IEEE Geoscience and Remote Sensing Magazine - March 2021 - 244

Incorporating 3D lidar observations across scales and ecosystems need to be extensively explored in future studies.
FROM 3D TO THE MULTIDIMENSIONAL
BIG DATA ERA
In conjunction with optical remote sensing and field surveys, lidar has played a crucial role in the development of
3D ecological observations and modeling. Lidar data have
allowed researchers to gain new insights and gather finer
details about ecological processes [256], [257]. Nevertheless, increasing demands have emerged to expand our ecological observations from 3D (i.e., X, Y, and Z) to multiple
dimensions (i.e., X, Y, Z and the spectral and temporal dimensions) to compensate for the weaknesses of single-type
or single-date remote sensing data.
MOVING TOWARD THE FUSION OF STRUCTURAL
AND SPECTRAL INFORMATION
Lidar has the known limitation of lacking spectral information. The intensity recorded by lidar sensors is highly influenced by signal attenuation and multipath effects [142], [296].
Although there have been significant efforts in developing
algorithms to calibrate lidar
WE NEED TO MONITOR
intensity [52], [297]-[301],
VEGETATION DYNAMICS
it is still difficult to generate
USING MULTITEMPORAL
physically meaningful specDATA SO THAT WE CAN
tral reflectance values or crossFURTHER UNDERSTAND
platform comparable intensity
values [296]. Moreover, most
ECOLOGICAL PROCESSES
current lidar systems have only
AND PREDICT POSSIBLE
one band in the near-infrared
ECOSYSTEMS CHANGES IN
wavelength range. Recent deFUTURE CLIMATE
velopments in dual-band and
SCENARIOS.
multiband lidar sensors show
the potential to address the
limitations of single-band lidar systems [57], [302]-[304], yet the costs of dual-band and
multiband lidar sensors are high and their capability to derive
cross-platform comparable reflectance values is still questionable [305]. Another solution for addressing this problem is to
fuse lidar data with multispectral and hyperspectral data to
compensate for the weakness of each type of data (i.e., lidar
data lack spectral information and multispectral and hyperspectral data lack structural information). Studies have proven
that by fusing lidar data with multispectral or hyperspectral
data, the resulting data can outperform the single type of data
in terms of species classification, vegetation mapping, biomass
estimation, and so on [163], [224], [305]-[308]. Therefore,
the simultaneous collection of multisource remote sensing
data remains the trend for future ecological observations.
MOVING TOWARD REGULAR TIME-SERIES
ECOLOGICAL OBSERVATIONS
Current available lidar data sets have been used to create valuable benchmarks of ecosystem attributes (e.g.,
244

vegetation attributes, snow depth, and tree line information) [163], [309]-[311], which can then be used as important baselines to evaluate ecosystem dynamics under the
background of global climate change. However, we need
to monitor vegetation dynamics using multitemporal data
so that we can further understand ecological processes and
predict possible ecosystems changes in future climate scenarios. Thus far, such studies have been dominated by the
use of time-series optical images and radar data across the
globe (e.g., Landsat TM, MODIS, Sentinel-1, and Sentinel-2)
[5], [312]-[314]. Although there are studies using multitemporal lidar data to investigate ecosystem dynamics
[169], [170], [172], [192], [315], collecting time-series lidar
data with a constant time interval has been sporadic due
to the paucity of spaceborne lidar sensors and the high cost
of collecting airborne and terrestrial lidar data. Recent developments in near-surface lidar platforms (e.g., backpack
lidar and UAV lidar) have greatly reduced the cost and increased the flexibility of lidar data collection. Moreover,
newly launched spaceborne lidar sensors (e.g., GEDI and
ICESat-2) can provide global-scale time-series lidar observations. This progress in lidar sensors has made the collection of time-series lidar data possible.
MOVING TOWARD THE ERA OF BIG DATA
It has been estimated that the data streaming from the
National Aeronautics and Space Administration active
space missions alone represent more than 1.7 gigabytes
[316]. With the accumulation of multisource, multitemporal, and multiscale remote sensing data, the amount of
ecological observations will grow explosively [317]. Remote sensing-based ecological observations are stepping
into the big data era.
Upon the transition of ecological observations from
3D to the multidimensional big data era, three common
challenges are being faced across ecosystems. First, how
should lidar data be fused with multispectral and hyperspectral data? The majority of existing studies have simply
compressed the 3D lidar data into 2D vegetation attributes
and then fused the derived vegetation attributes with those
from multispectral and hyperspectral data directly based
on spatial information [307], [318]. Further studies are
needed to explore the possibility of fusing lidar data with
multispectral and hyperspectral data in 3D so that the potential of multisource data can be fully explored. The use
of 3D radiative transfer models in the data fusion process
might be a potential solution to this question.
Second, how should changes from multitemporal lidar
data sets be detected? Due to the changes in lidar sensors,
viewing angles, vegetation conditions, and so on, lidar data
collected at different times and from different platforms
can hardly be directly compared [319]. The registration error of multitemporal lidar data may further increase the
difficulty of detecting ecosystem changes [41]. Finding a
method that can be used to compare lidar data across platforms and time is urgently needed.
IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE

MARCH 2021

IEEE Geoscience and Remote Sensing Magazine - March 2021

Table of Contents for the Digital Edition of IEEE Geoscience and Remote Sensing Magazine - March 2021