Autonomous detection of cryospheric change with hyperion on-board Earth Observing-1

Abstract

On-board detection of cryospheric change in sea ice, lake ice, and snow cover is being conducted as part of the Autonomous Sciencecraft Experiment (ASE), using classifiers developed for the Hyperion hyper-spectral visible/infrared spectrometer on-board the Earth Observing-1 (EO-1) spacecraft. This classifier development was done with consideration for the novel limitations of on-board processing, data calibration, spacecraft targeting error and the spectral range of the instrument. During on-board tests, these algorithms were used to measure the extent of cloud, snow, and ice cover at a global suite of targets. Coupled with baseline imaging, uploaded thresholds were used to detect cryospheric changes such as the freeze and thaw of lake ice and the formation and break-up of sea ice. These thresholds were used to autonomously trigger follow-up observations, demonstrating the capability of the technique for future planetary missions where downlink is a constrained resource and there is high interest in data covering dynamic events, including cryospheric change. Before upload classifier performance was assessed with an overall accuracy of 83.3% as measured against manual labeling of 134 scenes. Performance was further assessed against field mapping conducted at Lake Mendota, Wisconsin as well as with labeling of scenes that were classified during on-board tests.

Introduction

This paper describes the development, testing and application of science-event detection with ASE, in this case, a data classifier and triggering algorithms designed to identify changes in the cryosphere which the spacecraft can then autonomously respond to. The structure of the paper is as follows: after a description of the rationale behind ASE, and the desired goal of autonomously monitoring and reacting to changes in the cryosphere, we describe the operational environment for ASE with Hyperion on EO-1, the development of the cryosphere classifier and triggering algorithms suited for this specific application, testing and verification of this approach, with classification assessment done on the ground prior to upload, a field validation study conducted at Lake Mendota, Wisconsin and an assessment of each on-board test conducted, as well as discussion of future applications.

The cryosphere is the component of a planetary body composed of ice in the form of snow, permafrost, floating ice, and glaciers, which dynamically interacts with atmospheres, climates, lithospheres and hydrospheres. Ices, broadly defined as the solid form of substances that are moderately volatile under ambient conditions, are found throughout the solar system. These include the H₂O ice caps on Earth, the H₂O and CO₂ ice caps on Mars (Thomas et al., 1992), indications of ice in the permanently shadowed polar craters of Mercury (Slade et al., 1992) and the Moon (Feldman et al., 1998), and under the broadest definition of ice, the proposed metallic frosts in high altitudes on Venus (Brackett et al., 1995) and hydrocarbons in the Saturn system (Sagan et al., 1984). In addition, ice is a major component of moons in the outer solar system and undergoes dynamic behavior such as cryotectonism (e.g., Nimmo & Gaidos, 2002) and cryovolcanism (Geissler, 2000).

On-board detection of cryospheric change and autonomous spacecraft response was conducted as part of the Autonomous Sciencecraft Experiment (ASE) (Chien et al., 2003), operating on Earth Observing-1 (EO-1) (Ungar et al., 2003) with the Hyperion hyper-spectral visible/infrared spectrometer (Pearlman et al., 2003). This capability would enable future interplanetary missions to search for dynamic cryopsheric events, make optimal use of limited downlink resources, and autonomously task the spacecraft to collect further data with a quicker response time than possible by ground control from Earth.