BORIS Theses

BORIS Theses
Bern Open Repository and Information System

Quantification of vegetation effects on shallow landslide probability at regional scales

van Zadelhoff, Feiko Bernard (2023). Quantification of vegetation effects on shallow landslide probability at regional scales. (Thesis). Universität Bern, Bern

23vanzadelhoff_f.pdf - Thesis
Available under License Creative Commons: Attribution (CC-BY 4.0).

Download (10MB) | Preview


Shallow landslides are the cause of considerable direct and indirect losses to individuals, enterprises and society as a whole. An increase in shallow landslide occurrence is often related to the loss of the protective effect of trees by deforestation. Methods of targeted reforestation, such as silvopastoralism, can in turn reduce the risk associated with shallow landslides considerably. The effective implementation of such systems can benefit from guidance in terms of tree placement, planting density, planting pattern, tree species selection and tree size development prediction. For this guidance, computational models can help greatly, in addition to expert knowledge, especially on a regional scale. In this work a quantitative, physically-based, tool to assess the influence of vegetation on shallow landslide occurrence is presented. The focus is specifically on rainfall-induced translational shallow landslides on a regional scale. The results of this tool are ideally incorporated into holistic analyses of site-specific tree benefits and co-benefits. The tool as developed in this thesis is called SlideforMAP. It incorporates a probabilistic approach since the regional scale often poses challenges regarding heterogeneity in a range of parameters. Explicitly including this as uncertainty in the model can improve the model performance. Slope stability is computed by the limit equilibrium approach for a large number of randomly placed hypothetical landslide. The fraction of these hypothetical slides that are unstable, corresponds to local shallow landslide probability. The hydrological module is based on a topographic index and assumes a steady state sub-surface flux. All mechanical influences of vegetation are included with basal and lateral root reinforcement, being incorporated using the Root Bundle Model weibull approach on a single-tree basis. The model is calibrated and applied to three study areas (0.5 - 7.5 km2) in Switzerland. Tree density is 33 to 73 trees/ha on average, but all study areas consist of distinct forested and non-forested sections. Soil thickness, soil cohesion, and the ratio between precipitation intensity and hydraulic saturated conductivity are found to be the most sensitive model parameters. Performance is measured against a 2005 landslide inventory and the Receiver Operator Curve analysis. Area Under the Curve values are between 0.64 and 0.93. It is proven that the single-tree based inclusion of mechanical effects significantly improves model performance, as compared to a forest stand approach. To assess generalizability, SlideforMAP is applied in New Zealand silvopastoral systems in two small study areas (1.4 and 3.5 km2). Over these areas a full calibration and validation of SlideforMAP and an improved version of SlideforMAP is performed. The improved version includes interception, passive earth pressure, root compression, a non-steady state approach to the lateral flux and a runoff coefficient. The runoff coefficient is dependent on cumulative precipitation and relates tree presence to increased macropore presence and subsequently to an increased runoff coefficient. The availability of RADAR-based precipitation data enables a more realistic precipitation input by finding a representative rainfall intensity. The non-steady state approach, runoff coefficients, representative rainfall intensity and novelties. The inclusion of such novelties did not improve model performance, which could be due to flaws in its methodology, the performance measure or site-specific reasons. To place the application of SlideforMAP in context, it is compared to a statistical approach using binary logistical regression to shallow landslide susceptibility in New Zealand silvopastoral systems. For this a 2010 landslide inventory and a 2005 rainfall event are used. The model outputs are in agreement in approximately 70% of the study areas. Accountable for the remaining 30% are methodological differences, the heterogeneous input data in SlideforMAP and the difference in tree influence. SlideforMAP includes mechanical effects, where the statistical method implicitly includes all tree effects, but averaged over all trees. Practitioners are advised to prioritize tree planting on areas where both models are in agreement and predict high shallow landslide susceptiblity. SlideforMAP stands among other state of the art models focussed on vegetation effects on shallow landslide activity. This is due to the inclusion of single tree based lateral and basal root reinforcement and the tailoring to specific rainfall events. Further developments in SlideforMAP have enabled the application on large scales and for the analyses of specific components. All in all, land managers can benefit greatly by applying SlideforMAP to find the ideal targeted planting sites and stabilize the soil, as efficient as possible from a cost-benefit standpoint.

Item Type: Thesis
Dissertation Type: Cumulative
Date of Defense: 29 March 2023
Subjects: 900 History > 910 Geography & travel
Institute / Center: 08 Faculty of Science > Institute of Geography
Depositing User: Sarah Stalder
Date Deposited: 23 May 2023 08:03
Last Modified: 23 May 2023 08:54

Actions (login required)

View Item View Item