devchef87/seismic-lab

SeismicLab is a real-time earthquake monitoring dashboard built by a software engineer who got curious about seismology. The README is candid about this: the author is not a domain expert, and the experimental prediction models should not be used for real-world safety decisions. The project is open source partly to invite input from people who know seismology. The platform pulls data from 27 different sources across five categories: seismic monitoring stations, ocean buoy pressure sensors, solar and space weather feeds, geomagnetic indices, and measurements of ground deformation and thermal activity near volcanoes. It aggregates about 108 million data points with an 11-year historical archive from 2015 to the present, downloaded from a hosted dataset on Hugging Face. The full database is about 32 gigabytes. The live dashboard shows a real-time map of earthquake activity, waveforms from 53 seismic stations (24 of which are live-streamed), ocean buoy readings, volcanic activity, and an experimental forecasting panel. The forecasting panel scores each new earthquake as it arrives and estimates the probability that a larger event follows in the same location within seven days. The model works by analyzing the pattern of recent seismic activity at that location: how many events occurred, how their magnitudes relate to each other, and whether they follow recognizable escalation shapes. Patterns like a staircase (three or more consecutive magnitude increases) or a rumble (one large event towering over a cluster of smaller ones) are associated with much higher escalation rates. The model reports 0.87 AUC across 13 zones, compared to 0.54 for an earlier approach that polled geographic zones on a schedule. Environmental signals like solar wind and ocean pressure were tested as additional inputs but added almost no predictive value, so the final model uses only 25 earthquake catalog features. Setup requires Python, a download of the dataset from Hugging Face (about 32 gigabytes), and running a server command. A shell script starts the live real-time background services.