TemporalMixFormer: Spatiotemporal Spectral Unmixing for Multi-Date Hyperspectral Earth Observation Using Dynamic State-Space Networks
Keywords:
hyperspectral unmixing, spatiotemporal modeling, state-space networks, multi-date Earth observation, dynamic systems, deep learning, spectral mixture analysis, land cover change, remote sensing, data governanceAbstract
Multi-date hyperspectral Earth observation provides rich spectral and temporal information for land cover monitoring, but the spatial resolution of such sensors often results in mixed pixels where multiple materials contribute to a single spectral measurement. Spectral unmixing, the process of decomposing mixed pixels into constituent endmembers and their abundances, becomes significantly more challenging when temporal dynamics are introduced because land cover transitions, phenological cycles, and varying illumination conditions induce non-stationary mixing patterns over time. Existing unmixing approaches typically process each date independently or rely on simplistic temporal smoothing, failing to capture the underlying state-space structure governing the evolution of abundances. This paper presents TemporalMixFormer, a novel architecture that integrates state-space modeling with multi-head attention mechanisms to perform spatiotemporal spectral unmixing across multi-date hyperspectral imagery. TemporalMixFormer treats abundance trajectories as latent states that evolve according to a learned dynamic system, while a transformer-inspired mixing module adaptively weights spectral contributions from different acquisition dates. The system is designed as a scalable, modular framework suitable for large-scale operational deployment, with explicit considerations for computational efficiency, sensor interoperability, and temporal irregularity. We discuss architectural trade-offs between expressiveness and inference speed, the role of prior knowledge in regularizing abundance dynamics, and strategies for ensuring robustness under cloud cover, atmospheric variability, and missing observations. The paper further examines the broader implications of deploying such a model within Earth observation data cubes, including data governance, fairness across geographic regions, environmental sustainability of model training, and policy alignment with global monitoring initiatives such as the Sustainable Development Goals. By uniting state-space theory with modern deep learning components, TemporalMixFormer offers a principled path toward temporally coherent, physically interpretable unmixing at continental scales.
References
1. Keshava, N., & Mustard, J. F. (2002). Spectral unmixing. IEEE Signal Processing Magazine, 19(1), 44–57.
2. Bioucas-Dias, J. M., Plaza, A., Dobigeon, N., Parente, M., Du, Q., Gader, P., & Chanussot, J. (2012). Hyperspectral unmixing overview: Geometrical, statistical, and sparse regression-based approaches. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 5(2), 354–379.
3. Heylen, R., Parente, M., & Gader, P. (2014). A review of nonlinear hyperspectral unmixing methods. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 7(6), 1844–1868.
4. Eches, S., Dobigeon, N., & Tourneret, J.-Y. (2011). Bayesian unmixing of hyperspectral images with temporal regularization. IEEE Transactions on Geoscience and Remote Sensing, 49(12), 5060–5073.
5. Chen, J., Richard, C., & Honeine, P. (2013). Nonlinear unmixing of hyperspectral data based on a linear mixture of nonlinear endmembers. IEEE Transactions on Geoscience and Remote Sensing, 51(10), 4923–4935.
6. Gu, A., Goel, K., & Ré, C. (2022). Efficiently modeling long sequences with structured state spaces. In Proceedings of the International Conference on Learning Representations (ICLR).
7. Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., & Polosukhin, I. (2017). Attention is all you need. In Advances in Neural Information Processing Systems, 30.
8. Dobigeon, N., Tourneret, J.-Y., Richard, C., & Honeine, P. (2014). Nonlinear unmixing of hyperspectral images: Models and algorithms. IEEE Signal Processing Magazine, 31(1), 82–94.
9. Heinz, D. C., & Chang, C.-I. (2001). Fully constrained least squares linear spectral mixture analysis method for material quantification in hyperspectral imagery. IEEE Transactions on Geoscience and Remote Sensing, 39(3), 529–545.
10. Iordache, M.-D., Bioucas-Dias, J. M., & Plaza, A. (2011). Sparse unmixing of hyperspectral data. IEEE Transactions on Geoscience and Remote Sensing, 49(6), 2014–2039.
11. Palsson, B., Sigurdsson, J., Sveinsson, J. R., & Ulfarsson, M. O. (2017). Hyperspectral unmixing using a neural network autoencoder. IEEE Geoscience and Remote Sensing Letters, 15(2), 242–246.
12. Hong, D., Yokoya, N., Chanussot, J., & Zhu, X. X. (2021). Learning to propagate labels on graphs: An iterative multitask learning framework for semi-supervised hyperspectral image classification. IEEE Transactions on Geoscience and Remote Sensing, 59(4), 3356–3371.
13. Halimi, A., Altmann, Y., Dobigeon, N., & Tourneret, J.-Y. (2015). A multitemporal linear unmixing model for hyperspectral data. IEEE Transactions on Image Processing, 24(11), 4410–4423.
14. Xu, Z., Wang, Q., & Li, X. (2022). Spatiotemporal hyperspectral unmixing via recurrent neural networks. Remote Sensing, 14(8), 1847.
15. Long, Z., Zia, A., Fu, G., Rolland, V., & Zhou, J. (2026). WS-Net: Weak-Signal Representation Learning and Gated Abundance Reconstruction for Hyperspectral Unmixing via State-Space and Weak Signal Attention Fusion. arXiv preprint arXiv:2603.09037.
16. He, X., Chen, Y., & Lin, Z. (2023). Spatio-spectral transformer for hyperspectral unmixing. IEEE Transactions on Geoscience and Remote Sensing, 61, 1–15.
17. Cheng, Y., Wang, Y., & Li, J. (2020). Hyperspectral image unmixing using a convolutional neural network with spatial information. Remote Sensing, 12(6), 990.
18. Dao, T., Fu, D. Y., Saab, K., & Re, C. (2023). FlashAttention: Fast and memory-efficient exact attention with IO-awareness. In Advances in Neural Information Processing Systems, 36.
19. Li, J., Du, Q., & Sun, X. (2021). Spectral-spatial attention network for hyperspectral image classification. IEEE Transactions on Geoscience and Remote Sensing, 59(8), 6850–6864.
20. Rasti, B., Scheunders, P., Ghamisi, P., Licciardi, G., & Chanussot, J. (2018). Noise reduction in hyperspectral imagery: Overview and application. Remote Sensing, 10(3), 482.
21. Choromanski, K., Likhosherstov, V., Dohan, D., Song, X., Gane, A., Sarlos, T., Hawkins, P., Davis, J., Mohiuddin, A., Kaiser, L., Belanger, D., Colwell, L., & Weller, A. (2021). Rethinking attention with performers. In Proceedings of the International Conference on Learning Representations (ICLR).
22. Chen, T., Xu, B., Zhang, C., & Guestrin, C. (2016). Training deep nets with sublinear memory cost. arXiv preprint arXiv:1604.06174.
23. Zhu, X. X., Tuia, D., Mou, L., Xia, G.-S., Zhang, L., Xu, F., & Fraundorfer, F. (2017). Deep learning in remote sensing: A comprehensive review and list of resources. IEEE Geoscience and Remote Sensing Magazine, 5(4), 8–36.
24. Kendall, A., & Gal, Y. (2017). What uncertainties do we need in Bayesian deep learning for computer vision? In Advances in Neural Information Processing Systems, 30.
25. Ganin, Y., Ustinova, E., Ajakan, H., Germain, P., Larochelle, H., Laviolette, F., Marchand, M., & Lempitsky, V. (2016). Domain-adversarial training of neural networks. Journal of Machine Learning Research, 17(59), 1–35.
26. Burke, M., Driscoll, A., Lobell, D. B., & Ermon, S. (2021). Using satellite imagery to understand and promote sustainable development. Science, 371(6535), eabe8628.
27. Anderson, K., Ryan, B., Sonntag, W., Kavvada, A., & Friedl, L. (2017). Earth observation in service of the 2030 Agenda for Sustainable Development. Geo-spatial Information Science, 20(2), 77–96.
28. Strubell, E., Ganesh, A., & McCallum, A. (2019). Energy and policy considerations for modern deep learning research. In Proceedings of the AAAI Conference on Artificial Intelligence, 33, 13693–13696.
Downloads
Published
Issue
Section
License
Copyright (c) 2026 Journal of Advanced Artificial Intelligence Research
This work is licensed under a Creative Commons Attribution 4.0 International License.
This article is published under the Creative Commons Attribution 4.0 International License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
