Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
TRENDS AND NOISE REMOVAL USING A WAVELET BASED
MODIFICATION OF THE TREND FILTERING ALGORITHM:
APPLICATION TO WIDE-FIELD ASTRONOMICAL SURVEYS
DANIEL DEL SER BADIA
ICCUB WINTER MEETING
FEBRUARY 7, 2017
OUTLINE
• Trend Filtering Algorithm (TFA): basics and formulation • Introduction to Wavelets
• What is a Wavelet?
• Discrete Wavelet Transform (DWT) vs Stationary Wavelet Transform (SWT)
• Modified TFA (TFAW): application of the SWT • TFAW application to simulated data • TFAW application to ground-based wide-field surveys
• Telescope Fabra-ROA at Montsec (TFRM)
• The Evryscope
• TFAW application to space-based exoplanet surveys • CoRoT
• Kepler
• Conclusions
TREND FILTERING ALGORITHM (TFA): BASICS AND FORMULATION
• Many stars within a photometric survey suffer from the same systematic effects.
• Build an optimum filter from a template of reference stars:
• A(i) represents the noise- and trend-free signal to be found.
• TFA frequency analysis step:
• A(i) = <Y(i)>
• Most trends and systematics are removed Increased period detection probability.
• Signal distorted as we have assumed that the signal is constant.
TFA: Kovács, G., Bakos, G. & Noyes, R. W. 2005, MNRAS, 356, 557-567
∑
∑
=
=
−−=
=
N
i
M
jjj
iFiAiYD
iXciF
1
2
1
)]()()([
)()(
)()( iFiYY −=∧
TREND FILTERING ALGORITHM (TFA): BASICS AND FORMULATION
• Many stars within a photometric survey suffer from the same systematic effects.
• Build an optimum filter from a template of reference stars:
• A(i) represents the noise- and trend-free signal to be found.
• TFA signal reconstruction step:
• Iterative approximation of A(i).
• Can only be applied to periodic signals.
• A(i) computed through bin averaging the phase folded light curve. Although more complex methods are recommended.
∑
∑
=
=
−−=
=
N
i
M
jjj
iFiAiYD
iXciF
1
2
1
)]()()([
)()(
)()( iFiYY −=∧
TFA: Kovács, G., Bakos, G. & Noyes, R. W. 2005, MNRAS, 356, 557-567
WHAT IS A WAVELET?
• Highly localized impulses obtained from shifting and scaling the ‘mother wavelet’.
• Scaling and shifting allow to calculate the wavelet coefficients Correlation between the wavelet and a localized section of the signal.
Introduction to Wavelets
Biorthogonal 3.9 wavelet function
• Signal decomposed in dyadic blocks (i.e. shifting and scaling is based on a power 2).
• DWT calculated applying high and low pass filters to the signal Detail and Approximation coefficients.
• Signal can be iteratively decomposed into components of lower scale (higher resolution) Filter Bank.
• Maximum number of decomposition levels depends on the length of the data:
• Data length = 2N max_level == N
Discrete Wavelet Transform (DWT) vs Stationary Wavelet Transform (SWT)
Introduction to Wavelets
Discrete Wavelet Transform (DWT) vs Stationary Wavelet Transform (SWT)
• DWT is not a time-invariant transform very sensitive to the alignment of the signal in time.
• SWT is designed to overcome the lack of translation-invariance of the DWT.
• Same low and high pass filters are applied.
• NO subsampling Filters are upsampled versions of the previous ones (padded with zeroes) new sequences at each level have the same length as the original time series.
Introduction to Wavelets
Discrete Wavelet Transform (DWT) vs Stationary Wavelet Transform (SWT)
Introduction to Wavelets
Noise
High-frequency sinusoidal
Low-frequency sinusoidal Original signal
MODIFIED TFA (TFAW): APPLICATION OF THE SWT
• A(i) represents the noise- and trend-free signal to be found.
• Frequency analysis step is the same as with TFA (A(i) = <Y(i)>).
• Search for any significant period Light curve is phase folded.
• SWT of the phase folded light curve is computed.
• Noise and signal levels are selected.
• A(i) computed with those decomposition levels below signal level.
• New filter becomes:
∑
∑
=
=
−−=
=
N
i
M
jjj
iFiAiYD
iXciF
1
2
1
)]()()([
)()(
)()( iFiYY −=∧
−=∧
levelnoiseYISWTiFF _,)('
TFAW APPLICATION TO SIMULATED DATA
High SNR sinusoidal Low SNR sinusoidal
TFAW APPLICATION TO SIMULATED DATA
High SNR transit Low SNR transit
TFAW APPLICATION TO GROUND-BASED SURVEYS
THE TELESCOPE FABRA-ROA AT MONTSEC (TFRM)
• The Telescope Fabra-ROA at Montsec (TFRM) is a refurbished f/0.96 0.50m Baker-Nunn Camera for robotic CCD wide-FoV surveying purposes.
• The project is a joint collaboration between the Reial Acadèmia de Ciències i Arts de Barcelona - Observatori Fabra and the Real Instituto y Observatorio de la Armada (ROA).
• Installed at the summit of the Montsec d'Ares as part of the Observatori Astronòmic del Montsec.
TFAW application to ground-based surveys
TFRM: Fors, O. et al. 2013, PASP, 125, (927), 522
THE TELESCOPE FABRA-ROA AT MONTSEC (TFRM)
• TFRM is surveying a selected list of M-dwarfs for exoplanetary transits (TFRM-PSES) since its deployment at the end of 2011.
• TFRM-PSES benefits from the combination of its wide FoV (19.4 sq. deg.) for a 16 MPix camera, and from the 0.1% -precision 20-second-cadence light curves for every object brighter than V~16th magnitude.
TFAW application to ground-based surveys
THE TELESCOPE FABRA-ROA AT MONTSEC (TFRM)
• TFRM is surveying a selected list of M-dwarfs for exoplanetary transits (TFRM-PSES) since its deployment at the end of 2011.
• TFRM-PSES benefits from the combination of its wide FoV (19.4 sq. deg.) for a 16 MPix camera, and from the 0.1% -precision 20-second-cadence light curves for every object brighter than V~16th magnitude.
TFAW application to ground-based surveys
4.4 degrees aprox. 8.5 full Moons
THE TELESCOPE FABRA-ROA AT MONTSEC (TFRM)
• TFRM-PSES survey data used for the TFAW performance assessment comprises 2048 data points 11 SWT decomposition levels.
TFAW application to ground-based surveys
New Discovery!!!
2MASS10144313+5018191 NSVS 4921994
• The Evryscope is a University of North Carolina at Chapel Hill project.
• It is a telescope which covers the entire accessible sky above an airmass of 2. It consists of a single hemisphere containing twenty-three 61mm-aperture telescopes imaging a 691 MPix instantaneous 8000 sq. deg. FoV every 2 minutes.
TFAW application to ground-based surveys
Evryscope: Law, N. M. et al. 2015, PASP, 127, (949), 234
THE EVRYSCOPE
THE EVRYSCOPE
TFAW application to ground-based surveys
36K x 36K pixels, 100º x 100º FoV every 2 minutes down to g=16mag. 2 million objects detected per exposure.
• The Evryscope is located at Cerro Tololo Inter-American Observatory (CTIO), covering declinations between -90º and +10º.
• Currently delivering multi-year 1%-precision 2-minute-cadence light curves for every object brighter than g ~16, and millimagnitude-precision 16-minute-cadence light curves for every object brighter than g~12.
• BigData production: 230 TB/year (all archived in drives).
• Coadding allows deeper detections: glim~17.5(1hr) ; glim~19 (1 night)
TFAW application to ground-based surveys
THE EVRYSCOPE
2MASS13190996-8347115 1RXS J031713.9-853231
THE EVRYSCOPE • Evryscope data used for the TFAW performance assessment comprises 2048
data points.
TFAW application to ground-based surveys
New Discovery!!!
TFAW APPLICATION TO SPACE EXOPLANET SURVEYS
CoRoT
• Space observatory mission led by the French Space Agency (CNES) in conjunction with the European Space Agency (ESA).
• Launched 27 December 2006
• Mission objectives:
• Search for extrasolar planets with short orbital periods, particularly those of large terrestrial size.
• Perform asterosysmology by measuring oscillations in stars.
• 29 confirmed exoplanets, hundreds of candidates.
• De-orbited 17 June 2014.
• Systematics due to high energetic proton flux near the South Atlantic Anomaly (SAA), residuals at the CoRoT orbital period, spacecraft jitter, CCD long-term aging…
TFAW application to space exoplanet surveys
CoRoT • Data comprises 4096 points from light curves observed during the IRa0 run
of the CoRoT mission.
TFAW application to space exoplanet surveys
2MASS J06441118-0117536
KEPLER
• Kepler is a space mission led by NASA to discover Earth-size planets orbiting other stars.
• Launched on March 7, 2009 and still running.
• Designed to survey a portion of our region of the Milky Way to discover Earth-size exoplanets in or near habitable zones and estimate how many of the billions of stars in the Milky Way have such planets.
• 2331 confirmed planets (plus 144 from K2 mission) and over 5000 candidates.
• Systematics due to thermal effects, jitter, pixel sensitivity dropouts…
TFAW application to space exoplanet surveys
KEPLER • Data comprises 16384 points observed with the short cadence mode.
TFAW application to space exoplanet surveys
Kepler-19 b
CONCLUSIONS • TFA can remove trends and systematics but does not decouple noise from
signal.
• SWT allows to recover the noise-free signal without need of binning or fitting a model.
• TFAW is an algorithm that combines TFA trend and systematic removal with the SWT noise decoupling and filtering capabilities.
• TFAW is able to increase the SNR of periodic signals affected by systematics, trends and noise whithout modifying their shape, amplitude and phase.
• Also TFAW does not introduce any “fake” signatures in the data and improves the characterization of the frequency power spectra of the signals.
• TFAW is a totally generic algorithm that can be applied to ground- and space-based surveys provided that a sample of reference stars is available.
THANKS FOR YOUR ATTENTION
APPENDIX I
2MASS J06441118-0117536
TFAW application to space exoplanet surveys
APPENDIX I • Data comprises 4096 points from light curves observed during the IRa0 run
of the CoRoT mission.
CoRoT 102712679
APPENDIX I
CoRoT 102712679
APPENDIX I
Kepler-19 b
TFAW application to space exoplanet surveys
KEPLER • Data comprises 16384 points observed with the short cadence mode.
kplr003114661
APPENDIX I
kplr003114661
APENDIX II: MULTIPERIODIC SIGNAL
APENDIX III