Alberto Ardid

New paper! We use ML on seismic data to forecast eruptions at Steamboat Geyser, Yellowstone. A step toward anticipating hydrothermal blasts—like the July 2024 Biscuit Basin event—and managing risks in geothermal fields worldwide. doi.org/10.1029/2024JH… #geothermal #AI #Geysers

🚨🌋 Excited to share this interview with @NatureComms editor Sebastian Mueller, where we talk about our latest work on forecasting volcanic eruptions using transfer learning. 🎙️ Interview: youtube.com/watch?v=5NFC2r… 📄 Paper: nature.com/articles/s4146…

@AlbertoArdid and colleagues present new work in which they use transfer machine learning methods that help forecasting volcanic eruptions, also at less monitored volcanoes. nature.com/articles/s4146…

Graduado del @Dgf_uchile, Alberto Ardid (@AlbertoArdid), recibió Premio de Geofísica de Nueva Zelanda por sus contribuciones a la disciplina📌Accede a la noticia en: uchile.cl/ig225681

Our new AI forecast model uses data from 24 volcanoes, showing seismic patterns at well-monitored sites can help predict eruptions at data-scarce ones—potentially lowering risk mitigation costs. Open-access tools: doi.org/10.1038/s41467… @DavidEDempsey @scronin70 @CoCaudron

Un equipo internacional liderado por el geofísico chileno Dr. Alberto Ardid desarrolló un sistema de IA para predecir incendios forestales en tiempo real, mejorando un 47% la precisión respecto a métodos tradicionales. Aún en pruebas para adaptarse a distintos entornos @latercera

🚨🔥New AI-powered wildfire forecasting tool! doi.org/10.1071/WF24113 The model analyses weather data every 30 mins predicting wildfire potential days in advance—47% more accurate than current methods, offering new alternatives to manage fires🚒🌲 Exciting milestone for our group

Also I have a poster on 🔹 Teaching ML through Flood Forecasting (Poster) 📅 Tue, 10 Dec | 1:40-5:30 PM EST Looking forward to connecting with colleagues, discussing these projects, I’d love to catch up!

#AGU24 next week! I'll be sharing our latest research in forecasting natural hazards using time series and ML : 🔹 Forecasting Volcanic Eruptions (talk) 📅 Wed, 11 Dec | 9:20 AM EST 🔹 Forecasting Wildfires (talk) 📅 Thu, 12 Dec | 4:40 PM EST #timeseries #ML #PredictiveAI

@tut_ml @KirkDBorne @antgrasso @Ronald_vanLoon How is a decision tree a regression model?

Best ML Tutorials- mltut.com/best-courses-f… @KirkDBorne @antgrasso @ronald_vanloon #MachineLearning #DeepLearning #BigData #Datascience #ML #HealthTech #DataVisualization #ArtificialInteligence #SoftwareEngineering #GenAI #deeplearning #ChatGPT #OpenAI #GPT5 #AI #GPT4 #gpt24

@JudithGeology Hi. What’s that other place?

I will no longer be posting on this platform. To stay up to date about earthquakes, either subscribe to my blog, or find me in that other place, where the sky is blue.

hot off the press (just finished teaching)...here are the notes from my part of a 2nd-year course 'engineering-centric machine learning'.... was pretty interesting to develop & teach for me. Goal was to look at intersection of ML and traditional eng. github.com/omaclaren/open…

🚨 Curious about detecting and analyzing unusual seismic events like this? 🌍🔍 Join our PhD program @GeoUnipd! ⏳ Deadline: 21 November 2024 🔗 More info: sites.google.com/view/piero-pol… #Seismology #PhD #ResearchOpportunity #ClimateScience

@JudithGeology jaja you are right. I'm just asking the typical stuff that you get from annoying reviewers. Nothing that your nice plots already tell.

@AlbertoArdid ...that kind of data doesn't exist. Maybe we need machine learning experts to review machine-learning-seismology papers, who know what tests are needed to demonstrate effectiveness.

In a recent paper in Nature Communications, Dr. Társilo Girona and Dr. Kyriaki Drymoni claim that their machine learning algorithm can predict upcoming large earthquakes. We tested the method, and found that it does not work. Read more on our blog; link in my bio.

@JudithGeology Thanks for taking the time to review papers like this, @earthquakeinsights, @JudithGeology, and Kyle Bradley. I find it challenging to make time to even read them myself 😅.

@JudithGeology Finally, calculate the AUC to assess performance. This will allow you to test whether the model classifies pre-EQ data better than a random model (which has an AUC of 0.5).

@JudithGeology Thanks for taking the time to review papers like this, @earthquakeinsights @JudithGeology Kyle Bradley. I find it challenging to make time for even reading them myself 😅

Around one year ago with automatic global analysis of seismological data we identified this enigmatic signal. Now we solved the mystery science.org/doi/10.1126/sc…

@JudithGeology 🥲

If you've been hearing about this story about a mysterious seismic signal on the news and want to know more, this video by the authors is a great explainer! youtube.com/watch?v=60T9TK…