Yield Estimation of the August 2020 Beirut Explosion by Using Physics‐Based Propagation Simulations of Regional Infrasound

The August 2020 Beirut explosion is the largest single‐fired ammonium nitrate explosion documented in history. The massive explosion excited loud infrasound in the atmosphere, and clear waveforms were recorded by a regional infrasound array at an epicentral distance of 100 km, allowing for accurate measurements of explosion energy. We estimate the explosion size based on the infrasound waveform inversion. Unlike conventional inversions using empirical models, we perform full 3‐D finite‐difference simulations to obtain a physics‐based propagation model for the inversion. Accurate numerical modeling of infrasound is challenging as the propagation is substantially affected by the turbulent atmosphere. Instead of a single deterministic prediction, we provide a range of waveform predictions by running multiple simulations with stochastic weather forecast models, which allows for comprehensive uncertainty analysis of numerical modeling and estimated yields. Finally, we expand the yield estimation technique for seismoacoustic analysis and demonstrate the substantial advantage of the joint approach. Plain Language Summary On 4 August 2020, 2,750 metric tons of ammonium nitrate stored at the Port of Beirut in Lebanon exploded. It is the largest single‐fired ammonium nitrate explosion documented in history and produced unusually loud infrasound (low‐frequency sound below 20 Hz) up to about 6,000 km away. We estimate the size of the Beirut explosion based on the infrasound, whose waveforms were clearly recorded at a regional infrasound array at an epicentral distance of 100 km. These clear waveforms provide a rare opportunity to invert infrasound signals for the explosion yield. Unlike conventional inversions using a simple regression analysis, we use full 3‐D physics‐based numerical simulations for infrasound propagation to predict the observed waveforms. We calculate energy attenuation from the explosion site to the observing stations and then infer the explosion size based on the calculation. Numerical modeling of regional infrasound at this range is particularly challenging because of the highly turbulent atmosphere. We use a stochastic weather forecast model to capture the variability of weather condition and provide a quantitative uncertainty analysis of waveform prediction. Key Points Explosion yield of the 2020 Beirut explosion is estimated based on infrasound waveforms Full 3‐D physics‐based numerical simulations are performed to predict infrasound waveform