Publications by authors named "Thomas Vandal"
Article Synopsis
- Precipitation nowcasting is vital for effective flood and river management but has historically faced challenges that recent advances in deep generative modeling (DGM) aim to address.
- Traditional Numerical Weather Prediction (NWP) models, like those used by the Tennessee Valley Authority, have been inadequate in accurately predicting precipitation events, leading researchers to explore advanced machine learning methods, specifically physics-embedded approaches.
- The study highlights that NowcastNet, a advanced DGM method, significantly outperforms the latest NWP model (HRRR) in predicting heavy rainfall, suggesting that integrating these machine learning advancements could enhance flood response systems and water resource management.
A Framework for Deep Learning Emulation of Numerical Models With a Case Study in Satellite Remote Sensing.
IEEE Trans Neural Netw Learn Syst
July 2023
Numerical models based on physics represent the state of the art in Earth system modeling and comprise our best tools for generating insights and predictions. Despite rapid growth in computational power, the perceived need for higher model resolutions overwhelms the latest generation computers, reducing the ability of modelers to generate simulations for understanding parameter sensitivities and characterizing variability and uncertainty. Thus, surrogate models are often developed to capture the essential attributes of the full-blown numerical models.
View Article and Find Full Text PDFTemporal Interpolation of Geostationary Satellite Imagery With Optical Flow.
IEEE Trans Neural Netw Learn Syst
July 2023
Applications of satellite data in areas such as weather tracking and modeling, ecosystem monitoring, wildfire detection, and land-cover change are heavily dependent on the tradeoffs to spatial, spectral, and temporal resolutions of observations. In weather tracking, high-frequency temporal observations are critical and used to improve forecasts, study severe events, and extract atmospheric motion, among others. However, while the current generation of geostationary (GEO) satellites has hemispheric coverage at 10-15-min intervals, higher temporal frequency observations are ideal for studying mesoscale severe weather events.
View Article and Find Full Text PDFArticle Synopsis
- The increasing availability of Earth science data from climate simulations and satellite sensing provides new opportunities for research but poses computational challenges.
- Machine learning, particularly deep learning, shows promise for improving the efficiency of processing complex datasets like atmospheric corrections by acting as fast statistical models.
- This study introduces DeepEmSat, a deep learning emulator for atmospheric correction, and compares it to traditional physics-based models to demonstrate the potential benefits of using deep learning in satellite image processing.