Heavy rains and tropical storms often result in floods, which are expected to increase in frequency and intensity. Flood prediction models and inundation mapping tools provide decision-makers and emergency responders with crucial information to better prepare for these events. However, the performance of models relies on the accuracy and timeliness of data received from in-situ gaging stations and remote sensing; each of these data sources has its limitations, especially when it comes to real-time monitoring of floods. This study presents a vision-based framework for measuring water levels and detecting floods using Computer Vision and Deep Learning (DL) techniques. The DL models use time-lapse images captured by surveillance cameras during storm events for the semantic segmentation of water extent in images. Three different DL-based approaches, namely PSPNet, TransUNet, and SegFormer, were applied and evaluated for semantic segmentation. The predicted masks are transformed into water level values by intersecting the extracted water edges, with the 2D representation of a point cloud generated by an Apple iPhone 13 Pro LiDAR sensor. The estimated water levels were compared to reference data collected by an ultrasonic sensor. The results showed that SegFormer outperformed other DL-based approaches by achieving 99.55% and 99.81% for Intersection over Union (IoU) and accuracy, respectively. Moreover, the highest correlations between reference data and the vision-based approach reached above 0.98 for both the coefficient of determination (R2) and Nash-Sutcliffe Efficiency. This study demonstrates the potential of using surveillance cameras and Artificial Intelligence for hydrologic monitoring and their integration with existing surveillance infrastructure.

Coastal watersheds are vulnerable to compound flooding associated with intense rainfall, storm surge, and high tide. Coastal compound flooding (CCF) simulation, in particular for low-gradient coastal watersheds, requires a tight-coupling procedure to represent nonlinear and complex flood processes and interconnectivity among multidimensional hydraulics and hydrologic models. This calls for the development of a fully-coupled CCF modeling framework. Here, the modeling framework is centered around the development of interconnected meshes of the node-link-basin using the Interconnected Channel and Pond Routing (ICPR) model. The modeling framework has been built for a complex drainage network, consisting of tidal creeks, tidal channels, underground sewer networks, and detention ponds in Charleston Peninsula, SC. The floodplain dynamics of the urbanized peninsula are modeled by a high-resolution LiDAR-derived Digital Elevation Model (DEM) and Digital Surface Model (DSM), and overland flow is simulated by energy balance, momentum balance, or diffusive wave methods. The performance of the CCF model is tested for the 2015 SC major flood and 2021 tidal flood events. The momentum balance-based CCF model shows 98.35% efficiency in capturing street-level flooding location and the CCF model depicts that using the DSM potentially improves the simulation accuracy of the flood by 15-33% compared to LiDAR DEM. Moreover, it is found the momentum balance between surge arrival from a tidally influenced river and rainfall runoff plays an important role in flood dynamics in urbanized catchments. This study contributes to the existing knowledge of fine-scale floodplain dynamics in urban areas by enhancing the fully-coupled numerical representation of CCF processes.