Estimation of biophysical properties of woodland canopies using LiDAR and hyperspectral imagery

Janardhan, VB

Abstract

Vegetation canopies represent the interface between the atmosphere and the biosphere
and play an important role in determining energy and mass exchange. Accurate
characterisation of vegetation canopy characteristics is therefore important to
understand these processes, and remote sensing has been used for the past three
decades to retrieve information on forest structure and canopy variables. Most of this
work has employed passive optical sensors at a single date to provide two-dimensional
spectral information that may be correlated with a range of variables. In this research
three-dimensional multi-temporal data was extracted from discrete return airborne laser
scanner (ALS) and compared with data from an airborne hyperspectral sensor. The study
aimed to retrieve woodland stand height, LAI and canopy cover from the two sensor
types at Delamere Forest, located in North West England. The data were collected
between March 2008 and March 2009 in needle leaved evergreen, needle leaved
deciduous and mixed broad leaved deciduous stands to capture seasonal changes in
woodland structure. Ground data were collected at seven test plots to validate the
airborne data, including regular measurements of stand structure using a terrestrial laser
scanner (TLS) and upward-looking hemispherical photography. Comparison between the
plot-based and TLS-based estimates of tree height and canopy cover showed that the TLS
could be used to rapidly and accurately measure tree height, but measurements of
canopy cover were less reliable. A range of LiDAR metrics were applied to the test plots
across time with strong correlations for height and LAI and weaker correlations for cover.
The same metrics were applied to estimate height, cover and LAI at a large number of
plots, and temporal variations examined. Hyperspectral data were extracted for the same
plots and the LiDAR-derived variables correlated with a range of vegetation indices. The
results showed moderate relationships between LiDAR and the hyperspectral data.
Overall, the research has established that i) TLS and ALS can be used to estimate stand
height, but their effectiveness varied with stand characteristics and time ii) ALS can be
used to estimate LAI and their estimates also varied with stand characteristics and time
iii) Point based canopy cover estimates from either TLS or ALS are not reliable. This
research has demonstrated the retrieval of forest biophysical variables from three-dimensional LiDAR and two-dimensional hyperspectral imagery in heterogeneous forest
canopies highlighting their potential to capture forest canopy dynamics.

Citation

Janardhan, V. Estimation of biophysical properties of woodland canopies using LiDAR and hyperspectral imagery. (Thesis). University of Salford

• Thesis Type: Thesis
• Deposit Date: Aug 13, 2021
• Award Date: May 1, 2012