J/PASP/136/H4203 PGIR survey long period variables catalog (Suresh+, 2024)
An Automated Catalog of Long Period Variables using Infrared Lightcurves from
Palomar Gattini-IR.
Suresh A., Karambelkar V., Kasliwal M.M., Ashley M.C.B., De K.,
Hankins M.J., Moore A.M., Soon J., Soria R., Travouillon T.
<Publ. Astron. Soc. Pac. 136, 084203 (2024)>
=2024PASP..136h4203S 2024PASP..136h4203S (SIMBAD/NED BibCode)
ADC_Keywords: Stars, variable ; Photometry, infrared
Keywords: long period variable stars - near infrared astronomy -
asymptotic giant branch stars
Abstract:
Long Period Variables (LPVs) are stars with periods of several hundred
days, representing the late, dust-enshrouded phase of stellar
evolution in low to intermediate mass stars. In this paper, we present
a catalog of 154755 LPVs using near-IR lightcurves from the Palomar
Gattini-IR (PGIR) survey. PGIR has been surveying the entire
accessible northern sky (DE>-28°) in the J-band at a cadence of
2-3 days since 2018 September, and has produced J-band lightcurves
for more than 60 million sources. We used a gradient-boosted decision
tree classifier trained on a comprehensive feature set extracted from
PGIR lightcurves to search for LPVs in this data set. We developed a
parallelized and optimized code to extract features at a rate of
∼0.1s per lightcurve. Our model can successfully distinguish LPVs
from other stars with a true positive rate of 95%. Cross-matching with
known LPVs, we find 70369 (∼46%) new LPVs in our catalog.
Description:
We present a catalog of 155891 LPVs from the PGIR survey created
using a ML classifier. We assemble a training set of 4300 objects
spanning LPVs showing clean sinusoidal evolution in their lightcurves,
"Type II LPVs" showing non-sinusoidal but periodic lightcurves,
non-LPVs comprised of R Cor Bor stars, Young Stellar Objects and other
classes that display long timescale erratic variability, and
non-variables. We train a gradient-boosted decision tree classifier on
an artificially resampled data set using ADASYN upsampling and allKNN
downsampling to class-balance the training set, achieving a true
positive rate of 94.89% for the combined set of LPVs and Type II LPVs
and an overall weighted g-mean score of 0.95, indicating high
accuracy.
File Summary:
--------------------------------------------------------------------------------
FileName Lrecl Records Explanations
--------------------------------------------------------------------------------
ReadMe 80 . This file
lpv-cat.dat 1191 155891 A catalog of LPVs obtained from a decision tree
classifier trained on features extracted from
Palomar Gattini-IR J band
--------------------------------------------------------------------------------
Byte-by-byte Description of file: lpv-cat.dat
--------------------------------------------------------------------------------
Bytes Format Units Label Explanations
--------------------------------------------------------------------------------
1- 19 A19 --- Name Unique identifier for each LPV,
internal designation of the object
21- 40 F20.16 deg RAdeg Right ascension (J2000)
42- 61 F20.16 deg DEdeg Declination (J2000)
63- 67 F5.1 --- Ndet Number of detections
69- 86 F18.16 --- vonNScore von-Neumann score
88- 109 E22.17 --- gpScore Gaussian process regression fit score
111- 133 E23.17 --- slopemin Minimum slope of the lightcurve
135- 152 F18.16 --- slopemax Maximum slope of the lightcurve
154- 172 F19.16 mag ptp Peak-to-peak amplitude
174- 177 F4.1 --- Nflips Number of flips
179- 196 F18.16 --- maxrate ? Maximum point-wise slope
198- 219 E22.17 --- minrate ? Minimum point-wise slope
221- 239 F19.14 d lcdur Observation baseline
241- 258 F18.16 --- pdiff90-50 Difference between 90th and 50th
percentiles of J band magnitude
260- 277 F18.16 --- pdiff95-50 Difference between 95th and 50th
percentiles of J band magnitude
279- 301 E23.17 --- J Stetson J index
303- 320 F18.16 --- Ks Stetson K index, calculated as
single band data
322- 339 F18.16 --- Km Stetson K index, calculated as
multi-band data
341- 363 E23.17 --- Ls Stetson L index, calculated as
single band data
365- 387 E23.17 --- Lm Stetson L index, calculated as
multi-band data
389- 407 F19.16 --- phasechisq chi square of sinusiod fit to phase
folded lightcurve
409- 427 E19.16 --- phaseredchi Reduced chi square of sinusiod fit to
phase folded lightcurve
429- 450 F22.16 --- phaseaic akaike information criterion of sinusiod
fit to phase folded lightcurve
452- 472 F21.15 --- phasebic Bayesian information criterion of sinusiod
fit to phase folded lightcurve
474- 492 F19.14 d perBest ? Best fit Lomb-Scargle period
494- 512 F19.14 d perBest2 ? Second most likely Lomb-Scargle period
514- 532 F19.14 d perBest3 ? Third most likely Lomb-Scargle period
534- 554 F21.16 --- LSscore ? Lomb-Scargle score of best period
556- 576 F21.16 --- LSscore2 ? Lomb-Scargle score corresponding
to second best period
578- 598 F21.16 --- LSscore3 ? Lomb-Scargle score corresponding
to third best period
600- 617 F18.16 mag AmpJ ? J band amplitude
619- 641 E23.17 --- meanF ? Mean predicted value from Lomb-Scargle
fit
643- 665 F23.16 --- chi2 ? chi square of sinusoidal fit
to lightcurve
667- 687 F21.16 --- redchi2 Reduced chi square of sinusoidal fit
to lightcurve
689- 709 F21.16 --- rednullchi2 Reduced chi square of linear fit
to lightcurve
711- 734 F24.16 --- nullchi2 ? chi square of linear fit to lightcurve
736- 754 F19.14 --- maxperiod ? Maximum period from Lomb-Scargle fit
756- 778 E23.17 --- linslope ? Slope of linear fit to lightcurve
780- 798 F19.16 --- lininter ? y-intercept of linear fit to lightcurve
800- 818 F19.16 mag W1mag ? WISE W1 magnitude
820- 837 F18.16 mag e_W1mag ? WISE W1 magnitude error
839- 857 F19.16 mag W2mag ? WISE W2 magnitude
859- 876 F18.16 mag e_W2mag ? WISE W2 magnitude error
878- 896 F19.16 mag W3mag ? WISE W3 magnitude
898- 915 F18.16 mag e_W3mag ? WISE W3 magnitude error
917- 935 F19.16 mag W4mag ? WISE W4 magnitude
937- 954 F18.16 mag e_W4mag ? WISE W4 magnitude error
956- 961 F6.3 mag J2mag ? 2MASS J magnitude
963- 967 F5.3 mag e_J2mag ? 2MASS J magnitude error
969- 974 F6.3 mag H2mag ? 2MASS H magnitude
976- 980 F5.3 mag e_H2mag ? 2MASS H magnitude error
982- 987 F6.3 mag K2mag ? 2MASS K magnitude
989- 993 F5.3 mag e_K2mag ? 2MASS K magnitude error
995-1014 F20.16 --- LSratio Ratio of Lomb-Scargle score of two most
promiment peaks in Lomb-Scargle
periodogram
1016-1033 F18.16 --- chi2ratio Ratio of reduced chi square of sinusoidal
fit and reduced chi square of linear fit
to lightcurve
1035-1052 F18.16 --- periodratio Ratio of best fit Lomb-Scargle period and
maximum Lomb-Scargle period
1054-1072 F19.16 mag J-H 2MASS J-H colour index
1074-1092 F19.16 mag J-K 2MASS J-K colour index
1094-1112 F19.16 mag W1-W2 WISE W1-W2 colour index
1114-1132 F19.16 mag W3-W4 WISE W3-W4 colour index
1134-1152 F19.16 mag W1-W4 WISE W1-W4 colour index
1154-1171 F18.16 --- ProbLPVsum Machine-learning classifier LPV score
(obtained by summing probability of LPV
and type II LPV)
1173-1191 F19.16 mag Jmag Mean J band magnitude
--------------------------------------------------------------------------------
History:
Coped at https://zenodo.org/records/10929164
(End) Patricia Vannier [CDS] 11-Jun-2025