Analysis of the interaction between road network and gully erosion in Kinshasa, D. R. Congo

Details

Supervisors

Defourny, Pierre ; Bielders, Charles

Faculty

Faculté des bioingénieurs

Degree label

Master [120] en sciences agronomiques et industries du vivant, à finalité spécialisée: Geo-Information Science and Earth Observation for Environmental Modelling and Management

Abstract

Urban gullying poses serious environmental and socio-economic threats to rapidly urbanizing areas. The complex interaction between natural processes and urban expansion metrics triggers the formation and expansion of gullies in these megacities. This research aims to analyse the interaction between road network and gull erosion in the urban city of Kinshasa, D. R. Congo. The study area covers three communes - Ngaliema, Selembao, and Lemba where current gullying activities have been observed. The classification of these gullies into distinct typologies according to their interaction with the road network was conducted by integrating the DEM, road network dataset, gully inventory, and building dataset, in addition to the Google Earth satellite imagery. Through this qualitative analysis, gullies induced by topography, roads, and buildings were identified, which emphasizes the diverse factors influencing gully erosion in the study area. Using the road network dataset, the DEM of the region, and the existing gully inventory, the methodology relies on calculating the drainage direction and flow accumulation of the gully heads, identifying the contributing pixels, and extracting roads that feed into these pixels. By integrating the attributes of these roads, mainly the surface cover, and length, a logistic regression model is built to assess the correlation of these road features to gully erosion in the region. The results reveal that road length is positively correlated with gully occurrence and is significant while surface cover types unpaved and concrete surfaces correlate with gully occurrence though not significant. The performance of the logistic regression model showed a moderate performance in predicting gully erosion with 0.61 accuracy using only the length and surface cover of the roads. These findings emphasize the potential effectiveness of incorporating road network indicators into urban planning strategies to mitigate gully erosion risks.