Assistant Professor for Geo-Environmental Data Science
- Statistical and machine learning based data analysis
- Development of spatial sampling design frameworks
- Predictive mapping using a broad range of covariates
- Non-parametric validation statistics for prediction
- (spatial) statistical analysis using classical statistics and a large range of machine learning approaches
- Adaptation of statistical methods to soil and spatial data requirements and domain knowledge
- Geodata processing of very large datasets
- Visualisation of results e.g. as web-based GIS or interactive web graphics.
Fig. 1: Evaluation of a wide range of machine learning methods to accuratly predict soil texture for two study areas. Prediction methods for study areas with large uncertainties often smooth the distribution of the outcome. This might have an relevant influence on the end-use of soil texture maps, because the smoothing is even more pronounced for compositional data like sand, silt and clay (that have to sum up to 100 %).