Maximising Weather Forecasting Accuracy through the Utilisation of Graph Neural Networks and Dynamic GNNs
In one sentence. Compares graph neural networks with traditional ML for multivariate weather forecasting from heterogeneous sensors; the abstract reports no quantitative results.
Abstract
Weather forecasting is an essential task to tackle global climate change. Weather forecasting requires the analysis of multivariate data generated by heterogeneous meteorological sensors. These sensors comprise of ground-based sensors, radiosonde, and sensors mounted on satellites, etc., To analyze the data generated by these sensors we use Graph Neural Networks (GNNs) based weather forecasting model. GNNs are graph learning-based models which show strong empirical performance in many machine learning approaches. In this research, we investigate the performance of weather forecasting using GNNs and traditional Machine learning-based models.
Labels
- Topic
- Global medium-range NWP
- Type
- Evaluation or intercomparison of existing models
- Method
- Graph neural network, Classical ML (trees, GP, linear, kernels)
- Time scale
- 3–15 days
- Domain
- Global
- Data
- Other
- Status
- Research
- Identifiers
- arXiv:2301.12471 OpenAlex:W4318752007
- Added
- 2026-10-22
Labels and summary are generated by Claude and reviewed weekly. Wrong? Edit this entry.