Indian met office uses AI models for monsoon prediction

This year about 38m Indian farmers received monsoon forecasts generated by AI. Once trained, the models can run on a high-end laptop. This could lead to a “democratisation of weather forecasting”, says one expert

Weather forecasting has got to be one of the most complex projects we undertake, especially during critical seasons like the monsoon. Back in 2022, Nvidia began building the first AI-based weather models trained on decades of historical weather data. This was different from existing compute-heavy forecasting models that assign variables for pressure, temperature, windspeed, etc. and run countless simulations to provide forecasts. The AI isn't perfect; AI models, in their vacuum, still have trouble accounting for those 3-sigma weather events not well documented in the historical records. So this year, by situating two models representing the traditional and AI approaches in tandem (the latter based on Google’s NeuralGCM) 38m farmers in India got some of the most accurate and advanced forecasts for timing and severity on record. The best part: this is all being done in a fraction of the compute time, meaning poorer countries will have much better access to these tools. If this is happening with complex, multi-variable weather forecasting, it can happen with your payments. Turn that mountain of structured and unstructured payment (and payment-adjacent) data into insights on where, when, and how to optimize your payments. We just acquired Congrify to bring better, faster, AI- assisted modeling for you payments, check it out. #SaveNOAA AI models ace their predictions of India’s monsoon rains

𝐇𝐔𝐑𝐑𝐈𝐂𝐀𝐍𝐄 𝐌𝐄𝐋𝐈𝐒𝐒𝐀 𝐈𝐒 𝐁𝐑𝐄𝐀𝐊𝐈𝐍𝐆 𝐏𝐑𝐄𝐃𝐈𝐂𝐓𝐈𝐕𝐄 𝐌𝐎𝐃𝐄𝐋𝐒 Hurricane Melissa isn't just a weather event; it's a massive, real-time stress test for our most advanced predictive technologies. Meteorologists are on edge because the storm's erratic behavior is pushing our AI-driven forecasting systems to their absolute limits, exposing a potential weakness in our reliance on algorithms. This is a critical moment for computational meteorology. The data is pouring in, but the predictions are struggling to keep up with Melissa's unprecedented nature. ► 𝐖𝐡𝐚𝐭'𝐬 𝐜𝐡𝐚𝐥𝐥𝐞𝐧𝐠𝐢𝐧𝐠 𝐭𝐡𝐞 𝐭𝐞𝐜𝐡: → 𝑈𝑛𝑝𝑟𝑒𝑐𝑒𝑑𝑒𝑛𝑡𝑒𝑑 𝐷𝑎𝑡𝑎: The storm is generating novel atmospheric data patterns that current machine learning models were not trained on, causing significant forecast drift. → 𝐴𝐼 𝐶𝑜𝑛𝑓𝑖𝑑𝑒𝑛𝑐𝑒 𝐷𝑟𝑜𝑝: Major global weather prediction systems, which normally operate with high confidence, are flagging Melissa's trajectory with unusually high uncertainty margins. → 𝐶𝑜𝑚𝑝𝑢𝑡𝑎𝑡𝑖𝑜𝑛𝑎𝑙 𝐵𝑜𝑡𝑡𝑙𝑒𝑛𝑒𝑐𝑘: The sheer volume and velocity of sensor data are straining even supercomputer resources, forcing teams to prioritize models and sacrifice resolution. We're witnessing a fascinating battle between the raw power of nature and the logical precision of our best algorithms. It’s a stark reminder that even in the age of AI, the planet can still throw us a curveball that no one saw coming. What do you think is the next frontier for AI in predicting these "black swan" weather events? #AI #PredictiveAnalytics #ClimateTech #BigData #Meteorology

🧠 Learn with Mohammed – Revolution of AI ⚡ AI in Meteorology – Forecasting the Future with Precision 🌦️🌪️ Artificial Intelligence is transforming how we understand the skies. By analyzing millions of weather data points, AI can predict storms, rainfall, and climate shifts with higher accuracy than ever before — giving communities more time to prepare and protect lives. ⛈️ 💡 How AI Revolutionizes Weather Forecasting: • Predicts cyclones, floods, and droughts faster and more precisely 🌊 • Uses satellite imagery to analyze cloud movement and storm intensity ☁️ • Models long-term climate trends for sustainable planning 🌍 • Supports farmers with AI-based rainfall and crop predictions 🌾 • Enhances disaster preparedness through early alerts 🚨 📍 Real Examples: • IBM’s Watson analyzes global weather data to deliver minute-by-minute forecasts ⏱️ • Google’s DeepMind provides real-time rain predictions using radar images 🌧️ • The UK Met Office uses AI to simulate extreme weather patterns 🌀 💡 Why It Matters: AI in meteorology isn’t just about predicting weather — it’s about protecting humanity, supporting agriculture, and helping governments prepare for natural disasters before they strike. 🌤️ 📌 Your Takeaway Today: AI helps us read the sky like a scientist — turning uncertainty into preparedness. 🌈 ✨ Question for You: Would you trust an AI system to make weather-based decisions for your city’s safety? 👇 #LearnWithMohammed #AI #Day91 #Weather #Meteorology #Forecasting #ClimateTech #Innovation #Technology #DeepLearning #Satellite #ClimateChange #Sustainability #Storm #DataScience #Environment #Rain #Cyclone #Future #AIRevolution

Good read. From the article: This year, though, the Indian government tried something new. About 38m farmers received forecasts generated not by an NWP model, but by ones powered by artificial intelligence (AI) instead. These work in a different way: rather than going to the trouble of trying to simulate, equation by equation, exactly what is going on in the atmosphere, they mostly make their predictions by comparing the patterns they see in weather data with previous, similar patterns in the historical weather records on which they have been trained. The models aced their test. In some regions, they predicted when the rain would arrive 30 days ahead. They also forecast that rainfall would stall in the middle of the season—as it did, for 20 days—despite this not appearing in the NWP forecasts. Almost half of the farmers who paid attention to the messages later said that the information influenced their decisions on what to plant and when—though it is still too early to assess whether this will influence their eventual earnings. AI models ace their predictions of India’s monsoon rains https://lnkd.in/gk_mTFct

As wildfires intensify and spread unpredictably, new modelling technologies are becoming vital tools for prevention and response. From hybrid AI/physics models to rich datasets, researchers across Canada and California are pushing the boundaries of what wildfire prediction can achieve. In Waterloo, Firebird is building a state-of-the-art wildfire spread platform that combines satellite imagery, real-time weather data, and deep learning with physics-based components to forecast ignition zones and dynamic fire spread. Meanwhile, Professor Joshua Pulsipher’s lab is developing surrogate models using physics-informed neural operators that are both fast and accurate for estimating the uncertainty. Meanwhile in California, the game is being elevated across multiple fronts. For example, the California Wildfire Inventory (CAWFI) dataset has tens of millions of indicator points that support AI models that can correctly predict large wildfire events. This system helps allow early detection and more informed resource allocation to prevent wildfires. What does this means for communities and agencies? Faster, more precise forecasts, better risk mapping, and improved capacity to act before fires get out of hand. These technologies are increasingly applicable in practice for evacuation decisions, insurance risk assessment, land planning, and Firebird’s own mission. Curious how modelling tech is reshaping wildfire resilience? Check out recent work by Firebird, Dr. Pulsipher, and California datasets, such as CAWFI, to see their published studies and open datasets for in-depth analysis: Firebird: https://lnkd.in/g67DPqjp Pulsipher: https://lnkd.in/eafUJtmT  UW FRF: https://lnkd.in/eqEhcyUe  CAWFI: https://lnkd.in/ez8ZThV8 #Firebird #wildfire #aiintegration

The Ai Revolution in Weather Forecasting Is Here AI-driven advances revolutionize weather forecasting by enabling faster, more accurate predictions globally. New models harness deep learning to improve storm tracking and climate risk assessments, aiding governments and communities in better disaster preparedness and response. For more information: https://lnkd.in/d2WwhRg7 #AI #Global #Weather #Tech #Innovation

AI is driving a new frontier in climate intelligence. After transforming India’s monsoon forecasting and benefiting 38 million farmers, NeuralGCM is now being scaled to 30 countries. The University of Chicago has received support from the Gates Foundation to benchmark existing models over East and West Africa with a focus on rainy seasons and heatwaves. Pedram Hassanzadeh calls it “the beginning of an AI-driven, second revolution in weather forecasting.” Read how partnerships are redefining global resilience: https://lnkd.in/dcvFBsNQ

Today, Google is sharing our latest advancements for Google Earth AI—a fundamental leap in planetary understanding achieved through specialized AI models and agentic reasoning. This includes new research on Geospatial Reasoning, a framework powered by Gemini that lets AI automatically connect different Earth AI models—such as weather forecasts, population density, and satellite imagery—to answer complex, multi-modal questions. Personally, I'm excited by the many possibilities to apply this new tech in public health, crisis response, and resilience! We are already collaborating with partners like the Institute for Disease Modeling, University of Oxford, World Health Organization Regional Office for Africa and Cooper/Smith to evaluate how these geospatial models and agentic capabilities can improve infectious disease modeling and public health action. Read our Research Blog: https://lnkd.in/eUMcwE8H Check out the Earth AI website: https://ai.google/earth-ai

I’ve been often quoted in reference to challenges in health and healthcare as drowning in data and lacking in insights, now we can look for actionable intelligence with Google Earth AI.