RegisTree: a registration algorithm to enhance forest inventory plot georeferencing

The accuracy of remote sensing-based models of forest attributes could be improved by controlling the spatial registration of field and remote sensing data. We have demonstrated the potential of an algorithm matching plot-level field tree positions with lidar canopy height models and derived local maxima to achieve a precise registration automatically.

Context The accuracy of remote sensing-based estimates of forest parameters depends on the quality of the spatial registration of the training data.
Aims This study introduces an algorithm called RegisTree to correct field plot positions by matching a spatialized field tree height map with lidar canopy height models (CHMs).
Methods RegisTree is based on a point (field positions) to surface (CHM) adjustment approach modified to ensure that at least one field tree position corresponds to CHM local maxima.
Results RegisTree has been validated with respect to positioning errors and the performance of lidar-derived estimation of plot volume. Overall, RegisTree enabled to register field plots surveyed in a range of forest conditions with a precision of 1.5 m (± 1.23 m), but a higher performance for conifer plots, and a limited efficiency in homogeneous stands, having similar heights. Improved plot positions were found to have a limited impact on volume predictions under the range of tested conditions, with a gain up to 1.3%.
Conclusion RegisTree could be used to improve the forest plot position from field surveys collected with low-grade GPS and to contribute to the development of processing chains of 3D remote sensing-based models of forest parameters.

Keywords
Forest inventory, Lidar, Plot positioning, Registration algorithm, Forest parameter estimation

Publication
Fadili, M., Renaud, JP., Bock, J. et al. Annals of Forest Science (2019) 76: 30. https://doi.org/10.1007/s13595-019-0814-2

For the read-only version of the full text: https://rdcu.be/buDwl

Data availability
The datasets generated and/or analyzed during the current study are not publicly available due to ownership and funding constraints but are available from the author upon reasonable request and with permission of ONF and IGN. Code and sample data supporting the findings of this study are available in the Zenodo repository (Fadili et al. 2019).