LiDAR-Based Drainage Ditch Mapping for Peatland Restoration Across Low-Relief Coastal Landscapes in North Carolina

Abstract

Peatlands store approximately 25% of global soil carbon while providing critical ecosystem services, including water regulation, biodiversity support, and climate mitigation. In the North Carolina coastal plain, widespread construction of drainage ditches for agriculture, forestry, and development has lowered water tables, degraded peat soils, and increased risks of carbon loss, wildfire, flooding, and saltwater intrusion. Restoration efforts aim to reestablish natural hydrology by raising water tables, but a major barrier to effective planning is the lack of accurate, high-resolution data on drainage ditch networks. Existing national datasets often omit small, artificial ditches, limiting their use for site-level restoration design and prioritization. This project, conducted in partnership with The Nature Conservancy (TNC), addresses this gap by developing high-resolution drainage ditch mapping products and scalable workflows to support peatland restoration planning. The objectives were to 1) generate spatially explicit ditch datasets for priority sites, 2) develop transferable workflows for extracting ditch networks from LiDAR data, and 3) explore how ditch networks influence saltwater intrusion vulnerability. The project used publicly available USGS 3D Elevation Program (3DEP) LiDAR data, supplemented by drone LiDAR at select sites for validation. Drainage ditches were identified using a terrain analysis approach that detects linear depressions in high-resolution terrain models. This workflow was formalized into a custom ArcGIS Pro toolbox (DitchExtraction) to enable consistent, semi-automated mapping across sites. In parallel, a supervised machine learning approach using a Random Forest classifier trained on 14 features was developed to evaluate scalability for regional applications. A pilot saltwater intrusion vulnerability assessment was also initiated by integrating derived ditch networks with hydrologic modeling and terrain analysis. Results show that 3DEP LiDAR data are suitable for high-resolution ditch mapping across a range of peatland environments, performing comparably to higher-resolution drone data. Application of the DitchExtraction toolbox across approximately 383 km² of study sites produced 2,036 km of mapped ditches, generating the first high-resolution drainage network datasets for these landscapes. The Random Forest classifier demonstrated strong performance (96.96% overall accuracy; precision and recall of 0.97) at a test site, indicating that automated classification can effectively identify ditch features and holds promise for broader regional application. Early steps in the saltwater intrusion assessment established a foundation for evaluating how drainage ditch networks influence coastal vulnerability. This project establishes a replicable framework for high-resolution drainage ditch mapping and delivers actionable datasets to support TNC’s peatland restoration planning. The framework and datasets enable more informed, cost-effective restoration decisions and advance broader goals of climate mitigation, ecosystem restoration, and coastal resilience.