Module 4 – GeoMorpho

The main objective of this module will be to investigate the link between vegetation and geomorphic parameters and to provide spatially-distributed geomorphological data for the vegetation, soil and hydrogeological models (modules 2, 3 and 5).

Detailed tasks

Task 1 – Ground (surface) characterization and vegetation mapping – In order to understand the interactions between geomorphic variables and plant development, we will characterize the
ground surface and the subsurface at a high-resolution spatial scale and confront this geomorphic data with plant inventories and mapping, in collaboration with module 2. These detailed analyses will be carried out only at three specific field sites of limited spatial extent. The main geomorphic variables that will be investigated are 1) the type of landform, with GIS and field mapping; 2) the surface granulometry, by automatic mapping from drone surveys; 3) the ground characteristics (presence of permafrost, porosity, hydraulic conductivity), with geoelectrical surveys; 4) the ground surface temperature, with i-buttons; 5) the topography (slope gradient, aspect, curvature, etc.), by DEM manipulation; and 6) the process activity, in particular changes in ground cover over the last decades, by orthophoto comparison (digital photogrammetry).

Task 2 – Inference of geomorphological variables – Some of the data acquired in task 4.1 will be spatialized using geostatistical techniques in collaboration with module 1. Permafrost will be mapped based on Machine Learning algorithms developed in an ongoing SNF funded project.
Training dataset composed of permafrost evidences, ground cover characteristics data and topo-climatic data will be used to discover the functional dependencies between mountain permafrost and its controlling factors. These models are datadrive and require large amounts of data, provided by field investigations and the other project modules and centralized through the GeoDataHub. The permafrost evidences will come from: 1) geophysical and thermal data gathered in Task 4.1 and from 2) creeping permafrost features (typically rock glaciers), which will be inventoried and mapped for the entire study area, mainly by aerial photo interpretation. In addition to permafrost spatialization, detailed granulometry maps will be generated for the entire study area based on the drone-borne close-up photographs acquired in module 1.
These photographs will be georeferenced and analyzed using the Sedimetrics software, which comprises an algorithm able to automatically derive a granulometric curve through an image segmentation procedure. Using the granulometric curve as a signature will allow for identification of geomorphological facies and the production of geomorphological maps.

Task 3 – Semi-automated geomorphological mapping – This task will aim at developing a complete geomorphological map for the entire Vaud Alps, for use in the vegetation, soil and geohydrological models in modules 2, 3 and 5. As the study area covers a surface of 720 km2, establishing a classical geomorphological map (ie. by manual mapping on the field and on the computer) is not feasible. Based on the large amount of predictors accumulated in this project, we will adapt existing state-of-the-art semiautomated mapping methodologies. These methods are generally based on the extraction of geomorphological features from DEM and some of them are based on multiple-point geostatistics (Vannametee et al. 2014). To improve the accuracy of the geomorphic classification, we will develop a methodology for semi-automated mapping with ML algorithms, which uses not only high resolution DEM, but also orthorectified aerial photographies, vegetation and granulometry maps and a training map, in collaboration with module 1 for the geostatistical aspects related to the spatialization of the predictors. The training map will be a classical geomorphological map, which will be produced in the Vallon de Nant – Anzeindaz region, which covers around 47 km2.