Tuesday, 14 January 2025
Hall C (New Orleans Ernest N. Morial Convention Center)
Fog has a substantial economic and safety impact on coastal communities through disruptions of land, air, and sea transportation. This is the case for the northwest Gulf of Mexico, home of many airports and the majority of the top ten US ports by tonnage. A recent AI coastal fog prediction model, FogNet (Kamangir et al. 2021), showed improved performance over present operational models while using a 3D CNN architecture and fusing numerical weather output with satellite imagery of sea surface temperatures.
The application of eXplainable AI (XAI) to FogNet (Kamangir et al. 2022) provided information as to the impact on performance and the sensitivity of the model to model predictors and geographic locations. The XAI analysis identified the location of the prediction as a dominant factor for most predictors. This suggests that a simpler model solely focused on predictions at the coastal location of interest may approach the performance of the 3D CNN model.
The present work describes the architecture and performance of a variational autoencoder (VAE) developed for predicting visibility 12 hours in advance at 22 airports along the Gulf Coast as far inland as the city of Victoria and the Houston Intercontinental Airport. The hourly predictors are selected from the output of NOAA’s High Resolution Rapid Refresh (HRRR) model, which is compatible with potential operational application of the VAE predictions.
Recent results indicate that, for a given lead time, a single VAE model can be trained to predict visibility at all locations. The model architecture and hyperparameters are optimized using a Tree-Parzen Estimator and the relative importance of the selected predictors is estimated using Permutation Feature Importance (PFI). Three methods are compared to prune the model’s predictors: (1) an approach using the entire HRRR forecast variable set, (2) a subset of the available HRRR forecast variables with the addition of engineered features constructed from the same HRRR data by a domain expert, and (3) a multistep PFI-driven selection of the predictors.
The performance of the VAE model is evaluated against that of the HRRR model and The Weather Company’s Forecast on Demand visibility forecasts at 1, 3, and 5 mile thresholds, while the potential for further simplification, tests of spatial coherence, and operational deployment using the TAMUCC Semaphore software is discussed (Kastl et al. 2025).
Kamangir, H., Collins, W., Tissot, P., King, S. A., Dinh, H. T. H., Durham, N., & Rizzo, J. (2021). FogNet: A multiscale 3D CNN with double-branch dense block and attention mechanism for fog prediction. Machine Learning with Applications, 5, 100038.
Kamangir, H., Krell, E., Collins, W., King, S. A., & Tissot, P. (2022). Importance of 3D convolution and physics on a deep learning coastal fog model. Environmental Modelling & Software, 154, 105424.
Matthew Aiden Kastl, Texas A&M University - Corpus Christi, Corpus Christi, TX; and B. Estrada Jr., S. L. Stephenson, A. Beasley, F. Tissot, S. Nguyen, S. A. King, and P. E. Tissot Semaphore: A Software Platform for Operational Hosting of Coastal AI Models Presentation at the 15th Conference On Transition Of Research To Operations (15R2O), 2025 AMS Annual Meeting, New Orleans, LS.
The application of eXplainable AI (XAI) to FogNet (Kamangir et al. 2022) provided information as to the impact on performance and the sensitivity of the model to model predictors and geographic locations. The XAI analysis identified the location of the prediction as a dominant factor for most predictors. This suggests that a simpler model solely focused on predictions at the coastal location of interest may approach the performance of the 3D CNN model.
The present work describes the architecture and performance of a variational autoencoder (VAE) developed for predicting visibility 12 hours in advance at 22 airports along the Gulf Coast as far inland as the city of Victoria and the Houston Intercontinental Airport. The hourly predictors are selected from the output of NOAA’s High Resolution Rapid Refresh (HRRR) model, which is compatible with potential operational application of the VAE predictions.
Recent results indicate that, for a given lead time, a single VAE model can be trained to predict visibility at all locations. The model architecture and hyperparameters are optimized using a Tree-Parzen Estimator and the relative importance of the selected predictors is estimated using Permutation Feature Importance (PFI). Three methods are compared to prune the model’s predictors: (1) an approach using the entire HRRR forecast variable set, (2) a subset of the available HRRR forecast variables with the addition of engineered features constructed from the same HRRR data by a domain expert, and (3) a multistep PFI-driven selection of the predictors.
The performance of the VAE model is evaluated against that of the HRRR model and The Weather Company’s Forecast on Demand visibility forecasts at 1, 3, and 5 mile thresholds, while the potential for further simplification, tests of spatial coherence, and operational deployment using the TAMUCC Semaphore software is discussed (Kastl et al. 2025).
Kamangir, H., Collins, W., Tissot, P., King, S. A., Dinh, H. T. H., Durham, N., & Rizzo, J. (2021). FogNet: A multiscale 3D CNN with double-branch dense block and attention mechanism for fog prediction. Machine Learning with Applications, 5, 100038.
Kamangir, H., Krell, E., Collins, W., King, S. A., & Tissot, P. (2022). Importance of 3D convolution and physics on a deep learning coastal fog model. Environmental Modelling & Software, 154, 105424.
Matthew Aiden Kastl, Texas A&M University - Corpus Christi, Corpus Christi, TX; and B. Estrada Jr., S. L. Stephenson, A. Beasley, F. Tissot, S. Nguyen, S. A. King, and P. E. Tissot Semaphore: A Software Platform for Operational Hosting of Coastal AI Models Presentation at the 15th Conference On Transition Of Research To Operations (15R2O), 2025 AMS Annual Meeting, New Orleans, LS.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner