An automated geospatial framework to identify internal farm roadways with potential roadway runoff connectivity to waters at meso‑catchment scale

Internal farm roadway networks in temperate pastural systems are sub-components of the nutrient transfer continuum. These networks are a fresh and legacy nutrient source, which mobilises during rainfall and in some cases enters surface waters, impacting water quality. The objectives of this study were to develop an automated methodological framework to map internal farm roadways and waters (i.e., streams, open drainage ditches), identify roadway runoff pathways and locate their delivery to waters. A deep learning-based semantic segmentation was applied to aerial imagery to map farm roadways and to a Light Detection and Ranging (LiDAR)-derived digital terrain model (DTM) to detect waters at a grassland-dominated meso-catchment scale in south-east Ireland. Flow pathways were established using the DTM to determine farm roadway runoff pathways and delivery points to waters. To locate runoff pathways with high connectivity potential, the network index (NI) determined as the 95 th percentile of topographic wetness index (TWI) was used. A total of 39.4 km of roadways (of which 15.8 km were internal farm roadways) and 63.5 km of waters were identified and mapped. Within farms, 47 roadway sections (1.5 km or 9.6 % of the total mapped internal farm roadways) were found to accumulate runoff and have the potential to deliver roadway runoff (mean runoff pathway length of 132.5 m) to waters, of which 0.6 km (3.8 %) identified as having high connectivity. Field verification was conducted at several sites within the meso-catchment and confirmed the findings of the NI. This study presents a framework that enables the remote identification and assessment of internal farm roadways in grassland systems that are likely to deliver soiled runoff to waters. This can be used by landowners and catchment managers to locate where mitigation measures can be implemented.