Bay Area Faces Storms, Flooding, and Minor Quake
The San Francisco Bay Area is experiencing ongoing severe weather, with flood advisories issued for Marin, Napa, and Sonoma counties. Forecasts include hail, strong gusts, and potential snow in some areas. Separately, a small 2.8 magnitude earthquake struck near Bayview on February 17, serving as a reminder of the region's seismic risk.
- The current weather is part of a series of atmospheric rivers, a phenomenon that companies like HydroForecast are now predicting with greater accuracy by using physics-informed machine learning models to improve streamflow and flood forecasts. - Locally, the AI and climate-tech intersection is growing; Walnut Creek startup Gridware, co-founded by a UC Berkeley engineer, uses AI to monitor vibrations on utility poles, alerting power companies to hazards like falling tree branches in high winds to prevent grid damage and wildfires. - Machine learning has revolutionized seismology by automating the detection of tiny, previously invisible "microquakes." Stanford's "Earthquake Transformer" model, trained on a public dataset of 1.2 million seismic events, can find 10 times more quakes than traditional methods, revealing a clearer picture of fault behaviors. - Research is underway to improve the USGS ShakeAlert system by using deep learning to integrate sensor data from traditional seismic networks with crowdsourced data from smartphone apps like MyShake, which could potentially speed up earthquake detection by an average of 2.2 seconds. - In a recent project, Stanford researchers collaborated with Italian scientists, using an AI model to analyze seismic data from a volcanic region; the model detected four times as many earthquakes as previous methods, which allowed them to map previously unknown faults. - NOAA recently deployed a new generation of AI-driven global weather models, one of which is called AIGFS; a single 16-day forecast using this model requires only 0.3% of the computing resources of its traditional counterpart and finishes in about 40 minutes, a significant leap in efficiency for predicting large-scale weather patterns. - The field of earthquake forecasting is an active area of machine learning research, though AI models have not yet consistently outperformed traditional methods for prediction, they are excelling at expanding earthquake catalogs which helps illuminate the risks of future quakes. - Bay Area universities are central to AI-driven seismology research; a UC Berkeley professor's lab focuses on applying deep learning to large seismic datasets to discover hidden earthquake signals and better understand complex fault zone structures.
