Dynamically Optimizing Radar Sounder Sampling Based on Estimated Uncertainty in Bed Topography

Abstract:

While laser altimetry, interferometric synthetic aperture radar, and optical imagery are available at high spatial and temporal resolution, radar sounder measurements remain extremely sparse, both spatially and temporally, due to the relative logistical difficulty and cost of collecting this data through airborne field surveys compared to satellite-based instruments. Given the limited opportunities to collect additional radar sounder data each year, it is important to optimize the expected scientific value of the data collected. Current radar sounder surveys are manually planned in advance of the field season with limited opportunities to adapt as the data is collected. Anticipating advances in autonomous radar sounder aircraft and real-time data processing, we explore data-driven adaptive selection of radar sounder measurement locations.

One possible application of this approach is surveying to inform a mass conservation ice thickness map. We demonstrate a Bayesian approach to quantifying uncertainty in ice thickness under a soft mass conservation constraint, taking into account sources of uncertainty including measurement error, uncertainty in spatial variance, and model-data mismatch. We propose a simple algorithm to use this estimated uncertainty to adjust the flight pattern of a radar sounder-equipped aircraft in real-time and show that it produces reasonable flight paths in the presence of different sources of uncertainty.