Models capable of mapping floodplains, areas at risk of being inundated given a flood, are constantly evolving and improving. Conceptual models are advantageous in large-scale floodplain mapping due to their computational efficiency. One conceptual model is Height Above Nearest Drainage (HAND). HAND uses a digital representation of a terrain’s elevations, a Digital Elevation Model (DEM), to calculate the vertical distances between locations in the landscape and the surface of the nearest river or stream. This terrain descriptor has shown to be a good predictor of flood potential, and several studies record high accuracy of the modeled inundation compared to observed flood events. 

Although HAND-based floodplain mapping can be performed in any location given its DEM representation, accurate predictions presuppose the accurate selection of two input parameters. The choice of said parameters most often relies on data unavailable in ungauged locations, making global-scale modeling challenging. The first of these parameters is the drainage threshold, which determines what can be considered drainage channels in the terrain. However, delineating rivers and streams from a DEM is not always accurate and requires manual tuning of the said threshold. Using water labels from Land Use and Land Cover (LULC) data to identify rivers could result in a more accurate representation of the river network while alleviating the threshold selection process. The second parameter is the stage height, the amount the water surface elevation is modeled to rise. To make HAND useful in areas where historic stage heights or forecasts are unavailable, an approach that either alleviates challenges in the parameter selection process or altogether circumvents it is necessary. 

To this end, this thesis explores the effect of using land classification data to identify the river and streams in the terrain. After that, this thesis describes the results of a HAND-based floodplain sensitivity analysis which evaluates whether the floodplain can be found without historical knowledge of the river stage. For this purpose, a high-resolution (0.5 m) DEM and a LULC dataset provided by Maxar Technologies were used. Previous work has not explored using high precision and high-resolution DEMs, and land classification data from the same set of satellite imagery for HAND-based floodplain mapping. High precision and high-resolution data are expected to improve floodplain mapping accuracy and enable more detailed analysis. Therefore, this thesis also explores how land classification can bring more detail into the modeling in urbanized areas by considering buildings.

The HAND baseline and the proposed modification to the model were compared with floodplain maps generated by the Federal Emergency Management Agency (FEMA) for six study areas in the US. Furthermore, this thesis reports additional comparisons for one site where observed flood extent was available. The results show that river delineation from LULC data is a good alternative to the traditional methodology. The modeled flood extent gave better results when compared to an observed flood extent. The integration of buildings significantly affected the modeled inundation extent in densely urbanized areas, which shows the importance of accounting for buildings in the model and showcases how high-resolution data enables a more detailed floodplain analysis. Finally, the floodplain sensitivity analysis is able to describe the characteristics of the floodplains in the selected study areas. This was particularly evident in flat reliefs, where the geomorphological limits of the floodplains were visible. The HAND-based floodplain sensitivity analysis can provide useful information in understanding how a floodplain responds to changes in stage height and thus help evaluate what areas have a high risk of being inundated.