Spatio-Temporal Visual Saliency for Adaptive Weather Sensing Using Phased Array Radars

In this paper we present a novel analysis tool that identifies regions with elevated information content in weather radar images. The Weather Radar Spatio-Temporal Saliency (WR-STS) is based on information theory metrics and is applied to weather radar images with the goal of driving the implementation of Adaptive Weather Sensing (AWS) on Phased Array Radars (PAR). PARs can dynamically and adaptively change the beam pointing direction and associated acquisition parameters, which can potentially provide enhanced weather observations and lead to improved warnings and forecasts. In this paper we introduce the WR-STS, describe the impacts of various radar acquisition parameters, its behavior for different types of weather phenomena, and its application to AWS. We demonstrate advantages and limitations of the WR-STS using data collected with Next Generation Weather Radars (NEXRAD). A particularly complex severe weather event is used to validate the performance of WR-STS by correlating its results with warning polygons defined by operational forecasters during the event. This analysis supports the hypothesis that WR-STS can bring out regions with meteorologically important echoes. Furthermore, we illustrate the potential of WR-STS to drive the implementation of an AWS framework on phased arrays. The proposed Focused-Observations-by-Configuration- of-Update-times-using-Saliency (FOCUS) algorithm is described and its performance to dynamically define update times for different sectors of interest in the atmosphere is illustrated.