Using search-constrained inverse distance weight modeling for near real-time riverine flood modeling: Harris County, Texas, USA before, during, and after Hurricane Harvey

Flooding poses a serious public health hazard throughout the world. Flood modeling is an important tool for emergency preparedness and response, but some common methods require a high degree of expertise or may be unworkable due to poor data quality or data availability issues. The conceptually simple method of inverse distance weight modeling offers an alternative. Using stream gauges as inputs, this study interpolated stream elevation via inverse distance weight modeling under 15 different model input parameter scenarios for Harris County, Texas, USA, from August 25th to September 15th, 2017 (before, during, and after Hurricane Harvey inundated the county). A digital elevation model was used to identify areas where modeled stream elevation exceeded ground elevation, indicating flooding. Imagery and observed high water marks were used to validate the models’ outputs. There was a high degree of agreement (between 79 and 88%) between imagery and model outputs of parameterizations visually validated. Quantitative validations based on high water marks were also positive, with a Nash–Sutcliffe efficiency of in excess of .6 for all parameterizations relative to a Nash–Sutcliffe efficiency of the benchmark of 0.56. Inverse distance weight modeling offers a simple, accurate method for first-order estimations of riverine flooding in near real-time using readily available data, and outputs are robust to some alterations to input parameters.