An evidential reasoning geospatial approach to transport corridor susceptibility zonation

Abstract

Given the increased hazards faced by transport corridors such as climate induced extreme weather, it is essential that local spatial hot-spots of potential landslide susceptibility can be recognised. Traditionally, geotechnical survey and monitoring approaches have been used to recognise spatially landslide susceptibility zones. The increased availability of affordable very high resolution remotely-sensed datasets, such as airborne laser scanning (ALS) and multispectral aerial imagery, along with improved geospatial digital map data-sets, potentially allows the automated recognition of vulnerable earthwork slopes. However, the challenge remains to develop the analytical framework that allows such data to be integrated in an objective manner to recognise slopes potentially susceptible to failure. In this research, an evidential reasoning multi-source geospatial integration approach for the broad-scale recognition and prediction of landslide susceptibility in transport corridors has been developed. Airborne laser scanning and Ordnance Survey DTM data is used to derive slope stability parameters (slope gradient, aspect, terrain wetness index (TWI), stream power index (SPI) and curvature), while Compact Airborne Spectrographic Imager (CASI) imagery, and existing national scale digital map data-sets are used to characterise the spatial variability of land cover, land use and soil type. A novel approach to characterisation of soil moisture distribution within transport corridors is developed that incorporates the effects of the catchment contribution to local zones of moisture concentration in earthworks. In this approach, the land cover and soil type of the wider catchment are used to estimate the spatial contribution of precipitation contributing to surface runoff, which in turn is used to parameterise a weighted terrain accumulation flow model. The derived topographic and land use properties of the transport corridor are integrated within the evidential reasoning approach to characterise numeric measures of belief, disbelief and uncertainty regarding slope instability spatially within the transport corridor. Evidential reasoning was employed as it offers the ability to derive an objective weighting of the relative importance of each derived property to the final estimation of landslide susceptibility, whilst allowing the uncertainty of the properties to be taken into account. The developed framework was applied to railway transport earthworks located near Haltwhistle in northern England, UK. This section of the Carlisle-Newcastle rail line has a ii history of instability with the occurrence of numerous minor landslides in recent years. Results on spatial distribution of soil moisture indicate considerable contribution of the surrounding wider catchment topography to the localised zones of moisture accumulation. The degrees of belief and disbelief indicated the importance of slope with gradients between 250 to 350 and concave curvature. Permeable soils with variable intercalations accounted for over 80% of slope instability with 5.1% of the earthwork cuttings identified as relatively unstable in contrast to 47.5% for the earthwork embankment. The developed approach was found to have a goodness of fit of 88.5% with respect to the failed slopes used to parametrise the evidential reasoning model and an overall predictive capability of 77.75% based on independent validation dataset.