ERA5
PulseAugur coverage of ERA5 — every cluster mentioning ERA5 across labs, papers, and developer communities, ranked by signal.
10 day(s) with sentiment data
-
AI framework PARDIG enhances global precipitation estimates
Researchers have developed PARDIG, an AI framework utilizing a spherical graph-mesh transformer architecture to estimate global precipitation. Trained on diverse data including satellite imagery and atmospheric analyses…
-
Machine learning models improve weather front detection with reanalysis data
Researchers have developed machine learning models to detect weather fronts using satellite imagery, comparing a satellite-only approach with one that also incorporates reanalysis data. The model utilizing both satellit…
-
Machine learning weather models: distinguishing initial-condition and model error
A new research paper explores the distinction between initial-condition errors and model errors in machine learning weather models. The study evaluates gated and ungated optimizations in GraphCast and Pangu-Weather, mea…
-
New CARM framework enhances AI super-resolution for Earth observation data
Researchers have developed CARM (Conformal Adaptive Risk Masking), a new statistical framework designed to improve the reliability of deep learning-based super-resolution for Earth observation data. Unlike previous meth…
-
New ML framework automates weather front detection from satellite data
Researchers have developed SATFRONT, a machine learning framework designed to automatically detect atmospheric fronts from satellite imagery. This system is particularly useful for time-critical weather forecasting as i…
-
New Astra weather model uses less compute, beats ECMWF ensemble
Researchers have developed Astra, a novel probabilistic weather forecasting model that achieves high resolution with significantly reduced computational resources. Astra utilizes a two-sided transport attention mechanis…
-
New CD-FFM model enhances climate downscaling accuracy and speed
Researchers have developed a novel method called CD-FFM, a functional flow-matching model designed for climate downscaling. This model treats high-resolution climate data as a continuous random field, allowing for queri…
-
Astra model achieves high-resolution weather forecasts with reduced compute
A new probabilistic weather forecasting model named Astra has been developed, achieving high-resolution predictions with significantly reduced computational resources. Astra utilizes a novel two-sided transport attentio…
-
New model hindcasts wave spectra using data and physics guidance
Researchers have developed a data-driven and physics-guided hindcasting model for wave spectra, utilizing data from three Baltic Sea stations between 2017 and 2022. The model incorporates local wind, fetch, and depth, w…
-
New research proposes improved evaluation for time series generative models
A new research paper published on arXiv introduces a method to better evaluate generative models for multivariate time series data. The study highlights that standard evaluation techniques, which rely on pooled distribu…
-
New Transformer Model Enhances Weather Prediction with Sparse Data
Researchers have developed the Observation Assimilation Transformer (OAT), a novel neural network designed for probabilistic weather analysis. OAT refines numerical weather prediction fields by incorporating sparse obse…
-
GraphCast-Mamba enhances weather forecasts with Mamba correction branch · 2 sources tracked
Researchers have developed GraphCast-Mamba, a novel approach to weather forecasting that enhances existing models by incorporating a trainable Mamba correction branch. This new branch augments a pretrained GraphCast bac…
-
New DIAL method enhances weather forecast reliability by adjusting model randomness
Researchers have developed a novel method called Dynamic and Informed Adjustment of the Native Stochastic Law (DIAL) to improve the reliability of weather ensemble forecasts. DIAL works by dynamically adjusting the stoc…
-
AI weather models trained on synthetic data achieve competitive real-world skill
Researchers have developed a novel technique called "long-range distillation" to improve AI weather forecasting models. This method trains a "student" AI model using a massive synthetic dataset generated by a more compl…
-
New dataset and framework advance Earth system data-language alignment
Researchers have introduced EarthAlign, a large-scale dataset designed to bridge the gap between Earth system data and natural language. This dataset, spanning 2000-2024, combines expert-curated meteorological data with…
-
WeatherBridge adapter improves AI weather forecasts by bridging NWP and ERA5 data
Researchers have developed WeatherBridge, a novel adapter designed to improve the accuracy of pretrained weather forecasting models. This adapter bridges the gap between operational numerical weather prediction (NWP) in…
-
New ScaleCast framework aligns global and regional models for kilometer-scale weather prediction
Researchers have developed ScaleCast, a new framework for kilometer-scale regional weather prediction that aligns global and regional weather models. This framework uses a Global-Regional Conversion module to align repr…
-
New machine unlearning methods tackle bias and data deletion in time-series forecasting
Two new research papers introduce novel machine unlearning techniques tailored for time-series forecasting. The first paper, SSU-LSF, addresses bias in land surface forecasting models like Mamba by developing a framewor…
-
New ReRoute framework enables counterfactual predictions in scientific emulators
Researchers have developed ReRoute, a novel framework designed to make counterfactual predictions in scientific emulators without the need for controlled experiments. This method combines factual data with partial mecha…
-
Plant Pact tool predicts plant survival using weather data
An engineer has developed "Plant Pact," a tool designed to predict the likelihood of a gifted plant surviving based on its environment. The system analyzes plant species, the recipient's city, and specific window condit…