Deep Spatial Interpolation of Rain Field for U.K. Satellite Networks

Abstract

This paper presents two new state-of-the-art Spatial Rain Field Interpolation Convolutional Neural Networks (SRFICNNs), referred to as LD (Learned Deviation) and LI (Learned Interpolation) models, for predicting the point rain rate at finer spatial scales. The main contribution is the successful introduction of the prior-art deep learning technique into high resolution rainfall rate prediction with significant improvement in accuracy. This is very important for the effective implementation of fade mitigation techniques for both terrestrial and satellite networks. The comparison of the models' performances with ground truth (radar measurements) shows that the proposed models give an excellent mean square error (MSE) and Structural SIMilarity (SSIM) in rainfall fields reconstruction if the network depth falls in the range of 15~25 weight layers. The final model uses 20 layers for high resolution point rain rates prediction. Further study shows that the LD model offers a faster convergence and yields a more accurate rain rate prediction. In particular, this paper compares the rain rate exceedance distribution and Log-Normality property from the model estimates with values calculated from measured data. Results show that the LD model gives a highly accurate estimates of these two indices with corresponding Root Mean Square (RMS) error of 5.1709 × 10 -4 and 0.0013, respectively.