My research is at the intersection of machine learning, computer vision, earth sciences, and high-performance computing. I am interested in machine learning applications to our ever growing datastores of spatio-temporal earth science observations, forecasts, and projections to understand our environment. Recent projects have included developments of novel deep learning methods for spatio-temporal downscaling, physical model emulation, generating virtual sensors, and cloud tracking.
Thomas is a research scientist at the NASA Ames Research Center and Bay Area Environmental Research Institute within the NASA Earth Exchange (NEX) in Mountain View, CA. He is the Principle Investigator (PI) of a NASA ROSES grant studying optical flow methodologies for tracking atmospheric motion in satellite imagery. Thomas earned a Ph.D. in Interdisciplinary Engineering from Northeastern University in 2018 while working in the Sustainability and Data Sciences Lab advised by Auroop R. Ganguly. During graduate school his research won runner-up best paper award at SIGKDD 2017, outstanding graduate researcher at Northeastern University in 2018, and served as an elected student member of the committee on Artificial Intelligence Applications to Environmental Science for the American Meterological Society (AMS). Prior to graduate school he worked at startups in the Boston area, including the MIT Media Lab spin-out Affectiva. He completed his bachelors in mathematics at the University of Maryland College Park in 2012.
(June 2021) Paper titled “Spectral Synthesis for Geostationary Satellite-to-Satellite Translation” published in TGRS Paper.
(August 2020) Awarded best spotlight presentation at 1st ACM SIGKDD Workshop on Deep Learning for Spatiotemporal Data, Applications, and Systems.
(May 2020) Paper accepted to KDD 2020 research track titled “High-Dimensional Similarity Search with Quantum-Assisted Variational Autoencoder” Paper
(May 2020) NASA ROSES proposal selected by to develop optical flow methods for tracking atmospheric motion. Selections.
(May 2020) Journal paper on Data science and food security published in Frontiers Sustainable Food Systems Paper.
(March 2020) Pre-print posted “Temporal Interpolation of Geostationary Satellite Imagery with Task Specific Optical Flow”.
(July 2019) Presented at the Space Lidar Winds Working Group Meeting at the National Institute of Aerospace.
(April 2019) Presented at the 1st Workshop on Leveraging AI in the Exploitation of Satellite Earth Observations & Numerical Weather Prediction, Slides.
(July 2018) Our paper was invited for submission to the Sister’s Best Paper Track at IJCAI 2018.
(August 2018) Paper accepted to KDD 2018 research track on “Quantifying Uncertainty in Discrete-Continuous and Skewed Data with Bayesian Deep Learning”.
(December 2017) Research is featured in the Nature comment article The case for technology investments in the environment .
(August 2017) Paper accepted at KDD 2017 on DeepSD: Generating High Resolution Climate Change Projections through Single Image Super-Resolution.
Vandal, T. & Nemani, R. (2021). “Optical Flow for Intermediate Frame Interpolation of Multispectral Geostationary Satellite Data”. IEE Transactions on Neural Network and Learning Systems (TNNLS). Special issue on Special Issue on Deep Learning for Earth and Planetary Geosciences. [Link to come]
Vandal, T., McDuff, D., Wang, W., Duffy, K., Michaelis, A., & Nemani, R. (2021) “Spectral Synthesis for Geostationary Satellite-to-Satellite Translation”. IEEE Transactions on Geoscience and Remote Sensing. Paper
Vandal, T. & Nemani, R. (2020). “Optical Flow for Intermediate Frame Interpolation of Multispectral Geostationary Satellite Data”. In Proceedings of 1st ACM SIGKDD Workshop on Deep Learning for Spatiotemporal Data, Applications, and Systems (DeepSpatial ’20). paper
Gao, N., Wilson, M., Vandal, T., Vinci, W., Nemani, R., & Rieffel, E (2020). “High-Dimensional Similarity Search with Quantum-Assisted Variational Autoencoder,” Proceedings of the 26th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (Research Track, 16% acceptance rate). Paper
Vandal, T., Kodra, E., Ganguly, S., Dy, J., Nemani, R., & Ganguly, A (2018): “Quantifying Uncertainty in Discrete-Continuous and Skewed Data with Bayesian Deep Learning,” Proceedings of the 24rd ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 1663-1672. (Research Track, 18% acceptance rate)Paper, Code.
Vandal, T., Kodra, E., Ganguly, S., Michaelis, A., Nemani, R., & Ganguly, A. (2017). “DeepSD: Generating high resolution climate change projections through single image super-Resolution,” KDD 2017, Proceedings of the 23rd ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 1663-1672. (Runner-Up BEST PAPER Award in Applied Data Science Track, 9% oral acceptance rate). Paper [code].
Vandal, T., E. Kodra, and A. Ganguly, “Intercomparison of Machine Learning Methods for Statistical Downscaling: The Case of Daily and Extreme Precipitation.” Theoretical and Applied Climatology. September 2018. Paper.
Vandal, T., “Statistical Downscaling of Global Climate Models with Image Super-resolution and Uncertainty Quantification”, 2018. Northeastern University. Dissertation.