Single-station scalar moment estimation
Estimating earthquake size, with calibrated uncertainty, from one seismic station.
Magnitude estimation normally requires a network. Where the network is sparse — much of the southern hemisphere, the oceans, and anywhere a monitoring problem is hardest — you may have one station and a few minutes.
This work asks how well a single three-component record can constrain scalar moment, and more importantly how honestly the uncertainty on that estimate can be reported. A deep ensemble produces a predictive distribution rather than a point value, and the estimates are pooled across whatever stations happen to have returned data.
The comparison baselines are Mwp and mB_BB, both single-station methods in operational use. Neither reports an uncertainty. That absence is the argument: an estimate you cannot bound is an estimate you cannot act on.
A live record of prospective estimates, published before their reference solutions exist, is at /monitor/.
Presented at SSA 2026 (Pasadena) and AGU 2025. Manuscript in preparation; code and weights will be released with it.
No comments here, by design. If you have something to say — a correction, a different edition of the same source, results of your own — I would like to read it: hello@bmoyer.net.