Real-Time, In-Situ Detection of Dust Devils using Pressure Sensors

The Mars Pathfinder lander had a meteorology package that contained instruments that recorded near-surface temperature and pressure. Post-mission analysis of the pressure data indicated that dust devils passing over the lander could be detected as sudden, localized drops in pressure. This ground-based analysis identified 79 dust devils in the 83 sols of the Mars Pathfinder (MPF) mission (Murphy and Nelli, 2002). Based on the five sols in which continuous coverage for the entire sol was available, the average frequency of occurrence was estimated at 4.2 events per sol. If continuous coverage had been available for the entire mission, Murphy and Nelli estimated that 210-349 dust devils would have been expected. Consequently, 131-270 dust devils were missed due to bandwidth constraints and other factors that precluded continuous coverage. We have developed a method by which these events can be detected onboard a rover or lander that is equipped with a pressure sensor. We have experimented with various polynomial models of atmospheric pressure and found that a third-degree polynomial, with a ten-minute sliding context window, provides the best results. We have obtained 92% recall (8% misses) and 43% precision (57% false alarms), when evaluated against Murphy and Nelli's list of dust devils. This approach permits the in-situ detection of the dust devil events, which enables a higher degree of compression for the stretches of time in which no events occur. The 79 MPF events, including a 15-minute context window for each, consumed only 12% of the bandwidth used to send all observations back. In fact, the bandwidth allocation could have been used to transmit up to 680 detections, far more than are expected to have occurred. In the context of future missions, these on-the-spot detections will be an important complement to the distant dust devils detected in camera images, such as those currently collected by the Mars Exploration Rovers.