Automated Mapping of Surface Sedimentary Features in Mixed Sand-Gravel Tidal Inlets Using UAV, XGBoost and U-Net

Gong, J. and Burningham, H., 2024. Automated mapping of surface sedimentary features in mixed sand-gravel tidal inlet using UAV, XGBoost and U-Net. In: Phillips, M.R.; Al-Naemi, S., and Duarte, C.M. (eds.), Coastlines under Global Change: Proceedings from the International Coastal Symposium (ICS) 2024 (Doha, Qatar). Journal of Coastal Research, Special Issue No. 113, pp. 710–714. Charlotte (North Carolina), ISSN 0749-0208. Understanding the distribution of surface sedimentary features within tidal inlets is crucial for assessing their morphodynamic response to waves and tides. Sediment size significantly influences sediment transport and deposition dynamics. However, in mixed sand-gravel tidal inlets, classifying sediment distribution poses unique challenges. Previous research has relied on labour-intensive field sampling, while the rapid spatial changes in the surface features challenge traditional surveying methods and reduce the potential for frequent monitoring. Although satellite imagery offers regular observations, low resolutions hinder the accurate classification of detailed surface characteristics. This study integrates consumer-grade UAV technology with XGboost and U-Net (ResNet34) model to develop automated high-resolution mapping models for surface features in a mixed sand-gravel tidal inlet at the mouth of the Deben estuary, based on the RGB images. The results show that both XGBoost and U-Net have good performance and high potential to classify surface sediments and map these at the pixel level in mixed sand-gravel systems, with relatively high accuracy in the prediction of gravel, sand and vegetation cover. These combined methods demonstrate the potential for regular UAV monitoring of tidal inlets over short- and long-term scales, which can enhance our morphodynamic understanding and contribute to the coastal monitoring and management.