High-resolution source estimation of volcanic sulfur dioxide emissions using large-scale transport simulations

Abstract

High-resolution reconstruction of emission rates from different sources is essential to achieve accurate simulations of atmospheric transport processes. How to achieve real-time forecasts of atmospheric transport is still a great challenge, in particular due to the large computational demands of this problem. Considering a case study of volcanic sulfur dioxide emissions, the codes of the Lagrangian particle dispersion model MPTRAC and an inversion algorithm for emission rate estimation based on sequential importance resampling are deployed on the Tianhe-2 supercomputer. The high-throughput based parallel computing strategy shows excellent scalability and computational efficiency. Therefore, the spatial-temporal resolution of the emission reconstruction can be improved by increasing the parallel scale. In our study, the largest parallel scale is up to 1.446 million compute processes, which allows us to obtain emission rates with a resolution of 30 min in time and 100 m in altitude. By applying massive-parallel computing systems such as Tianhe-2, real-time source estimation and forecasts of atmospheric transport are becoming feasible.