Detection of forest gaps due to

lightning based on LiDAR technology

Ali Fakour
MSc graduate in Gorgan University of Agricultural Sciences and Natural Resources
[email protected]

Abstract
The skyrocketing rise in spatiotemporal changes in composition,
structure and function of forest ecosystems during recent years, it
makes inevitable to use remote-sensed data as the main complementary
sources. Parallelly, procedures for forest quantification by light
detection and ranging (LiDAR) data have been developed at a rapid
pace. At this research, the airborne LiDAR data is employed to create
Canopy Height Model (CHM). After applying necessary corrections,
the final raster map is produced which suggests existent gaps owing to
lightning in the forest study area. The result map could be used to detect
the temporal change in order to predict the sensitive sites and
considering them in future protection plans. This research also
illustrates one of the diversified applications of LiDAR method in forest
management which is the fundamental goal.

Keywords: Forest, LiDAR, GIS, Remote sensing, Lightning

 1. Introduction
Some people argue that lightning does not damage the same place twice, however only one beat is
adequate for crucial damage. In addition to killing about 24000 fellows every year, they cause forest
fires, power outages and structural destruction (European Commission website). Forests are of
substantial protectors of the world. What is important is that only large-scale damages are evident.
Thousands of trees have been demolishing due to lighting (Taylor, 1974). Indeed, forests have seen
lightning damages through the history and the biological and physical impacts have been interesting
for scientists (e.g., Stone, 1903; also, Rakov & Uman, 2003). Global warming, changing the local
weather and demolishing of species are among the foremost results of forest fires (Alkhatib et al,
2014). Estimation of tree morality seems to be an essential factor especially in forest modeling (P.
Yanoviak et al, 2017). Unfortunately, being remote and filled by trees have made it difficult to assess
the changes. Whereas field-based surveys greatly have spatial limit, the air borne and satellite born
methods have played an important role in the damage assessment in the larger scales. They have been
employed for detection of changes owing to the wildfires. The cons are affecting the quality due to the
atmospheric conditions, high expenditures and therefore the low accessibility (Alkhatib et al, 2014). It
has to be impossible to measure gaps via a ground-based survey on a large scale because of the logical
dilemmas of intensified forests. Light detection and ranging (LiDAR) is a state-of-the-art technology
which presents direct measurements of ground and canopy elevations separately (Zhang, 2006).
Airborne Laser Scanning (ALS), as an example, is more developed to assess the structural forest
parameters and niches for forest regeneration (Stupariu et al, 2020) (Pătru-Stupariu et al, 2017). Due to
this method, a 3-D dataset will be extracted that includes data such as canopy height, crown dimension
and vertical structure. Indeed, ALS can aid in mapping the forest canopy height and therefore supplies
information about the stand that addresses many challenges in the terrestrial method thanks to the
visibility of the crown tops from the above.
The main goal of this research is to apply a LiDAR-based RS data to be implemented as the basis data
in the gap’s detection of the forest owing to the lightning. In other words, the study targets innovative
approaches for LiDAR application in forestry, which are limited at the present time.

2. Study area
Targeted site is located by the Shark River which is a major distributary of Harney River in the
southwestern portion of Everglades National Park. It is located in Monroe County, Florida, United
States (Figure 1&2). The river is entirely sea level. The mouth of the river is at Ponce de Leon Bay,
part of the Gulf of Mexico. The river is only 9.5 miles (15.3 km) long (U.S. Geological Survey, 2011).
Figure 3 shows an aerial picture of the site study. The center of study plot’s coordinates is Easting:
491626.77 m E, Northing: 2804956.11 m N, in zone 17R.

Figure 1. Study area.

 Figure2. Satellite image of the study plot Figure3. Aerial view of Shark River,
Produced from Google Earth in 2018 Everglades National Park, 2007-08-14

3. Method
3.1. Creation of LAS dataset
Forest data collection is done by an air-borne LiDAR instrument. This technology has some
brilliant advantages, such as relatively faultless recording of geospatial data and the ability of
penetration into the foliage and thus surveying of the understory. This means, the airborne laser
scanner can record 3D data of the vertical structure. Also, it can provide us with the auxiliary data to
investigate on the canopy cover. After radiation of the laser wave, backscatters from diverse sections
of the tree are registered by the device. The produced 3D point cloud by Lidar can be converted in
various file formats including LAS and XYZ, to make it possible to be imported and used in spatial
software (Figure 4).
Figure 4. LAS data file
 Among them, Esri's ArcGIS Desktop 3D Analyst™ tools have powerful functions to work with
the exported LiDAR data. To manage the data and connect them with the geospatial data for
processing, ArcGIS10.5® software will be employed in this research. Firstly, the 3D point cloud which
illustrates the surrounding terrain is added to the software in the format of LAS file.
After the quality checking, it should be and georeferenced, so it can combine with the other
spatial data in the software environment. The process can be done using ground control points
surveyed by GPS or other georeferenced images. Then is converted to a dataset. This action is done by
“Create LAS dataset” function, which accessible form the data management folder in the toolbox
section. The output file is stored in “LASD” format.

3.2. Creation of DTM, DSM and CHM maps

Double clicking on the exported dataset, and from the layer properties, the Filter tab can be
accessible. At the column namely Classification codes, there are some codes assigned to various
objects by the operator just after the creation of LAS data file. At the center column, returns tab can be
seen. As aforementioned, after radiation of laser wave from the LiDAR, they hit to the different
surfaces and backscatters then are returned to the sensors. Return 1, are the rays radiated from the
nearest surfaces, which in in forest are the leaves. The last return will be the soil surface which is the
furthest object to the sensors. The highest numbers, the furthest object (Figure 5).

Figure 5. LAS filter tab

Figure 5. LAS data file

The next step is creation of Digital Terrain Model (DTM), which is the area surface
topography without considering trees and Digital Surface Model (DSM) that includes the above
ground objects such as trees or buildings. By subtraction of them, the gaps occurred in forest can be
detected. To create DTM, in classification code section, the ground class should be checked, while at
the other section all returns should return checked. Then, the tool “LAS dataset to raster” is used.
What is important in the opened window, is that at the cell assignment type section, elevation should
be selected and at the void fill method section, natural neighbor is preferable (Figure 6).
What is more, DSM map should be composed from the LAS file. At the filter tab (Figure 5),
all classes and return 1 option should be check. Because we need to visualize what is above the
ground, also the nearest object, so we choose return 1, which in this research is leaves of trees. Again,
the tool “LAS dataset to raster” is employed (Figure 7).
 Figure 6. DTM map Figure 7. DSM map

Next, to determine the forest canopy height, the subtraction of the created maps should be
calculated. The result will be Canopy height model (CHM). This function is done by raster calculator
tool in ArcGIS toolbox. The formula is as the following:
CHM = DSM – DTM

3.3. Reclassification and making the binary map

To begin estimation of the pixels represented gaps, 3D analyst extension is employed in
ArcGIS 10.5. Firstly, a linear profile is created using “interpolate line” which creates a 3D line by
interpolating heights from the selected functional surface. Then clicking on “profile graph” button in
the toolbar, a profile graph of the selected 3D line. To detect the value of gap pixels, a line sketched so
that it embraces white and black pixels, and the procedure should be repeated to estimate the value of
gap pixels. It can be seen from a histogram that gaps approximately valued between zero and five
(Figure 8).

Figure 8. Linear Profile Graph

 Therefore, the next step is creating a new binary which zero value represent old values that are
less than five. Raster calculator tool is used again and a conditional query is written in the related
window as the following. It can also be done sing “reclassify” tool.
Con("chm" <= 5,1,0)

3.4. Corrections
After that, some irregular anomalies such as noises should be removed, called Thinning. As an
illustration, in this study plot, rivers and lake should be eliminated. Besides, tiny pixels could be
neglected as noises. To apply necessary corrections, ArcScan extension is used. In ArcGIS 10.5, it is
provided as an extension that can be added from the toolbar section, unlike the previous versions
which it was implemented as an independent software. Since ArcScan is designed to work with
editing, you must start an edit session to activate the toolbar. This means that all the editing tools and
commands can be used in conjunction with the ArcScan tools and commands. ArcScan uses editing
settings, such as the snapping environment and target templates and layers. ArcScan can vectorize any
raster format supported by ArcGIS so long as it is represented as a bi-level image. This requires that
you symbolize raster layers with two unique colors. You can use either the ArcMap Unique Values or
Classified rendering options to separate the raster into two colors. Most scanned documents tend to
consist of two colors that delineate the foreground and background values. Typically, the foreground is
represented as a dark color, such as black, and the background is represented as a light color, such as
white. However, these colors can be reversed or represented by different values. As long as the two
colors possess unique values, ArcScan supports vectorization for the current foreground raster cells.
After adding ArcScan toolbar, the option “select connected cells” is chose from the cell selection
menu. This function selects raster features by querying cell size area or length. This often used to
select noise or holes in raster features so they may be erased during a raster cleanup session. Selected
cells may also be vectorized or saved to a new raster layer. In the opened window, we choose
foreground to be corrected of the connected cells. The value which is occupied the majority of pixels
accounted for background and the other value has to be foreground. It seems that the operator equal or
less than 4 is appropriate for raster pixel for this research. In other words, connected cells with value 1
which are of foreground, and are equal or less than 4, will be selected as noises. From the “raster clean
up” menu, we can erase the selected cells.
There are also some noises into the gaps. This means tiny pixels with value 0 which are inside the
foreground. They should be changed to foreground values. In the aforementioned window, at this time,
we choose background to select these pixels. From “cell selection menu”, “fill selected cells” function
is used to implement these changes (Figure 9).

Figure 9. Linear Profile Graph

Before After
 As the next correction, the pixel corners are smoothed. It can be applied by “closing” function in
raster cleanup section. The appropriate number of pixels to be entered here is 1. This function is a
morphological operation which dilates and immediately erode raster features and is useful for
smoothing rough linear features in the raster and filling in small gaps between raster foreground
objects.
Last but not least, huge phenomena such as rivers that is wrongly is in foreground should be
deleted. From “connected cells” in “cell selection” menu in ArcScan, we targe to select pixels from the
foreground with greater than 600 raster pixels. Then these pixels should be erased as it cited before.
What is remained are pixels represented gaps in the forest site which have the value 1. It can be
then converted to vector polygon based on the user aim.

4. Results
4.1. Created CHM map
The produced Digital Terrain Model (DTM) map which shows the topography of the study plan
is illustrated in Figure 6. In addition, the result Digital Surface Model (DSM) map that unfolds the
surface model is suggested in Figure 7. It can be seen from the processed images that trees have a
brighter color which represent the higher value than the other objects including the earth surface and
water. As aforementioned, the subtraction of the models is calculated which is the Canopy height
model (CHM) (Figure 10). It can be witnessed that two last maps are very similar.

4.2. Created binary map

In the produced binary map the value 1 represents trees and waters and the value 0 represents
other old objects embracing the soil surface. Figure 11 shows the result primary binary map. Green
pixels are trees and black pixels are for others.

Figure 10. CHM map Figure 11. Binary map

Figure 10. CHM map

After the related corrections such as filling and erasing selected connected cells which are
explained at method section, the final map will be composed illustrating the gaps owing to lightning in
the forest site (Figure 12). The green pixels that have the value 1 represents the gaps between trees in
the forest after the lightning.
 Figure 12. Final map

5. Conclusion
To conclude, LiDAR is a promising technology which used for characterization of the forest
vertical structure, tree height, biomass and canopy cover. Thanks to this state-of-the-art method,
researches are able to investigate on the spatio-temporal changes in forest canopy covers owing to
various phenomena such as lightning. LiDAR technology gives much more accurate information for
accurate ecological and land use classification and can be used to map the forests by measuring the
vertical structures of the canopy. To process exported 3D points, some spatial software is adapted to
manage LAS data. In this research, ArcGIS 10.5 is applied to process the LAS data. The processed
information can be used to assess the changes and also predict the future trends. Furthermore, the most
sensitive regions could be unfolded to consider them in the prospective protection programs. Totally, the
main aim of the research is the investigation of LiDAR application in forest management, which is
affirmed in this research.
References
Alkhatib, Ahmad. (2014). A Review on Forest Fire Detection Techniques. Hindawi Publishing Corporation
International Journal of Distributed Sensor Networks Volume 2014, Article ID 597368, 12 pages.
http://dx.doi.org/10.1155/2014/597368.

Bouquegneau, C., & Rakov, V. (2010). How dangerous is lightning? Mineola, NY: Dover Publications.

K. Zhang. (2006). Identification of Lightning Gaps in Mangrove Forests Using Airborne LIDAR Measurements.
American Geophysical Union, Fall Meeting 2006, abstract id. B41A-0175.

Laser technology to protect critical infrastructure from lightning strikes project. (2020). European Commission
website.

Pătru-Stupariu, I., Stupariu, M. S., Stoicescu, I., Peringer, A., Buttler, A., & Fürst, C. (2017). Integrating geo-
biodiversity features in the analysis of landscape patterns. Ecological Indicators, 80, 363-375.

P. Yanoviak, Stephen. M. Gora, Evan. M. Burchfield, Jeffrey. M.Bitzer. Phillip. And Detto, Matteo. (2017).
Willey Ecology and Evolution. DOI: 10.1002/ece3.3095. April 2017.

Rakov, V. A., & Uman, M. A. (2003). Lightning: Physics and effects. Cambridge: Cambridge University Press.

Stone, G. E. (1903). Injuries to shade trees from electricity. Hatch Experiment Station of the Massachusetts
Agricultural College, 91, 1–21.

Stupariu, M. S., Pleșoianu, A. I., Pătru-Stupariu, I., & Fürst, C. (2020). A Method for Tree Detection Based on
Similarity with Geometric Shapes of 3D Geospatial Data. ISPRS International Journal of Geo-Information,
9(5), 298.

Taylor, A. R. (1974). Ecological aspects of lightning in forests. Proceedings of the Tall Timbers Fire Ecology
Conference, 13, 455–482.

U.S. Geological Survey. National Hydrography Dataset high-resolution flowline data. The National Map,
accessed April 18, 2011.