Georg Huettenegger’s Post

Lawrence Berkeley National Laboratory scientists are using AI and historical US Geological Survey maps to locate undocumented and often hazardous orphaned oil and natural gas wells across the United States. Their AI algorithm can recognize oil and gas well symbols on maps, and the potential well locations can then be verified through satellite imagery and field surveys. #berkeley #ai #maps #orphanwells #oilandgasindustry

Berkeley researchers have created an AI tool to find lost oil and gas wells in the US. There are between 300,000 and 800,000 undocumented wells across America, but they could all be located by using the AI tool to search 19th and 20th century maps. https://lnkd.in/emxRHKAj

Here's an important AI development that could save lives and help me decide which t-shirt to wear. https://lnkd.in/g-dfBarR

Incredible! #AI 🤖 and #climate #science 🌍! “Here we introduce Aurora, a large-scale foundation model for the Earth system trained on over a million hours of diverse data. Aurora outperforms operational forecasts for air quality, ocean waves, tropical cyclone tracks, and high-resolution weather forecasting at orders of magnitude smaller computational expense than dedicated existing systems.”

Yesterday, two investors asked me how AI will affect our business model for earlier and more accurate forecasts. We use AI, so that is the starting point. And training AI with models that integrate our data will beat the pants off of AI models that do not integrate our data. AI helps us and is not a threat to what we are doing, however… The problem I foresee, and my biggest concern, is AI is only as good as the data you feed it. I think there could be a point where we decrease the investment in weather collection (sensors, satellites, stations, etc) in favor of AI and if we do that, we will be in big trouble. #ClimateChange is making the weather more difficult to predict. Training AI on past data and not continuously integrating hard, measured data as things are becoming increasingly volatile could be a disaster. Last, it’s not simply about forecasting weather. One has to apply it in a practical way that can give businesses and industries an edge. https://lnkd.in/eE_fChec

Water supply forecasting 💧 is a major issue in the Western U.S., with snow patterns and temperatures changing in ways we haven't seen before. Water managers need better tools for planning and operating large supply reservoirs, anticipating heavy precipitation events and storing water efficiently for the season. Join us for a webinar to learn how machine learning and AI methods produce highly accurate and easy-to-use forecasts, and for a discussion on the big AI opportunities in forecasting that we think you should know about. There will be time for Q&A at the end of the session. Register here 👉 https://hubs.la/Q02qzHC40 #hydrology #AI #machinelearning #watermanagement #waterutilities #climateadaptation #energytransition

Google's AI-powered weather model 'NeuralGCM' is 100,00 times more efficient that other models for simulating the atmosphere. It combines methods of machine learning and neural networks with traditional physics-based modeling. What an incredible leap forward to help the insurance industry with risk and resilience planning! The model is also open-source and can run on a laptop- super cool! https://lnkd.in/gAUzhZBM #google #AI #weather #machinelearning #insurance #insurancenews

One of the nicest things to be proud of working at Google, is that you realize there are teams within the org. working on hardest problems of society & science... one of this is #Bellwether (a.k.a, how to predict critical weather scenarios), now an "official" X.Moonshot (so 'it's no longer a lab project, but a real initiative) If curious scientist/engineers can help the world with #Data & #AI, anyone can help their company/community also with #Data & #AI... and if you need help, go get help from any local #Googler https://lnkd.in/gHigjgvJ

Exploring Super Resolution Data Assimilation (SRDA) with Antoine Bernigaud At #OceanPredict2024, Antoine from Nansen Environmental and Remote Sensing Center (NERSC) is presenting groundbreaking work on Super Resolution Data Assimilation (SRDA), a novel approach developed at NERSC for Arctic and North Atlantic modelling. This method combines a "low"-resolution coupled model (TOPAZ 18-km) with a neural network (a residual U-Net) to emulate high-resolution results akin to the TOPAZ-6km model—at a fraction of the computational cost. By forecasting with the LR model, interpolating outputs to HR grids, and refining them with a neural network, the approach assimilates high-resolution observations, delivering computationally efficient, HR-quality insights. An exciting step forward in precision ocean modelling! 🌎 #NECCTON #DataAssimilation #OceanModelling