Land cover and land use mapping of the iSimangaliso Wetland Park, South Africa: comparison of oblique and orthogonal random forest algorithms

A study comparing oblique and orthogonal random forest algorithms for land cover classification in the iSimangaliso Wetland Park, South Africa, found that both methods achieved high accuracy (>80%) using WorldView-2 image data. While no significant difference was observed between the two algorithms, both struggled to accurately classify indigenous forests. The research highlights the potential of oblique random forest (oRF) as a robust tool for mapping land cover and use in heterogeneous protected areas. This approach leverages supervised models to optimize tree splits, offering a promising alternative or complement to traditional random forest (RF) methods.

Bassa, Z., Bob, U., Szantoi, Z., & Ismail, R. (2016). Land cover and land use mapping of the iSimangaliso Wetland Park, South Africa: comparison of oblique and orthogonal random forest algorithms. Journal of Applied Remote Sensing, 10(01), 1–22. https://doi.org/10.1117/1.jrs.10.015017