Today some of our EAGLE students presented their internship and innovation laboratory projects. Very interesting topics and they obviously applied and deepened their remote sensing knowledge a lot.
Julia Sauerbrey: Prediction of Organic Matter Content from Sentinel-2 Imagery in Germany
Marina Reiter: Conservation in Tasmania: Green Leaf Project and World Class Reserve Systems
Sarah Nolting: Regional Scale Statistical Mapping of Snow Avalanche Likelihood and its Combination with an Optical Remote Sensing Based Avalanche Detection Approach – First Attempts for the Province of South Tyrol (Italy)
The course “from field work to spatial data” by Tobias Ullmann and Martin Wegmann is covering the whole range of field campaign planning and especially training all necessary methods such as GPS handling, coordinate systems, setting waypoints or finding locations. In the next weeks the EAGLE students will collect field data, link it to attribute tables, import it into a GIS and create a land cover classification. The final project will be an actual field campaign in the Steigerwald area which they have to design, plan, coordinate and analyze it.
Jakob Schwalb-Willmann just started his M.Sc. thesis titled “A deep learning movement prediction model using environmental data to identify movement anomalies”. He will combine animal movement and remote sensing data in order to develop a generic, data-driven DL-based model that predicts movements from movement history alongside environmental covariates in order to detect movement anomalies. He will establish simulated, controlled environments that allow precise adjustments of the model inputs to test the model’s feedbacks and its variability. It can be considered as a precursor study for the model’s deployment on real data and to only experimentally apply it on such due to the given constraints (time and content) of his M.Sc. thesis. The first supervisor is Dr. Martin Wegmann.
Today our EAGLE students applied data munging, pipes, plotting and statistics using colour distribution of sweets. They specifically used the dplyr, ggplot, kableExtra and others to compute derivatives, rearrange the data, plot it and run statistics on colour distribution differences between different sweet packages. After this initial task of the course “spatial prediction and modelling” actual spatial prediction of in-situ point data combined with remote sensing data will be performed. This course will cover a wide range of approaches and packages to run spatial models and predict them spatially and temporally.
The final project presentations of the spatial coding course by the EAGLE students revealed quite some impressive analysis achieved within the last couple of months. All analysis were done using R and presentations created within R using knitr. The aim was to run a variety of remote sensing analysis using R coding in order to increase the coding and remote sensing skills.
The project topics ranged from landcover change detection to sensitivity analysis of thresholds, covariates or training data concerning supervised classifications. All students added their scripts to a git and some even provided dedicated R packages for their specific analysis.