Mapping internal farm roadways to identify runoff accumulation areas using an integrated GIS, aerial imagery and deep learning approach in grassland farms

This research addresses a critical knowledge gap in agricultural environmental management by developing automated methods to map internal farm roadway networks on grassland farms and assess their potential for generating phosphorus-enriched runoff that could impact water quality.

Key Contributions

• Deep learning model development: Evaluated three model architectures (U-Net, PSPNet, and DeepLab V3+) for automatic detection and mapping of internal farm roadways from high-resolution aerial imagery
• Model performance: PSPNet with ResNet-50 backbone achieved best results with 0.90 overall accuracy, 0.86 recall, and 0.82 F1 score
• Large-scale mapping: Successfully extracted 34.6 km of internal farm roadways across 10 grassland farms
• Runoff risk assessment: Identified roadway sections with high (8.3-20%) and very high (0.6-4.9%) runoff susceptibility using Topographic Wetness Index (TWI)
• New pollution hotspots: Discovered previously unidentified phosphorus flow delivery pathways not present in existing national datasets

Significance

In temperate regions, farm roadway networks facilitate the movement of dairy and beef animals between grazed paddocks and farmyards. During animal movement, excreta deposited on roadway surfaces becomes a source of phosphorus-enriched runoff during rainfall events, potentially causing water quality degradation when mobilized to waterbodies.

While national roadway networks have been mapped using deep learning approaches, internal farm roadway mapping has remained a significant knowledge gap. This study provides the first automated, scalable solution for:

1. Precision mapping of internal agricultural infrastructure at farm scale
2. Risk assessment of runoff accumulation and delivery to water bodies
3. Environmental management support for targeted mitigation strategies
4. Water quality protection through identification of critical source areas

Methodology

The integrated workflow combines:

• High-resolution aerial imagery analysis
• Deep learning semantic segmentation (PSPNet architecture)
• GIS-based topographic analysis using TWI
• Runoff accumulation modeling

This approach enables efficient, automated assessment of farm roadway networks and their associated environmental risks, providing an important tool for sustainable agricultural management and water quality protection.

Open Access

This article is published under a Creative Commons license, making it freely available for research and educational purposes.

Recommended citation: Sifundza, L.S., Murnane, J.G., Adams, R., Daly, K., Habib, W., Tuohy, P., & Fenton, O. (2025). "Mapping internal farm roadways to identify runoff accumulation areas using an integrated GIS, aerial imagery and deep learning approach in grassland farms." International Journal of Applied Earth Observation and Geoinformation, 134, 104896.
Download Paper