Global long-term sub-daily reanalysis of fluvial floods through high-resolution modeling

Flood is one of the most devastating natural disasters of severe societal, economic, and environmental consequences. Understanding the characteristics of floods, especially at fine spatial and short temporal scales, can be critical for improving forecast and risk management efforts. Due to the limited availability, in-situ observations have been inadequate for meeting the challenges at global extent. Existing global flood modeling efforts also lack the sufficient spatial/temporal resolutions for capturing rapid/local flood events, e.g., those developed in less than a day. Here we implement a carefully-designed modeling framework to reconstruct global river discharge at very high resolution (5-km and 3-hourly for runoff calculation and ~2.94 million river reaches derived from 90-m DEM for river routing) for 40 years (1979-2018). The Variable Infiltration Capacity (VIC) model with calibrated parameters, is coupled with the Routing Application for Parallel computation of Discharge (RAPID), serving as the core of the modeling framework. The state-of-the-art merged precipitation product, Multi-Source Weighted-Ensemble Precipitation (MSWEP) and flowlines vectorized from the MERIT Hydro are used. Pixel-level model calibration and distributional bias correction are performed against global runoff characteristics derived from observations and machine learning. Skill assessments are carried out both globally at daily sale and over contiguous U.S. (CONUS) at 3-hourly scale, using both general discharge performance metrics (Kling-Gupta Efficiency and it three components) and sub-daily flood-specific metrics (probability of detection, false alarm rate, flood volume error, peak magnitude error, timing error, etc.). The work here aims to provide some first-time understanding of local scale rapid flooding over the global domain. We also expect to learn more about the modeling tools developed for analyzing/monitoring fine scale flooding globally – their efficacy and lack thereof, why, and where to improve.