Augmenting a 2D soil map with a 3D perspective: A case study of soil nitrogen in Burkina Faso
The purpose of this student project was to use our own creativity to develop a map as part my coursework for the International Master Cartography at Technische Universität Wien (TUW), Austria.
- Completed: April 2023
- Context: M.Sc. Student Project at TUW, Austria
- Supervisor: Dipl.-Ing. (FH) Manuela Schmidt
- Tools: GEE, R Programming, QGIS
Study Area
The study area is Burkina Faso as shown in the map. This was selected as a result of availability of data from ISRIC-World Soil Information WoSIS Snapshot 2019 dataset (License CC BY-NC 3.0). The yellow dots are the Total Nitrogen soil sample locations.
WORKPLAN
The methodology is as shown in the image.
MACHINE LEARNING TRAINING AND PREDICTION
I trained a Random Forest regression model on over 400 covariates across Climate, Remote Sensing Reflectances and Relief with the Total Nitrogen as the dependent variable.
The model achieved an accuracy of 50% which was used to make predictions at a 1km x 1km grid intervals in Burkina Faso. The result is a 2D map as shown.
BUILDING 3D MAP FROM THE 2D MAP
I used the rayshader package in R to develop a 3D map based on the predicted 2D GeoTIFF image. The results is as shown.