Title: Acceleration of PLUME TRACKER Volcanic Emission Analysis Software
Presenting Author: Alexander Berk
Organization: Spectral Sciences, Inc.
Co-Author(s): Vincent Realmuto and Chona Guiang

Abstract:
Acceleration of PLUME TRACKER Volcanic Emission Analysis Software Plume Tracker (PT), formerly MAP_SO2, is robust, extensively validated, analyst driven software for retrieving gaseous emissions from volcanic releases. While Plume Tracker provides very accurate results, analyzing multi-spectral imagery typically takes several days of analyst time. The major bottleneck in the PT processing is the relatively slow but accurate FORTRAN-based MODTRAN gas plume and atmospheric radiative transfer (RT) model. This paper describes five physics/algorithm/software based approaches which together accelerate the Thermal Infrared (TIR) data analysis of PT analysis by in excess of 1000-fold. In particular, PT now includes (1) a Hash Table to recognize redundant MODTRAN runs; (2) an improved best fit algorithm based on Brent Minimization and a spectral angle mapper metric; (3) analytic expressions for modeling variations in surface temperatures and spectral emissivities; (4) an OpenMP implementation to exploit multiple core systems; and (5) utilities for eliminating redundant calculations within MODTRAN runs. Processing speed up results will be presented. In addition, PT retrievals of SO2 column density maps through its analysis of ASTER data are validated against two independent measurement campaigns: (1) USGS Hawaiian Volcano Observatory and the University of Hawaii ground station measurements of the Kilauea Volcano plume in Hawaii, and (2) University of Costa Rica tethered-sonde measurements at Turrialba volcano.