Impact of background representation on Bayesian radioxenon source estimation using IMS measurements

Radioxenon detection plays a critical role in verifying compliance with the Comprehensive Nuclear-Test-Ban Treaty (CTBT), yet remains challenging due to radioxenon's highly variable and ever present background and the sparse nature of the International Monitoring System (IMS) network. Radioxenon background signals are however arising from numerous, sometimes poorly characterized, civilian sources and are strongly influenced by long-range atmospheric transport, introducing significant uncertainty in background predictions. To address these issues, we develop a Bayesian inverse modeling framework for unknown source term estimation that integrates IMS measurements, atmospheric transport modeling, and multiple representations of radioxenon background. We test three background configurations: no background, summed background per sensor, and source-specific background contributions per sensor. Using data from the 1st Nuclear Explosion Signal Screening Open Inter-comparison Exercise 2021, combined with radioxenon background forecasts, we demonstrate the effectiveness of Bayesian modeling for estimating unknown source terms in simulated nuclear test scenarios. Notably, the use of source-specific background contributions per sensor leads to improved accuracy in the estimated releases.