High-Resolution Intensity Field Fusion of Array Weather Radar on GPU

Abstract

Array Weather Radar (AWR) is a new type weather radar. A distributed phased array technology is utilized to the AWR. With offering very high spatiotemporal resolution, the AWR can obtain the small-scale weather systems with high changing speed. The AWR system provides networked and coordinated observations of the same target from at least three radar front ends. This paper introduces a fast and high-resolution intensity field fusion algorithm for the AWR based on Graphics Processing Unit (GPU). By utilizing the feature of asynchronous execution between CPU and GPU, the fusion algorithm is parallelized and optimized on Compute Unified Device Architecture (CUDA). First, based on the intensity data from the AWR front end volume scans, resolution expansion coefficients in azimuths and elevations are calculated in each GPU thread. Then, the vertical and horizontal filling is completed simultaneously. Finally, the fusion of three front end intensity data is achieved in a common coordinate system. The entire process has been implemented on NVIDIA GPU architecture. The result shows that the GPU-based method for the AWR greatly improves the speed of calculation compared with the conventional CPUbased implementation, which can satisfy the requirements of meteorological applications in real-time.