Data-intensive computing in radiative transfer modelling

Abstract

The operational processing of remote sensing data requires high-performance radiative transfer model (RTM) simulations. To date, considerable success has been achieved in dimensionality reduction techniques as well as in heterogeneous multi-CPU/GPU computing for highly intensive parallel computations. We have developed several techniques for accelerating the radiative transfer solver. They include (1) analytical methods which allow to compute set of atmospheric scenarios in one RTM call; (2) dimensionality reduction of the datasets, and (3) GPU-computing using CUDA framework. These techniques provide almost 300x cumulative speed-up for the RTM with respect to the original single-threaded CPU code. In this paper, we analyze the applicability of the proposed methods to a practical problem of total ozone column retrieval from UV-backscatter measurements.