|dc.description.abstract||Natural hazards cause catastrophic damages to both population and economy. In the U.S., floods are the costliest hazard. In 2017, Hurricane Harvey made landfall along the Texas coast on August 25th and lasted for five days. It was one of the most destructive hurricanes in the history of the state. In order to enhance emergency response and management, it is essential to have a better understanding of the flood status, risks and conditions.
In flood modeling, conventional data sources include remote sensing, high water marks (HWMs) from field survey, and stream gauges are generally used. The availability of Volunteered Geographic Information (VGI), such as tweets and crowdsourced data, empowered the researchers to model flood (e.g. Water Depth (WD)) in near-real-time by integrating multi-sourced data available. Nevertheless, the quality of VGI and its reliability for flood analysis is not well understood and validated by empirical data. Therefore, the primary objective of this study was to evaluate the quality of multiple VGI data sources, especially the multimedia that include pictures and videos, against authoritative data for inundation mapping. This study collected the geospatial data from multiple sources to analyze the changing WD during Hurricane Harvey in Harris County, Texas. First, WD was generated from three VGI data modalities: (1) text, (2) pictures, and (3) videos, and they were compared against each other using Friedman test and Chi-square. Then, the VGI-derived WD was synthesized and consolidated to reconstruct the time-series of WD in Harris County. Finally, the quality of synthesized WD and VGI was validated against remote sensing (RS) and two authoritative data: (1) water level records from stream gauges at discrete locations, and (2) modeled depth grids by Federal Emergency Management Agency (FEMA) using paired t-test. The results showed that there was no statistically significant differences among all VGI data modalities in terms of precision, while it showed significant difference in terms of spatial and temporal characteristics. In addition, the results showed that there was a statistically significant difference between VGI WD and RS WD. Finally, the analysis revealed that there was no statistically significant difference between VGI data and water records from the stream gauges, while it showed a statistically significant difference when VGI were compared with the depth grids from FEMA.||