Keywords

• Sediment fluxes and sediment budget
• Landslide volume assessment
• Structure-from-Motion (SfM) and very-high resolution DSM extraction
• Alpine environments

Research project

Advisors: Veerle Vanacker and Kristof Van Oost.

Context

Sedimentary fluxes exported to oceans come largely from tectonically active mountain ranges and particularly from small river catchments whose mean elevation is higher than 3,000 meters and whose mean slope gradient is high (Milliman, 1995). Recently, it has been quantified that more than 50% of the global denudation occur on the steepest 10% of continental surfaces (Larsen et al., 2014; Willenbring et al., 2014).  Sedimentary dynamics in mountainous environments is controlled by natural climatic (Crozier, 2010) and anthropic perturbations (Glade, 2003; Vanacker et al., 2003). Besides, knowledge in the field of sedimentary dynamics is increasing and the concept of sediment budget is one method to model these phenomena (Hinderer, 2012). A sediment budget is defined as the accounting of sources, sinks and redistribution pathways of sediments in a unit of region over unit time (Slaymaker, 2003). However, no unifying methodological concept exists because of the wide range of temporal and spatial scales and the different transport modes and properties of matter. Three approaches can be used to handle the sediment budget concept (Hinderer, 2012). (1) Sediment fluxes can be quantified with several techniques of sediment fingerprinting, depending on temporal and spatial scale (D’Haen et al., 2012), or by GPS measurements (Schwab et al., 2007). Regarding sediment fingerprinting techniques, it is worth to mention fallout radionuclides (i.e. ¹³⁷Cs, ²¹⁰Pb and ⁷Be) to study erosion and soil redistribution at the catchment scale over short time ranges (Du and Walling, 2011; Van Oost et al., 2005) and cosmogenic radionuclides (i.e. ¹⁰Be) which are useful for assessing long-term denudation rates because of their longer half-life (von Blanckenburg, 2005). (2) Sediment volume assessment and monitoring can be achieved by topographic reconstruction of earth surfaces based on ground or remote acquisition techniques. These topographic surveys are typically conducted using a real-time kinematic (RTK) GPS (e.g. Brasington et al., 2000; Squarzoni et al., 2005), total stations (e.g. Fuller et al., 2003; Moss et al., 1999), terrestrial laser scanning (e.g. Brasington et al., 2012; Travelletti et al., 2008) and aerial laser scanning (e.g. Dewitte et al., 2008; McKean and Roering, 2004; Notebaert et al., 2009). While these acquisition techniques are increasingly used in geomorphology and result in rich observational datasets, they are often time-consuming and costly. Recent developments in image processing, with the application of computer vision algorithms as Structure-from-Motion (SfM), and the availability of reliable, low-cost and lightweight unmanned aerial vehicles (UAVs), permit to overcome some of these drawbacks, i.e. for landslide monitoring (Niethammer et al., 2012). (3) Earth surface processes can be embedded in physically-based spatial numerical models to explore quantitative linkages of sediment production and sediment routing in complex routing systems and to understand the evolution of landscapes and mountain ranges (Tucker and Slingerland, 1997; Tucker et al., 2001).

Mass movements exert a primary control on the landform development, incision history and sediment discharge of mountainous watersheds (i.e. Hovius et al., 1997; Korup et al., 2004). However, temporal mismatch between sediment production by landslide, storage on slopes and export to river streams is often not yet integrated into current concepts of feedback mechanisms between uplift and erosion in tectonically active mountain belts (Korup et al., 2010). In this thesis, we will focus on the multi-temporal assessment of sediment fluxes in a landslide-affected river catchment located in the Central Swiss Alps. In other words, we aim at quantifying residence time of sediments on hillslopes before being transported into the channel network at different temporal scales. At short-term (5 years), slip rates and sediment volumes will be assessed by monitoring a landslide based on a time series of very-high resolution Digital Elevation Models (DEM). At intermediate time scale, integration of erosion rates datasets from previous studies will be done, i.e. dendrochronology (Savi et al., 2013) and classic aerial photogrammetry (Schwab et al., 2008). Finally, long-term sediment fluxes will be derived from the quantification of ¹⁰Be cosmogenic radionuclides in river sediments.

One aspect to highlight in this research is the monitoring of an active earth flow in the Central Swiss Alps to infer erosion rates at short term. To this end, a low-cost and lightweight UAV (Unmanned Aerial Vehicle) platform equipped with a standard reflex camera is developed in order to repeatedly acquire aerial photographs through time. Based on these datasets, very-high resolution earth topography representations can be reconstructed based on SfM algorithms (Eisenbeiss and Sauerbier, 2011; Niethammer et al., 2010).

Research objectives

The main goal of the research project is to get a better understanding of the sedimentary dynamics in mountain environments where landslides are an important erosion controlling factor. Therefore, we aim at quantifying the impact of landslides on the global sediment budget of the catchment at several temporal scales.

The first section of this thesis is about the performance of pre-processing techniques on semi-automatic landslide detection on high resolution satellite data, i.e. ASTER images. This part explores the impact of atmospheric and topographic corrections of remote sensing data on the spatial pattern of detected landslides based on NDVI threshold values in the mountainous tropical region in northern Vietnam. In fact, improvements in spatial mapping of landslides are useful to better indirectly derive sediment fluxes based on the area-volume relationship characterizing landslides.

 Related research question

• What is the performance of the atmospheric and topographic corrections applied to high resolution ASTER images for the semi-automatic landslide detection based on NDVI values? Part 1

The second section of this thesis consists in the monitoring of an active earth flow based on a time series of very-high resolution digital surface models (DSMs), over 5 years. As already mentioned, it will be achieved by 3D topographic reconstructions based on aerial photographs taken from an UAV. It will allow us to quantify erosion rates at short term and assess whether the landslide sediment supply is directly exported to the channel network.

 Related research questions

 Are earth topography reconstructions based on the combination of UAV and SfM algorithms reproducible? What are the parameters that influence the accuracy and the precision of this workflow? Part 2

• Are hillslopes and channels coupled regarding sediment transport at short term within the Schimbrig landslide? Will landslide sediment supply stay on hillslopes or will it be directly exported to the channel network? Part 3

The third section of the thesis will explore the sediment dynamic within the catchment at intermediate time scale, i.e. decadal to centennial perspective. The aim is to integrate results about slip rates and chronology spatial pattern of mass movements, from two existing studies based on dendrochronology (Savi et al., 2013) and on classic aerial photogrammetry (Schwab et al., 2008).

 Related research question

• Are the sediment fluxes assessed at intermediate time scale consistent with the values derived from very-high 3D topographic reconstructions at short term over the Schimbrig landslide? Part 4

The fourth section concerns the erosion rates over thousand years, which can be derived based on ¹⁰Be cosmogenic radionuclides quantification, in the river catchment. These measurements will be achieved within the Schimbrig catchment in order to explore whether modern erosion rates are consistent with long term rates. The latter will be also compared to long term erosion rates derived in a river catchment which is not affected by landslides.

Related research questions

• Are long term erosion rates consistent with modern erosion rates? Do landslides affect the sediment budget of a river catchment over long time scales? Part 5

Finally, a last section is considered to conclude the entire project by summarizing the sediment dynamic over the different studied time scales, by including the effect of landslides on the sediment budget. Part 6