Integrating Machine Learning for Planetary Science: Perspectives for the Next Decade

Abstract

† We would like to recognize the extraordinary effort which this decadal has taken and the members of our community who were unable to participate in this work. We would also like to acknowledge conversations with additional white paper teams on data management, automation, and other machine learning relevant contributions and we encourage you to review these data science relevant papers submitted to the survey.

Classification of Cassini’s Orbit Regions as Magnetosphere, Magnetosheath, and Solar Wind via Machine Learning

“…As noted by several studies (Azari et al, 2020;Hook et al, 2020;Theiling et al, 2021;Vandegriff et al, 2021), current missions are already facing severe downlink constraints and more data-intensive sensors. Without increased capabilities in on-board storage and deep space communications, missions may ultimately require the use of on-board autonomy to sift through the deluge of collected data to prioritize the most relevant observations for downlink or optimize the science collection of the sensors for the environment the spacecraft or lander is currently inhabiting.…”

“…At present, the detection and cataloging of such events is done primarily by visual inspection of the data sets by domain experts. Yet, as the current and near-future space missions continue to fly evermore data-intensive sensors, the space physics community is rapidly approaching a point in which the data volume vastly exceeds the analysis capacity of the domain experts (Azari et al, 2020). Additionally, manual detection and cataloging of the events embeds the bias of the individual observer into the curated catalog, consequently precluding the inter-comparison of results from two independent observers.…”

DataSheet1.pdf

“…As noted by several studies (Azari et al, 2020;Hook et al, 2020;Theiling et al, 2021;Vandegriff et al, 2021), current missions are already facing severe downlink constraints and more data-intensive sensors. Without increased capabilities in on-board storage and deep space communications, missions may ultimately require the use of on-board autonomy to sift through the deluge of collected data to prioritize the most relevant observations for downlink or optimize the science collection of the sensors for the environment the spacecraft or lander is currently inhabiting.…”

“…At present, the detection and cataloging of such events is done primarily by visual inspection of the data sets by domain experts. Yet, as the current and near-future space missions continue to fly evermore data-intensive sensors, the space physics community is rapidly approaching a point in which the data volume vastly exceeds the analysis capacity of the domain experts (Azari et al, 2020). Additionally, manual detection and cataloging of the events embeds the bias of the individual observer into the curated catalog, consequently precluding the inter-comparison of results from two independent observers.…”