2013). This explanatory capacity may, however, be key in the
sense of explaining a proportion of variance in stand volume
not explained by the height metrics.
Conclusions
An analysis of the effects of reference height thresholds on
metrics one-by-one revealed effects, which were concealed
in studies evaluating the performance of final models. Our
results showed that a bad choice of height thresholds dam-
ages the predictive potential of most metrics, and thresholds
should therefore be chosen for each metric individually.
Hence, we advise that reference heights should be optimized
to each metric, or type of metric, rather than subjectively
fixed for all the metrics collectively, which is common prac-
tice. We propose an optimization method based on
MIC
, since
it accounts for non-linear relationships and may therefore be
best for non-parametric estimation such as nearest neighbor
imputation.
Acknowledgments
This paper is part of the research program developed by
GET-LiDAR, a research group on lidar forest applications at
the Department of Forest Sciences in the University of São
Paulo, Brazil. The airborne data acquisition and the field
data collection were funded by Fibria S. A. and EUCFLUX
project. The Coordenação de Aperfeiçoamento de Pessoal
de Nível Superior, CAPES supported with a scholarship, the
ForEAdapt (FP7-PEOPLE-2009-IRSES) supported with a visit
to UEF/Finland during data processing and FAPESP (process
2014/13381-5) supported with a visit to ESALQ/Brazil during
the paper revision.
References
Andersen, H.E., R.J. McGaughey, and S.E. Reutebuch. 2005.
Estimating forest canopy fuel parameters using LIDAR data,
Remote Sensing of Environment
, 94(4):441–449. doi:10.1016/
j.rse.2004.10.013.
Figure 3. Maximal Information Coefficient (
MIC
) obtained for different reference heights (minimum height,
MinH
) by increas-
ing percentiles of ALS return heights distribution. They are grouped into those computed from (a) all returns, or (b) first
returns only. See Table 2 for metric definitions.
Figure 4. Exemplification of the influence of reference heights (
MinH
, in this case) in metrics: mean of the ALS return heights
(black solid line), their 25
th
percentile of the height returns (lower gray dashed line) and percentile 75
th
of the height returns
(upper gray dotted line).
348
May 2017
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
