Learning Robust Precipitation Forecaster by Temporal Frame Interpolation (2311.18341v2)
Abstract: Recent advances in deep learning have significantly elevated weather prediction models. However, these models often falter in real-world scenarios due to their sensitivity to spatial-temporal shifts. This issue is particularly acute in weather forecasting, where models are prone to overfit to local and temporal variations, especially when tasked with fine-grained predictions. In this paper, we address these challenges by developing a robust precipitation forecasting model that demonstrates resilience against such spatial-temporal discrepancies. We introduce Temporal Frame Interpolation (TFI), a novel technique that enhances the training dataset by generating synthetic samples through interpolating adjacent frames from satellite imagery and ground radar data, thus improving the model's robustness against frame noise. Moreover, we incorporate a unique Multi-Level Dice (ML-Dice) loss function, leveraging the ordinal nature of rainfall intensities to improve the model's performance. Our approach has led to significant improvements in forecasting precision, culminating in our model securing \textit{1st place} in the transfer learning leaderboard of the \textit{Weather4cast'23} competition. This achievement not only underscores the effectiveness of our methodologies but also establishes a new standard for deep learning applications in weather forecasting. Our code and weights have been public on \url{https://github.com/Secilia-Cxy/UNetTFI}.
- 3d u-net: Learning dense volumetric segmentation from sparse annotation. In Sébastien Ourselin, Leo Joskowicz, Mert R. Sabuncu, Gözde B. Ünal, and William M. Wells III, editors, MICCAI, volume 9901 of Lecture Notes in Computer Science, pages 424–432, 2016.
- Deep ordinal regression network for monocular depth estimation. In CVPR, pages 2002–2011. Computer Vision Foundation / IEEE Computer Society, 2018.
- Weather4cast at neurips 2022: Super-resolution rain movie prediction under spatio-temporal shifts. In NeurIPS 2022 Competition Track, pages 292–313. PMLR, 2022.
- The capacity and robustness trade-off: Revisiting the channel independent strategy for multivariate time series forecasting. arXiv preprint arXiv:2304.05206, 2023.
- Benchmarking neural network robustness to common corruptions and perturbations. arXiv preprint arXiv:1903.12261, 2019.
- Shruti Jadon. A survey of loss functions for semantic segmentation. In CIBCB, pages 1–7. IEEE, 2020.
- Super-resolution probabilistic rain prediction from satellite data using 3d u-nets and earthformers. arXiv preprint arXiv:2212.02998, 2022.
- Peter Lynch. The origins of computer weather prediction and climate modeling. Journal of computational physics, 227(7):3431–3444, 2008.
- Weatherfusionnet: Predicting precipitation from satellite data. arXiv preprint arXiv:2211.16824, 2022.
- U-net: Convolutional networks for biomedical image segmentation. In MICCAI, pages 234–241. Springer, 2015.
- Domain generalization strategy to train classifiers robust to spatial-temporal shift. arXiv preprint arXiv:2212.02968, 2022.
- Convolutional lstm network: A machine learning approach for precipitation nowcasting. NIPS, 28, 2015.
- Metnet: A neural weather model for precipitation forecasting. arXiv preprint arXiv:2003.12140, 2020.
- Generalised dice overlap as a deep learning loss function for highly unbalanced segmentations. In Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support - Third International Workshop, DLMIA 2017, and 7th International Workshop, ML-CDS 2017, Held in Conjunction with MICCAI, volume 10553 of Lecture Notes in Computer Science, pages 240–248. Springer, 2017.
- mixup: Beyond empirical risk minimization. In ICLR. OpenReview.net, 2018.