ICG2022-536, updated on 10 Jan 2024
https://doi.org/10.5194/icg2022-536
10th International Conference on Geomorphology
© Author(s) 2024. This work is distributed under
the Creative Commons Attribution 4.0 License.

Identification of wall collapses in abandoned agricultural terraces using Object Based Image Analysis and High-Resolution Topography

Gonzalo Fernández-Olloqui1,2, Jorge Lorenzo-Lacruz3, Noemí Lana-Renault3, and Fernando Pérez-Cabello1,2
Gonzalo Fernández-Olloqui et al.
  • 1Department of Geography and Regional Planning, University of Zaragoza, 50009 Zaragoza, Spain.
  • 2Institute of Environmental Sciences (IUCA), University of Zaragoza, 50009 Zaragoza, Spain.
  • 3Department of Human Sciences, University of La Rioja, 26004 Logroño, Spain.

Wall collapses, sometimes accompanied by mass movements, is one of the main degradation processes in long-abandoned agricultural terraces and critically affects the hydrological and geomorphological behaviour of terraced hillslopes (e.g., development of new sediment sources, increasing soil erosion, increasing hydrological connectivity, re-establishment of natural drainage pathways). An efficient monitoring of the spatial and temporal dynamics of wall collapses is fundamental to understand all these processes. However, the identification of terrace wall collapses in the field is a time-consuming task with spatial limitations. In this study, Object Based Image Analysis (OBIA) applied on High-Resolution Topography was used (and evaluated) for detecting wall collapses in terraced hillslopes. The approach, which includes image segmentation and Support Vector Machine classification, was implemented in the Cidacos Valley (Iberian Range, Spain), extensively affected by the abandonment and degradation of agricultural terraces. Point clouds extracted from three different remote data sources (airborne LiDAR, terrestrial LiDAR and airborne photogrammetry) were used to obtain three different digital terrain models (DTM). The low spatial resolution of the DTM derived from airborne LiDAR (1 m pixel size) was not sufficient to detect any terrace wall collapse, which had a median size of 10 m2. The results showed that slope, Sky-View factor, topographic openness and curvature DTM-derivatives produced the best segmentations. Field inventories of wall collapses were used to train the classification algorithm (Support Vector Machine) and validate the results of the OBIA approach. Terrace wall collapses were identified with an Overall Accuracy of 0.80 and a Producer Accuracy of 0.50. The results of the approach and further improvements are discussed. Although promising, the detection of wall collapses in terraced hillslopes using OBIA is challenging, especially when it is compared to the detection of larger scale hillslope geomorphological processes (e.g.: landslides).

How to cite: Fernández-Olloqui, G., Lorenzo-Lacruz, J., Lana-Renault, N., and Pérez-Cabello, F.: Identification of wall collapses in abandoned agricultural terraces using Object Based Image Analysis and High-Resolution Topography, 10th International Conference on Geomorphology, Coimbra, Portugal, 12–16 Sep 2022, ICG2022-536, https://doi.org/10.5194/icg2022-536, 2022.