Responsive Onboard Science for Europa Clipper Kiri L

Responsive Onboard Science for Europa Clipper Kiri L. Wagstaff1, Ashley Davies1, Gary Doran1, Srija Chakraborty2, Saadat Anwar2, Diana L. Blaney1, Steve Chien1, Philip R. Christensen2, and Serina Diniega1 1 Jet Propulsion Laboratory, California Institute of Technology, 2 Arizona State University Outer Planets Assessment Group (OPAG) Meeting, September 11, 2018 Mission: 40-50 flybys of Europa Current Status • Some locations will only be visited once; limited downlink • Successful detection of thermal anomalies (<190 K in simulated E-THEMIS data) • Cannot predict when/where interesting phenomena • Successful detection of spectral anomalies and minerals of interest will occur (e.g., plumes) in Galileo NIMS observations of Europa; experiments with simulated • Want to maximize return of scientifically valuable data MISE data in progress

New Capability: Analyze observations onboard Next Steps (FY'19) and assign downlink priorities based on content • Develop onboard analysis methods for other Clipper instruments: • Initial focus on two instruments: • PIMS: (Plasma Instrument for Magnetic Sounding) • E-THEMIS (Europa Thermal Imaging System) Detect ionosphere/magnetosphere boundary JosephPIMS Westlake [7] and the PIMS Team • MISE (Mapping Imaging Spectrometer for Europa) • Enable in situ adjustment of energy range to OPAG: 24 August 2015 • Detect and increase priority for: maximize temporal resolution • Thermal anomalies (hotspots) • EIS: (Europa Imaging System): Detect interesting morphology Credits: NASA/JPL-Caltech • Minerals of interest • Enable extra collection of narrow-angle images with • Spectral anomalies (unexpected minerals) subsequent prioritization Space • Evaluate resource consumption on RAD750 processor 18 EIS [5]

E-THEMIS: Thermal anomaly detection MISE: Spectral anomaly (mineral) detection Onboard operations (Europa Thermal Imaging System, 7-70 µm) (Mapping Imaging Spectrometer for Europa, 0.8-5.0 µm) • Analysis can be done by instruments or by main spacecraft CPU Simulated MISE data Europa thermal Simulated E-THEMIS data • Prioritize data based on science content for downlink inertia [3] (hyperspectral cube) (3-band IR observations) E-THEMIS 1700 nm (update priorities) Inject simulated DNs anomaly (via lookup table) • Generate detection alerts

Simulated surface nm 800 • Generate summary products / maps Band 1 7-14 µm rAAAB/nicdVDJSgNBEO1xjXGLiicvjUHwYugJRpODEPDiMYJZIAmhp1NJmvQsdNeIYQj4K148KOLV7/Dm39hZBBV9UPB4r4qqel6kpEHGPpyFxaXlldXUWnp9Y3NrO7OzWzNhrAVURahC3fC4ASUDqKJEBY1IA/c9BXVveDnx67egjQyDGxxF0PZ5P5A9KThaqZPZ1x1JL2gL4Q4TKeBESzEYdzJZlmNTUJYrMLd05lqSL7BSkVF3bmXJHJVO5r3VDUXsQ4BCcWOaLouwnXCNUigYp1uxgYiLIe9D09KA+2DayfT8MT2ySpf2Qm0rQDpVv08k3Ddm5Hu20+c4ML+9ifiX14yxV2wnMohihEDMFvViRTGkkyxoV2oQqEaWcKGlvZWKAddcoE0sbUP4+pT+T2r5nMty7vVptpyfx5EiB+SQHBOXnJMyuSIVUiWCJOSBPJFn5955dF6c11nrgjOf2SM/4Lx9AvFolW8= i =ice-rich R = Ar +(1 A)p temperature AAAB/nicbVDLSgMxFL1TX7W+RsWVm2ARKmKZEUEXCi1uXFaxD2iHIZOmNTSTGZKMUIaCv+LGhSJu/Q53/o1pOwttPXC5h3PuJTcniDlT2nG+rdzC4tLySn61sLa+sbllb+80VJRIQusk4pFsBVhRzgSta6Y5bcWS4jDgtBkMrsd+85FKxSJxr4cx9ULcF6zHCNZG8u29O5+hK1SVph2jkntSPYp95ttFp+xMgOaJm5EiZKj59lenG5EkpEITjpVqu06svRRLzQino0InUTTGZID7tG2owCFVXjo5f4QOjdJFvUiaEhpN1N8bKQ6VGoaBmQyxflCz3lj8z2snunfhpUzEiaaCTB/qJRzpCI2zQF0mKdF8aAgmkplbEXnAEhNtEiuYENzZL8+TxmnZdcru7VmxcpnFkYd9OIASuHAOFbiBGtSBQArP8Apv1pP1Yr1bH9PRnJXt7MIfWJ8/UaSTHA== i i i Europa Downlink surface Bulk Data System (BDS) priority Band 2 14-28 µm table Spacecraft temperature Albedo = A position → model Filter FSW viewing AAAB/nicbVDLSgNBEJz1GeNrVTx5GQyCF8NuEPQiBLx4jGAekCxhdtJJhszuLDO9YlgC/ooXD4p49Tu8+TdOHogmFjQUVd10d4WJFAY978tZWl5ZXVvPbeQ3t7Z3dt29/ZpRqeZQ5Uoq3QiZASliqKJACY1EA4tCCfVwcD326/egjVDxHQ4TCCLWi0VXcIZWaruHSVvQK9pCeMBMcDhLlNKjtlvwit4E9If486RAZqi03c9WR/E0ghi5ZMY0fS/BIGMaBZcwyrdSAwnjA9aDpqUxi8AE2eT8ET2xSod2lbYVI52ovycyFhkzjELbGTHsm3lvLP7nNVPsXgaZiJMUIebTRd1UUlR0nAXtCA0c5dASxrWwt1LeZ5pxtInlbQgLLy+SWqnoe0X/9rxQLs3iyJEjckxOiU8uSJnckAqpEk4y8kReyKvz6Dw7b877tHXJmc0ckD9wPr4B1YaVWg== pi = ice-poor Band 3 28-70 µm Unfiltered data Instrument files Filesystem packets (in priority bins) Select (4Gb) Transmit Basis spectra Reflectance Derived from NIMS high/low albedo areas CPU Onboard Science:

DNAAAB/HicbZDLSsNAFIZP6q3WW7RLN4NFEISSiKALhYIuXEkVe4E2hMl00g6dZMLMRCihvoobF4q49UHc+TZO2yy09YeBj/+cwznzBwlnSjvOt1VYWl5ZXSuulzY2t7Z37N29phKpJLRBBBeyHWBFOYtpQzPNaTuRFEcBp61geDWptx6pVEzED3qUUC/C/ZiFjGBtLN8uX9/6DF2ivs/uDRwj4TPfrjhVZyq0CG4OFchV9+2vbk+QNKKxJhwr1XGdRHsZlpoRTselbqpogskQ92nHYIwjqrxsevwYHRqnh0IhzYs1mrq/JzIcKTWKAtMZYT1Q87WJ+V+tk+rw3MtYnKSaxmS2KEw50gJNkkA9JinRfGQAE8nMrYgMsMREm7xKJgR3/suL0Dypuk7VvTut1C7yOIqwDwdwBC6cQQ1uoA4NIDCCZ3iFN+vJerHerY9Za8HKZ8rwR9bnDy10kyU= = g R + o Galileo SSI i i i i Ground Analyze data and archive 25ESDARKBP01 update priorities / Instruments Experimental results Experimental results generate new products • Multiple random trials • Multiple random trials • Inject synthetic anomaly: linear mixture of pixel + • Inject synthetic anomaly: library spectrum for selected mineral (e.g., bloedite) linear mixture of pixel + blackbody curve) • Detection: Use anomaly detection methods to rank all pixels Coordinated science scenario and determine how quickly the anomaly is found • Detections by one instrument can inform • Detection: Compare observations • RX [1]: Deviation from background prioritizations made by another instrument at two wavelength ranges (shown in blue and green) with t-test • DEMUD [6]: SVD-based detection • For example: E-THEMIS → MISE to detect anomalous rise at longer wavelengths [2]. • Matched filter [3]: Search for known (injected) mineral E-THEMIS detects Band 1 (7-14 µm) Band 2 (14-28 µm) Band 3 (28-70 µm) a thermal anomaly Galileo NIMS - Europa E-THEMIS Large/hot during approach (easy) (hyperspectral cube, • Larger anomalies are easier to detect • Matched filter performs best 0.7-5.2 µm) (as expected) • DEMUD and RX find bloedite

Faster detection without any training

Small/cold (hard) of detection Probability MISE increases priority Larger anomaly MISE Detection sensitivity in simulated E-THEMIS data with synthetic anomalies Visualization of NIMS data Novelty detection results for NIMS of an observation injected, as a function of anomaly size (pixel fraction, x-axis) and temperature (14ENSUCOMP01A). Red dot shows an data with synthetic bloedite (sulfate) at the same location (y-axis). Band 1 shows the highest sensitivity to thermal anomalies, as expected. example randomly selected pixel where a anomalies as a function of anomaly Results obtained from 5000 Monte Carlo trials. synthetic anomaly was injected. fraction (x-axis).

References: [1] Chiang and Chiang, "Anomaly detection and classification for hyperspectral imagery," IEEE Trans. on Geoscience and Remote Sensing, 40(6), 2002; [2] Doran et al., "Onboard detection of thermal anomalies for Europa Clipper," Proc. of European Planetary Science Congress (EPSC), 2018; [3] Nasrabadi et al., "Hyperspectral target detection," IEEE Signal Processing Magazine, 31(1):34-44, 2014; [4] Rathbun et al., "Galileo PPR observations of Europa: Hotspot detection limits and surface thermal properties," Icarus, 210:763-769, 2010; [5] Turtle et al., "EIS: Europa Imaging System," Outer Planets Assessment Group, 2015; [6] Wagstaff et al., "Guiding scientific discovery with explanations using DEMUD," Proc. of the 27th AAAI Conference on Artificial Intelligence, p. 905-911, 2013; [7] Westlake and the PIMS Team, "Plasma Instrument for Magnetic Sounding (PIMS)", Outer Planets Assessment Group, 2015.

© 2018, California Institute of Technology. Government sponsorship acknowledged. Funded by the JPL Research and Technology Development Program. We thank Europa Clipper project members for many informative discussions. This work was performed in part at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with NASA.