Critical judgment in AI-enhanced weather forecasting

The rollout of various new AI weather models over the last year has been something of a blur and, now that the excitement of a cold winter is behind us, we thought it would be time to offer some thoughts from our unique perspective as a leading voice in the energy markets. 1. The AI models are quite useful, but are still not as good, in aggregate, as the better legacy NWP models, especially when looking at fields like 500 mb GPH. Discussions with our operational forecasters, who are in the trenches every day, suggest that the AI models are still used secondarily to the legacy models - "I don't use it other than a gut check/reference". My personal experience is that I still do not consult the AI models nearly as much as a good high-resolution NWP model/ensembles. Perhaps that will evolve with time, but that is the current perspective from those with an extreme level of skin in the game, those who are highly motivated to produce an accurate forecast. 2. However, there are many situations where the legacy models are still severely flawed, especially for 2-meter temperatures, where the AI models add considerable value. We know that the calculation of 2-meter temperatures in the legacy NWP models is a complex process involving highly imperfect parameterizations of surface energy exchanges/fluxes, which is especially complicated and difficult at night. Given that AI models are effectively very mathematically sophisticated analog models, trained on actual observations, they are not crippled by the same biases/errors that the legacy NWP models are. Further, there are certain well-known situations where even the best legacy models do poorly, such as southward-moving shallow and dense cold air masses in the lee of the Rockies and Appalachians, and we've seen multiple instances this past winter where AI models do astoundingly well, while legacy models can be 20-30 degrees off with mistimed cold fronts, etc. 3. The value of AI models relative to legacy models decreases with forecast horizon. An examination of forecast accuracy suggests that AI models can outperform legacy models in the 1-7 day window, but fall off considerably behind that. This applies when comparing both deterministic and ensemble mean solutions. In summary, we are excited to see the continued investment in this space, and are continuing to follow developments as we work to optimally integrate the new models into our product suite. However, we do caution that these new models are a complement, not a replacement, for legacy NWP models, at least for now. #atmosphericg2 #ai #weather

Human forecasters augmented by GenAI improve performance by 23% and vastly outperform AI-only predictions. Fascinating new research has uncovered important lessons, not just on Humans + AI forecasting, but more generally AI-augmented thinking. 🔮Human forecasters provided an LLM with a 'Superforecaster' prompt substantially improved their prediction performance. 📊In contrast to studies in other domains, the improvement was consistent across more and less skilled forecasters. 🔄Even the use of biased models improves performance to a similar degree, showing that the value was in providing additional perspectives to be assessed by human judgment. 💬Back-and-forth interaction is critical to value creation. Simple Humans + AI thinking processes such as incorporating predictions is of limited use. Forecasters using the models through their thinking process is high value. 🌈Prediction diversity is not degraded by use fo LLMs, with users not letting the models homogenize their thinking. 🚀Forecasting is an excellent use case and example for AI-augmented thinking. High-level human decision-making is highly complex and cannot be delegated to machines, but LLMs, used well, can substantially improve outcomes. The 'Superforecaster' prompt used in the study and a link to the pre-print paper are in the post. #foresight #forecasting #humansplusai #augmentedintelligence

🌎 𝗜𝘀 𝗔𝗜 𝘁𝗵𝗲 𝗿𝗲𝘃𝗼𝗹𝘂𝘁𝗶𝗼𝗻 𝗼𝗳 𝘄𝗲𝗮𝘁𝗵𝗲𝗿 𝗳𝗼𝗿𝗲𝗰𝗮𝘀𝘁𝗶𝗻𝗴? In my last talk I explored both sides: ✅ Advantages: speed, new data sources, hyperlocal predictions ⚠️ Risks: data biases, black-box models, overconfidence, rare/extreme events 👉 My conclusion: it’s not so much 𝗿𝗲𝘃𝗼𝗹𝘂𝘁𝗶𝗼𝗻 as an 𝗲𝘃𝗼𝗹𝘂𝘁𝗶𝗼𝗻. Numerical weather prediction and AI/data-driven models are not mutually exclusive — they’ll benefit a lot from each other. One example is #MieAI, a data driven approach to accelerate the mie back scattering effect of cloud aerosol developed by Gholam Ali Hoshyaripour and his team for the #ICON-ART weather model. Nevertheless the value of #AI weather forecast is limited because it is not fundamentally doing different things or solving problems related to weather. The results of the GraphDOP model that is only trained with real observations instead of reanalysis shows how important it is to have Knowledge about the mid-troposphere. The models performs well in the first 24-hours, afterwards the performance decreases compared to the IFS model. 🚀 Startups & Innovators to watch * Atmo – AI-driven forecasts that are more detailed, faster, more accurate * Tomorrow.io – real-time weather intelligence via IoT, radar satellites, interesting how they build up their own data * Skyfora – turning telecom towers into sensors for AI forecasts * Brightband - the Team around Daniel Rothenberg and the Scientist Ryan Keisler * Beyond Weather – long-range forecasting and impact modelling * alitiq - Forecasting - developing Data driven weather models for energy industry 🧭 My Take 🔵Synergy > replacement: physics-based + AI = the strongest path 🔵Transparency and robustness remain key challenges 🔵Hybrid models (physics + AI + data assimilation) look most promising 💡 I’d love to hear your thoughts: where do you see the biggest gains (or risks)? And do you think “evolution” captures it — or is there still room to call it a “revolution”? #AI #WeatherForecasting #ClimateTech #NumericalWeatherPrediction #MachineLearning

Update on AI and Decision-Making from the Harvard Business School: “AI can help leaders work faster, but it can also distort decision-making and lead to overconfidence. If you’re integrating AI tools into forecasting or strategy work, use these safeguards to stay grounded. 1) Watch for built-in biases. AI presents forecasts with impressive detail and confidence and tends to extrapolate from recent trends, which can make you overly optimistic. To counter this, make the system justify its output: Ask it for a confidence interval and an explanation of how the prediction could be wrong. 2) Seek peer input. Don’t replace human discussion with AI. Talk with colleagues before finalizing forecasts. Peer feedback brings emotional caution, diverse perspectives, and healthy skepticism that AI lacks. Use the AI for fast analysis, then pressure-test its take with your team. 3) Think critically about every forecast. No matter where advice comes from, ask: What’s this based on? What might be missing? AI may sound authoritative, but it’s not infallible. Treat it as a starting point, not the final word. 4) Set clear rules for how your team uses AI. Build in safeguards, such as requiring peer review before acting on AI recommendations and structuring decision-making to include both machine input and human insight.”  Posted July 11, 2025, on the Harvard Business Review’s Management Tip Of The Day. For more #ThoughtsAndObservations about #AI and the #Workplace go to https://lnkd.in/gf-d2xXN #ArtificialIntelligence #DecisionMaking