Jaime C Revenga , Stefan Oehmcke, Mana Gharun, Flurin Sutter, Katerina Trepekli, Fabian Gieseke, Nina Buchmann, Alexander Damm

Forest structure analysis and biomass prediction systems are key tools for advancing
forest trait-based ecology and management. Surveys using Unmanned Aerial Vehicles
(UAV) and Light Detection and Ranging (LiDAR) systems have contributed
to this field with increased accuracy in tree phenotyping. Moreover, methods combining
UAV LiDAR surveying and machine learning (ML) have also emerged to enhance
estimates of single tree traits. Here, we utilized
a UAV LiDAR system to survey a Norway spruce forest in Davos, Switzerland, where a
detailed field-based inventory served as ground truth data. Our objectives were (i) to
gain insights into variation and gradients of tree height and (ii)
to evaluate whether such insights may prove useful as contextual information to
improve predictions of stem diameter and tree-level biomass. We segmented the point
cloud data scene into individual canopies and treated the LiDAR derived tree height as
the variable of interest.
We then used local indicators of spatial association to detect the significant local
context, and defined tree neighborhoods within the forest. Then,we extracted metrics
fromthe neighborhoods and introduced them in a ML regression experiment to
evaluate predictions of individual tree diameter.
The focus was on comparing performance of tree diameter predictions between
regression models that either consider neighborhood metrics (i.e. context-aware
models), or not. Next, AGB was estimated from the tree height derived
from theUAV LiDAR survey, the predicted tree diameter and allometry. The benefits of
context awareness were assessed in terms of accuracy gained in estimating AGB. We
obtained results of different machine learning methods
(i.e. AdaBoost, Lasso and Random Forest) and evaluated these based on nested
cross-validation. We applied this approach to two separate tree data sets within the
same site, one being clustered and continuous, the other discontinuous
and scattered in separate sampling plots. In both cases, we found evidence of
enhanced AGB prediction performance in context-aware regressions, where the RMSE
was reduced by 4.0% and by 9.1%, respectively.
These findings indicate that gradients in tree heights across the ecosystem may proxy
for local microclimate, edaphic conditions and biotic factors that influence tree growth,
which can be leveraged to enhance predictions of AGB. The method proposed is fully
native to UAV LiDAR data.