Featured Collection Introduction: National Flood Interoperability Experiment I

The ribbon cutting ceremony in May of 2015 of the National Weather Service's (NWS) new National Water Center (NWC) provided the necessary facilities for the National Oceanic and Atmospheric Administration (NOAA) to carry out its vision for a weather-ready nation (OWP, 2016). The idea of a national center was to compliment the 13 existing River Forecast Centers (RFCs) by transforming the national capability for hydrologic prediction with a seamless national water model that could enhance both temporal and spatial forecasting resolution. The NWC provides a hub, whereby the NWS can engage the academic community and other federal institutions involved in the variety of data, modeling, visualization, and dissemination of forecast products activities. Of the NWC Dr. Kathryn Sullivan said, “The National Water Center will bring researchers, forecasters, and stakeholders together to accelerate progress towards the improved water forecasts needed to manage our precious water resources wisely.” This research-to-operations mentality is what led to the idea of a National Flood Interoperability Experiment or NFIE as an inaugural event at the NWC. Proposed by Dr. David Maidment (Maidment et al., 2014) as a way to kick start activities in support of the NWS vision, it also intends to train a new generation of scientists who can contribute to it. The overarching goal of the first NFIE was to demonstrate that a continental-scale flood forecast modeling system was feasible, a task that for all of the data and computational resources required had not been achieved. Maidment outlines five basic building blocks that were used to describe conceptually and break the monumental task into smaller parts that could be tackled separately and then brought together. These blocks included the geofabric or NFIE-GEO, based principally on the scale of the NHD-Plus 2.7 (Horizons Systems Corporation, 2011) million river reaches, NFIE-Hydro the hydrologic model, NFIE-River the local scale river flooding, NFIE-Response with a focus on how to communicate flood information to local responders, and finally NFIE-Services the foundational computation and delivery cyber infrastructure. Maidment's paper introduces the concept of NFIE-GEO and then the remaining five papers in this edition of the NFIE Special Collection addresses the other four building blocks. Tavakoly et al. (2016) describes how the Routing Application for Parallel computation of Discharge (RAPID) routing model can be used over the Mississippi Basin to connect surface water hydrology to the vector-based NHD-Plus for output. The paper concludes that continental-scale river routing at the level of the NHD-Plus was feasible. It also discusses the limitations, including the importance of modeling reservoir and other storage in order to accurately predict streamflow downstream of any storage structure. This article is followed by two others which examine the possibility of using forecasted discharge to develop flood inundation maps. In their paper, Follum et al. (2016) describe an automated flood delineation technique, using AutoRoute in series with RAPID to efficiently compute raster inundation maps, using standard 10-meter digital elevation models. While it does not consider the full dynamic hydraulics or as previously stated with respect to RAPID instream storage from dams, levees, and other structures, it does provide a method that can estimate flood impact in areas where resources for more sophisticated hydraulic analysis can be justified and a means whereby a continental-scale inundation can be computed. They conclude that it is a feasible method for given data and computational resources and that it works best in steeper topography and the more extreme discharges. A second paper on flood mapping by Buahin et al. (2017) examines the application of the Hydrologic Engineering Center River Analysis System (HEC-RAS) at a more localized level where the necessary data are available. Besides using more detailed channel geometry in a HEC-RAS model, they develop a probabilistic map using the European Center for Medium-range Weather Forecasts (ECMWF) ensemble forecasts. The next paper by Du et al. (2016) focused on the heterogeneity of human responses to flood forecast information and stressed the importance of understanding and developing warning systems that humans are responsive to, and not just improved accuracy of the forecast itself. They examine both heterogeneity of human responses and residential density with an agent-based modeling framework coupled with a transportation model to simulate evacuation. The final paper in the collection discusses the importance of computational cyber-infrastructure that is necessary for carrying out something as aspirational as a continental scale flood forecasting system. Christensen et al. (2016) describe a cloud-based implementation of a High Throughput Computing (HTC) system that was developed to support a demonstration national-scale forecast, using the ECMWF ensemble products. Every 12 hours a 51-member ensemble forecast is generated for the entire continental United States on an approximate 32 kilometer-squared grid. In order to produce runoff at the NHD-Plus level each forecast must be mapped from the grid to basin level and routed with RAPID. The only way the entire set of forecasts can be computed is through parallel computing of the ensembles over several regions, which required a “pool” of computing resources. In their paper, they describe a set of python scripts that enable the requisition and use of a cluster of cloud computing nodes that can be used for this and other applications.