Fast high-quality three-dimensional reconstruction from compressive observation of phased array weather radar

Phased array weather radar (PAWR) is capable of spatially and temporally high resolution observation. This means that a PAWR generates a huge amount of observation data, say 500 megabytes in every 30 seconds. To transfer this big data in a public internet line, this paper proposes a fast 3D compressive sensing method for PAWR. The proposed method reconstructs the original data, from compressed data, as the minimizer of a convex function which evaluates the local similarity in the spatial domain and the sparsity in the frequency domain. By combining blockwise optimization with Nesterov's acceleration, we obtain an approximate solution of the above convex optimization problem at high speed. Numerical simulations show that the proposed method outperforms conventional reconstruction methods.