Near–Real-Time Atmospheric River Detection and Forecasting with NASA GEOS-FP

OverviewAtmospheric Rivers (ARs) are critical drivers of the hydrological cycle in the western United States, accounting for 50–70% of annual precipitation. In Southern California, approximately 70% of post-fire debris flows are associated with AR events. Furthermore, shifts in AR trajectories can induce hot, dry, and windy conditions that exacerbate mega-fire risk.This showcase presents an automated, near-real-time (NRT) and predictive global AR detection system utilizing NASA’s GEOS-FP data. While the core algorithm has been validated using reanalysis products (e.g., ERA5, MERRA2, etc), this effort adapts the tool for NASA model outputs to complement and integrate with existing operational weather products for disaster response and potential mission support. Showcase Products * Visualization key: Globe, CONUS, WUS (Western U.S.), and EUS (Eastern U.S.) regions are shown. All times are given in UTC. For reference, Pacific Standard Time (boreal winter) is UTC−8. Purple (maps)/Magenta (maps in the 6-panel plot) Contour: Detected AR object. Angle Label: Orientation relative to the horizontal axis (critical for landfalling impact). Black Arrows: Major and minor axes of AR object determined via Principal Component Analysis (PCA). 6-panel plots: Maps and PDFs of GEOS-FP based ARs, IVT, total precipitation, and surface wind speed for the selected regions. 1. Analysis Showcase (NRT Detection)Two recent events demonstrate the performance of the methodology using near–real-time GEOS-FP data. December 19–25, 2025: An extreme precipitation event in the Pacific Northwest. The visualization illustrates Integrated Vapor Transport (IVT) and cumulative precipitation and surface wind speed across British Columbia and the U.S. West Coast. February 6–12, 2026: A moderate AR event showing rainfall impacts along the California coast. Validation: Comparison plots (6-hourly) are provided for the December 19–25, 2025 case, showing strong agreement in AR detection between GEOS-FP and ECMWF ERA5. 2. Forecasting (fcst) Showcase February 13–20, 2026: This component demonstrates the potential for NRT forecasting. The example shows a 7-day forecast (3-hourly intervals) of IVT, AR detection, total precipitation, and surface winds, initialized on February 13, 2026 and valid for February 13 to 20, 2026. Methodology Global objective AR detection ARs are identified using the objective detection algorithm developed by Ryoo and Park (2023). The algorithm uses vertically integrated vapor transport (IVT), calculated by vertically integrating the product of horizontal wind and specific humidity from the surface (1000 hPa) to the upper troposphere (250 hPa).The algorithm determines AR geometry (length and width) by applying Principal Component Analysis (PCA) to the detected objects. This unsupervised machine learning technique preserves data variations while providing a low-dimensional representation of the AR axes. Detailed methodology can be found in Ryoo and Park (2023) and Ryoo et al. (2025), along with their supplementary materials. Keywords: Atmospheric Rivers, NASA GEOS-FP, Near-Real Time, Forecasting, IVT, Precipitation, Surface Wind Speed, Disaster Response, Flooding, Wildfire Hazards. Note: For inquiries about data processing, research applications, regional use cases, or algorithm usage, please contact the Author.