Compressed Sensing Based Scan Strategy Selection in Weather Radar

An important problem in weather radar is reducing data acquisition time. For the current generation of weather radar, reductions in data acquisition time are needed to obtain faster updates on fast evolving weather hazards such as tornadoes. For the next generation of weather radar, additional reductions in data acquisition time will be needed to free up time for the radar to perform multiple functions, such as aircraft and drone surveillance and tracking. The importance of the problem have led to the development of a number of strategies for speeding data acquisition. These include: the SAILS VCP for reducing low-altitude update times in the NEXRAD system during tornadoes; ADAPTS for reducing volume update times in phased-array radars operating in isolation; CASA (Collaborative Adaptive Sensing) for reducing volume updates in densely deployed, networked radar systems; and BMX (beam multiplexing) for reducing dwell times in phased-array radars. The above techniques reduce update times by visiting certain beam positions more often than others (SAILS), preferentially visiting those beam positions where weather had been detected during previous volume updates while not visiting those where no weather was detected (ADAPTS, CASA), or by reducing the dwell time at a given beam position (BMX). As an alternative, or adjunct to any of these techniques, this paper explores scan strategies based on compressed sensing. Compressed sensing provides theory and techniques for reducing data acquisition cost in time, computation, and/or bandwidth, by reducing the number of samples that need to be collected to obtain a good reconstruction of a signal of interest. In particular, if a signal is sparse (the majority of its entries are zero) then compressed sensing can recover the signal exactly from a number of samples that is significantly smaller than the dimension of the signal. Because the number of samples is much smaller than the dimension of the signal, the signal is said to have been sampled in compressed form. For signals, such as weather radar reflectivity, which are not truly sparse, but are only weakly sparse, reconstruction from a small set of observations, though not exact can be nearly indistinguishable from the field obtained from a complete sampling of the signal. By reducing the number of range bins sampled at a beam position, computation at the radar and bandwidth from the radar can be reduced. By reducing the number of beam positions, volume update time can be reduced. By reducing both, cost reductions in time, computation, and bandwidth are achieved. After explaining compressive sensing and its applicability to weather radar data acquisition, this paper presents a number of examples to illustrate how compressive sensing based scan strategies can reduce data acquisition costs in single and networked radar systems.