LWI Region 3 Modeling Overview

Machine learning techniques were used on NAIP imagery to classify key land cover types, such as trees, water, buildings, and crops, supplemented with road, wetland, and building footprint data from national sources. For parts of the watershed in Arkansas, where 2D ROM modeling was implemented, the team used the 2019 NLCD dataset due to lack of NAIP coverage (Figure 3-4). The Normalized Difference Vegetation Index was calculated using the red and near-infrared bands to separate vegetated and nonvegetated features, providing a strong basis for supervised classification.

Training samples were collected across the watershed, and a maximum likelihood algorithm with image segmentation was used to classify the imagery into meaningful land use categories. LEARN MORE: B_2.1.3 + This enhanced land cover layer informed Manning’s roughness distribution in the 2D model, with spatial variation applied across cell faces for improved accuracy.

NLCD

Remote Sensing

Figure 3-4: Comparison between land cover dataset developed from remote sensing (left) and the NLCD (right).

13

Hydrologic and Hydraulic Modeling Methodology