MNRAS 437 3812ndash3823 (2014) doi101093mnrasstt2178Advance Access publication 2013 December 4

The unexpected 2012 Draconid meteor storm

Quanzhi Ye1lsaquo Paul A Wiegert1 Peter G Brown12 Margaret D Campbell-Brown1

and Robert J Weryk1

1Department of Physics and Astronomy the University of Western Ontario London ON N6A 3K7 Canada2Centre for Planetary Science amp Exploration the University of Western Ontario London ON N6A 5B8 Canada

Accepted 2013 November 7 Received 2013 November 5 in original form 2013 August 16

ABSTRACTAn unexpected intense outburst of the Draconid meteor shower was detected by the CanadianMeteor Orbit Radar on 2012 October 8 The peak flux occurred at sim1640 UT on October 8with a maximum of 24 plusmn 03 hminus1 kmminus2 (appropriate to meteoroid mass larger than 10minus7 kg)equivalent to a ZHRmax asymp 9000 plusmn 1000 using 5-min intervals using a mass distribution indexof s = 188 plusmn 001 as determined from the amplitude distribution of underdense Draconidechoes This makes the outburst among the strongest Draconid returns since 1946 and thehighest flux shower since the 1966 Leonid meteor storm assuming that a constant power-lawdistribution holds from radar to visual meteoroid sizes The weighted mean geocentric radiantin the time interval of 15ndash19 h UT 2012 October 8 was αg = 2624 plusmn 01 δg = 557 plusmn 01(epoch J20000) Visual observers also reported increased activity around the peak time butwith a much lower rate (ZHR sim 200) suggesting that the magnitudendashcumulative numberrelationship is not a simple power law Ablation modelling of the observed meteors as apopulation does not yield a unique solution for the grain size and distribution of Draconidmeteoroids but is consistent with a typical Draconid meteoroid of mtotal between 10minus6 and10minus4 kg being composed of 10ndash100 grains Dynamical simulations indicate that the outburstwas caused by dust particles released during the 1966 perihelion passage of the parent comet21PGiacobini-Zinner although there are discrepancies between the modelled and observedtiming of the encounter presumably caused by approaches of the comet to Jupiter during1966ndash1972 Based on the results of our dynamical simulation we predict possible increasedactivity of the Draconid meteor shower in 2018 2019 2021 and 2025

Key words comets individual21PGiacobini-Zinner ndash meteorites meteors meteoroids

1 IN T RO D U C T I O N

The October Draconid meteor shower (DRA or IAUMDC 009usually referred to as lsquoDraconidsrsquo some older literature may alsorefer to it as the lsquoGiacobinidsrsquo) is an annual meteor shower producedby comet 21PGiacobini-Zinner and is active in early October Itwas first observed in the 1920rsquos (Davidson 1915 Denning 1926)and produced two spectacular meteor storms in 1933 and 1946 (cfJenniskens 1995) Aside from the two storms and a few outbursts ithas been quiet in most years with hourly rates no more than a fewmeteors per hour

A review of the historic observations and studies of the Draconidsmay be found in our earlier work (Ye et al 2013) in this introduc-tory section we mainly address the strong returns in 2011ndash2012The outburst of the 2011 Draconids was predicted by a number ofresearchers (Maslov 2011 Vaubaillon et al 2011 to name a few)

E-mail qye22uwoca

and analyses made by various observing teams indicate that the pre-dictions were quite accurate in both timing and meteor rates (suchas Kero et al 2012 Koten et al 2012 Sato Watanabe amp Ohkawa2012 Vaubaillon et al 2012 and many others)

In contrast no significant Draconid outburst was predicted for2012 Maslov (2011) suggested encounters with the material ejectedby comet Giacobini-Zinner in 1959 and 1966 (in the form of the1959 and 1966 trails) at 15ndash17 h UT on 2012 October 8 but henoted that lsquovisually-detectable activity is unlikelyrsquo predicting themaximum Zenith Hourly Rate (ZHR) to be 05 and 02 respectivelyfor each encounter The prediction by Maslov is summarized inTable 1

In fact an intense outburst of the Draconid meteor showerwas subsequently detected by the Canadian Meteor Orbit Radar(CMOR) on 2012 October 8 (Brown amp Ye 2012) As seen byCMOR the meteor rate started increasing at sim15 h UT quicklyreached a maximum around 1640 UT and returned to the back-ground rate around 19 h UT Preliminary analysis indicated that thepeak ZHR was well above the storm threshold (1000 meteors per

Ccopy 2013 The AuthorsPublished by Oxford University Press on behalf of the Royal Astronomical Society

at University of W

estern Ontario on February 12 2014

httpmnrasoxfordjournalsorg

Dow

nloaded from

The unexpected 2012 Draconid meteor storm 3813

Table 1 The Draconid 2012 outburst prediction by Maslov(2011)

Trail Time αg δg vg ZHR(October 8UT) (J2000) (km sminus1)

1959 1622 1654 2627 +558 210 051966 1537 ndash ndash 02

hour) Unfortunately this was mostly a radar event in addition tothe poor timing of the peak (which only favoured Siberia and Cen-tral Asia in terms of darkness) visual and video observers werelargely caught unprepared The quick look analysis carried out bythe International Meteor Organization (IMO) revealed a peak ofZHR = 324 plusmn 66 centred at 1651 UT based on observations fromonly four observers1 The camera set up by the Petnica MeteorGroup in Serbia also detected some activity at 17ndash18 h UT2

Due to the lack of observations by other techniques the radar ob-servations are essential for the study of this outburst We will mostlyfollow the methodology used in our earlier study of the 2011 event(Ye et al 2013) to analyse the 2012 event as recorded by CMORThe goals of this paper are to estimate the basic observational char-acteristics of the outburst such as mean radiant mass distributionand flux variation and to apply the Campbell-Brown amp Koschny(2004) meteoroid ablation model to try to constrain the physicalcharacteristics of the meteoroids We will also investigate the causeof this outburst through numerical modelling of the stream

2 IN S T RU M E N TAT I O N A N D DATAR E D U C T I O N

The instrumental and data reduction details of this work are similarto that given in Ye et al (2013) as applied to CMOR observationsof the 2011 Draconids Technical details of the entire system mayalso be found in Hocking Fuller amp Vandepeer (2001) Jones et al(2005) Brown et al (2008) and Weryk amp Brown (2012) Here weonly summarize a few key concepts

CMOR is a six-site interferometric radar array located near Lon-don Ontario and Canada It operates at 1745 2985 and 3815 MHzbut only the 2985 MHz data are used in this study The 2985 MHzcomponent transmits with 12 kW peak power with a pulse repeti-tion frequency of 532 Hz The radar lsquoseesrsquo a meteor by detectingthe reflection of the outgoing beam from an ionized meteor trailtherefore radar meteors are detected 90 away from their radiants(such a reflection condition is called specular) As such when theradiant is close to the zenith it becomes difficult for the radar todetect echoes from a particular stream as the echoes are at lowelevation and large ranges when the radiant is close to the zenithDetected echoes are processed in an automatic manner to eliminatequestionable detections correlate the same events observed at dif-ferent sites and calculate meteor trajectories Usually raw streameddata are not permanently kept but we intentionally saved the rawdata from the 2012 Draconid outburst for further analysis

Meteor echoes seen by radar may be classified as either under-dense or overdense echoes based on their apparent amplitude-timecharacteristics The deciding factors are many (see Ye et al 2013for a more theoretical introduction) but for a given shower over-dense echoes are typically associated with larger meteoroids than

1 httpwwwimonetlivedraconids20122 httpwww meteorirswordpressi-ove-godine-pojacana-aktivnost-drakonida

Figure 1 Examples of meteor echoes as seen by CMOR Top underdensetrail middle overdense trail bottom overdense wind-twisted trail

underdense echoes The underdense echoes are always character-ized by a rapid rise and an exponential decay in amplitude (Fig 1a)making them relatively easy to identify automatically Ideally theoverdense echoes are characterized by a rapid rise a plateau and asteady decay in amplitude (Fig 1b) but in reality trails distorted byupper atmospheric winds and exhibiting multiple reflection pointsare often seen (Fig 1c) making them difficult to characterize withautomatic algorithms

Noting the advantages and weakness of the automatic routineswe take two approaches for the initial data reduction

(i) For the first approach we try to separate Draconid echoesfrom other echoes in the automatically processed data This is donein several steps First we select all meteor radiants measured withtime-of-flight (tof) velocities within 10 of the apparent commonradiant (ie an acceptance radius of 10 around the apparent centreof the cluster of meteor radiants) and 20 per cent of the annual geo-centric velocity vg = 204 km sminus1 (Jenniskens 2006) The velocityrestriction is intentionally broad reflecting the spread due to mea-surement error and the high decelerations of Draconid meteoroidsa consequence of their fragility (eg Jacchia Kopal amp Millman1950 Campbell-Brown et al 2006) All these echoes are manu-ally inspected to remove data with problematic time picks (ie animproperly automatically chosen time of occurrence of the spec-ular point) determined by the automatic algorithms and to ensurethe trajectory solution is correct We do not revise the time picksas we need to keep the procedure repeatable by automatic algo-rithms in order to estimate the uncertainty The uncertainties areestimated with a Monto Carlo synthetic echo simulator developedby Weryk amp Brown (2012) allowing the derivation of a weightedmean common radiant We then explore the ideal radiant size forthis outburst event by varying the acceptance radiant interval tosearch for the acceptance radius at which the number of radiantsreaches an asymptote approaching the background level which inthis case is sim5 (Fig 2) therefore we reduce the acceptance radius

at University of W

estern Ontario on February 12 2014

httpmnrasoxfordjournalsorg

Dow

nloaded from

3814 Q-Z Ye et al

Figure 2 Radiant densities centred at λ minus λ = 53 β = +79 in sun-centred ecliptic coordinates as a function of radiant size on the outburst date2012 October 8 Shown for comparison is the average of the same eclipticsun-centred radiant location on 2012 October 7 and 9 when no significantDraconid activity was present It can be seen that the radiant size on October8 and the background rate meets at sim5

Table 2 Characteristics of the 2012 Draconid radiant as observed byCMOR The mean radiants are calculated using error weightings accord-ing to the Monte Carlo error simulations by the synthetic echo simulator(see Section 21 of this paper as well as Weryk amp Brown 2012 Ye et al2013)

Time period αg δg Nradiants

(J2000)

October 8 (all day) 26232 plusmn 010 +5567 plusmn 005 576October 8 (15ndash19 h UT) 26235 plusmn 010 +5567 plusmn 005 545

from 10 to 5 We identify 576 Draconid echoes in this mannerand refer to this data set as the complete data set (Table 2)

(ii) For the second approach we try to separate all overdenseDraconid echoes As noted before overdense echoes are some-times difficult to identify automatically therefore we separate theseechoes by manually inspecting the raw data from the main site be-tween 15ndash19 h UT on 2012 October 8 Trajectories of these meteorechoes are manually determined when remote site observations areavailable to determine whether they are Draconids In all some 240overdense Draconids are identified this way and will be referred asthe overdense data set

Since the velocities in the complete data set are determined usingthe tof method which depends on trail geometry and is uncer-tain in part due to height averaging we also measure the speedof the meteoroids using the Fresnel phase-time method and Fresnelamplitude-time method (simplified as lsquopre-t0rsquo and lsquoFresnelrsquo methodhereafter see Ceplecha et al 1998 sections 461 and 462 for anoverview) The advantage of these two methods is that they onlydepend on the observation from the main site which does not intro-duce the geometry or height averaging effects of the tof techniqueIn total we gathered 360 echoes with pre-t0 speeds and 25 echoeswith Fresnel speeds where 13 echoes have both pre-t0 and Fresnelspeeds Numerically the two speeds are equivalent as the decelera-tion of the meteoroid is minimal during its passage over the Fresnelzones (amounting to sim1ndash2 km of trail length Fig 3) we thereforeassume that the pre-t0 or Fresnel speed is the meteoroidrsquos speed at

Figure 3 The Fresnel velocities versus pre-t0 velocities for the 13 high-altitude echoes having both techniques for velocities measurable

Figure 4 The raw number of Draconid echoes detected by CMOR inthe complete and the overdense data set binned in 10-min intervals Thebackground level is normally 1 Draconid per 10-min bin

its specular height For the 13 echoes where both pre-t0 and Fresnelvelocities are measured preference is given to the one with smalleruncertainty

The raw number of echoes as a function of time on 2012 October8 in each data set is given in Fig 4

3 O B S E RVAT I O NA L R E S U LT S

31 Radiant and orbit

The weighted mean radiant of all meteors in the complete data set isαg = 2623 plusmn 01 δg = +557 plusmn 01 (J2000 epoch) (Fig 5) Theweighted mean radiant of meteors detected within 15ndash19 h UT isαg = 2624 plusmn 01 δg = +557 plusmn 01 (J2000 epoch)

The derivation of true orbit requires additional effort in datareduction as the fragility of Draconid meteoroids introducesan underestimation of the deceleration correction as noted by

at University of W

estern Ontario on February 12 2014

httpmnrasoxfordjournalsorg

Dow

nloaded from

The unexpected 2012 Draconid meteor storm 3815

Figure 5 Distribution of individual Draconid radiant of the complete dataset

Campbell-Brown et al (2006) and Ye et al (2013) which leadsto a reduction in estimating the true out of atmosphere speed of themeteoroids To solve this problem we make use of the echoes de-tected at higher heights as they do not penetrate very deep into theatmosphere and therefore suffer less deceleration We also need toselect the echoes with pre-t0 or Fresnel velocity measured to min-imize the error caused by deceleration undercorrection We onlyconsider the echoes detected within 15ndash19 h UT (ie close to thepeak time) to minimize the possible contamination of backgroundDraconid meteors not associated with the outburst We then choose98 km as the cut-off height as it is the highest possible cut-offwith a statistically significant sample (only two echoes are detectedat ge99 km) Taking all these constraints into account a total of14 echoes are selected from the complete data set We note that asignificant fraction (514) of these have local zenith angle η above50 which suggests that the undercorrection effect is again signif-icant due to the longer atmospheric path of the meteoroid enteringthe atmosphere at such a shallow angle Hence we remove thesefive echoes and create an additional subset and present both resultstogether (Table 3)

We note that the mean orbit derived from the nine-meteor sampleis consistent with the orbit of 21PGiacobini-Zinner at its 1959 and1966 apparitions Additionally the vg derived from the nine-meteorsample agrees with that derived from the multistation video obser-vations of the 2011 Draconid outburst (vg = 209 plusmn 10 km sminus1 asnoted by Jenniskens Barentsen amp Yrjola 2011) within uncertaintyThis agreement is reassuring as the video observations have morecomplete coverage of the trail length of each meteoroid comparedto radar allowing the deceleration correction for individual me-teoroids to be established with higher confidence along with thepre-atmosphere velocity (vg) and orbit We believe that the orbitderived from the nine-meteor sample represents the best mean orbitfor the 2012 Draconid outburst as observed by CMOR

32 Mass index

The mass index can be determined using either underdense echoes oroverdense echoes For underdense echoes the cumulative number ofechoes with amplitude greater than A follows the relation N prop Aminuss+1

(McIntosh 1968) while for diffusion-limited overdense echoesN prop τminus34(sminus1) (McIntosh 1968) where τ is the echo duration

We first use the underdense echoes detected between 15ndash19 hUT in the complete data set to determine s Echoes within a rangeinterval between 110ndash130 km from the main site are selected toavoid contamination from overdense transition echoes (see Blaauwet al 2011 for discussion) We determine the mass index to be188 plusmn 001 by fitting the linear portion of the data in Fig 6 Theuncertainty given here is the fitting uncertainty only and could beseveral times smaller than the real one due to the small sample size

We then use the overdense data set as a check on the mass indexvalue determined using the amplitude distribution alone (Fig 7)The trail of electrons can fall to an undetectable density eitherthrough diffusion or chemical recombination The latter is morelikely for very dense trails produced by larger meteoroids at lowaltitudes The possible turnover between diffusion- and chemistry-limited regime or the lsquocharacteristic timersquo can be seen at about25 s in Fig 7 which agrees with the value derived for 2011 (27 sYe et al 2013) Using the few data points in the possible diffusion-limited regime the mass index can be estimated to be around 17ndash18 However we should note that this result is doubtful due to thepresence of a sudden steep rise in cumulative number at τ lt sim2 sThe possibility of underdense contamination can be ruled out byexamining the theoretical underdense region in the heightndashdurationdistribution (Fig 8) A possible explanation of this behaviour is thelack of long overdense echoes due to the radiant geometry at thetime of the outburst as well as abundance of smaller meteoroids ofthis event preventing us from getting enough statistics for longerduration echoes Alternatively the lack of a power-law fit at thehigh-mass end of the meteoroid distribution (ie overdense echoes)may indicate that the size distribution within the stream does notfollow a power law in contrast to the behaviour seen in 2011 wherea clear fit existed to the smallest overdense echoes durations In thiscase the upper upturn below τ lt sim2 indicates an overabundanceof smaller Draconids in the outburst

33 Flux

The flux can be calculated from the number of echoes detectedper unit time divided by the effective collecting area of the radarfor the Draconid radiant The effective collecting area of CMORis calculated following the scheme described in Brown amp Jones(1995) Simply put the collecting area for a given radiant is theintegration of the magnitude of the gain over the lsquoreflecting striprsquoof the radar wave (a 90 great circle perpendicular to the radiantdirection) which takes into account both the radiant geometry andmass index The ZHR (ie the number of meteors that an averageobserver would see in 1 h given that the sky is clear and darkand the radiant is at the zenith) can then be calculated by usingits relationship to the actual meteoroid flux (Koschack amp Rendtel1990) as computed by the radar and mass index using the fact thatthe limiting meteor radio magnitude of CMOR is sim85

As shown in Fig 9 the main peak based on 5-min averagedCMOR data occurred at 1638 UT (solar longitude λ = 195622)October 8 with maximum flux of 24 plusmn 03 hminus1 kmminus2 (ap-propriate to meteoroid mass larger than 10minus7 kg) equallingZHRmax = 9000 plusmn 1000 The IMO data show several lsquopeak-letsrsquofrom 1625ndash1655 UT with ZHRmax around 500ndash600 The exis-tence of these lsquopeak-letsrsquo is likely a statistical oddity due to thelimited number of observers (only one observer for each of theselsquopeak-letsrsquo) For example the time range around 1640 UT (whichcontained three observers reporting 10 Draconids total observed)corresponds to a ZHR of 185 plusmn 55

at University of W

estern Ontario on February 12 2014

httpmnrasoxfordjournalsorg

Dow

nloaded from

3816 Q-Z Ye et al

Table 3 High-quality Draconid radiants and speeds (top) and orbits (bottom) from the 2012 outburst as detected by CMOR selected on the basis of theirhigh ablation heights where atmospheric deceleration is minimal Here we show the relative accuracy of individual speed measurements by comparing thetime-of-flight speed as determined from time differences between specular points at three or more station with independent speeds determined by the pre-t0phase fitting method (Cervera Elford amp Steel 1997) and Fresnel amplitude post-t0 oscillations (Ceplecha et al 1998) The echoes with leading dagger indicateradiant zenith angles η lt 50 where deceleration should be negligible

Time αg δg vfresnel vpreminust0 vtof Specular heightOctober 8 (UTC) (J2000) (J2000) (km sminus1) (km sminus1) (km sminus1) (km)

CMOR observations 152139682 26142 plusmn 058 5523 plusmn 058 ndash 2313 plusmn 184 2188 plusmn 014 980152519024 26175 plusmn 028 5489 plusmn 028 ndash 2273 plusmn 108 2288 plusmn 035 989152559484 26131 plusmn 040 5561 plusmn 040 ndash 2143 plusmn 049 2205 plusmn 012 981154934396 26190 plusmn 125 5576 plusmn 125 ndash 2210 plusmn 077 2207 plusmn 026 983155210086 26071 plusmn 054 5551 plusmn 054 2292 plusmn 024 ndash 2230 plusmn 011 990

dagger 160952576 26151 plusmn 078 5550 plusmn 078 ndash 2188 plusmn 098 2213 plusmn 020 990dagger 161646118 26268 plusmn 087 5593 plusmn 087 ndash 2629 plusmn 190 2365 plusmn 104 990dagger 161750582 26473 plusmn 115 5599 plusmn 115 ndash 2641 plusmn 166 2133 plusmn 138 983dagger 162556203 26395 plusmn 064 5639 plusmn 064 ndash 2388 plusmn 086 2225 plusmn 058 989dagger 162804019 26298 plusmn 047 5593 plusmn 047 ndash 2365 plusmn 017 2234 plusmn 027 987dagger 162825902 25888 plusmn 046 5450 plusmn 046 ndash 2331 plusmn 269 2304 plusmn 010 989dagger 170035201 26279 plusmn 044 5467 plusmn 044 ndash 2237 plusmn 055 2300 plusmn 050 986dagger 171041590 26249 plusmn 147 5444 plusmn 147 ndash 2412 plusmn 037 2333 plusmn 060 981dagger 172022399 26319 plusmn 141 5265 plusmn 141 ndash 2415 plusmn 037 2297 plusmn 060 983

21PGiacobini-Zinner (1959) ndash ndash ndash ndash ndash ndash21PGiacobini-Zinner (1966) ndash ndash ndash ndash ndash ndash

Weighted mean ndash 26191 plusmn 039 5525 plusmn 025 ndash ndash ndashWeighted mean η lt 50 ndash 26231 plusmn 056 5522 plusmn 037 ndash ndash ndash

Time a e i ω q vg

October 8 (UT) (au) (au) (km sminus1)

CMOR observations 152139682 2922 plusmn 0887 0660 plusmn 0104 30851 plusmn 2568 195569 plusmn 0001 171323 plusmn 0950 0994 plusmn 0001 20076 plusmn 2103152519024 2870 plusmn 0500 0653 plusmn 0061 30071 plusmn 1521 195571 plusmn 0001 171970 plusmn 0506 0995 plusmn 0001 19615 plusmn 1241152559484 2266 plusmn 0137 0561 plusmn 0027 28503 plusmn 0757 195572 plusmn 0001 172100 plusmn 0598 0996 plusmn 0001 18111 plusmn 0575154934396 2465 plusmn 0302 0596 plusmn 0050 29521 plusmn 1194 195588 plusmn 0001 172286 plusmn 1742 0996 plusmn 0002 18894 plusmn 0893155210086 2783 plusmn 0141 0643 plusmn 0018 30630 plusmn 0382 195590 plusmn 0001 171136 plusmn 0733 0994 plusmn 0001 19838 plusmn 0275

dagger 160952576 2414 plusmn 0326 0588 plusmn 0056 29137 plusmn 1454 195602 plusmn 0001 172032 plusmn 1140 0995 plusmn 0001 18639 plusmn 1141dagger 161646118 7109 plusmn 6378 0860 plusmn 0126 34850 plusmn 2261 195607 plusmn 0001 172274 plusmn 1101 0995 plusmn 0001 23627 plusmn 2099dagger 161750582 7295 plusmn 5974 0864 plusmn 0112 35089 plusmn 1994 195608 plusmn 0001 172835 plusmn 1340 0995 plusmn 0001 23761 plusmn 1832dagger 162556203 3298 plusmn 0578 0698 plusmn 0053 32059 plusmn 1158 195613 plusmn 0001 173324 plusmn 0848 0996 plusmn 0001 20938 plusmn 0973dagger 162804019 3211 plusmn 0153 0690 plusmn 0015 31637 plusmn 0279 195614 plusmn 0001 172662 plusmn 0613 0996 plusmn 0001 20676 plusmn 0193dagger 162825902 3108 plusmn 1536 0681 plusmn 0158 30858 plusmn 3598 195614 plusmn 0001 169638 plusmn 0916 0992 plusmn 0002 20285 plusmn 3067dagger 170035201 2783 plusmn 0261 0642 plusmn 0034 29482 plusmn 0771 195637 plusmn 0001 172785 plusmn 0598 0996 plusmn 0001 19213 plusmn 0635dagger 171041590 4062 plusmn 0700 0755 plusmn 0042 31684 plusmn 0710 195644 plusmn 0001 172123 plusmn 1861 0995 plusmn 0002 21212 plusmn 0418dagger 172022399 4971 plusmn 1026 0800 plusmn 0041 31048 plusmn 0687 195650 plusmn 0001 172036 plusmn 1703 0995 plusmn 0002 21240 plusmn 0418

21PGiacobini-Zinner (1959) ndash 3453 154 0728 953 308976 1967326 1728373 0935 967 ndash21PGiacobini-Zinner (1966) ndash 3449 669 0729 394 309382 1966674 1729153 0933 501 ndash

Weighted mean ndash 2876 plusmn 0440 0672 plusmn 0025 30909 plusmn 0517 195602 plusmn 0001 172047 plusmn 0240 09951 plusmn 00003 2028 plusmn 045Weighted mean η lt 50 ndash 3223 plusmn 0610 0710 plusmn 0032 31446 plusmn 0689 195620 plusmn 0001 172256 plusmn 0351 09952 plusmn 00003 2080 plusmn 056

We note that these ZHRs assumed a single power-law fit fromfainter radar meteors to equivalent radio magnitude of +65 wherethe ZHR is defined As noted earlier an apparent deviation is notablebetween larger visual meteoroids and smaller radar meteoroids froma pure power law hence the effective ZHR quoted from CMORobservations are likely upper limits

34 Ablation modelling

To model the structure of the Draconid meteoroids we used the me-teoroid ablation model developed by Campbell-Brown amp Koschny(2004) The velocities and electron line densities as measured byCMOR are used as constraints to fit the model The velocities usedhere are appropriate to the echo specular height (ie pre-t0 or Fres-nel velocities) the electron line densities are computed from theobserved amplitude value using the formula appropriate to eitherunderdense echoes (when q lt 24 times 1014 mminus1) or overdense echoes(when q gt 24 times 1014 mminus1) The methodology is essentially the

same as for the 2011 outburst (Ye et al 2013) except that we donot model individual events for deceleration as no such events arefound for the 2012 outburst We use the parameters suggested bytwo previous studies of the Draconids Borovicka Spurny amp Koten(2007) and Ye et al (2013) while other parameters are left fixed assuggested in Campbell-Brown amp Koschny (2004) The parametersare summarized in Table 4

We first tried to find the optimal combination of grain massand grain number by fixing the total mass at 10minus5 kg and vary-ing these two variables We choose 10minus5 kg here as this value isa reasonable compromise between the detection limit (sim10minus7 kg)and the underdenseoverdense transition level (sim10minus4 kg appropri-ate to meteors with q asymp 1014 mminus1 or peak visual brightness of+4 mag) We used the transition level as our upper limit as most ofthe echoes in the complete data set are underdense echoes Finallywe look for a modelling fit that minimizes the deviation betweenthe model and trend of the data points with respect to the heightHowever from Fig 10 we cannot identify a unique fit the range

at University of W

estern Ontario on February 12 2014

httpmnrasoxfordjournalsorg

Dow

nloaded from

The unexpected 2012 Draconid meteor storm 3817

Figure 6 Cumulative echo numbers as a function of amplitude for Draconidechoes with ranges between 110ndash130 km The mass index is determined tobe 188 plusmn 001 by fitting the linear portion of the curve as shown

Figure 7 Cumulative echo numbers as a function of echo duration foroverdense Draconid echoes The possible characteristic time can be seennear tc = 25 s

of optimal fit seems to lie somewhere between ngrain = 10 and10 000

We then plot the modelling fit for mtotal = 10minus6ndash10minus4 kg and forngrain = 10ndash10 000 to further examine the goodness of each pa-rameter set (Fig 11) Again we do not see an obviously uniquesolution but it seems that the ngrain = 10 and 100 scenarios producea better qualitative fit to both velocities and electron line densitiesThis generally agrees with our previous finding for the 2011 out-burst which find the optimal ngrain = 100 it also agrees with thephotographic result for the 2005 outburst to some extent (Borovickaet al 2007) but we note that the fits are not well constrained fromour observations therefore they can only be used to broadly estab-lish the fact that the 2012 Draconid meteoroids were not radicallydifferent in physical makeup from those detected during the 2011outburst

Figure 8 The height range of the selected overdense echoes in the over-dense data set The shaded area marks the underdense region defined byMcKinley (1961) If this population were mainly underdense a clear dura-tion versus height trend would be presented in the graph

4 DY NA M I C M O D E L L I N G

To better understand the 2012 Draconids outburst numerical simu-lations were performed of the parent comet 21PGiacobini-ZinnerThe simulations included a Solar system of eight planets whoseinitial conditions were derived from the JPL DE405 ephemeris(Standish 1998) with the EarthndashMoon represented by a single par-ticle at the EarthndashMoon barycentre The system of planets parentand meteoroids was integrated with the RADAU method (Everhart1985) with a time step of 7 d used in all cases

The comet orbital elements used in these simulations are de-rived from Marsden amp Williams (2008) where orbital elements areprovided for each appearance of 21PGiacobini-Zinner from its dis-covery in 1900 through 2005 This extensive data set is importantas the parent being a Jupiter-family comet is extensively perturbedand has time-varying non-gravitational parameters

An initial survey of the meteoroid complex used the catalogueorbit for 21PGiacobini-Zinnerrsquos 2005 perihelion passage as astarting point Simulations of meteoroids released during each of21PGiacobini-Zinnerrsquos perihelion passages back to 1862 were ex-amined for clues as to the origin of the material that produced the2012 outburst These simulations revealed that the 1966 apparitionwas the source of the 2012 event and as a result the orbital elementsgiven for the 1966 perihelion passage were used in subsequent sim-ulations These orbital elements are listed in Table 5

No non-gravitational forces due to outgassing (eg MarsdenSekanina amp Yeomans 1973) were applied to the comet in our fi-nal simulation results Some test simulations that did include non-gravitational forces using the values for 21PGiacobini-Zinner fromYeomans (1986) as listed in Marsden amp Williams (2008) were runbut these did not produce any noticeable differences in the results

The simulations were run with the two-stage refinement proce-dure described in Wiegert et al (2013) Briefly in the first stagethe comet orbit is integrated backwards to the desired starting pointin this case back approximately 150 yr to 1862 The comet is thenintegrated forward again releasing meteoroids at each perihelionpassage as it does so All meteoroids which pass sufficiently closeto the Earthrsquos orbit during the simulation are collected this is our listof lsquobullrsquos-eyesrsquo Bullrsquos-eyes are those that have a minimum orbitalintersection distance (MOID) of no more than 002 au and that are

at University of W

estern Ontario on February 12 2014

httpmnrasoxfordjournalsorg

Dow

nloaded from

3818 Q-Z Ye et al

Figure 9 The ZHRflux profile of the 2012 Draconid outburst as observed by CMOR (binned in 5-min intervals error bars represent Poisson errors only)comparing to the visual profile reported by the visual observers to the IMO The radar flux is appropriate for meteoroids with m gt 10minus7 kg the visual flux isappropriate for meteors with V lt 2minus3 mag

Table 4 Input parameter for the ablation model as used in Ye et al (2013)

Parameter Value Unit

Deceleration corrected apparent velocity vinfin 2327 km sminus1

Zenith angle η 55 degBulk density ρbulk 300 kg mminus3

Grain density ρgrain 3 000 kg mminus3

Heat of ablation qheat 3 times 106 J kgminus1

Thermal conductivity κ 02 J mminus1 sminus1 Kminus1

at their MOID with the Earth no more than plusmn7 d from when ourplanet is there

In the second stage the integration of the comet forward in timeis repeated However in this case instead of releasing particlesrandomly as determined by a cometary ejection model particlesare released only near initial conditions in which a bullrsquos-eye wasproduced in the same time step in the first simulation These secondgeneration particles are released at the same position (that of thenucleus taken to be a point particle) but are given a random changein each velocity component of up to plusmn10 per cent of that of theoriginal bullrsquos-eye These second generation meteoroids inevitablycontain far more particles that reach our planet Of this samplethose which pass closest to the Earth in space and time will beconsidered to constitute the simulation outburst

In our initial simulations at each perihelion passage a numberM = 104ndash105 particles is released in each of the four size rangeswith radii r from 10minus5 to 10minus4 m 10minus4 to 10minus3 m 10minus3 to 10minus2

m and 10minus2 to 10minus1 m the whole range spanning from 10 microm to10 cm in diameter They are chosen so that the distribution ofparticle radii is flat when binned logarithmically by size A power-law size distribution with mass index s can be recovered after thesimulation is complete by giving each particle a weight proportional

Figure 10 Sensitivity test for different number of grains ngrain at a fixedtotal mass m = 10minus5 kg

at University of W

estern Ontario on February 12 2014

httpmnrasoxfordjournalsorg

Dow

nloaded from

The unexpected 2012 Draconid meteor storm 3819

Figure 11 Sensitivity test for different number of grains ngrain but for varying mgrain (ranges from 10minus10 to 10minus5 kg) The corresponding mtotal range is from10minus6 to 10minus4 kg

to rminus3s + 3 (Wiegert Vaubaillon amp Campbell-Brown 2009) if desiredHowever we do not apply such a weight to the results reported herebecause the outburst is found to consist of such a narrow range ofparticles (radii almost exclusively from 100 microm to 1 mm) that theapplication of such a weighting is likely inappropriate

Post-Newtonian general relativistic corrections and radiative (iePoyntingndashRobertson) effects are also included The ratio of solarradiation pressure to gravitational force β is related to the particleradius r (in microm) through β = 19r following Weidenschilling ampJackson (1993) though our expression assumes a particle mass

at University of W

estern Ontario on February 12 2014

httpmnrasoxfordjournalsorg

Dow

nloaded from

3820 Q-Z Ye et al

Table 5 A list of the orbital elements used for comet 21PGiacobini-Zinner for dynamic modelling of the Draconid meteoroid stream formationFrom Marsden amp Williams (2008) CMOR observations are extracted from the nine-meteor solution in Table 3

Name Perihelion q (au) e a (au) ω i

21P 2005 July 27605 10379 14 0705 691 3526 613 1725429 1954301 31810921P 1966 March 282844 0933 501 0729 394 3449 669 1729153 1966674 309382Sim 09968 plusmn 00003 0725 plusmn 0003 362 plusmn 004 17275plusmn009 19565plusmn004 3179plusmn002

Obs 09952 plusmn 00003 0710 plusmn 0032 322 plusmn 061 17226 plusmn 035 195620 plusmn 0001 3145 plusmn 069

density ρ = 300 kg mminus3 for Draconid meteoroids as was reportedby Borovicka et al (2007)

The comet is considered active (that is simulated meteoroids arereleased) when at a heliocentric distance of 3 au or less during thefirst simulation stage the simulated parent releases particles at a uni-form rate in this part of its orbit While active particles are releasedwith velocities from the prescription of either Crifo amp Rodionov(1997) or the revised Whipple model of Jones (1995) We use anucleus radius for 21PGiacobini-Zinner of 1000 m (Tancredi et al2000) but in the absence of other details we assume a Bond albedofor the nucleus of 005 a nucleus density of 300 kg mminus3 and an activefraction of the cometrsquos surface of 20 per cent where needed in theabove ejection models The Brown amp Jones (1995) model was foundto reproduce the duration of the outburst as observed by CMORslightly better and the results reported here use the Jones model

A supplementary integration of the comet orbit backwards for athousand years allowed a determination of the Lyapunov exponentusing the algorithm of Mikkola amp Innanen (1999) The e-foldingtime-scale is approximately 30 yr not unexpected for a Jupiter-family comet with a node near that giant planet Our primary resultthat the outburst originated from the 1966 perihelion passage ofthe parent comet is thus two e-folding times into the past and thusnot unduely affected by chaotic effects The short Lyapunov timeis a result of the frequent close encounters that 21P suffers withJupiter The one of most relevance here is the only close encounteroccurring in the 1966ndash2012 time frame a close approach to withinless than 16 Hill radii in 1969 This encounter strongly affects boththe parent and the meteoroids released during the 1966 perihelionpassage

5 C O M PA R I S O N O F SI M U L AT I O N SW I T H O B S E RVAT I O N S

The simulation output for the year 2012 is presented in Fig 12which shows the nodal intersection points of the simulated mete-oroids overplotted on the Earthrsquos orbit The dots shown representthose test particles which are within 002 au of Earthrsquos orbit andreaching that point within plusmn7 d of our planet being at the samesolar longitude These are colour-coded by the perihelion of theparent that released them The black points represent those whichare closest to intersecting the Earth within 13r = plusmn0002 au and13t = plusmn1 d these will be taken to constitute the simulated outburst

Only one perihelion passage has a significant population of par-ticles that meet the outburst criteria 1966 The location of the peakcorresponds to the time at which CMOR detected the outburstthough the nodal footprint is slightly off the Earthrsquos orbit this willbe discussed further below

51 Why was there no prediction of an outburst

The largest difference between the simulations and observationsis the fact that the simulated meteors approach but do not quite

intersect the Earthrsquos orbit This likely contributed to the absence oftheoretical predictions of the outburst Maslov3 did predict activityfor the 2012 Draconids at this time in particular in the radio meteorrange which he attributed to the 1959 and 1966 trails which oursimulations confirm

In order to investigate the question of why the simulated nodalfootprint of the 1966 streamlet does not intersect the Earthrsquos orbitwe performed additional simulations with higher ejection velocitiesto determine the conditions required for these meteoroids to reachthe Earth Additional ejection velocities around 50ndash100 m sminus1 areneeded to move the simulated 1966 stream to the Earthrsquos orbit cross-ing Particles in our model are ejected with a range of velocities Vejwhich depends on their size and heliocentric distance Particles from10minus5 to 10minus4 m have Vej = 210 plusmn 80 m sminus1 (average plusmn one standarddeviation) 10minus4 to 10minus3 m Vej = 66 plusmn 25 m sminus1 10minus3 to 10minus2 mVej = 21 plusmn 8 m sminus1 and 10minus2 to 10minus1 m Vej = 66 plusmn 25 m sminus1The three largest size ranges can be brought to Earth intersectionby increasing the ejection velocity to 50ndash100 m sminus1 while 30 m sminus1

proves just insufficient Thus the discrepancy in the nodal posi-tion (which amounts to roughly 0001 au or about half the distanceto the Moon) could be accounted for by relatively high ejectionvelocities (as Maslov also noted) However we note that the closeencounter with Jupiter suffered by 21PGiacobini-Zinner and itsdaughter particles in 1969 as well as uncertainties in the cometorbit itself [ie its non-gravitational parameters changed markedlyin the 1966ndash1972 time frame Yeomans (1986)] make it difficult topinpoint a single cause We conjecture that inevitable errors in theJupiter family cometrsquos orbit coupled with its frequent approachesto Jupiter produce uncertainties in the location of the streamrsquos nodelarge enough to confound attempts to model the shower accurately

Nonetheless we can conclude with some confidence that theoutburst observed by CMOR was produced by the 1966 trailetThis is because a comparison of the properties of these particles isconsistent to a high degree with the Draconid outburst observed byCMOR This includes the timing of the outburst its radiant velocityand the overall orbital elements which we present below

52 Timing and radiants

Fig 13 shows the timing of the shower The radar and simulationresults are offset slightly with the simulations showing activitypeaking about 003 of solar longitude (about 45 min) earlier thanwas observed though the precise width and peak of the simulatedshower is somewhat sensitive to our choice of 13r and 13t

The radiants are shown in Fig 14 The locations of the CMORradiant are indicated by the error bars The ellipse indicates onestandard deviation of the simulated radiant The observed and sim-ulated radiants thus have αg and δg which are consistent to within one

3 httpferajnarodruRadiantsPredictions1901-2100engDraconids1901-2100predenghtml

at University of W

estern Ontario on February 12 2014

httpmnrasoxfordjournalsorg

Dow

nloaded from

The unexpected 2012 Draconid meteor storm 3821

Figure 12 The nodal footprint of the Draconid meteor stream for 2012 from simulation The most populous perihelion passages are indicated by specificcolours all others are in grey Black is reserved for those particles which meet the most stringent encounter criteria and are deemed to comprise the simulatedshower Only the 1966 perihelion passage has a substantial number of such particles The timing of the observed CMOR radar peak is indicated by a dottedline See the text for more details

Figure 13 The timing of the Draconid outburst of 2012 The black his-togram shows the simulation the grey band indicates the times where CMORsaw significant shower activity and the white line indicates the radar peak

Figure 14 The simulated radiant The grey points indicate those particlespassing within plusmn002 au and plusmn7 d of Earth The black points indicate thosemeeting the most stringent criteria (13r = 0002 au 13t = plusmn1 d) and aredeemed most likely to have comprised the material producing the outburstThe green error bars indicate the CMOR radiant the red ellipse indicatesthe one standard deviation position of the simulated radiant

at University of W

estern Ontario on February 12 2014

httpmnrasoxfordjournalsorg

Dow

nloaded from

3822 Q-Z Ye et al

sigma The geocentric velocity (far from Earth) of simulated me-teors is 2093 plusmn 003 km sminus1 These agree with the radar-measuredvelocities within their respective uncertainties

53 The lsquoabsentrsquo storm for visual observers

The size distribution of the outburst in the simulation is concentratedin a narrow range of meteoroid sizes from about 01 to 07 mmcorresponding to a typical β asymp 5 times 10minus3 This size selection arisesessentially from the timing constraint Particles of different sizestend to have different arrival times at Earth even if released fromthe parent at the same time and with the same ejection velocitybecause a particlersquos orbital period is a function of β The need toarrive at the correct time to produce the outburst thus will restrict β

to a narrow range This size distribution is qualitatively consistentwith CMOR observations Because of the strong size selection forthis outburst the concept of a power-law distribution of sizes mustbe used with great care in this case

The orbital elements for the core of the simulated outburst arelisted in Table 5 Given the match in the timing and width of theoutburst as well as consistency between the radar-derived sizes andradiants and those of the simulations we conclude that the 2012Draconids outburst consisted primarily of particles released duringthe 1966 perihelion passage of comet 21PGiacobini-Zinner

The meteoroids which reach the Earth from the 1966 perihelionpassage show no concentration of release times in the pre- or post-perihelion orbital arc The β and ejection velocities in the simulationare shown in Fig 15 only a narrow segment of the particles ejectedfrom the comet reach the Earth during the 2012 outburst

The extensive compilation of historical comet observations ofKronk (1999) was examined to see if the 1966 perihelion passage of21PGiacobini-Zinner displayed any behaviour that might explainthe outburst The apparition was a faint one and poorly observedbut there are no indications of an outburst or other unusual dustproduction during its 1966 apparition

Figure 15 The phase space of β and ejection speed sampled by the simu-lation (grey) and those that are part of the simulated shower (black) demon-strating the reason for dynamical filtering of the outburst to a narrow rangeof meteoroid sizes

Table 6 Predicted future Draconid activity

Year Trail Start Peak End Max r (mm) Min r (mm) Offset

2018 1953 1952 1954 1956 100 04 Yes2019 1979 1940 1942 1945 2 lt01 No2021 1979 1935 1938 1945 025 lt01 No2025 1933 1940 1950 1953 10 03 No

54 Future outlook

We extended our simulations to determine whether any further out-bursts of the Draconids are likely between now and 2025 This pro-cess is fraught with difficulties since we have already concludedthat our knowledge of the parentrsquos orbit is insufficient to modeleven the observed 2012 outburst in full detail However the desireto better understand this peculiar shower leads us to the attemptnonetheless

Our earlier discussion of the 2012 outburst revealed insufficientaccuracy either in our knowledge of the cometrsquos dynamical parame-ters (ie orbital elements and non-gravitational parameters) or in ourability to simulate the cometrsquos evolution To attempt to get aroundthese restrictions we run simulations based on each of the sets ofosculating orbital elements listed in Marsden amp Williams (2008)for all 14 appearances of 21PGiacobini-Zinner from its discoveryin 1900 through 2005 inclusive Each is integrated backwards for100 yr and then forwards again producing dust at each perihelionpassage until the year 2025 The initial simulations and refinementstages are performed in all of these cases as described earlier In theabsence of observation uncertainty and the chaos produced by closeencounters with Jupiter these simulations would all produce thesame results Most ndash but not all ndash simulations produced similar pre-dictions However observational uncertainty and chaos are presentand thus each of the 14 orbits is only an approximation to the trueorbit at some instant in time and we find that shower predictionsare very sensitive to such small variations in the cometrsquos orbit Theresults reported here for future activity represent a somewhat sub-jective selection of those years where a majority of the simulationsproduce appreciable activity

To better understand the situation we first compared the simula-tions with known outbursts of the Draconids (2005 2011 2012) inthe 2001ndash2012 range Here we were heartened to be able to repro-duce the occurrence and absence of outbursts as observed with onecaveat that the simulated nodal footprint of the 2005 outburst wasdisplaced slightly outside the Earthrsquos orbit similarly to that of 2012(the 2011 outburst footprint did not require adjustment)

Given these results we provide here a list of years when thesimulations show the possibility of increased Draconids activitywhere many (though sometimes not all) of the nodal foot prints areon or near the Earthrsquos orbit These years are 2018 2019 2021 and2025 The strongest activity is expected from the year 2018 but itsnodal footprint is offset slightly from the Earth a precise activitylevel is hard to calculate but an outburst similar to 2012 is possibleThe other shower years are not offset from the Earthrsquos orbit butshow lower rates and smaller particle sizes (Table 6)

6 C O N C L U S I O N

We have reported the analysis of the unexpected intense outburst ofthe Draconid meteor shower detected by CMOR on 2012 October8 The peak occurred at sim1640 UT (λ = 195622) with a ZHRmax

asymp 9000 plusmn 1000 The weighted mean radiant around the peak time(15ndash19 h UT) was at αg = 2624 plusmn 01 δg = 557 plusmn 01 (J2000)

at University of W

estern Ontario on February 12 2014

httpmnrasoxfordjournalsorg

Dow

nloaded from

The unexpected 2012 Draconid meteor storm 3823

The mass distribution index was determined to be s = 188 plusmn 001lower than the value found in 2011 indicating that the outburstwas dominated by faint meteors compared to 2011 We then usedthe meteoroid ablation model developed by Campbell-Brown ampKoschny (2004) to model the structure of the Draconid meteoroidswhich does not reveal unique solutions but suggests that the numberof grains in each meteoroid falls within the range 10ndash100

Visual observations also showed increased activity around thepeak time but with a much lower activity than indicated by radar(ZHR sim 200) We interpret this to indicate a strong size selection inthe outburst centred near radar meteoroid sizes a result consistentwith simulations which predict a similar narrow range in smallparticle sizes The concept of a single power-law distribution inmeteoroid sizes may not be valid for this event spanning radar tovisual sizes indicating that ZHR and flux estimates are similarlyuncertain

Dynamical simulations of this event show with some confidencethat this outburst was originated from particles released during the1966 perihelion passage of the parent body 21PGiacobini-Zinneralthough there are some uncertainties of the exact timing of the en-counter Further simulations showed that possible increased activityof the Draconid meteor shower may occur in 2018 2019 2021 and2025

AC K N OW L E D G E M E N T S

We thank Dr David Asher for his very thorough comments andDr Rainer Arlt for his helps on the IMO visual data We alsothank Zbigniew Krzeminski Jason Gill and Daniel Wong for help-ing with CMOR analysis and operations Funding support fromthe NASA Meteoroid Environment Office (cooperative agreementNNX11AB76A) for CMOR operations is gratefully acknowledgedQY thanks 21PGiacobini-Zinner for giving him a totally unex-pected surprise which makes his study and research more cheerful

Additionally we thank the following visual observers for theirreports which are essential as a confirmation of this outburstBoris Badmaev Igor Bukva Jose Vicente Diaz Martinez Svet-lana Dondokova Francisco Jose Sevilla Lobato Ljubica GrasicIlija Ivanovic Javor Kac Julia Mangadaeva Olga MasheevaAleksandar Matic Jovana Milic Galina Muhutdinova FranciscoOcana Gonzalez Rolanda Ondar Vladimir Osorov Galina OttoSasha Prokofyev Maciej Reszelski Alejandro Sanchez De MiguelNikolay Solominskiy Jakub Koukal Michel Vandeputte SalvadorAguirre Jurgen Rendtel Terrence Ross Branislav Savic AlexandrMaidik Koen Miskotte Snezana Todorovic Sogto TsurendashievKristina Veljkovic Allen Zhong Milena Zivotic and BimbaZurbaev

R E F E R E N C E S

Blaauw R C Campbell-Brown M D Weryk R J 2011 MNRAS 4122033

Borovicka J Spurny P Koten P 2007 AampA 473 661Brown P Jones J 1995 Earth Moon Planets 68 223Brown P Weryk R J Wong D K Jones J 2008 Icarus 195 317

Brown P Ye Q 2012 Cent Bur Electron Telegrams 3249 1Campbell-Brown M D Koschny D 2004 AampA 418 751Campbell-Brown M Vaubaillon J Brown P Weryk R J Arlt R 2006

AampA 451 339Ceplecha Z Borovicka J Elford W G Revelle D O Hawkes R L

Porubcan V Simek M 1998 Space Sci Rev 84 327Cervera M A Elford W G Steel D I 1997 Radio Sci 32 805Crifo J F Rodionov A V 1997 Icarus 127 319Davidson R M 1915 J Br Astron Assoc 25 292Denning W F 1926 MNRAS 87 104Everhart E 1985 in Carusi A Valsecchi G B eds Dynamics of Comets

Their Origin and Evolution an Efficient Integrator that Uses Gauss-Radau Spacings Kluwer Dordrecht p 185

Hocking W K Fuller B Vandepeer B 2001 J Atmos Sol-Terr Phys63 155

Jacchia L G Kopal Z Millman P M 1950 ApJ 111 104Jenniskens P 1995 AampA 295 206Jenniskens P 2006 Meteor Showers and their Parent Comets Cambridge

Univ Press CambridgeJenniskens P Barentsen G Yrjola I 2011 Cent Bur Electron Telegrams

2862 1Jones J 1995 MNRAS 275 773Jones J Brown P Ellis K J Webster A R Campbell-Brown M Krze-

menski Z Weryk R J 2005 Planet Space Sci 53 413Kero J Fujiwara Y Abo M Szasz C Nakamura T 2012 MNRAS 424

1799Koschack R Rendtel J 1990 WGN J Int Meteor Organ 18 44Koten P Vaubaillon J Toth J Zenden J McAuliffe J Koschny D Pautet

D 2012 LPI Contrib 1667 6225Kronk G W 1999 Cometography Vol 1 Cambridge Univ Press Cam-

bridgeMarsden B G Sekanina Z Yeomans D K 1973 AJ 78 211Marsden B G Williams G V 2008 Catalogue of Cometary Orbits 17th

edn IAU Central Bureau for Astronomical Telegrams Minor PlanetCenter Cambridge MA

Maslov M 2011 WGN J Int Meteor Organ 39 64McIntosh B A 1968 in Kresak L Millman P M eds Proc IAU Symp

33 Physics and Dynamics of Meteors Reidel Dordrecht p 343McKinley D W R 1961 Meteor science and engineering McGraw-Hill

New YorkMikkola S Innanen K 1999 Celest Mech Dyn Astron 74 59Sato M Watanabe J Ohkawa T 2012 LPI Contrib 1667 6319Standish E M 1998 Technical Report Planetary and Lunar Ephemerides

DE405LE405 NASA Jet Propulsion Laboratory CaliforniaTancredi G Fernandez J A Rickman H Licandro J 2000 AampAS 146

73Vaubaillon J Watanabe J Sato M Horii S Koten P 2011 WGN J Int

Meteor Organ 39 59Vaubaillon J et al 2012 LPI Contrib 1667 6280Weidenschilling S J Jackson A A 1993 Icarus 104 244Weryk R J Brown P G 2012 Planet Space Sci 62 132Wiegert P Vaubaillon J Campbell-Brown M 2009 Icarus 201 295Wiegert P A Brown P G Weryk R J Wong D K 2013 AJ 145 70Ye Q Brown P G Campbell-Brown M D Weryk R J 2013 MNRAS

436 675Yeomans D K 1986 QJRAS 27 116

This paper has been typeset from a TEXLATEX file prepared by the author

at University of W

estern Ontario on February 12 2014

httpmnrasoxfordjournalsorg

Dow

nloaded from