Detecting Water Features over Diverse Land Surfaces Using Uncrewed Aerial System (UAS) Imagery and the Normalized Difference Water Index (NDWI)

While the normalized difference water index (NDWI) was initially developed to define open water features in satellite imagery, the advent of aerial imagery from uncrewed aerial systems (UAS) for hydrologic operations and research has broadened the applications of the approach. In basic terms, the NDWI compares the visible green and near-infrared spectral bands (NDWI = [Green – NIR] – [Green + NIR]) to detect water bodies due to the difference in reflectance profiles of the two wavelengths over water and land/vegetation. With a range of +/- 1, positive values of NDWI relate to water while negative values relate to bare land; however, a major drawback to the approach is the sensitivity of the NDWI algorithm to built structures, such that asphalt rooftops and roadways can be interpreted as water due to their highly absorptive properties in the visible wavelengths. At lower spatial resolution this issue can be considered minimal, especially over areas where built structures comprise a small percentage of an image pixel; however, with high resolution UAS imagery, this issue can become important as roadways and buildings can be interpreted as water bodies while in-stream vegetation can cause water features to appear disjointed. When accumulated over a larger region, such issues can lead to substantial errors in estimated inundated area. To that end, this project will use UAS-based imagery over two diverse landscapes, including Jackson, MS (built-up urban) and Yazoo City, MS (rural agricultural) to quantify uncertainty in defining water bodies using the NDWI approach. The imagery was collected using Group 3 UAS platforms as part of an on-going project related to operational hydrologic support and prediction; therefore, understanding issues that impact the estimation of inundated area and recognition of waterways is critical to proper application of the imagery in an environment focused on decision support and/or situational awareness. By defining physical limitations to the NDWI using the high-resolution imagery, work can be done to refine the post-processed data to better reflect diverse landscapes and features when producing land-water masks for hydrologic analysis.