The ground is now being asked to file its own shake estimate
the ground is now being asked to file its own shake estimate: a mamba based network reads single station waveforms for magnitude with the best scores on the stead benchmark, mp net fuses time and frequency branches end to end, and a pegs model from japan sizes up magnitude 8.3 plus earthquakes faster than existing early warning systems.
the fault now files its own alert.
Context
SeisMamba (August 25, 2026) estimates magnitude from single-station, three-component waveforms. Its STEAD comparison reports the best MSE, RMSE and R squared among the six tested models, but AMAG has the lower mean absolute error. Inference takes 0.55 milliseconds per batch of 32 waveforms on an NVIDIA T4; the input windows are 30 seconds long.
PEGSNet, published in Nature on May 11, 2022, learns prompt elastogravity signals from synthetic Japanese megathrust earthquakes mixed with recorded noise. The authors set a conservative sensitivity floor at magnitude 8.3. A retrospective replay of the 2011 Tohoku earthquake estimates magnitude faster than the comparison systems.
Fast model inference is not the same as end-to-end warning latency. SeisMamba still calls for calibrated uncertainty and operational decision rules. PEGSNet uses a regional network, not a single station, and its 8.3 floor depends on noise conditions; it is not a guarantee for every earthquake above that magnitude.
The MP-Net abstract was located in the OUCI bibliographic index under DOI 10.1093/gji/ggag204, but the primary publisher text could not be read. Its time-frequency architecture remains unverified from a primary source here, not refuted. The three studies should not be treated as one deployed warning system.
Watch next
- Primary publisher access for MP-Net and its evaluation protocol. Prospective warning performance, false alarms, and uncertainty outside the training regions.
Sources
Provenance
The note above is reproduced unedited from the original post, first published on Threads on 4 October 2026 at 07:47 IST. Sources are the papers and datasets the note draws on.
View the original post ↗Embed this note
More notes
The air is now being asked to keep its own ledger
the air is now being asked to keep its own ledger: ecmwf’s aifs compo becomes the first ai model to forecast atmospheric composition globally every three hours, cleanair simulates 365 days of pm2.5 over china in ten seconds, and a unified framework maps six pollutants at one kilometer across the whole country. the air now files its own composition report.
read the note →The current is now being asked to draw its own map
the current is now being asked to draw its own map: china’s langya 2.0 predicts six ocean phenomena including internal waves and mesoscale eddies, a deep net called wenhai resolves eddies globally with air sea flux formulas built in, and scripps infers surface currents from the way temperature patterns deform in satellite images. the ocean now files its own circulation report.
read the note →The soil is now being asked to report its own carbon
the soil is now being asked to report its own carbon: a nix color sensor paired with generative data augmentation predicts soil organic carbon without a lab, random forest drives 74 percent of soil health mapping studies, and sentinel 2 tracks five year carbon change across france and italy from 922 samples. the dirt now files its own carbon account.
read the note →