The first contribution of the GeoDataHub to the overall project will be the acquisition and gathering of remote sensing data for the region studied. A second task of the GeoDataHub is to enable the flow of data and information across the interface the other modules. As such, the GeoDataHub will constitute the main platform for the transfer of interdisciplinary data throughout the project.
Task 1 – Identification, gathering and self-acquisition of remote sensing data – The gathering of remote sensing data will focus on two specific scales: the first one is the large-scale, low resolution
data, relying on satellite images, aerial images and aggregated products (cryoland.eu data for snow cover, gridded Meteoswiss data for rainfall, temperature and solar radiation). The satellite images used will mostly consist of Landsat-7 and Landsat-8 data, freely available every 8 days at resolutions ranging from 15m (panchromatic) to 100m (thermal). The variables that will be collected at the large scale are albedo, surface temperature and surface reflectance in a variety of
spectral bands. In addition, detailed RGB and infrared aerial images will be used for the local study area (Nant- Anzeindaz). The other scale is the small-scale, high-resolution data that will be acquired for specific focus sites (in collaboration with module 3) using equipment that is already operational as at the University of Lausanne and will be brought to the project as in-kind (initially one, then several QuadH2O and HexH2O drones, http://www.quadh2o.com, which are equipped with sensors able to measure point rainfall, i.e. mobile rain gauges, as well as visible and thermal
cameras). On focus sites, drones will be used to measure point rainfall, detailed soil structure imagery and localized day/night temperature changes.
Task 2 – Data organization and spatialization – This task will involve setting up a data server (GeoDataHub) based on the NAS computer infrastructure provided by the University of Lausanne. This server will host and share the entire project the raw data. More importantly, the GeoDataHub will translate all data to a common denominator, which will consist in a series of gridded layers that will serve as inputs to modules 1 to 4. A high-precision digital elevation model for the entire region (swissTLM3D, available at UNIL) will also be included in the GeoDataHub. The production of unified data layers will involve geostatistical interpolation methods and upscaling/downscaling techniques that will be based on training samples acquired at the focus sites (module 3). An important challenge will be to address the spatialization of non-stationary variables, i.e. presenting trends related to altitude changes, which can be very important for some variables . To this end, specific geostatistical methods will be used.
Task 3 – Data expansion: developing new variables from existing ones – A fundamental role of data integration is to make sense of multiple data such that the final product is more informative
than the sum of its parts. This will be accomplished by processing the layers produced in the above task. The following additional layers will at least be produced and made available to the other modules. Using the Landsat images, maps of predictors for the distribution of ecosystems will be produced. Based on albedo values, the presence or absence of snow cover will be derived on a sub-monthly basis for the last 43 years (the Landsat perspective), and used in models of vegetation, soils, permafrost and hydrology. The resulting time-series will also allow directly retrieving spatial information on permafrost, vegetation (e.g. using NDVI) and presence of water bodies (e.g. NDWI). Pansharpening will be used to integrate satellite data with existing databases that are available at a higher temporal resolution.
Based on drone surveys, close-up photographs will be available of the sediment cover. In collaboration with module 4, these photographs will be used to derive spatialized granulometric curves based on morphological properties, which will be validated against the soil data (module 3) and the vegetation data (module 2). In collaboration with modules 2 and 5, soil moisture maps will be derived from intra-day surface temperature changes. This will be done using the principle of thermal inertia which can allow identifying the thermal properties of soils, which is highly related to the water content. The same input data (e.g. surface temperatures,
albedos) will in turn allow estimating land surface evapotranspiration using an energy balance model to be used in modules 2 and 5.
Automated geomorphological mapping based on spatialized predictors and detailed granulometric signatures determined at focus sites (see details in module 4).