This project is in collaboration with Open Climate Fix, and will be co-supervised by Prof. Robert Mullins from the University of Cambridge and Jacob Bieker from Open Climate Fix.
Decades of steady improvements in the quantity and types of observational data, better modeling techniques, and more computational power have resulted in increasingly accurate weather forecasts and growing adoption of Numerical weather prediction in real-world applications. There is also now a rising interests of using Machine Learning models to help on the weather prediction problem.
Recent work (Forecasting Global Weather with GNNs) has demonstrated a new possibility of weather forecasting using graph neural networks. The use of message-passing GNNs helps to handle the spherical geometry of the earth in a more straightforward manner.
Open Climate Fix has been working on collecting large-scale weather data and building open weather models. They have recently also provided an open-source implementation of this weather GNN model.
Forecasting Global Weather with Graph Neural Networks
openclimatefix (Open Climate Fix)
GitHub - openclimatefix/graph_weather: PyTorch implementation of Ryan Keisler's 2022 "Forecasting Global Weather with Graph Neural Networks" paper (https://arxiv.org/abs/2202.07575)
MetNet: A Neural Weather Model for Precipitation Forecasting
The GNN architecture used for weather forecast may not be the optimal architecture given the problem setup. The base of the project would be manually explore other alternative network architectures on the same dataset and compare their performances with several baseline models. The goals would be:
Re-implement the CNN baseline and also GNN used in the paper (Forecasting Global Weather with GNNs), and compare their performances.
Explore other advanced GNN architectures, an exhaustive list of the architectures can be found in the following link.
Try to build a GNN with ‘meta-layers’. Each meta-layer can be a set of different best-performing GNN layers, averaged together in a form of:
$y = \frac{1}{M}\sum_{m=0}^{M}f_{m}(x)$, where each $f_m$ can be a different GNN layer.
Network Architecture Search for Weather Prediction. The student would have to try to re-implement the following paper on the weather modeling data.
End-to-end model optimization. The GNN model outputs are normally fed to networks such as the MetNet to make further predictions. The student will investigate how to design novel AutoML or NAS techniques on the combined model architecture.