High-Resolution Source Estimation of Volcanic Sulfur Dioxide Emissions Using Large-Scale Transport Simulations

Liu, M.; Chen, P.; Huang, Y.; Huang, C.; Hoffmann, Lars; Heng, Y.

doi:10.1007/978-3-030-50420-5_5

Contribution to a conference proceedings/Contribution to a book

FZJ-2020-03185

High-Resolution Source Estimation of Volcanic Sulfur Dioxide Emissions Using Large-Scale Transport Simulations

Liu, M. ; Huang, Y. ; Hoffmann, L.FZJ* ; Huang, C. ; Chen, P. ; Heng, Y. (Corresponding author)

2020
Springer Cham

Computational Science – ICCS 2020
International Conference on Computational Science 2020, ICCS 2020, Amsterdam, The Netherlands, 3 Jun 2020 - 5 Jun 2020 Cham : Springer, Lecture Notes in Computer Science 12139, 60-73 (2020) [10.1007/978-3-030-50420-5_5]

This record in other databases:

Please use a persistent id in citations: http://hdl.handle.net/2128/25731 doi:10.1007/978-3-030-50420-5_5

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.

Contributing Institute(s):