A New Dark Vortex on Neptune

Abstract

An outburst of cloud activity on Neptune in 2015 led to speculation about whether the clouds were convective in nature, a wave phenomenon, or bright companions to an unseen dark vortex (similar to the Great Dark Spot studied in detail by Voyager 2). The Hubble Space Telescope (HST) finally answered this question by discovering a new dark vortex at 45 degrees south planetographic latitude, named SDS-2015 for "southern dark spot discovered in 2015." SDS-2015 is only the fifth dark vortex ever seen on Neptune. In this paper, we report on imaging of SDS-2015 using HST's Wide Field Camera 3 across four epochs: 2015 September, 2016 May, 2016 October, and 2017 October. We find that the size of SDS-2015 did not exceed 20 degrees of longitude, more than a factor of two smaller than the Voyager dark spots, but only slightly smaller than previous northern-hemisphere dark spots. A slow (1.7–2.5 deg/year) poleward drift was observed for the vortex. Properties of SDS-2015 and its surroundings suggest that the meridional wind shear may be twice as strong at the deep level of the vortex as it is at the level of cloud-tracked winds. Over the 2015–2017 period, the dark spot's contrast weakened from about to about , while companion clouds shifted from offset to centered, a similar evolution to some historical dark spots. The properties and evolution of SDS-2015 highlight the diversity of Neptune's dark spots and the need for faster cadence dark spot observations in the future.

Full text

A New Dark Vortex on Neptune
Michael H. Wong
1,11
, Joshua Tollefson
1,12
, Andrew I. Hsu
1,13
, Imke de Pater
1,11,12
, Amy A. Simon
2
, Ricardo Hueso
3
,
Agustín Sánchez-Lavega
3
, Lawrence Sromovsky
4
, Patrick Fry
4
, Statia Luszcz-Cook
5
, Heidi Hammel
6
, Marc Delcroix
7
,
Katherine de Kleer
8
, Glenn S. Orton
9
, and Christoph Baranec
10
1
University of California, Berkeley, CA 94720, USA; mikewong@astro.berkeley.edu
2
NASA Goddard Space Flight Center, 8800 Greenbelt Road, Greenbelt, MD 20771, USA
3
Universidad del Pais Vasco Bilbao, Spain
4
University of Wisconsin, Madison, WI 53706, USA
5
American Museum of Natural History and Columbia University, New York NY, USA
6
Association of Universities for Research in Astronomy, Washington DC 20004-1752, USA
7
Societe Astronomique de France, France
8
California Institute of Technology Pasadena CA 91109, USA
9
Jet Propulsion Laboratory, 4800 Oak Grove Drive, California Institute of Technology, Pasadena, CA, USA
10
Institute for Astronomy, University of Hawai‘iatMānoa, Hilo, HI 96720-2700, USA
Received 2017 November 21; revised 2017 December 22; accepted 2017 December 29; published 2018 February 15
Abstract
An outburst of cloud activity on Neptune in 2015 led to speculation about whether the clouds were convective in
nature, a wave phenomenon, or bright companions to an unseen dark vortex (similar to the Great Dark Spot studied
in detail by Voyager 2). The Hubble Space Telescope (HST)finally answered this question by discovering a new
dark vortex at 45 degrees south planetographic latitude, named SDS-2015 for “southern dark spot discovered in
2015.”SDS-2015 is only the fifth dark vortex ever seen on Neptune. In this paper, we report on imaging of SDS-
2015 using HST’s Wide Field Camera 3 across four epochs: 2015 September, 2016 May, 2016 October, and 2017
October. We find that the size of SDS-2015 did not exceed 20 degrees of longitude, more than a factor of two
smaller than the Voyager dark spots, but only slightly smaller than previous northern-hemisphere dark spots. A
slow (1.7–2.5 deg/year)poleward drift was observed for the vortex. Properties of SDS-2015 and its surroundings
suggest that the meridional wind shear may be twice as strong at the deep level of the vortex as it is at the level of
cloud-tracked winds. Over the 2015–2017 period, the dark spot’s contrast weakened from about
-
7% to about
-
3%, while companion clouds shifted from offset to centered, a similar evolution to some historical dark spots.
The properties and evolution of SDS-2015 highlight the diversity of Neptune’s dark spots and the need for faster
cadence dark spot observations in the future.
Key words: astrometry –hydrodynamics –planets and satellites: atmospheres –planets and satellites: gaseous
planets –planets and satellites: individual (Neptune)–techniques: high angular resolution
1. Introduction
Voyager 2 provided the first high-resolution glimpse of
Neptune in 1989, observing two dark spots. We assume these
features are anticyclonic vortices based on their behavior
(Smith et al. 1989), although there has never been an actual
measurement of their internal circulation. Since Voyager, only
Hubble has discovered dark vortices: two in the 1990s, and one
in 2015 (Table 1). Ground-based facilities lack the resolution to
detect these low-contrast features at blue optical wavelengths,
while infrared observations do not detect the dark spots
themselves, only their bright companion features. This is not
quite the case for Uranus, where we have seen dark features at
red wavelengths in Hubble images (Hammel et al. 2009)and at
near-IR wavelengths in Keck H-filter images (Sromovsky
et al. 2015). The five Neptune dark spots exhibited surprising
diversity, in terms of size, shape, companion cloud distribution,
oscillatory behavior, meridional drift rates, and meandering.
Neptune’s dark vortices come and go on much shorter
timescales compared to similar anticyclones on Jupiter, which
evolve over decades (e.g., Ingersoll et al. 2007; Shetty &
Marcus 2010; Simon et al. 2014). Many questions remain as to
how dark vortices originate, what controls their drift and
oscillation, how they interact with the environment, and how
they eventually dissipate. Vortex behavior also provides insight
into the structure and dynamics of the surrounding atmosphere.
The discovery of SDS-2015 was enabled by the annual
cadence of the Outer Planet Atmospheres Legacy (OPAL)
program, which acquires maps covering two consecutive
rotations of each giant planet, annually. The Neptune
observations are part of OPAL’s large time-domain survey of
atmospheric evolution (Simon et al. 2015), which began in
2014 with observations of Uranus (Irwin et al. 2017).
Companion clouds are usually associated with dark vortices
and are thought to result from ambient air being diverted
upward as flow is perturbed over the vortex (Smith et al. 1989;
Stratman et al. 2001). The appearance and evolution of
companion clouds has been different for all of the dark
vortices observed to date, with SDS-2015 (Figure 1)having
persistent companions that may exhibit longitudinal oscilla-
tions in their mean position (Hueso et al. 2017). The brightness
of companion clouds may be related to the vortex top altitude; a
The Astronomical Journal, 155:117 (9pp), 2018 March https://doi.org/10.3847/1538-3881/aaa6d6
© 2018. The American Astronomical Society.
11
Astronomy Department.
12
Earth and Planetary Sciences Department.
13
Physics Department.
Original content from this work may be used under the terms
of the Creative Commons Attribution 3.0 licence. Any further
distribution of this work must maintain attribution to the author(s)and the title
of the work, journal citation and DOI.
1

global circulation model of the GDS (Stratman et al. 2001)
found that cloud opacity weakened as the top of the vortex
reached higher into the tropopause region. Their deepest model
vortex had the highest companion cloud opacity.
The Voyager observations revealed many different types of
oscillations and drifts in the behavior of dark vortices GDS and
DS2 (Sromovsky et al. 1993). The center of the GDS oscillated
in longitude, and also drifted slowly equatorward, with some
intermittent poleward meandering. It also exhibited a “rolling”
appearance, where the aspect ratio oscillated in phase with a
rotation of the ellipse axes. LeBeau & Dowling (1998)modeled
these oscillations and identified a common physical mechanism
explaining the drift of both terrestrial hurricanes, and the GDS
on Neptune. In both cases, the spinning of the vortices advects
the background atmospheric potential vorticity, creating a weak
dipole of vorticity aligned approximately north–south. LeBeau
& Dowling (1998)postulated the destruction of these vortices,
as they drift equatorward and experience either Rossby-wave
radiation, or loss of material as they merge with regions of
similar potential vorticity. The actual origin or dissipation of a
Neptunian dark vortex has never been conclusively documen-
ted. Discovery dates are listed in Table 1because origin dates
are unknown. Likewise, the demise or dissipation of these
vortices has never been precisely determined. End dates in the
table are estimates based on extrapolated positions, the
disappearance of bright companions, and absence in subse-
quent observations.
The meridional drift of vortices acts as a probe of Neptune’s
zonal flow, which is puzzling in many ways (Limaye &
Sromovsky 1991; Sromovsky et al. 1995; Hammel &
Lockwood 1997; Sromovsky et al. 2001; Fitzpatrick et al.
2014; Sánchez-Lavega et al. 2018; Tollefson et al. 2018).
Neptune has a single retrograde equatorial jet with a speed of
∼400ms
−1
, and two prograde circumpolar jets with speeds
around 250ms
−1
. Dispersions above 200ms
−1
have been
measured between cloud features at the same latitude on
Neptune. This observed velocity dispersion may be explained
by temporal variability in the wind field, vertical wind shear,
rapid evolution of cloud morphology, and/or longitudinal
variability. Wave propagation, rather than mass transport, may
also determine the observed motion of cloud features.
Persistent, well-separated features at different latitudes have
been observed to move at a common zonal rate, invisibly
linked (Sromovsky et al. 2001; Martin et al. 2012). LeBeau &
Dowling (1998)found that the rate of north–south drift of the
vortex is very sensitive to the background potential vorticity
gradient (i.e., the horizontal wind shear). Thus, measuring
meridional drift may reveal fine structure in the zonal wind
profile that cannot be derived by cloud-tracking, while
oscillation may reveal meandering in the background jet
(Karkoschka 2011).
Here, we report the size, shape, motion, and contrast of SDS-
2015, based on observations spanning 2015–2017.
2. Observations
Observations reported here were taken with the UVIS
detector of the WFC3 instrument aboard the Hubble Space
Telescope (HST). Spectral responses of WFC3/UVIS filters
listed here are given in Dressel (2016).
2.1. Cadence
Table 2lists four epochs of observations of the dark vortex.
In Figure 1, we show images and maps of Neptune at each
epoch. The original discovery of SDS-2015 in OPAL data led
to a supplementary Midcycle program, focusing on the blue
wavelengths where the vortex itself is most evident. Only
16 days before the OPAL 2015 observations, WFC3 imaging
of Neptune taken in support of VLA observations (GO-14044)
barely missed the dark vortex. Companion clouds were seen,
but the exposures were acquired some 20–30 minutes before
SDS-2015 rotated onto the disk.
2.2. Detector Readout
We used 512×512 subarrays to minimize read time and
instrument memory transfer overhead (buffer dumps), except
for exposures using the methane-band quad filters (FQ727N,
FQ619N), which generate 2K×2K subarray frames, like all
quad-filter exposures. Rather than pairing FQ727N methane-
band observations with the FQ750N continuum filter, we used
F763M. This filter’s wider bandpass includes both continuum
and CH
4
-band contributions, preserving information content
(and vertical resolution)while still permitting the advantageous
512×512 subarray readout. Although visual comparison of
FQ727N and FQ750N images is more straightforward, using
F763M enables more time-on-target during an HST orbit.
2.3. Photometric Calibration
For the FQ727N frames, we applied corrections to remove
photometric errors from fringing, an effect expected to be in the
2%–5% range based on analysis of flat field data (Wong 2010).
The actual magnitude of the correction across the regions of the
Table 1
Characteristics of Neptune’s Dark Vortices
Size
Name Discovery Demise Latitude
a
Lon.×Lat. Meridional Drift References
GDS 1989 1990 -20 38°×15°+0°.11/day equatorward Smith et al. (1989)
DS2 1989 <1994 -55 39°×6°±2°.4 oscillation Sromovsky et al. (1993)
NDS-1994 1994 1998–2000 +32 35°×10°0°/day Hammel et al. (1995), Sromovsky et al. (2001)
NDS-1996 1996 1997–1998 +15 22°×12°0°/day Sromovsky et al. (2001)
SDS-2015 2015 >2017 -49 15°×5°-2.5/year poleward Wong et al. (2016)
Note. Latitude, size, and meridional drift rates are representative values. Actual values were known to vary in time, except for meridional drift of NDS-1994 and NDS-
1996, which were stable in latitude. The actual demise of a dark vortex has never been observed, so values listed here are estimates (see the text). The northern spots
have been renamed in this paper to establish a consistent nomenclature for post-Voyager dark spots.
a
Planetographic latitude is used throughout this work.
2
The Astronomical Journal, 155:117 (9pp), 2018 March Wong et al.

detector where Neptune was observed was 0.5%–1.5%, a
smaller effect than predicted. This is because part of the overall
2%–5% fringing amplitude in this filter bandpass is corrected
by the pipeline flat fields (obtained using a continuum light
source), leaving just the error due to the difference in spectral
energy distribution between the calibration source and
Neptune’s scattered solar spectrum. We also minimized the
fringing amplitude by placing Neptune on an area of the
detector where fringing is relatively smooth over the 50 pixel
size scale of Neptune’s disk. In some areas, fringing maxima/
minima are separated by only 10 pixels.
In the 2015–2017 time range, the WFC3 instrument team at
Space Telescope Science Institute implemented a set of
significant revisions to their photometric calibration pipeline,
collectively termed “UVIS 2.0.”The revisions included new flat
fields and normalization procedures (Mack et al. 2016), and new
sensitivity values for each filter (Deustua et al. 2016,2017).All
Neptune data presented here, although acquired by HST prior to
Figure 1. Gallery of SDS-2015 images and maps. [A, B, C, D]Red/green/blue (RGB)composite images at four epochs (see Table 2), co-added from multiple frames
using the frydrizzle rotational dithering approach described in Section 2.5.[E, F, G, H]RGB cylindrical maps of the dark vortex at each epoch. [I, J, K, L]Blue
(F467M)filter images show the dark vortex itself most clearly. The brightness scale is the same for all epochs. The F763M data (red channel in color maps)span I/F
0.08 to 0.35 on a square root stretch. The F547M data (green channel in color maps)span I/F 0.48 to 0.60 on a linear stretch. The F467M data (blue channel in color
maps, greyscale in panels I–L)span I/F 0.63 to 0.76 on a linear stretch (see colorbar).
3
The Astronomical Journal, 155:117 (9pp), 2018 March Wong et al.

these revisions, have been recalibrated using the UVIS 2.0
photometry. The UVIS 2.0 improvements have not yet been
implemented for WFC3/UVIS quad filters.
2.4. Navigation and High-level Science Products (HLSPs)
We navigated the images by aligning synthetic limb-
darkened planetary disks to the data (Lii et al. 2010), giving
a navigation precision of 0.0825 pixels. We estimate a
navigation accuracy of about 0.18 pixels for the F467M filter
images, leading to disk-center latitude/longitude uncertainties
of about 0°. 3. Global maps of Neptune acquired by the OPAL
program are available as HLSPs on the MAST archive,
14
as
described in Simon et al. (2015,2016).
Navigated single-frame image data for all programs listed in
Table 2can also be obtained from MAST
15
as HLSPs ending in
the extension “nav.fits.”Each file is a multi-extension FITS
file, similar to HST “drz.fits”files but with additional
processing and data. The processing consists of corrections
for cosmic ray hits using the LA Cosmic method (van
Dokkum 2001)and for fringing in long-wavelength narrow-
band filters (Wong 2011). The additional data includes FITS
keywords containing target ephemeris data provided by JPL
Horizons,
16
and four FITS extensions giving the planeto-
graphic latitude, west longitude, emission angle, and solar
incidence angle for each pixel on the planet’s disk. In addition
to the STScI-supplied PHOTFLAM inverse sensitivity key-
word that converts image data units from e
−
s
−1
to
ergcm
−2
s
−1
Å
−1
, we include a PHOTIF keyword that
converts e
−
s
−1
to planetary reflectivity units of I/F, and an
associated SIG_PHOT keyword giving the photometric
uncertainty in the I/Fcalibration.
2.5. Rotational Dithering
Relative to OPAL program observations, the midcycle
program was intended to improve image quality and short-
wavelength coverage, specifically focusing on the dark vortex.
Short-wavelength coverage was improved by adding blue
filters and dropping two longer wavelength filters, and image
quality was improved by taking a larger number of frames in
the blue filters. We then co-add the map data, accounting for
the planet’s rotation and the longitudinal drift of the dark spot.
Astronomical images of static targets can be improved by
combining individual frames with drizzle methods (resampling
and co-adding), which has the added benefit of being able to
reject cosmic ray events and other image defects (e.g., Fruchter
& Hook 2002). We would also like to co-add Neptune images,
to improve sampling of the HST/WFC3 PSF and to average
over pixels with nonuniform response, but, each exposure in
our program differs due to Neptune’s rotation as well as
advection by atmospheric winds. In order to combine frames,
additional steps must be taken, following an algorithm known
informally as frydrizzle. The frydrizzle algorithm is fully
described in Section 3.1 of Fry et al. (2012). The principal steps
of our slightly modified procedure (we add limb-darkening
treatment that was not necessary at the wavelengths used in Fry
et al. 2012)are:
1. Navigate each frame to associate each element of the
image data with the appropriate latitude, longitude, and
emission/incidence angles.
2. Apply a limb-darkening correction to normalize reflec-
tances to a common viewing geometry.
3. If images separated by a significant amount of time, shift
(or “advect”)each latitude using an a-priori zonal wind
field, to correct for differential rotation.
4. Co-add data in latitude/longitude space rather than image
x,yspace.
5. If creating a co-added disk image on the sky (rather than
co-added map in longitude-latitude space), then restore
limb-darkening effects and re-map reflectances onto the
globe.
3. Results
3.1. Morphology
Measurements of the size and location of SDS-2015 are
based on albedo contrasts inside and outside the vortex, which
appears as a 5%-level contrast feature at blue wavelengths. We
estimate the boundary of the vortex to be located where the
contrast reaches its half-power point, and accurate measure-
ment of this location relies on determining the effect of limb
darkening, to distinguish between the intrinsic albedo contrast
of the vortex itself, and apparent contrasts due to variation in
emission angle across the span of the dark spot.
Table 2
HST/WFC3 Observations of Neptune in the SDS-2015 era
Times Program Filters
(UTC)ID (number of frames)
2015.71
(2015 Sep 18 16:48 to 2015 Sep 19 06:19)
OPAL GO-13937
(PI Simon)
F467M (4), F547M (4), F657N (4), F763M (4),
F845M (8), FQ727N (4), FQ619N (4)
2016.37
(2016 May 16 03:47 to 04:16)
Midcycle GO-14492
(PI Wong)
F336W (1), F390M (1), F410M (1), F467M (6),
F547M (6), F763M (3), FQ727N (2)
2016.76
(2016 Oct 03 11:04 to 2016 Oct 04 04:59)
OPAL GO-14334
(PI Simon)
F467M (4), F547M (4), F657N (4), F763M (4),
F845M (8), FQ727N (4), FQ619N (4)
2017.76
(2017 Oct 06 05:49 to 2017 Oct 07 03:15)
OPAL GO-14756
(PI Simon)
F467M (8), F547M (8), F657N (8), F763M (8),
F845M (16), FQ727N (8), FQ619N (8)
Note.Observing programs listed here mapped Neptune over at least one full rotation, but times listed apply only to exposures in which SDS-2015 is on the face of the
disk. Additional data from program GO-14044 (PI de Pater), 2015.67, are not listed but were used for background atmosphere analyses (e.g., Table 3).
14
OPAL archive page: https://archive.stsci.edu/prepds/opal,10.17909/T9G593.
15
Dark vortex archive page: https://archive.stsci.edu/prepds/sds-2015,10.17909/
T9G67P.
16
JPL Horizons server: http://ssd.jpl.nasa.gov/?horizons.
4
The Astronomical Journal, 155:117 (9pp), 2018 March Wong et al.

We use the Minnaert equation to convert observed
I
F
obs. to
limb-darkening corrected
I
F
corr.:
mm=--()
()
IF IF ,1
kk
corr. obs. 0
1
where
m0
is the cosine of the incidence angle and μis the
cosine of the emission angle.
I
F
corr. can also be called the
disk-center I/F, or the Minnaert albedo. This limb-darkening
function was originally used to describe lunar reflectivity
(Minnaert 1941). Although simple functions of this type do not
perfectly capture limb-darkening behavior of Neptune’s
vertically and horizontally inhomogeneous atmosphere, we
still use Equation (1)to achieve a first-order brightness
correction, especially at emission/incidence angles less than
75°. For each wavelength, we derived the appropriate value of
kby fitting to data at longitudes excluding the dark vortex. Best
fit parameters based on four data sets (two in 2015 and two in
2016)are listed in Table 3.
Maps in Figure 1show reflectivity with this first-order limb-
darkening correction applied. East–west cuts through these
maps are shown in Figure 2. These photometric profiles, from
co-added F467M maps, were used to estimate widths and
positions of SDS-2015 at the four observational epochs, with
results listed in Table 4.
The largest source of uncertainty in the size and position
estimates of Table 4comes from the influence of the
omnipresent companion clouds. Comparison of the RGB and
F467M maps in Figure 1consistently show reduced contrast
within the vortex at locations where companion cloud opacity
is high. Thus, we have defined error bars that encompass the
directly observed vortex bounds (half-power points in contrast
profiles like those in Figure 2)and the potentially larger size of
the vortex including a qualitative compensation for vortex area
masked by the companion cloud.
3.2. Contrast
The contrast of the vortex with respect to the surroundings
was determined for wavelengths where the dark spot is visible,
and reported in Figure 3. The contrast of the vortex in F467M
has been decreasing since its discovery, with a very low
contrast detected in the 2017 observations.
Contrasts were determined using the darkest location in the
vortex (not necessarily the center, where companion cloud
reflectivity was sometimes significant). At wavelengths longer
than 547 nm, the dark spot is actually brighter than its
surroundings, because the contrast is dominated by aerosols
associated with the companion cloud. The relative background
at the darkest location was found by linear interpolation, fitted
to background levels outside the vortex. As with Sromovsky
et al. (2002),wefind maximum contrast in the F467M
bandpass, although the SDS-2015 contrasts are consistently
larger than most contrast measurements of NDS-1994 and
NDS-1996 reported by Sromovsky et al. (2002).
3.3. Drift Rates
Over the time that SDS-2015 has been observed, it drifted
polewards until it reached
-
49 S(Figure 4). There is no large
statistical difference between two models of its meridional
drift: (1)a linear poleward drift of 
2
.5
/year in 2015–2016,
with no change between 2016.76 and 2017.76; or (2)a
poleward drift of

1.7
/year over the full period of the
observations. Model 1 (3-point fit)gives a drift rate very close
to the change in companion cloud position of ~
2.4/year based
on all 2015 measurements in Hueso et al. (2017), specifically
their Figure 19.
Longitudinal drift rates for SDS-2015 and its companion
clouds are listed in Table 5. Data are derived from Hueso et al.
(2017)for the entire 2015B period (2015.51–2015.95), and
from OPAL data for the remaining three epochs. The 2016.37
epoch covered only a single Neptune rotation, so the temporal
baseline was not long enough to measure longitudinal drifts
with any certainty.
Zonal mean velocities in Table 5, from the 6th-order
polynomial fit of Sromovsky et al. (1993), were measured by
tracking discrete cloud features, presumably composed of CH
4
,
located at pressures of 1.4 bar or less (de Pater et al. 2014).
Dark vortex companion clouds are also thought to be CH
4
Table 3
Limb Darkening Parameters Used for Image Analysis
HST WFC3/UVIS Filter k(global)k(35°–55°S)
F336W 0.804±0.011 0.804±0.012
F390M 0.825±0.005 0.807±0.009
F410M 0.836±0.007 0.813±0.017
F467M 0.867±0.016 0.823±0.026
F547M 0.798±0.041 0.765±0.074
Note. Limb darkening is modeled using the Minnaert equation (Equation (1)).
Values of kwere found by linearizing image data where emission and
incidence angles <75°, and uncertainties reflect scatter from instrumental noise
and from actual spatial/temporal variation. Data from epochs 2015.67,
2015.71, 2016.37, and 2016.76 were used to derive limb-darkening
coefficients.
Figure 2. Photometric cuts through SDS-2015 used for size, position, and
contrast measurements. Top panel: east–west cuts are shown in the highest-
contrast filter (F467M)at the four epochs. Vortex center positions are as listed
in Table 4. Bottom panel: east–west cuts for multiple filters are shown, for the
2016.37 epoch. The latitude of these cuts is slightly to the north of the
estimated dark spot center (compare F467M trace with that in top panel, and
refer to map in Figure 1(J). This was done to measure the maximum contrast at
other wavelengths where the companion cloud contribution diminishes contrast
at the center of the vortex. Limb darkening corrections have been applied.
5
The Astronomical Journal, 155:117 (9pp), 2018 March Wong et al.

clouds (Smith et al. 1989). Since the companion clouds are
within the same altitude range of clouds used to derive zonal
wind profiles, it might be expected that they would drift at the
mean zonal speed, but Table 5shows that they do not. This can
be explained by their dynamical connection to the dark vortex,
which itself had a 20–30ms
−1
velocity offset from the
zonal mean.
Measurement of companion cloud drift rates was done by
manually tracking the center of light in individual images and
dividing by image offset times to obtain velocities. For the
2016 and 2017 dark spot data, drift rates were determined by
shifting longitudinal photometric profiles (as in Figure 2)using
different rates, until the half-power points in the profiles were
best aligned by eye. This method could not be used for the
2015.71 data, because SDS-2015 was positioned so close to the
limb in some data frames that part of the vortex became
indistinguishable from the surrounding atmosphere. To com-
pensate for this effect, we manually determined the vortex
position in each frame in the 2015.71 epoch, using WinJUPOS
software.
17
For all measurements using the HST data, we
estimate an uncertainty of about 10ms
−1
, including error
contributions from image navigation uncertainty, noise in the
image, and feature centroiding.
Table 4
Morphology of SDS-2015
Epoch Longitudinal Width (a
2
)Latitudinal Width (b
2
)Central Longitude (
l
0)Central Latitude (f0)Aspect Ratio (a/b)
2015.71 16°.8±4°.0 (5120 ±1230 km)5°.5±0°.6 (2300 ±270 km)9°.1±1°.0 45°.6±0°. 2 S 2.2±0.6
2016.37 19°.5±1°.5 (5770 ±440 km)6°.4±3°.8 (2690 ±1580 km)140°.3±3°.3 46°.3±1°. 0 S 2.1±1.3
2016.76 11°.3±2°.7 (3190 ±770 km)5°.8±3°.1 (2450 ±1310 km)142°.5±1°.1 48°.9±0°. 9 S 1.3±0.8
2017.76 13°.0±2°.7 (3720 ±770 km)3°.8±0°.5 (1580 ±210 km)134°.8±0°.9 49°.0±0°. 2 S 2.4±
0.6
Note. Centers and widths are determined by measuring half-power points in brightness profiles (Figure 2top)in the F467M data. Multiple images are combined per
epoch to achieve higher signal to noise; this requires combining data after correction for limb darkening (using Equation (1)and values in Table 3). Aspect ratios are
calculated using linear dimensions (km)rather than longitude/latitude. Uncertainties include the potential obscuration of part of the dark spot by the bright companion
clouds (see last paragraph of Section 3.1).
Figure 3. Spectral contrast between SDS-2015 and its surroundings. The
contrast Cis defined as, =-()(
)C
IF IF IF
DS surr. surr. , where
I
FDS is
defined at
l
0,f0(see Table 4), and
I
F
surr. is a linear fit to data at longitudes
well separated from the spot, at the same latitude of the vortex. Error bars
include random noise, systematic photometric uncertainty, and spatial variation
across the dark spot. Data points are averages of multiple frames (Table 2).
Data points for F467M and F547M data are horizontally offset for clarity, but
in all cases the horizontal error bars give the filter widths taken from Table 6.2
of Dressel (2016).
Figure 4. Poleward drift of SDS-2015. A linear drift rate of -2.5/year fits
SDS-2015 positions in the 2015–2016 time period (with no change in latitude
between 2016.76 and 2017.76). A drift rate of -1.7/year fits SDS-2015
positions over the full 2015–2017 period. Neither fit is strongly favored over
the other statistically: cc==
nn
() ()3 points 1.5; 4 points 1.6
22
.
Table 5
Eastward Drift Rates
Eastward Velocity (ms
−1
)
Date Lat. (deg)Zonal Mean SDS-2015 Companions
2015B −40.1 −142 L−94
2015.71 −45.6 −69 −90 L
−41.0 −131 L−95
2016.76 −48.9 −22 −40 L
−46.5 −56 L−42
2017.76 −49.0 −20 −39 L
−47.5 −42 L−35
Note. The first epoch of measurement, 2015B, gives the drift rate of the
companion cloud determined from ground-based and HST observations
spanning the 2015.51–2015.95 time range from Hueso et al. (2017). Other
epochs are based on two consecutive Neptune rotations, as described in
Table 2. The formal uncertainty of the 2015B drift rate is only 0.1ms
−1
, due
to the long temporal baseline and high density of measurements. Uncertainties
in HST-derived drift rates are about 10ms
−1
. Zonal mean velocities are
calculated using the sixth order polynomial profile from Sromovsky
et al. (1993).
17
WinJUPOS is available from http://jupos.org.
6
The Astronomical Journal, 155:117 (9pp), 2018 March Wong et al.

4. Discussion
4.1. Drifts and Oscillations
Dark vortices on Neptune have exhibited a range of
meridional motions (Table 1), but SDS-2015 is the only one
seen to drift poleward. Drift rates of SDS-2015 (Figure 4)are
about an order of magnitude smaller than the equatorward drift
rate of the GDS seen by Voyager 2, and two orders of
magnitude smaller than the oscillatory drift of DS2 (Hammel
et al. 1989; Sromovsky et al. 1993). Our observations cannot
distinguish between several descriptions of the meridional
motion of SDS-2015: poleward drift over the full 2015–2017
period, poleward drift from 2015 to 2016 followed by rest at
49°S, meandering, or poorly sampled oscillation. Latitudinal
oscillation in the dark spot, if any, is not reflected in the
behavior of bright companion clouds (Hueso et al. 2017),to
within a precision of 1°. 5, over the 2015 July–December period.
The latitudinal stability of the companion clouds is robust,
whether measurements are limited to large telescopes (HST,
Keck, Lick, Calar Alto, Palomar)or combined with more
frequently obtained amateur data. Most anticyclones in the
vortex simulations of LeBeau & Dowling (1998)drifted
equatorward, although some remained at a constant latitude
and some experienced periods of poleward drift.
Longitudinal drift rates were measured separately for SDS-
2015 and for its companion clouds (Table 5). At all epochs, the
vortex system moved at a different rate compared to the zonal
mean, and the companion clouds shared a common long-
itudinal drift rate with the dark spot.
Oscillations in shape were one of the most remarkable
findings regarding the GDS (Smith et al. 1989).Voyager
imaging sequences showed the GDS aspect ratio to vary by
about a factor of 3 over a 290 hr period (assuming sinusoidal
behavior). Table 4lists aspect ratios for the shape of SDS-2015.
The aspect ratio values lack the precision or temporal
resolution that would be needed to demonstrate any oscillation
in shape on such timescales. One significant limitation to
determining the shape of the spot is set by the companion
cloud, which locally reduces the photometric contrast needed to
establish the dark spot boundary (Figure 1). A constant aspect
ratio of 2 is consistent with the 2015–2017 measurements.
4.2. Implications for the Background Wind Field
As discussed in Section 1, Neptune’s wind field may have
complexities beyond the smooth Voyager profile of Sromovsky
et al. (1993). The behavior and properties of SDS-2015 and its
surroundings can provide insights into these complexities.
In the meridional direction, SDS-2015 originated in a dark
band centered near
-
45 latitude, then drifted southward
toward a bright band centered near
-
55 . No relationship
between these bands and the zonal wind structure has ever been
established. Albedo patterns are variable on decadal timescales
at these latitudes. In 1989, Voyager found a broad dark band
from 45°to 65°S(Smith et al. 1989). In 2003, HST/STIS
observations instead found a narrower dark band centered at
60°S(Karkoschka & Tomasko 2011); the 45°–55°S range had
grown bright. Twin dark bands are present in the 2015–2017
epoch, centered at 45°S and 60°S, with an intervening bright
band at 51°S(Figure 1). To summarize, the dynamic area on
Neptune is bounded near 65°S and 40°S in all three epochs,
although the patterns of bright and dark between these bounds
varied over time.
Jupiter has many alternating bright and dark bands, along
with alternating eastward and westward jets. The jets are
largely stable in position and magnitude, while coloration is
variable and sometimes reverses (e.g., Rogers 1995; Fletcher
et al. 2011; Tollefson et al. 2017b). Future research may
determine whether any common processes operate on Neptune
and Jupiter, where horizontal bands may remain stable in
position, but be variable in color/brightness. Zonal winds on
Jupiter seem to set up fixed boundaries for light and dark
bands, although the winds cannot be the only control on
coloration because band colors sometimes change even while
winds remain stable. In Neptune’s case, the smooth Voyager
zonal wind profile provides no dynamical boundaries to
constrain the dark bands evident in Figure 1, but there may
be fine structure not currently resolved by the observations.
Vertical wind shear is expected to be present on Neptune,
given the vertical and latitudinal variation in temperatures
derived from Voyager data (Conrath et al. 1991; Fletcher
et al. 2014). Observationally, the dispersion among zonal
speeds of individual cloud tracers in ground-based imaging
data (Fitzpatrick et al. 2014)may be a result of vertical wind
shear. Dark vortices lie deeper in the atmosphere than other
cloud tracers (Tollefson et al. 2017a), so their behavior may
reveal differences between the deeper wind field and the zonal
wind field based on Voyager observations of bright cloud
tracers.
Although flow within the vortex cannot be directly
measured, we can estimate the tangential velocity Vof the
vortex, based on assumptions regarding its vorticity. The spot
resides in a latitude with anticyclonic vorticity. The relative
vorticity z==-´
-
du dy 3.4 10
5
s
−1
at the latitude of
SDS-2015. The Coriolis parameter =- ´ -
f1.6 10 4s
−1
,so
the absolute vorticity z+=- ´ -
f1.9 10 4s
−1
. From Jupiter
and Saturn we know that typically the intrinsic vorticity z0of
vortices is halfway between ζand f(Sromovsky et al. 1993;
Legarreta and Sánchez-Lavega 2005; García-Melendo et al.
2007), so a rough estimate is z=- -
10
04s
−1
.Defining for an
elliptical circuit z=VL
A
e
0, where L
e
is the length of the
ellipse (p=--(())
L
ae Oe21 4
e24
),eis the eccentricity,
and p=
A
ab is the area. Then,
z
=-=- -
() ()Vb
e21 4 68 m s , 2
0
2
1
using b=1100 km and =- =eba10.87
22 as typical
values from Table 4.
Some time ago it was proposed that Neptune’s dark spots
could be Kida type vortices (Polvani et al. 1990). A stationary
Kida vortex obeys the relationship
zz l
ll
=-
+() ()
1
1,3
0
where
l
=ba
is the reciprocal aspect ratio. Taking
l
=0.5 as
a representative value from Table 4,zz =0.67
0for a Kida
vortex, which is greater than the value of zz =0.34
0derived
above from the zonal flow and assumed intrinsic vorticity
values.
The disagreement between these two estimates of
zz
0may
carry information regarding the deeper wind field to which the
dark vortex may be sensitive. Specifically, our first estimate of
zz
0relied on the value of
∣
∣du dy at the level of the cloud-
tracked winds. But if SDS-2015 resides at a deep level where
7
The Astronomical Journal, 155:117 (9pp), 2018 March Wong et al.

=´
∣
∣∣∣du dy du dy2
deep Voy., then the two methods for
estimating
zz
0would be exactly consistent.
4.3. The Demise of Dark Spots
An early goal of our observing program had been to check
for equatorward drift of the vortex, in the hope of capturing
widespread equatorial cloud activity resulting from the
destruction of the vortex (LeBeau & Dowling 1998). The
equatorward drift of the GDS may have ended in such an event,
but there was no capability in 1990 to conduct observations
with the sensitivity and resolution to detect it. The evolution of
“Berg”on Uranus may have been an example of this mode of
vortex demise (de Pater et al. 2011). If not for the annual
cadence of OPAL observations, it might be thought that the
eruption of widespread equatorial cloud activity in June–July
2017 (Molter et al. 2017)could potentially be related to the
violent demise of SDS-2015.
But other dark spots did not seem to drift equatorward
(Table 1). DS2 oscillated in latitude but had no known overall
meridional drift; the end of this vortex was completely
unconstrained. The northern vortices NDS-1994 and NDS-
1996 appeared to have zero meridional drift (Sromovsky
et al. 2001). These features eventually disappeared several
years after discovery, but NDS-1994 curiously became
invisible, while companion clouds at the same latitude
remained for some time. The OPAL 2017 data show a
persistent compact cloud at
-

47 . 5 latitude, nearly centered
over a barely discernible dark spot with a much lower contrast
than at the earlier times.
It seems that three out of the five known dark vortices (DS2,
NDS-1994, and SDS-2015)are ending in the same way: with
centered companion clouds. Companion clouds can make it
harder to see the underlying vortex at blue wavelengths at HST
resolution. At high resolution, Voyager images showed that
unlike the GDS, DS2ʼs companion clouds did not form at the
periphery of the spot, but over the center. It would be
interesting to model these data by degrading the Voyager
images to HST resolution and assessing the effect of the
centered companion clouds on the measured contrast of DS2.
The endings of NDS-1994 and NDS-1996 were not
conclusively witnessed. Persistent cloud features remained at
the latitude of NDS-1994, even after the dark vortex itself had
ceased being directly visible. SDS-2017 may be following a
similar evolutionary path. The companion clouds are centered,
and the contrast of the dark spot is reduced to the point of being
barely detectable.
4.4. Cadences for Planetary Observations
The importance of the time domain in observations of the
planets in our solar system is now widely accepted. We are no
longer in the era where the results of a flyby mission, or a
single telescopic observation at a new wavelength (or with a
higher spectral or spatial resolution)will automatically lead to
major discoveries. But, discoveries remain to be made.
The OPAL program (Simon et al. 2015)represents a step
into this new era. Approved by the STScI director as a
discretionary program, this effort showed a commitment by a
major astrophysical observatory to contribute to time-domain
solar system science. The discovery of SDS-2015 is a direct
result of that commitment. The discovery of dark vortices on
Neptune is one example of science that has a timescale
perfectly suited to OPAL’s annual cadence of global map pairs.
Other observatories have contributed time-domain observations
of the atmospheres of Neptune and other giant planets. Kepler K2
observations produced long-duration, short-cadence photometry
(Simon et al. 2016). The resulting light curve data are valuable for
comparison with brown dwarf and exoplanet variability, because
the K2 Neptune observations were tied to resolved imaging data
to interpret the simple photometric variation. A similar effort
was done using Spitzer photometry (Stauffer et al. 2016).Twilight
adaptive-optics imaging programs are building up a new
campaign of frequent snapshot observations at Lick and Keck
observatories (Molter et al. 2017), providing a higher-resolution
complement to ground-based amateur coverage (Hueso et al.
2017).
Hypothetically, if OPAL 2017 observations had been too late
to detect a small companion cloud centered over a very low-
contrast dark spot, there would be a major open question: is the
massive equatorial cloud activity in 2017 (Molter et al. 2017)
related to the demise of SDS-2015? Proving such a link would
require dense temporal sampling over throughout 2017, with
1000 km resolution, at visible blue wavelengths. Obtaining
such coverage with the oversubscribed HST is very difficult,
and a dedicated solar system space observatory is not yet
available (Bell et al. 2015). Imaging with robotic visible-light
adaptive optics, Robo-AO (Baranec et al. 2014), can fulfill the
monitoring cadence and angular resolutions (~0.
1
)required to
measure the dynamics of major vortex and cloud activity on
Neptune. A more capable Robo-AO system is being prepared
for deployment on the UH 2.2 m telescope at Maunakea, HI,
and part of its mission is to enable planetary science not
feasible elsewhere.
Clearly, there is much room in the discovery space of solar
system time domain science. There is room in this discovery
space for exploration by a dedicated solar system space
telescope, a network of ground facilities, and cadence programs
at astrophysical observatories with advanced capabilities.
5. Summary
A new dark spot was discovered on Neptune in 2015 and
observed through 2017. We draw the following conclusions
and inferences:
1. The vortex drifted poleward, with a drift rate in the range
of 1°.7–2°.5/year (Figure 4).
2. The vortex faded over time (Figure 3), with a 467 nm
contrast of
-
6.6% 0.8% in 2015.71 and
-
2.9%
0
.4% in 2017.76. Bright companion clouds also grew
increasingly centered over the dark spot with each
successive observation (Figure 1). These evolutionary
changes may be part of a vortex dissipation sequence,
because aspects of these changes were seen in previous
dark spots: DS2 had centered companion clouds during
the Voyager 2 encounter and was never seen again, and
companion clouds of NDS-1994 persisted for some time
after the dark spot itself faded from view.
3. If we make reasonable (but poorly constrained)assump-
tions about the vorticity of SDS-2015 and its surround-
ings (described in Section 4.2), then we find that the
horizontal wind shear
d
ud
y
at the deep altitude of the
vortex must be about twice the value measured by
Voyager cloud-tracked winds.
8
The Astronomical Journal, 155:117 (9pp), 2018 March Wong et al.

4. Companion clouds create errors in measurements of the
dark spot’s boundaries, leading to uncertainties in the
feature’s size, shape, drift, and position that are large
compared to other sources of error. Stated uncertainties in
Table 4include the effect of companion clouds on
measurements of dark spot morphology.
5. Additional details of dark vortex behavior and evolution may
be gleaned in the future, when faster cadence observations
(with high spatial resolution at optical wavelengths)enable
exploration of the time-domain discovery space.
Based on observations associated with programs GO-13937,
GO-14044, GO-14334, GO-14492, and GO-14756, with
support provided by NASA through grants from the Space
Telescope Science Institute (operated by the Association of
Universities for Research in Astronomy, Inc., under NASA
contract NAS 5-26555). Data were obtained from the Data
Archive at the Space Telescope Science Institute. Support was
provided by the National Science Foundation through grant
AST-1615004 and by the NASA Earth and Space Science
Fellowship through grant NNX16AP12H to I.dP. and J.T., and
by the Alfred P. Sloan Foundation and the National Science
Foundation through grant AST-1712014 to C.B. We acknowl-
edge support from grants AYA2015-65041-P (MINECO/
FEDER, UE), Grupos Gobierno Vasco IT-765-13, and UPV/
EHU UFI11/55 to A.S.L. and R.H.
We thank an anonymous reviewer for their quick and
constructive review.
Facility: HST(WFC3).
Software: Astroconda, IDL, L.A.Cosmic, WinJUPOS, JPL
Horizons Ephemerides.
ORCID iDs
Michael H. Wong https://orcid.org/0000-0003-2804-5086
Amy A. Simon https://orcid.org/0000-0003-4641-6186
Christoph Baranec https://orcid.org/0000-0002-1917-9157
References
Baranec, C., Riddle, R., Law, N. M., et al. 2014, ApJL,790, L8
Bell, J. F., Schneider, N. M., Brown, M. E., et al. 2015, in LPI Contributions
1829, Kuiper: A Discovery-Class Observatory for Outer Solar System Giant
Planets, Satellites, and Small Bodies, 6043
Conrath, B. J., Flasar, F. M., & Gierasch, P. J. 1991, JGR,96, 18931
de Pater, I., Fletcher, L. N., Luszcz-Cook, S., et al. 2014, Icar,237, 211
de Pater, I., Sromovsky, L. A., Hammel, H. B., et al. 2011, Icar,215, 332
Deustua, S. E., Mack, J., Bajaj, V., & Khandrika, H. 2017, WFC3/UVIS
Updated 2017 Chip-Dependent Inverse Sensitivity Values, Tech. rep.,
WFC3 ISR 2017-14 (Baltimore, MD: STScI)
Deustua, S. E., Mack, J., Bowers, A. S., et al. 2016, UVIS 2.0 Chip-dependent
Inverse Sensitivity Values, Tech. rep., WFC3 ISR 2016-03 (Baltimore, MD:
STScI),3
Dressel, L. 2016, Wide Field Camera 3 Instrument Handbook, Version
8.0 (Baltimore, MD: STScI),http://documents.stsci.edu/hst/wfc3/
documents/handbooks/cycle24/wfc3_cover.html
Fitzpatrick, P. J., de Pater, I., Luszcz-Cook, S., Wong, M. H., & Hammel, H. B.
2014, Ap&SS,350, 65
Fletcher, L. N., de Pater, I., Orton, G. S., et al. 2014, Icar,231, 146
Fletcher, L. N., Orton, G. S., Rogers, J. H., et al. 2011, Icar,213, 564
Fruchter, A. S., & Hook, R. N. 2002, PASP,114, 144
Fry, P. M., Sromovsky, L. A., de Pater, I., Hammel, H. B., & Rages, K. A.
2012, AJ,143, 150
García-Melendo, E., Sánchez-Lavega, A., & Hueso, R. 2007, Icar,191, 665
Hammel, H. B., Beebe, R. F., de Jong, E. M., et al. 1989, Sci,245, 1367
Hammel, H. B., & Lockwood, G. W. 1997, Icar,129, 466
Hammel, H. B., Lockwood, G. W., Mills, J. R., & Barnet, C. D. 1995, Sci,
268, 1740
Hammel, H. B., Sromovsky, L. A., Fry, P. M., et al. 2009, Icar,201, 257
Hueso, R., de Pater, I., Simon, A., et al. 2017, Icar,295, 89
Ingersoll, A. P., Dowling, T. E., Gierasch, P. J., et al. 2007, in Jupiter, ed.
F. Bagenal, T. E. Dowling, & W. B. McKinnon (Cambridge: Cambridge
Univ. Press),105
Irwin, P. G. J., Wong, M. H., Simon, A. A., Orton, G. S., & Toledo, D. 2017,
Icar,288, 99
Karkoschka, E. 2011, Icar,215, 759
Karkoschka, E., & Tomasko, M. G. 2011, Icar,211, 780
LeBeau, R. P., & Dowling, T. E. 1998, Icar,132, 239
Legarreta, J., & Sánchez-Lavega, A. 2005, Icar,174, 178
Lii, P. S., Wong, M. H., & de Pater, I. 2010, Icar,209, 591
Limaye, S. S., & Sromovsky, L. A. 1991, JGR,96, 18941
Mack, J., Dahlen, T., Sabbi, E., & Bowers, A. S. 2016, Chip-Dependent Flats,
Tech. rep., WFC3 ISR 2016-04 (Baltimore, MD: STScI),4
Martin, S. C., de Pater, I., & Marcus, P. 2012, Ap&SS,337, 65
Minnaert, M. 1941, ApJ,93, 403
Molter, E. M., de Pater, I., Alvarez, C., Tollefson, J., & Luszcz-Cook, S. H.
2017, in AGU Fall Meeting Abstracts, P31D-2856
Polvani, L. M., Wisdom, J., Dejong, E., & Ingersoll, A. P. 1990, Sci,249,
1393
Rogers, J. H. 1995, The Giant Planet Jupiter (Cambridge: Cambridge Univ.
Press)
Sánchez-Lavega, A., Sromovsky, L., Showman, A., et al. 2018, in Zonal Jets,
ed. P. Galperin & B. Read (Cambridge: Cambridge Univ. Press)
Shetty, S., & Marcus, P. S. 2010, Icar,210, 182
Simon, A. A., Rowe, J. F., Gaulme, P., et al. 2016, ApJ,817, 162
Simon, A. A., Wong, M. H., & Orton, G. S. 2015, ApJ,812, 55
Simon, A. A., Wong, M. H., Rogers, J. H., et al. 2014, ApJL,797, L31
Smith, B. A., Soderblom, L. A., Banfield, D., et al. 1989, Sci,246, 1422
Sromovsky, L. A., de Pater, I., Fry, P. M., Hammel, H. B., & Marcus, P. 2015,
Icar,258, 192
Sromovsky, L. A., Fry, P. M., & Baines, K. H. 2002, Icar,156, 16
Sromovsky, L. A., Fry, P. M., Dowling, T. E., Baines, K. H., & Limaye, S. S.
2001, Icar,149, 459
Sromovsky, L. A., Limaye, S. S., & Fry, P. M. 1993, Icar,105, 110
Sromovsky, L. A., Limaye, S. S., & Fry, P. M. 1995, Icar,118, 25
Stauffer, J., Marley, M. S., Gizis, J. E., et al. 2016, AJ,152, 142
Stratman, P. W., Showman, A. P., Dowling, T. E., & Sromovsky, L. A. 2001,
Icar,151, 275
Tollefson, J., de Pater, I., Marcus, P. S., et al. 2018, Icar, submitted
Tollefson, J., Luszcz-Cook, S. H., Wong, M. H., & de Pater, I. 2017a, in AAS/
DPS Meeting 49 Abstracts, 205.03
Tollefson, J., Wong, M. H., de Pater, I., et al. 2017b, Icar,296, 163
van Dokkum, P. G. 2001, PASP,113, 1420
Wong, M. H. 2010, Amplitude of fringing in WFC3/UVIS narrowband red
filters, Tech. rep., WFC3 ISR 2010-04 (Baltimore, MD: STScI),4
Wong, M. H. 2011, in Proc. 2010 Space Telescope Science Institute
Calibration Workshop, ed. S. Deustua & C. Oliveira (Baltimore, MD:
STScI),22
Wong, M. H., Fry, P. M., & Simon, A. A. 2016, CBET, 4278, http://www.
cbat.eps.harvard.edu/cbet/004200/CBET004278.txt
9
The Astronomical Journal, 155:117 (9pp), 2018 March Wong et al.