Emerging flood model validation frameworks for street-level inundation modeling with StormSense

Abstract

Technological progress in flood monitoring and the proliferation of cost-efficient IoT-enabled water level sensors are enabling new streams of information for today's smart cities. StormSense is an inundation forecasting research initiative and an active participant in the GCTC seeking to enhance flood preparedness in the Hampton Roads region for flooding resulting from storm surge, rain, and tides and demonstrating replicability of the solution. Herein, we present street-level hydrodynamic modeling results at 5m resolution with conventional flood validation sources alongside new emergent techniques for validating model predictions during three prominent recent flooding events in Hampton Roads during Fall 2016: Hurricane Hermine, Tropical Storm Julia, and Hurricane Matthew. Emerging validation techniques include: (1) IoT-water level sensors, (2) crowd-sourced GPS maximum flood extent measurements, and (3) geospatial flooded area comparisons with drone-surveyed flood extents via ESRI's Drone2Map. Model uncertainty was validated against 5 newly-established tide gauges within the domain for an aggregate vertical root mean squared error of ±8.19 cm between the sensor observations and model predictions. Also, geospatial uncertainty was assessed using mean horizontal distance difference as ±4.97 m via 206 crowd-sourced GPS flood extents from the Sea Level Rise App.