratio of f/2.8 in both channels. Infront of each detector a lens pair is placed in order to pro-vide the necessary focal reduction and the same effectivefocal length in each channel. A set of Johnson-Cousins R-and I-band filters is used so that calibrated photometricmeasurements can be made. The choise of this set of filtersalso assures direct comparison with other NEO studies al-ready conducted (e.g. Suggs et al. (2014)). Fig. 2 presentsthe filter response functions and the dichroic beam splittertransmission curves.Field corrector lenses (cid:48) . Appropriate anti-reflective coat-ings were applied to the faces of the prisms. In order toavoid vignetting, we specified a minimum clear aperture of100 . × . ± . −
730 nm with a throughput of > −
950 nm, with an identicalthroughput to its counterpart (see Fig. 2).
3. The Lunar Imager camera system
Systems designed, so far, for monitoring lunar impactevents mostly use black-and-white CCD video cameras.These are rather small size detectors (about 700 ×
500 pix-els) operating at 25 to 30 frames per second rate producinginterlaced 8-bit frames. In this way, the effective rate is dou-bled, but a half frame (including either the odd or the even rows) is recorded after each exposure. For the NELIOTAproject large size sCMOS (scientific Complementary Metal-Oxide Semiconductors) detectors are utilized.The twin imaging system for NELIOTA includes a pairof identical detectors (Zyla 5.5 sCMOS) providing simul-taneous observations in two photometric bands. These de-tectors were selected since they satisfy the project require-ments for a fast-frame rate, high sensitivity and resolution,and a light and compact design (being in the prime focusthey had to be small enough in size in order not to causevignetting of light from the source) making NELIOTA thefirst, to our knowledge, astronomical system to use sCMOSdetectors at a fast-frame rate. The active chip is 22 mm indiagonal with 2560 × µ m in size. With the NELIOTA setup (see Table2) these cameras provide images at a rate of 30 frames persecond, offering as small as a sub-microsecond interframegap. Their quantum efficiency is 60 −
50% in the R-band and30 −
20% in the I-band). Andor’s Zyla 5.5 sCMOS offersa high-accuracy hardware-generated time-stamp on eachframe, essential for providing high precision time measure-ments of the detected lunar impact events. For NELIOTAobservations a 2 × − , well under the typical see-ing measured on site, a FOV of 17.0 (cid:48) × (cid:48) (see Fig. 3),and also well sampled frames in the time domain.The sCMOS sensor has a highly parallel read-out ar-chitecture. Each of the 2560 columns possess a separateamplifier and analogue to digital converter, at both the topand bottom of the column. While all columns are read outin parallel, the read-out direction of each column is split inthe center into the signal from the top and bottom halves.There are two different pixel read-out rates, slow read at200 MHz (100 MHz × × ×
280 MHz, using the Gain settings of Low noise& High well capacity (16-bit) and the Global shutter. Thelatter offers the capability of all pixels being exposed simul-taneously and allows for easier synchronization of the twocameras. The specifications of the cameras are summarizedin Table 2. An extensive evaluation of sCMOS cameras forastronomical purposes can be found in Qiu et al. (2013).
4. The NELIOTA system
The NELIOTA system has the ability to plan and ensureacquisition of the necessary data (Moon observations, cal-ibration stars, flat-field and dark frames), to manage thedata recording to high capacity storage systems, to processthe stored datasets and, finally, to detect possible impactevents. Furthermore, the system organizes and stores thedetected events in the archive and, finally, provides usefulinformation about the events on the dedicated NELIOTAweb-based interface .The NELIOTA system is organized as a sequence offour functional domains, namely, “observation”, “detec-tion”, “archiving”, “information”, each of them controllingparticular subsystems and assuring that the necessary in-formation is shared among them (see Fig. 4). The “obser- vation” domain controls the external hardware (telescope,cameras, GPS and meteorological station) and monitorstheir status, ensuring that the desired observing plan is exe-cuted and that the raw (uncalibrated) frames and metadataare stored in real time. The “detection” domain providespost-observation data processing of the raw frames and ap-plications of the impact detection algorithm (see Sect. 5.2).As a result of this process, candidate impact flashes areautomatically detected and calibrated. They are then au-tomatically forwarded to and treated within the “archiv-ing” domain, where they are stored in a database-drivenarchive, and further evaluation and validation by expertscientists is performed. Finally, the “information” domainensures web-based publishing of all the relevant informa-tion about the validated impact events and provides thegeneral public with access to the NELIOTA results. In ad-dition to this, registered users have indirect, read-only, ac-cess (downloads) to the detected event files through thearchiving system service, allowing for independent analysisof the datasets. In particular, the registered user can down-load a data cube (in FITS format) of the detected eventaccompanied by the seven frames of the observation beforethe beginning of the event and after the end of the eventso that a direct comparison of the local background can bemade. For each observing run the detected events (if any)are published in the NELIOTA database within 24 hoursfrom the beginning of the observation. Each domain includes a top-level software component whichcontrols particular hardware subsystems through a dedi-cated machine-to-machine interface and/or processes thedata flow as necessary. These operations are controlledand monitored through a graphical user interface (GUI).Particular technologies have been used for the softwaredevelopment and implementation of the processing algo-rithms. With the exception of the “information” module,all others are developed in Java 8 SDK platform with itswidget toolkit “JFC/Swing” applied for creating the GUIs.The Eclipse Integrated Development Environment (IDE)has been used to write, to test and to package the threeJava-based modules as JAR-executables.The “observation” component is developed almost ex-clusively in Java, with a small interface in C++ to control,through a TCP socket, the two sCMOS cameras utilizingthe supplier’s library (Andor-SDK3). The telescope is con-trolled through a TCP/IP link to the Telescope ControlSystem (DFM’s WinTCS), exploiting the system’s remotecontrol commands. The GPS (ntp-service) and the weatherstation services are passively accessed through correspond-ing NELIOTA system interfaces. The GUI utilizes two 4Kultra-HD monitors, one used for the operations control andfor the monitoring of the status of the NELIOTA sys-tem, and the other for real-time display of the acquiredframes. The observation system software is packaged in aexecutable JAR file.The “detection” system is fully developed in Java andimplements the algorithm for automated detection of thecandidate impact events (see Sect 5.2), their characteriza-tion and location on the Moon as well as the re-formatting MET GPSCAMTEL ARCSTO
MetIF GpsIFTelIF CamIFC++ &Cam DLLsTCP/IP
ODSTO
Java IO ArcIFftp ArcPubIFWin.Serv . ARCSTO
ObStoIFJava IO ArcStoIFMySQL ServerEvents RecordsObStoIF esa.neliota.observation esa.neliota.detection esa.neliota.archiving esa.neliota.information
ObsOperHMI Observation Chunks & logs DetOperHMI Events ReportsArcOperHMI ExpOperHMI PublOperHMI WebIFHTTP+CMSEvents Data
ObsOper DetOper ArchOper ExpOper PublOper
Public Users
Fig. 4.
The NELIOTA system architecture and communication block diagram. The four software modules (observation,detection, archiving and information), the basic hardware telescope (TEL), camera system (CAM), time server (GPS),meteorological station (MET), high-performance storage (ODSTO), archiving storage (ARCSTO) and their interfaces(IF), as well as the human-machine-interfaces (HMI) are shown.
Camera RED Camera NIRObservation Server
Fiber OpticConnection Fiber OpticConnection USB3.0-to-FO USB3.0-to-FOController A Controller B
Storage Array
Storage Pool A (RealTime-RED) Storage Pool B (RealTime-NIR) Storage Pool C (Events)
Detection Server
Camera ControllerCamera Controller
Telescope Control System – TCS
Telescope PrivateNetwork (TCP/IP) Management Private Network(TCP/IP)iSCSI Network
Fig. 5.
The configuration of the NELIOTA physical datasystem at the Kryoneri site.of the raw frames into FITS format. An interface based ona FTP-client then transfers the detected event files to theremote archiving system.The “archiving” system (also written in Java) includesan interface to the FTP-server that allows the data of thecandidate impact events to be transfered to the data-basedriven archive (Microsoft SQL Server). Its GUI providesthe appropriate tools so that the expert scientist can vi-sually inspect the data and validate the detected events.Furthermore, the “archiving” system utilizes a WindowsService and a Web Service for interfacing with the “infor-mation” domain. Through this interface the bulk of infor-mation of the new events are being transfered in a properformat, ready to be published on the web. The archive also,indirectly, services users requests to the “information” sys-tem for events, supplying the latter with properly packagedevent records for further downloading. The executable JARis activated as a non-stop service.The “information” system’s top-level component is awebsite which implements its functionality as an ASP.net web application based on the Model-View-Controller(MVC) architectural pattern. Most components are writ-ten in C pages and to service downloads. It is the only domain towhich public users have limited access. The system imple-ments a Windows Service to manage downloading requestsand a Web Service to get the necessary content from thearchiving system. The functional requirements set above for the NELIOTAsystem (Section 4) impose special characteristics to thehardware in order (1) to assure synchronized operation ofthe telescope and the two cameras, (2) to achieve high per-formance storage capabilities in terms of speed, capacity,security and hardware-failure safe operations, (3) for stor-age virtualization, and (4) for high-performance data pro-cessing resources. For practical reasons that have to do withthe location of Kryoneri Observatory, the NELIOTA sys-tem had to be split into two subsystems in order to bemost efficient. One of the subsystems is physically presentat Kryoneri Observatory and the other is hosted at thecomputer center of IAASARS/NOA located in the north-ern suburbs of Athens, with the two of them connectedthrough a private 11 Gbps RF-link. The “observation” andthe “detection” systems are deployed at the Kryoneri siteon a cluster of two nodes, both using a common exter-nal high-performance storage array of a total of 38.4 TBthrough a dedicated switch (Table 3). The storage “subsys-tem” is linked through 8 × Table 3.
Specifications of the NELIOTA hardware systemsat the Kryoneri site.
Specifications of the two servers (per server)Model HP ProLiant DL380 Generation9Processor Intel Xeon E5-2660v3 CPU(2.66GHz, 10 cores, 20 threads, 25MB Cache)Memory 64GB RAM (8 × × × × Fig. 6.
Instrument performance for the two channels (I-and R-band; left and right panels respectively). In eachpanel the measured magnitudes of stars in the open clusterNGC 1960 are plotted against their measured uncertainty( σ mag ) for three different exposure settings [10 sec (circles),1 sec (asteriscs) and 0.023 sec (crosses)]. The curves passingthrough the photometric measurements are the model pre-dictions of Eq. 1 providing information on the magnitudelevel that can be reached at a certain σ mag level (see textfor more details). The horizontal line in each panel indicatethe 0.1 mag noise level. Table 4.
Specifications of the NELIOTA virtual serverhardware systems.
Virtual server specificationsProcessor QEMU V-CPU v. 1.1.2 2.27GHzMemory 5GB (Archiving server), 8GB (Information server)Storage 100GB (Archiving server), 50GB (Information server)System Windows Server 2012 R2, x64 servations” and “detection” systems, are mirrored on thetwo nodes, so in the case of a failure, only the reconnec-tion of the camera USB links from one node to the other isneeded for resuming the observation. The NELIOTA phys-ical data system at Kryoneri is presented in Fig. 5.The NELIOTA “archiving” and “information” systemsare deployed in the IAASARS/NOA node on two virtual
Fig. 7.
Scintillation noise ( σ scint ) as a function of airmass.The top panel shows the scatter of the instrumental R- andI-band brightness of a photometric starndard star (SA 113-475) observed with an exposure time of 23 msec at fourdifferent airmasses ( X =1.26, 1.7, 2.2, 2.75; small circles)along with the mean values at each airmass (large circles).In the bottom panel the standard deviation of the bright-ness of the source, in each airmass, is plotted against theairmass with the linear fits to the data provided (see thetext for more details).servers. The two servers are almost identical apart from thesize of the dedicated memory and storage (Table 4). Thehardware of the server system has been implemented onthe Virtual Machines Cluster of the Network InformationCenter (NIC) of the National Observatory of Athens. The built system configuration achieves a throughput of5 Gbps/camera (camera to storage array). The two con-trollers of the storage array have been programmed to op-erate in parallel and independently manipulate the twostreams/cameras, routing the related data streams to thestorage array (red and green solid lines in Fig. 5). Followingthis concept the storage array is divided into two parts(storage pools “A” and “B” organized in RAID-5 parity)and each controller directs the related stream to a separateset of physical hard disks. In addition, each controller oper-ates as a backup of the other one ensuring the maintenanceof operation in cases of hardware failure (dashed lines inFig. 5). On the other hand, a storage pool “C” has beencreated with RAID-1 parity in order to secure the data withthe candidate events, produced by the detection system.Based on the observation setup with the two cameraseach acquiring 30 frames per second with a 2 × Fig. 8.
Field distortion map of the I-band (left panel) and the R-band (right panel) frames. Each arrow shows theexpected position on the sky (beginning of arrow) and the measured position (end of arrow). For displaying purposesthe arrows are artificially enlarged by a factor of 50 with a scale indicator (4 ‘enlarged’ pixels) shown in the upper leftpart of each plot.keep observed raw frames for at least two months. On theother hand, the storage pool “C” capacity provides enoughstorage to hold detected event records for a few years ofNELIOTA operation.
5. The NELIOTA software
A dedicated software system has been developed in orderto control the telescope, schedule observations, reduce in-dividual frames, and perform automatic detection of theimpact events. In addition, an entry in our web-based,user-friendly, database is automatically created when a newevent is detected (within 24 hours of the event observation).The software (presented in detail elsewhere; Fytsilis et al.in prep) has been developed in JAVA using ECLIPSE IDEand JAVA RE8 and is running on a Windows Server. Itprovides a computer interface between all individual com-ponents of the NELIOTA system.
The NELIOTA observations are scheduled and performedusing the observation planner and acquisition functions ofthe software. The observation planner produces an observ-ing plan for each night according to the visibility of theMoon and the observability of the photometric calibrators.The observations are performed in the non-sunlit part of theMoon when the illuminated fraction is between ∼ − ◦ above horizon is set due to the dome clear-ance. The Moon is observed at a frame rate of 30 frames per second for a 15 minute interval or “chunk” before thisdataset is stored on the server. A selection of photometriccalibrator stars from Landolt (1992) is made by the plan-ner according to their proximity to the Moon and observedin between lunar observing “chunks”. The exposure time ofthe calibrators ranges from 0.05 to 4 sec depending on theirapparent magnitude. In the case of an impact event the cal-ibrator observed closest in time to the event is selected toconduct the photometric analysis. Ancillary observations(flat-field and dark frames) are obtained on each night. Skyflat-field frames are observed during twilight with exposuretimes set automatically by the software (typically rangingbetween 0.04 to 2 sec) in order to make the best use ofthe dynamical range of the detectors. Dark frames are ob-tained before and after the end of lunar observations. TheNELIOTA software commands the telescope through theTCS interface, while the cameras are controlled by usingthe software development kit (SDK) version 3.11 providedby Andor. Due to the large amount of data collected, a pipelinehas been developed in order to detect candidate impactevents, which are later evaluated by an expert scientist.The pipeline is based on two successive steps: (i) calibra-tion of the raw data, and (ii) application of the detectionalgorithm.At the observed rate of 30 frames per second the actualon-source integration time is 0.023 sec followed by a read-out time of 0.010 sec. Events that are detected in a singleframe are considered to have a duration of 0.033 sec (upper t=−33 ms t=132 mst=0 ms t=33 ms t=66 ms t=99 ms
Fig. 9.
Time sequence of the I-band brightness of the impact detected by the NELIOTA system at 04:35:09.967 UT onDecember 14 th , 2017 (event ID 30). The magenta contours show the 500 ADUs level, corresponding to ∼ ×
85 arcseconds.limit), while for the events that are detected in more thanone frames, the total duration is calculated by summingthe integration and read-out times of the successive framesuntil the end of the event.Calibration of the observed frames is done in a stan-dard way by performing a dark current subtraction us-ing the median-combined dark frames and also flat-fieldcorrections using the median-combined sky flat-frames ob-served. The calibrated lunar frames are then backgroundsubtracted to remove the inhomogeneous surface of theMoon and earthshine. To calculate the background B t foreach individual frame I t observed at time t , an adaptivebackground is created by combining all individual framesof the Moon observed until t − t ). These frames are weighted in timeas B t = α × I t − + (1 − α ) × B t − with α having valuesin the range (0, 1) depending on the degree of importanceof the most recent frames. The value that we use in theNELIOTA setup is α = 0 .
35. This ensures that at the ob-served rate of 30 frames-per-second, the main contributionin the background comes from the most recent observations(closer to time t ), which reduces the effects of seeing as wellas possible telescope tracking deviations, but on the otherhand produces a robust estimate of the global background.The background subtracted image of interest D t = I t − B t is then subjected to a threshold detection processusing a high and a low threshold value. The high thresholdvalue is primarily used to define pixels with large devia-tions from the local background, while a low value is sub-sequently used to better define these deviations by lookingat the fainter levels. When a candidate event is flagged, theconnected components are extracted so that the event canbe treated as a single object and its extent can be defined.With this criterion, objects that are smaller than a mini-mum size (10 pixels) are considered to be noise and are thusrejected. Objects that are large enough are then consideredas candidate events and are recorded. This same procedureis then repeated for the subsequent frame ( t + 1) with theexception that in the calculation of the new background,the area extracted by the connected component analysis isnot considered. This secures detection of the same event insubsequent frames as the event fades.For each candidate event that is detected by the soft-ware, visual inspection is performed in order to validate it.The twin camera system can easily rule out cosmic rayssince they only appear in one of the two frames. Otherartifacts, such as satellites or airplanes crossing the field-of-view can also be easily ruled out due to their elongatedtrajectory. Once an event is validated, photometric calibra-tion is performed by an expert scientist. The photometric calibrator closest (in time and angular distance) to the ob-servation of the event is then used to derive the photometriczero-point and thus calculate the apparent magnitude of theevent. Events that are only detected in the I-band resem-bling flashes are also stored and flagged as “suspected”.The location of the flashes on the Moon is derived in asemi-automatic way using a dedicated tool of the NELIOTAdetection software and a high detailed image of the Moonfrom the Virtual Moon Atlas 6.0 (VMA6) software. Usingthe options available for the latter (i.e. date, time observ-ing site, libration) it is feasible to export a highly detailedimage of the Moon as seen from the observing site at theexact time when the flash is detected. The localization toolstacks the images contained in the data cube of each event(see Section 4) in order to produce a higher contrast im-age of the Moon in which the lunar features can easily bespotted. Subsequently, the user imports the detailed imageproduced by the VMA6 and identifies at least three lunarfeatures (e.g. small craters) by selecting them in both theVMA6 image and the stacked image. After the successfulcross match, the software exports the initial VMA6 imagein which the location of the flash is marked. The exact se-lenographic coordinates can then be extracted. The error inthe location determination, taking into account the typicalseeing values for the site and also the error in corss match-ing features on the Moon between the observations and theVMA6 map, is roughly estimated to be 6 . (cid:48)(cid:48) ( ∼ . within 24 hours after the obser-vation.
6. Instrument performance
We have performed observations of the open clusterNGC 1960 (RA : 05 h m s , DEC : +34 ◦ (cid:48) (cid:48)(cid:48) ) inorder to assess the instrument performance in terms ofdepth and instrument/photonic noise. The open cluster wasobserved simultaneously in the two bands on the nightof March 11 th , 2018 with exposure times of 10, 1, and0.023 sec (the last being the fixed exposure time setting forthe NELIOTA observations) and four different airmasses( X = 1 . , . , . , . phot task in IRAF . The photometry was per-formed within an aperture of 3 pixels radius, corresponding http://ap-i.net/avl/en/start IRAF is distributed by the National Optical AstronomyObservatories, which are operated by the Association of9ilouris et al.: The NELIOTA lunar monitoring system to 2 . (cid:48)(cid:48) , while the sky background was measured in an an-nulus centered at each star with an inner radius of 10 pixelsand an outer radius of 15 pixels. In each 10 sec exposure weused 20 relatively bright and isolated stars in the field tocalculate the aperture corrections needed for an apertureof 10 pixels in order to account for the stellar light missingin the smaller, 3 pixel, aperture used in the photometry ofthe cluster. The mean sky background standard deviation( ST D bkg ) was measured to be 12, 14, 38 counts in 0.023,1, 10 sec, respectively, in the I-band and 12, 13, 30 countsin 0.023, 1, 10 sec, respectively, in the R-band. The un-certainty on the brightness of the source ( σ mag ) was thencalculated as σ mag = 1 . /SN R (1)with the signal-to-noise ratio ( SN R ) defined as:
SN R = N (cid:113) σ source + σ bkg + σ RN (2)(see, e.g., Everett & Howell (2001)). Here, N is the sourceflux in electrons, σ source the source shot noise, σ bkg thebackground noise and σ RN the read-out noise. In particular, N = t × G × − . mag − zpt ) (3)with t being the exposure time (in sec), mag , the mag-nitude of the source and zpt the zero point in the mag-nitude scale (see below for the computed values in theR- and I-bands ( r and i respectively)). The source shotnoise is defined as σ source = √ N , the background noise as σ bkg = ST D bkg × G × (cid:112) n source (1 + n source /n bkg ) and theread-out noise as σ RN = RN × (cid:112) n source (1 + n source /n bkg ).In this case G = 0 . − per A/D count, and RN = 5 . − (see Table 2 for the specifications of the detectors), while n source is the number of pixels inside the photometry aper-ture (radius of 3 pixels) and n bkg the number of pixelsinside the background annulus (inner radius of 10 pixelsand outter radius of 15 pixels). Given the values above, n source /n bkg = 0 . r , i ), of atmospheric extinction coefficients ( r , i )and of the color terms ( r , i ) in the following photometricequations: r = R + r + r X + r ( R − I ) i = I + i + i X + i ( R − I ) , where r and i are the instrumental magnitudes of the stars, R and I the reference magnitudes, and X the airmasses(common for both bands) at the time of the observationof the cluster. In our photometric calibration run we found r = 23 .
01 mag, i = 23 .
07 mag, r = 0 .
11 mag/airmass, i = 0 .
06 mag/airmass, r = 0 .
30 mag and i = 0 .
06 mag.The inverted transformation equations were then used to
Universities for Research in Astronomy, Inc., under cooperativeagreement with the National Science Foundation.
Fig. 10.
Positions of impacts on the Moon observed duringthe first year of the NELIOTA campaign. A total of 31“validated” events have been recorded so far with 27 beingsporadic (yellow points) and 4 possibly originating fromthe Geminid stream observed during the period of 12-14December 2017 (green points). The index associated witheach impact corresponds to the flash ID (1 st column inTable 5). During this period a total of 17 flashes (not shownhere) were classified as “suspected” since they were onlydetected in the I-band.calibrate the instrumental magnitudes and their associateduncertainties of 25 stars in the 0.023 sec exposure, 325 starsin the 1 sec exposure and 650 stars in the 10 sec observationscommon in the I- and R-bands.The results are shown in Fig. 6 with the I-band observa-tions in the left panel and the R-band observations in theright panel (see the caption of Fig. 6 for the explanationof different symbols and lines). For both bands the 0.023sec, the 1 sec and the 10 sec observations are shown alongwith the theoretical predictions imposed by Eq. 1. Whatis evident is that the model predictions follow nicely themeasured instrumental magnitudes and associated errors.Using the model, the magnitudes reached at these expo-sure times (at a certain noise level) can be derived. We findthat observations through KPFI of stellar-like objects on adark night in the I-band can reach up to 13.05 mag, 17.07mag and 18.76 mag in 0.023, 1 and 10 sec, respectively, ata 0 . × σ mag level (corresponding to 0.25 SN R detections).At 0 . × σ mag level ( SN R = 10) the magnitudes reached are11.43, 15.45 and 17.21 in 0.023, 1, and 10 sec, respectively.R-band observations can reach up to 12.81 mag, 16.82 magand 18.74 mag in 0.023, 1 and 10 sec, respectively, at a0 . × σ mag level, while the magnitudes that can be reachedat a 0 . × σ mag level are 11.19, 15.21 and 17.18 in 0.023, 1,and 10 sec, respectively.Additionally to the sources of uncertainties discussedabove there is another kind of noise which becomes signifi-cant at short exposures (like in this case) and can dominatethe error in the measurements especially at large airmasses. Fig. 11.
The distribution of the peak magnitudes of the‘validated’ events in the I-band (left panel) and in the R-band (right panel). For comparison we have plotted thedistribution of the events observed by Suggs et al. (2014)in the R-band (black-line histogram in the right panel). Itis evident that the NELIOTA campaign is detecting flashesabout two magnitudes fainter than what was previously de-tected by Suggs et al. (2014), a result of the use of a largeraperture telescope (see text for details). The difference inthe number of the events is mainly due to the different cam-paign periods [one year of NELIOTA observations versusfive years of observations in Suggs et al. (2014)].This is the scintillation noise ( σ scint ) which depends onmany parameters, mainly on the stability of the atmosphereon the night of oservation. As a first attempt to explorethis effect we follow the description described in (Suggset al. 2014) by performing observations of a photometricstandard star (SA 113-475, RA : 21 h m s , DEC :+00 ◦ (cid:48) (cid:48)(cid:48) ; Landolt (1992)) on the night of 7 July, 2018.We have observed this star in both bands at four differentairmasses ( X =1.26, 1.7, 2.2, 2.75) with an exposure timeof 23 msec and for 30 times per airmass. The results ofthe aperture photometry that we performed, using the phot task in IRAF , are shown in the top panel of Fig. 7 withthe instrumental magnitudes (small circles) in both R- andI-bands (red and blue symbols respectively) plotted againstthe airmass. The large circle in each airmass group of pointsis the corresponding mean value of the measured counts inmagnitudes. It is evident that there is a signifficant scatterin brightness, mainly due to scintillation, around the meanvalue which gets larger with airmass. This is shown moreclearly in the bottom panel of Fig. 7 with the scintillationnoise ( σ scint ), calculated as the standard deviation of theobservations in each airmass group, plotted against the air-mass. The standard deviation was calculated as the 1-sigmavariation of the measured fluxes (in counts). Linear fits tothe data give an estimate of the actual effect of scintilla-tion with airmass. In particular, we find that in the R-band σ scint changes as 0 .
034 + 0 . X with airmass and in theI-band this relation becomes − .
011 + 0 . X . We can seethat at large airmasses this can be the dominant source ofuncertainty reaching up to ∼ . σ = (cid:113) σ mag + σ scint (4) although, in a subsequent paper (Liakos et al. in prep), weplan to carry out a more thorough analysis in a subsequentpaper taking into account various seeing conditions and arange in brightness so that more realistic measurements ofthe scintillation noise can be obtained. Despite the simple design of the KPFI, the passage of lightthrough the optical elements will cause some distortion.The knowledge of the level of distortion in the two chan-nels of the KPFI is of great importance for NELIOTA sci-ence since it could introduce uncertainties in the determi-nation of the exact position of the NEO impact. We ob-served the dense stellar field of the open cluster NGC 6811(RA : 19 h m s , DEC : +46 ◦ (cid:48) (cid:48)(cid:48) ), on the nightof October 13 th , 2017, at 10 sec exposure, in order to pro-duce the distortion maps in the R- and I-bands. This wasdone by calculating the astrometric differences of the starsin the field using the ccmap task in IRAF that providesthe plate solution using a list of matched pixel and celestialcoordinates, the latter obtained from Janes et al. (2013).A total of 30 bright stars located throughout the field wereused in both bands to compute the plane solution by fitting3 rd order polynomials in both the x- and y-axis. The mapsare presented in Fig. 8 for the I-band (left panel) and forthe R-band (right panel). The arrows indicate the expectedposition of the star (beginning of the arrow) and the ob-served position (end of arrow). For displaying purposes thearrows are artificially enlarged by a factor of 50 with a scaleindicator (4 ‘enlarged’ pixels) shown in the upper left partof each plot. It is evident that the distortion is very small inthe center of the field, while it increases towards the edges.The rms of the fit was found to be 60 mas in the I-bandand 55 mas in the R-band. Nevertheless, the maximum dis-tortion is of the order of the pixel scale and, in any case,less than the seeing effects, therefore it does not affect thelocalization of NELIOTA flashes (accurate to 6 . (cid:48)(cid:48) on theMoon; see Sect. 7).
7. The NELIOTA observations
During the first year of the NELIOTA campaign, February2017 - February 2018, 31 NEO impacts on the Moon havebeen successfully detected and recorded, at a frame rate of30 fps, simultaneously in two photometric bands (R and I;Table 5). About half of them fade out very fast [within theduration of one single frame (33 msec)], while for the restwe were able to follow their brightness variation with time(see Table 5). An example of such a multi-frame event isshown in Fig. 9. In this time sequence the light variation ofthe event observed at 04:35:09.967 UT on December 14 th ,2017 (event ID 30) is shown in the four I-band frames inwhich it was detected. For comparison the frame just be-fore the event occurred and right after the event finished arealso presented. The time at the beginning of the exposurein each frame is also given in each panel (t=0 msec corre-sponds to the frame where the event was initially detected).In one case (event ID 21) the event was captured duringits rise with the I-band brightness measured in the firstframe (8.49 ± NELIOTA observations began in February during commis-sioning phase but the official campaign started in March 2017.11ilouris et al.: The NELIOTA lunar monitoring system
Fig. 12.
The background standard deviation (
ST D bkg )distributions in I- and R-band (left and right panels re-spectively) for observations made in lunar phase of 0.118 onMarch 1 st , 2017. Each measurement comes from an aper-ture of radius of 11 pixels with 352 such regions randomlyselected throughout the Moon’s surface observed. Fig. 13.
Measured magnitudes versus the associated mag-nitude uncertainty ( σ mag ) for both stars and lunar impactflashes in the I- and R-bands (left and right, respectively).In both panels the stellar measurements are presented asstars, while the ’validated’ impact flashes are shown as cir-cles. For the I-band the ‘suspected’ impact events (cyancrosses) are also shown. For the stars, the predicted model(Eq. 1) is also presented (solid line; see Fig. 6). The dashedand dotted lines in each panel indicate the model predic-tion when the minimum and maximum value, respectively,of the background noise ( ST D bkg ) are considered for thecase of the observations of March 1 st , 2017 with the lowerlunar phase ∼ . ± . ± .
05 mag. For comparison, in Fig. 11we have also included the R-band magnitude distributionof the events detected by Suggs et al. (2014). This showsthat the NELIOTA setup was successful in detecting fainterflashes (about two magnitudes fainter than was done be-fore). The primary reason for this is the larger aperturetelescope that is now being used but also the higher ef-ficiency detectors that are implemented in the NELIOTAsystem. Since these fainter flashes are more frequent [see,e.g., Suggs et al. (2014)] the NELIOTA system is more sen-sitive in detecting those, compared to brighter flashes al-ready detected by other experiments. The difference in thenumber statistics between the two samples (Suggs et al.(2014) and NELIOTA) comes from the fact that the firstone accounts for a five year campaign compared to the oneyear operation of NELIOTA.Taking into account the total number of impacts (31),the total exposure time during which these events were col-lected (54 hours of Moon observations), as well as the aver-age surface observed per night, we estimate a detection rateof 1.96 × − events km − h − . This is about a factor oftwo higher than what has been reported before [1.03 × − events km − h − (Suggs et al. 2014) and 1.09 × − eventskm − h − (Rembold & Ryan 2015)]. This is a direct resultof the use of a larger aperture telescope (compared to previ-ous campaigns) allowing for the detection of fainter events(see Fig. 11), which are more frequent (see, e.g. Drolshagenet al. (2017) and references therein). In our attempt to quantify the expected performance ofthe NELIOTA system, we explored how the analysis per-formed on stellar photometry (Sect. 6.1) can be appliedto predict the limiting magnitudes of the impact flashesthat the NELIOTA system can reach. One parameter thatgreately affects the detection of such flashes is the back-ground noise ( σ bkg ; see Eq. 2) which in the case of theMoon, in contrast to the uniform sky background observedin stellar objects, varies significantly depending on the lu-nar phase. As the noise of the background ( σ bkg ) scales withthe square root of the brightness of the background, in thiscase the illumination of the Moon, we expect that σ bkg ob-tains its smallest values on observations obtained when theside of the Moon visible from the earth is illuminated theleast. For NELIOTA scheduling this corresponds to phase ∼ .
1. Even in this case, since the Moon is not uniformly
Table 5.
Basic parameters of the impact flashes observed with the 1.2 m Kryoneri telescope as part of the NELIOTAproject.
Flash ID Date UT Airmass Lunar phase Latitude Longitude R ± σ R I ± σ I Duration(degrees) (degrees) (mag) (mag) (msec)1 2017-02-01 17:13:57.863 1.66 0.234 -1.5 -29.2 10.15 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± Notes.
R- and I-band peak magnitudes (columns 8 and 9) and duration of the impact flashes (column 10) observed with the1.2 m Kryoneri telescope for the NELIOTA project. The UT date and time at the beginning of the observation (columns 2 and3 respectively), the airmass (column 4), the selenographic coordinates (Latitude and Longitude; columns 6 and 7) and the lunarphase (column 5) are also presented. Each flash is assigned with an ID (column 1). The error in the selenographic coordinates isestimated to be ∼ . illuminated, σ bkg is expected to vary throughout its surface(gradually becoming brighter from the limb to the center).In order to explore the effect of σ bkg on the impact flashdetections, we calculate the standard deviation ( ST D bkg )in 352, randomly selected, regions throughout the Moon’ssurface, in both bands (R and I). The background was mea-sured within an aperture of a radius of 11 pixels, whichroughly corresponds to an area similar to the one used tocalculate the local background of the stars in the cluster.We used the observations performed on March 1 st , 2017(exposure time of 23 msec) that corresponds to the lowestlunar phase (0.118) at which obsrevations have been ob-tained. The measured values for ST D bkg are shown in thetwo histograms in Fig. 12 corresponding to the I-band (leftpanel) and R-band (right panel). We find that the values of
ST D bkg range from ∼
30 to 50 counts in the I-band, withthe most frequent value at ∼
35 counts and from ∼
22 to37 counts in the R-band, with the most frequent value at ∼
25 counts.In order to estimate the NELIOTA detection limit weplotted in Fig. 13 the observed magnitudes and their cal-culated uncertainties for both the impact flashes and thestars (see Fig. 6) at the fixed NELIOTA exposure time of0.023 sec. In both bands the observed impact flashes at agiven uncertainty level are brighter than their respective stellar analogues, a result, mainly due to the significant in-crease in σ bkg compared to the sky background as discussedabove. Furthermore, the scatter in brightness of the impactflashes at a given uncertainty is, mainly, due to the varia-tion in σ bkg , which, as mentioned earlier, greatly depends onthe lunar phase. In Fig. 13, we have also plotted the modelpredictions (Eq. 1) for the stars (the solid line; also seen inFig. 6), corresponding to a ST D bkg value of 12 counts (seeSect. 6.1). Adjusting Eq. 2 for different background noiselevels we can then plot the model for the range of
ST D bkg values observed for ∼ . ST D bkg (30 and 50 counts inthe I-band and 22 and 37 for the R-band). For the I-bandwe find that the NELIOTA detections at the
SN R = 2 . ST D bkg (as shown in Fig. 12). The 10
SN R detections range between 10.35 mag and 10.83 mag.Similarly, in the R-band, the
SN R = 2 . SN R = 10detections range between 10.39 mag and 10.83 mag.We furthermore explored the detection completeness asa funtion of the brightness of the impact flash in the R-band. To achieve this we created mock impact flashes ofcertain brightnesses on a real observation of the Moon (in
Fig. 14.
Detection completeness as a function of thebrightness of the impact flashes in the R-band. For eachmagnitude bin of 0.2 mag in the R-band 50 mock flasheswere placed in random positions throughout a real obser-vation of the Moon (in both the R- and I-band images)with a color index of R-I=1 mag. The completeness is thencomputed by forming the ratio of the ones detected with
SN R above 2.5 ( σ mag < . SN R above 2.5 ( σ mag < . st ,2017 (see above) and placed 50 mock flashes in ran-dom places throughout the Moon’s surface using the task mkobjects in IRAF . Each flash was made to resemble pointlike gaussian distributions of FWHM of 2.4 (cid:48)(cid:48) , the mean see-ing value of all the nights with NELIOTA observations andwere grouped in bins of 0.2 mags in brightness ranging from9.4 to 13.4 mag in the R-band. The produced image ofthe impact flashes on the Moon was then background sub-tracted with the procedure described in Sect. 5.2 and usedas input to the phot task in IRAF in order to count the de-tections showing a SN R greater than 2.5. The completenesswas then formed as the ratio of the mock flashes fulfillingthe above criteria to the initial number of flashes created(50). The results of this procedure are presented in Fig. 14with the detection completeness presented as a function ofR-band brightness. The drop in the source completenessbegins at 10.4 mag (0.98) and it gets practically zero (0.02)at 12.4 mag, well in accordance with our previous analysisfor the limiting magnitudes (Fig. 13), with a 50% detectioncompleteness at 11.4 mag. In a subsequent paper we planfor a more detailed analysis on the NELIOTA detectioncompleteness taking into account all the parameters thatinfluence such calculations (e.g. seeing, source morphology,variable background level, and variable color index) in or-der to properly define the intrinsic luminosity function ofrelatively faint Lunar impact flashes.
8. Conclusions
We present the design and performance of KPFI, a wide-field, high-cadence, twin lunar monitoring system at theprime focus of the 1.2 m Kryoneri telescope that is usedfor the NELIOTA project. The project aims at detectingfaint flashes on the Moon’s surface produced by impacts ofNEOs. The optical design, the detectors, the control sys-tem as well as the dedicated software for the detection of the NEO impacts are discussed, while the instrument per-formance and highlights of the first scientific results areshown. The novelty of the NELIOTA system is the use ofa large aperture telescope (larger than ever used before forthis purpose) and the high-cadence achieved by two detec-tors observing simultaneously at a rate of 30 frames-per-second in two optical bands (R and I). The noise modelpredicts that the NELIOTA system can detect NEO impactflashes at a
SN R = 2 . ∼ . × − events km − h − .The ability to monitor the impact flashes in two bands atthe same time provides unique and significant constraintson the temperatures produced during the impact.The wide-field, high-cadence and simultaneous multi-color abilities of KPFI make it a unique instrument thatcan be used by the community for a variety of astronomyprojects such as occultations (e.g. Sicardy et al., in prep),exoplanet transit light curves, monitoring of early super-nova light curves [e.g. Bonanos & Boumis (2016)], as wellas transient follow up [e.g. Wyrzykowski et al. (2017)] andother NEO science.The reader is encouraged to visit the dedicatedNELIOTA website for further information. Acknowledgements.
The authors wish to thank the anonymous ref-eree and the Editor (Thierry Forveille) for constructive comments thathelped improving the paper. We are also greatful to A. Georgakakisfor stimulating discussions and to E. Palaiologou for useful guidanceon the field distortion calculations. The authors greatfully acknowl-edge financial support by the European Space Agency under theNELIOTA program, contract No. 4000112943. This work has madeuse of data from the European Space Agency NELIOTA project,obtained with the 1.2 m Kryoneri telescope, which is operated byIAASARS, National Observatory of Athens, Greece.
References
Ait Moulay Larbi, M., Daassou, A., Baratoux, D., et al. 2015, EarthMoon and Planets, 115, 1Artem’eva, N. A., Kosarev, I. B., Nemtchinov, I. V., Trubetskaya,I. A., & Shuvalov, V. V. 2001, Solar System Research, 35, 177Bellot Rubio, L. R., Ortiz, J. L., & Sada, P. V. 2000, ApJ, 542, L65Bonanos, A. Z., Avdellidou, C., Liakos, A., et al. 2018, A&A, 612,A76Bonanos, A. Z. & Boumis, P. 2016, A&A, 585, A19Bouley, S., Baratoux, D., Vaubaillon, J., et al. 2012, Icarus, 218, 115Brown, P. G., Assink, J. D., Astiz, L., et al. 2013, Nature, 503, 238Drolshagen, G., Koschny, D., Drolshagen, S., Kretschmer, J., &Poppe, B. 2017, Planet. Space Sci., 143, 21Everett, M. E. & Howell, S. B. 2001, PASP, 113, 1428Goode, P. R., Qiu, J., Yurchyshyn, V., et al. 2001, Geophys. Res. Lett.,28, 1671Harris, A. W. & D’Abramo, G. 2015, Icarus, 257, 302Ivanov, B. A., Neukum, G., Bottke, Jr., W. F., & Hartmann, W. K.2002, The Comparison of Size-Frequency Distributions of ImpactCraters and Asteroids and the Planetary Cratering Rate, ed. W. F.Bottke, Jr., A. Cellino, P. Paolicchi, & R. P. Binzel, 89–101Janes, K., Barnes, S. A., Meibom, S., & Hoq, S. 2013, AJ, 145, 7Jeffries, R. D., Naylor, T., Mayne, N. J., Bell, C. P. M., & Littlefair,S. P. 2013, MNRAS, 434, 2438Landolt, A. U. 1992, AJ, 104, 340Madiedo, J. M., Ortiz, J. L., Morales, N., & Cabrera-Ca˜no, J. 2014,MNRAS, 439, 2364Moser, D. E., Suggs, R. M., Swift, W. R., et al. 2011, in Meteoroids:The Smallest Solar System Bodies, ed. W. J. Cooke, D. E. Moser,B. F. Hardin, & D. Janches, 142