weather-forecasting

Microsoft introduces 'Microsoft Research Focus', a new video/podcast digest covering AI agents, everyday AI usage patterns, BioEmu2's public preview, and weather forecasting research.

NVIDIA announced that organizations across Europe, including AXA, Vaisala Xweather, University of Manchester, and TotalEnergies, are deploying its Earth-2 open AI weather and climate models, with the new StormScope EU regional nowcasting model joining the family at GTC Berlin.

本文评估了五个天气基础模型作为冻结编码器在增强岩石风化（ERW）碳去除量化中的迁移能力，其中基于 Aurora 1.5 构建的 TerraNova-ERW 在未见农田上表现最佳，支持少样本预测并可显著减少土壤采样需求。

Ai2 推出 HiRO-ACEv2 气候模拟系统，可在 3 km 分辨率下高精度复现极端降雨、风速和气压事件，成本仅为 X-SHiELD 等传统高分辨率大气模型的约 1/300。

微软研究院发布并发表于 Nature 的机器学习框架 Probabilistic Bias Correction (PBC)，通过学习现有天气模型的系统性误差并加以校正，显著提升两到六周的次季节天气预报准确性，在全球竞赛中夺冠，并已用于肯尼亚季风预报等实际场景。

MIT researchers and collaborators published a Nature paper describing a machine-learning model that corrects errors in existing weather forecasts, doubling accuracy in the two-to-six-week range and beating climatological averages for the first time. The approach also improves prediction of extreme weather events like flooding and cold spells, placing first in an international forecasting competition.

WindBorne Systems won a U.S. Air Force contract worth up to $49 million for atmospheric data collected by its fleet of ~700 autonomous weather balloons, which also feeds its WeatherMesh AI forecasting model. The deal was awarded without competitive bidding and marks the startup's largest government agreement to date.

Google DeepMind的AI天气模型提前11天预测到飓风Isaias，提前4天锁定登陆位置，并在提前3天时准确预测其峰值强度，可能为社区疏散和防灾争取更多时间。

提出 Tucker Bottleneck Attention（TuBA），利用低秩张量结构实现多维序列的高效全局 token 混合，在视频预测和全球天气预报上以显著更低的计算成本取得更优精度，最高减少 85.1% 计算量并实现 4.27 倍加速。

A study investigating post-training quantization (PTQ) on pre-trained AI weather emulators (DLWP and FourCastNet) to accelerate autoregressive inference and enable deployment on edge hardware, finding that quantized configurations retain qualitatively meaningful short-range forecasts.

Show HN: Nightwatch is an open-source macOS menu-bar app that forecasts clear-sky windows for astrophotography, plans observing targets for tonight and the week ahead, and integrates with telescopes like the DWARF Mini for imaging.

The paper introduces STCFormer, an adaptive spatio-temporal Transformer that dynamically clusters weather stations per temporal patch using a Cluster-Guided Attention Block combining local intra-cluster attention with global regional summaries, achieving the lowest 24-hour MSE across all eight temperature and wind forecasting tasks.

M2Weather introduces a benchmark for joint multi-station and multi-variable weather forecasting, collecting 2,809 stations across France, Europe, and Global scales, and proposing a plug-and-play adapter that improves forecasting by modeling station-variable relationships.

Researchers from KAIST, Kookmin University, and NYU propose a proper scoring rule-based diffusion framework for probabilistic weather forecasting that adds auxiliary conditional denoising tasks to CRPS-trained models, enabling improved forecast accuracy and calibration at longer horizons while keeping single-forward-pass ensemble generation.

StationPDE is a station-oriented surface PDE learning model for multi-station multivariate weather forecasting that constructs terrain-aware continuous fields and models physical dynamics to outperform state-of-the-art baselines with a 9.6% MSE reduction.

The article introduces the TACLS system, which uses machine learning and satellite data to improve flash flood warnings, potentially saving lives by providing earlier alerts.

Google highlights AI advancements to accelerate science and improve lives, including supporting 300 languages, mapping genetic changes, and enhancing weather prediction models.

In this podcast episode, Peter Battaglia and Fry explore WeatherNext 3, an AI model for weather forecasting, discussing its applications in hurricane tracking and renewable energy grid optimization, along with probabilistic forecasting and future outlooks.

WeatherNext 3 is DeepMind's most advanced global weather AI model, capable of generating hourly forecasts with 5-10km resolution for applications like renewable energy and integration into Google products.

WeatherNext 3 is an advanced AI weather forecasting model from Google that incorporates real-time satellite data, hourly refreshes, and precise precipitation forecasting, now integrated into Google services like Search, Gemini, Maps, and Cloud.