This repository contains the R-scripts used in the master thesis "Development of a new spatial cross-validation strategy and its application for machine learning based modelling of ecological data" by Jan Linnenbrink. The thesis is partly based on "kNNDM: k-fold Nearest Neighbour Distance Matching Cross-Validation for map accuracy estimation" by Jan Linnenbrink, Carles Milà, Marvin Ludwig and Hanna Meyer. The manuscript has been submitted to the journal Geoscientific Model Development.

The repository is structured as follows:

• simulation_virtualSpecies: This folder contains the code to reproduce the first simulation.

  • figures.Rmd: R-Markdown file to reproduce the figures.
  • code: contains the R-code to reproduce all analyses.
  • data: contains the data used in the simulation.
  • results: contains the generated results.
  • figures: contains the figures.

• simulation_AGB_OCS: This folder contains the code to reproduce the second simulation.

  • summary.R: The R-Script to generate the figures of simulation
  • R: contains the R-scripts to run the deBruin simulations.
  • material: contains the resulting tables and figures.

• global_map: This folder contains the code to reproduce the global map.

  • code: contains the R-code to reproduce the mapping study.
  • results: contains the reproduced results based on random CV.
  • results: contains the reproduced results based on kNNDM CV.
  • figures: contains the figures.