Channel Roughness, Morphology, Bankfull Discharge and Hydraulic Modeling (FIM)

Flood Inundation Mapping (FIM) is a critical component in the flood forecasting and planning paradigms, translating rainfall and streamflow predictions and observations to flood inundation extent over an impacted area. FIM forecasting combines predictions by a hydrological model, which estimates streamflow and overland runoff, and a hydraulic model which predicts the movement of water over the landscape. Hydraulic solvers tend to be computationally demanding, both in terms of processing (run) time and data needs. They are therefore difficult to apply over large geographic domains. The NOAA Office of Water Predictions (OWP) developed an operational FIM framework that employs a simplified hydraulic solver based on the so-called Height Above Nearest Drainage (HAND) approach. The OWP HAND translates streamflow predictions by the OWP National Water Model (NWM), its operational hydrological forecasting framework, to water level at each of the NWM prediction locations using a location-specific rating curve (a conversion plot from streamflow to water height). The accuracy of the OWP HAND methodology, as well as more complex hydraulic solvers, was shown to be highly sensitive to the representation of the shape and properties of river channels and their surrounding landscape (floodplains). Channel attributes are also important for estimating the flow conditions above which flooding will commence (referred to as Bankfull Discharge). Knowledge about bankfull discharge is particularly important for operational flood forecasting as it is used as a trigger for taking action. Estimation of bankfull discharge remains highly elusive, even for small river systems, and is typically based on calculating the likelihood (e.g. one in a hundred years) of extreme flow conditions from a long streamflow record (or predictions). These probabilistic approaches were shown to be highly inadequate. This project addresses the limitation in the representation of river channel shape and hydraulic characteristics by developing new data-driven estimates for the entire river network across the Contiguous United States (CONUS), leveraging and developing new datasets, and utilizing state-of-the-art Machine Learning approaches. Outcomes from these tasks will be a new CONUS-wide dataset of river channel width, depth, shape, bankfull discharge, and roughness for over 2.5 million river reaches. Parallel with these efforts, the project will address well-documented limitations in the OWP HAND FIM approach by testing a suite of hydraulic solvers at varying degrees of complexity against a first-of-its-kind large-scale observational FIM dataset. The observational dataset will be based on satellite image processing techniques. The field of Earth Observing Satellite Remote Sensing has seen great advances in recent years including the emergence of higher-resolution sensors and Machine Learning based image processing tools. These will be leveraged in this project to generate an open observational FIM portal which is envisioned to become a focal point for the broader flood data providers and users communities.