Light detection and ranging (LiDAR) has become a common tool for
generating remotely sensed forest inventories. However, regional
modeling of forest attributes using LiDAR has remained challenging
due to varying parameters between LiDAR datasets, such as pulse
density. Here we develop a regional model using a three dimensional
convolutional neural network (CNN). We then apply our model to
publicly available data over New England, generating maps of
fourteen forest attributes at a 10 m resolution over 85 % of the
region. Attributes include aboveground biomass (kg), total biomass
(kg), tree count (#), percent conifer (%), basal area (m^2), mean
height (m), quadratic mean diameter (cm), percent spruce/fir (%),
percent white pine (%), inner bark volume (m^3), merchantable volume
(m^3), and spruce/fir volume (m^3. All values correspond to the
amount per pixel cell (I.E. kg of biomass found within that pixel).
Map/model performance was assessed using the USFS’s FIA inventory,
which constituted an independent dataset free from spatial
autocorrelation. More data can be found in the following pre-print:
Ayrey, E., Hayes, D. J., Kilbride, J. B., Fraver, S., Kershaw, J.
A., Cook, B. D., & Weiskittel, A. R. (2019). Synthesizing
Disparate LiDAR and Satellite Datasets through Deep Learning to
Generate Wall-to-Wall Regional Forest
Inventories. bioRxiv, 580514.
