Evacuation Route Selection Based on Tree-Based Hazards Using Light Detection and Ranging and GIS

TL;DR: A novel methodology for automating the tree threat identification process by using airborne laser altimetry data and a geographical information system (GIS) has the potential to be used for selecting the best possible evacuation routes based on tree hazards.

Abstract: Falling trees pose a great hazard to the safe and uninterrupted use of the road transportation system during storm events. The present process of manually identifying potentially hazardous trees is laborious and inefficient. This paper presents a novel methodology for automating the tree threat identification process by using airborne laser altimetry data and a geographical information system (GIS). This methodology has the potential to be used for selecting the best possible evacuation routes based on tree hazards. The proposed method harnesses the power of spatial analysis functionality provided by existing GIS software and high-quality, three-dimensional (3D) data obtained from an airborne laser scanning system. This paper highlights the benefits related to using: (1) Height calculation of tall objects; (2) identification of hazardous objects; and (3) object identification from irregular 3D light detection and ranging point data over the currently employed manual methods.

Summary

INTRODUCTION

• Wooded areas flank the primary and secondary road networks in many states.
• Besides disrupting traffic flow, such incidents endanger the lives of drivers and emergency crews, hinder service restoration, and impede effect ive emergency response.
• To avoid this, trees that may be a threat should be identified, and measures should be considered to prevent them from falling on overhead lines and thoroughfares.
• A manual approach to such a problem is not viable, because of the time and resources necessary for the init ial assessment and the need for periodic updates.
• In formation System (GIS) and Remote Sensing (RS), an automated hazard identification tool can be developed for the aforeme ntioned problem that is more economical, efficient, reliable, and safer than a manual approach.

BACKGROUND

• Hurricanes, other strong winds, and ice storms cause tremendous damage across the United States (US) including flooding, wind-based destruction of property, and blockage of major road networks from fallen objects.
• The blockage of roads due to fallen trees and utility lines during severe weather is a significant problem (Draper 2003).
• The problem of tree-based hazards is well recognized, and each year millions of dollars are spent annually to identify and cut or trim such trees (Act 2003).
• In most states, identificat ion of potentially hazardous tall objects has, to date, been done manually by road crews on an ad hoc basis.
• Light Detection And Ranging , a recent advancement in RS technologies, offers the potential of automated and expeditious data collection , and GIS provides a means to analyze and process such data.

LiDAR and Its Characteristics

• LiDAR is an active remote sensing technology that is used to collect topographic data (NOAA 2003).
• The difference in time is measured from when a laser pulse is emitted from a sensor to when the target objects in the path of the laser reflect back the pulse.
• The end product is accurate, geographically reg istered longitude, latitude, and elevation from the mean sea level (x, y, z) positions for every data point (Baltsavias 1999).
• As it does not provide feature informat ion, clever and efficient algorithms are required for feature identification depending upon the desired application.

GIS and Its Characteristics

• GIS can be defined as a computer-based tool set for collecting, storing, retriev ing, transforming, and displaying spatial data from a discipline-specific domain for a particular set of purposes (Burrough and McDonnell 1998).
• The true strength of GIS lies not its ability to store and process data but in its spatial representation and spatial analysis (Rasdorf et al. 2000).
• Spatial analysis involves examin ing the geographic patterns in data and observing relationships between geographical features.
• GIS-T refers to the principles and applications of applying geographic information technologies to transportation problems (Miller and Shaw 2001).
• For road safety hazard management, GIS can he lp to identify hazards posed by specific man-made or natural objects.

EXISTING HAZARDOUS TREE IDENTIFICATION METHODS

• Because of budget restrictions, across the US there is no generalized, method to identify hazardous trees, despite the acknowledged expenses that they generate.
• These operators are also asked to look for roadway hazards and to report them to the Highway Maintenance Supervisor.
• A survey of 20% of the DOTs, focusing mainly on states with large tree growth and severe weather problems, established that most states were highly reactive in their approach and did not have the resources to do more than respond to calls from the public o r from maintenance workers, who happened to notice a potential hazard (Table 1).
• Every 3 years, a computer program is used to randomly generate 3 segments (each 1/16 km in length) for a given section of highway, if it is less than 16.1 km long, in terms of its paving type and general characteristics.
• An alternate approach to casual visual observation to manually identify the heights of the tree is to use a hypsometer.

PROPOSED METHODOLOGY

• The research described herein adopts a unique technique to identify and assess potentially hazardous tall objects: (a) buffer calculation, (b) elevation characterization, (c) shortest distance determination, and (d) feature identificat ion.
• To achieve these four tasks, the proposed method employs a LiDAR dataset, a centerline GIS data set of the road network, and multi-spectral imagery.

Study Area, Data Characteristics and Software used

• The area was selected primarily due to the availability of all the required data for the project: LiDAR data, centerline road data, and mult ispectral aerial imagery.
• The bare -earth data are required to have a vertical RMSE of 20-cm for coastal counties and 25- cm for inland counties, computed after discarding the worst 5% of the checkpoints to account for un -cleaned artifacts (NCFPM 2003).
• The centerline road data were collected and produced by NCDOT.
• The projection is NAD83, North Caro lina State Plane, and the original data units are in feet.
• The fo llowing provided functions were employed: Buffer Pointdistance Table Query ing and Manipulation Identify.

Buffer Calculation

• Airborne LiDAR systems are capable of producing extremely detailed informat ion of even small scanned areas (Baltsavias 1999), but to achieve a high level of detail requires substantial processing time.
• An effective measure to counter such resource intensiveness, while not losing the desired level of detail, is to reduce the dataset size.
• In order to select the relevant data points, a buffer zone can be created around the road network; should fringe lane data be available, instead of centerline data, an even more accurate evaluation is possible.
• Only points that lie within the defined buffer zone are selected.
• The criterion for a 120m buffer zone was based on the assumption that the maximum height of any tree is 60m; this parameter is easily changed to reflect local vegetative growth patterns.

Comparative Elevation Calculation

• Once a buffer zone is established, the LiDAR points within the buffer zone need to be analyzed.
• One dataset represents a reflective surface elevation model [also known as Digital Surface Model (DSM)], which contains elevation data from the ground, as well as from vegetation, tree canopy, and other features such as buildings.
• The other dataset represents “bare earth” ground points, which excludes additions to the terrain.
• Height informat ion above bare ground surface is important to determine the potential hazard posed by any tall object within the buffer zone.
• The height column is the elevation difference for LiDAR points 0001-0004, whereas the distance column reflects the respective shortest distance value to their nearest road segments.

Shortest Distance Calculation

• After the height calculation is completed, the shortest distances between the LiDAR points (within the buffer zone) and their adjoining road are calculated by subtracting half the road distance from the centerline position location(Table 2).
• GIS based applications can determine the shortest distance between a point feature and a line or polygon feature.
• ArcInfo TM from ESRI is used to perform the shortest-distance calculations.
• The distance between the LiDAR point and its adjoining road centerline is stored as attribute data for the corresponding LiDAR point (Table 2).

Vulnerability Analysis

• The objective of the vulnerability analysis determines those LiDAR points that represent potentially hazardous trees.
• If the elevation of any LiDAR point represents a height that is greater than its calculated, shortest distance to the adjacent road, the object representing the corresponding point may pose a threat to the road, should the tall object fall perpendicular to the road.
• Li, then object = hazard Table 3 shows that in this data subset only points A and D are at elevation values greater than their shortest distance values.
• The analysis is relatively complicated as it would have to be repeatedly done for every section, where a tree is present.
• Thus, the relatively simple approach presented here can be considered more economical.

Risk Assessment Prioritization

• Hazard identification is the first step to risk assessment prioritization.
• The percentage of the road pavement that might be impacted should actually be considered the sweep over which the tree may fa ll (eqn 4) [Figure 5b] subtracted by the non-pavement portion (eqn 4) to assess the full area of vulnerable pavement (eqn 5).
• The traffic volume could be considered, and whe ther or not alternate routes exist could be incorporated as part of the evaluation process.
• In order to identify the shape and dimension of objects, an algorithm is required.
• The approach uses buffering to group LiDAR points belonging to a hazard area.

RESULTS

• In December of 2003, two inspectors were sent to the study area to conduct visual hazard identificat ion assessments.
• Because of tree density, cluster identification was not possible.
• A reas with 1 or 2 trees were listed as hazardous as those with 6 or 8; further system refinement could include algorithms to identify how many hazardous items exist along a specified road length.
• As such procedures become routine and the cost of technology decreases, this automated approach may eventua lly serve as a cost effective method by itself.
• By integrating growth trends for local vegetation, the values for H i can be automatically augmented annually, thereby precluding the need for frequent flyovers for data collection.

CONCLUSIONS

• An automated tree-hazard pred iction methodology that utilizes GIS and high quality remotely sensed data offers substantial advantages over existing manual process in terms of safety, speed, and efficiency.
• This method avoids the extensive use of labor in surveying long lengths of highways.
• Since the identification process is automated and computationally efficient, the process is many times faster than manual approach and can be achieved in a systematic and automated fashion.
• Thus, the deployed method demonstrates the capability of 3D LiDAR data combined with GIS-based software to identify potentially hazardous tall objects adjacent to major roads.
• Such a system can render substantial help to optimize evacuation route selection.

Figures

TABLE 1 . Tree Identification Methods Employed by Various DOTs

TABLE 2 . Attribute Table of LiDAR Points for Height and Distance Identi fication

FIG. 6. Proposed Buffer to Account for Tree Crowns.

FIG. 2. Algorithm Flowchart.

FIG. 1. Study Area (Figure to Scale).

FIG. 5a. Measurement of Tree across the Road. FIG. 5b. Impact of the Downed Tree (Plan View).

FIG. 7. Hazard Posing Tree with Defined Crown Diameter and Background Removed.

FIG. 4. LiDAR Points and Road Segments with 120 m Buffer from Road Centerline.

TABLE 3 . Vulnerability Analysis

FIG. 3. LiDAR Points and Road Segments without Buffer Zone.

Full text

Provided by the author(s) and University College Dublin Library in accordance with publisher
policies. Please cite the published version when available.
Title Evacuation route selection based on tree-based hazards using LiDar and GIS
Authors(s) Laefer, Debra F.; Pradhan, Anu
Publication date 2006-04
Publication information Journal of Transportation Engineering, 132 (4): 312-320, Technical papers
Publisher American Society of Civil Engineering (ASCE)
Link to online version http://dx.doi.org/10.1061/(ASCE)0733-947X(2006)132:4(312)
Item record/more information http://hdl.handle.net/10197/2307
Publisher's version (DOI) 10.1061/(ASCE)0733-947X(2006)132:4(312)
Downloaded 2022-08-10T07:51:54Z
The UCD community has made this article openly available. Please share how this access
benefits you. Your story matters! (@ucd_oa)
© Some rights reserved. For more information, please see the item record link above.

1
Evacuation Route Selection Based on Tree-Based
Hazards Using LiDAR and GIS
Debra F. Laefer,
1
M. ASCE and Anu R. Pradhan
2
ABSTRACT
Falling trees pose a great hazard to the safe and uninterrupted use of the road transportation system, during storm
events. The present process of manually identifying potentially hazardous trees is laborious and inefficient. This
paper presents a novel methodology for automating the tree threat identification process by using airborne laser
altimetry data and a Geographical Information System (GIS). This methodology has the potential to be used for
selecting the best possible evacuation routes based on tree hazards. The proposed method harnesses the power of
spatial analysis functionality provided by existing GIS software and high-quality, three-dimensional (3D) data
obtained from an airborne laser scanning system. This paper highlights the benefits related to using (a) height
calculation of tall objects, (b) identification of hazardous objects, and (c) object identification from irregular 3D
Light Detection And Ranging (LiDAR) point data over the currently employed manual methods.
KEYWORDS: Geographic Information Systems, Roadside Hazards, Highway Maintenance, Trees,
Transportation Management, Highway Safety, Lasers, Evacuation
1
Lecturer, Department of Civil Engineering, University College Dublin, Earlsfort Terrace, Room 115, Dublin 2, Ireland. Email:
debra.laefer@ucd.ie
2
Graduate Research Assistant, Department of Civil and Environmental Engineering, PH 119, Carnegie Mellon University, Pittsburgh, PA. Email:
arpradha@andrew.cmu.edu

2
INTRODUCTION
Wooded areas flank the primary and secondary road networks in many states. During severe weather incidents,
fallen trees knock down power lines, endanger telephone lines, and block roadways. Besides disrupting traffic flow,
such incidents endanger the lives of drivers and emergency crews, hinder service restoration, and impede effective
emergency response. To avoid this, trees that may be a threat should be identified, and measures should be
considered to prevent them from falling on overhead lines and thoroughfares. Before preventive measures can be
considered, however, these trees must be identified. A manual approach to such a problem is not viable, because of
the time and resources necessary for the initial assessment and the need for periodic updates. New technologies offer
the possibility of automating such an identification process. With advances in Geographical Information System
(GIS) and Remote Sensing (RS), an automated hazard identification tool can be developed for the aforementioned
problem that is more economical, efficient, reliable, and safer than a manual approach. The ideal application of such
an evaluation is in the selection of the most reliable evacuation routes. The goal of this research is to describe the
development of such algorithms and the resources necessary to automate identification of trees that may endanger
transportation routes.
BACKGROUND
Hurricanes, other strong winds, and ice storms cause tremendous damage across the United States (US) including
flooding, wind-based destruction of property, and blockage of major road networks from fallen objects. The direct
cost of such incidents can be in the billions of dollars, as was the case with hurricane Floyd in 1999 in eastern North
Carolina ($5.45 billion in damage costs) [Herbert et al. 1997], as well as indirect costs for economic losses due to
closed businesses and lost productivity. The blockage of roads due to fallen trees and utility lines during severe
weather is a significant problem (Draper 2003). According to the North Carolina Division of Emergency
Management, after hurricane Isabel, Bertie County alone required tree debris removal from all roads in the county
except three, totaling 780 km of road from which 52,865 cubic meter or 43,245.27 kg of debris was removed, at a
cost of $1.6 million, over the course of 84 days (Canty 2004). Beyond economics, posing a great risk to motorists,
and interfering with utility service, roadway obstacles also disrupt traffic flow, and thus hamper evacuation and
rescue operations.

3
The problem of tree-based hazards is well recognized, and each year millions of dollars are spent annually to
identify and cut or trim such trees (Act 2003). In most states, identification of potentially hazardous tall objects has,
to date, been done manually by road crews on an ad hoc basis. The task is significant. In North Carolina (NC) alone,
there are 125,580 km of state maintained roads, not including 34,680 km of county and city roads (NC-DOT 2003)
covering an area of 136,523 km
2
(State Library of NC 2004). Across the US, the national interstate system alone is
comprised of 75,150 km of highway (FHWA 2003). As such, a manual approach to hazardous tree identification is
cost prohibitive as well as time-consuming, especially since the effort would require periodic updates because of tree
growth and road expansion. Thus, an automated process that is fast and potentially cost-effective, during initial
identification and later updating, is critically needed.
At the most elementary level, for an automated tool to be effective it should be able to (a) identify the roads of
interest, (b) locate tall objects, (c) calculate the heights of potentially hazardous objects along the roads, and (d)
compare the height of the object to its distance from the roadway. The success of such an automated tool depends on
both the quality of available data and a robust application that analyzes and processes the given data. As highly
detailed data must be collected along the entire length of road network, data gathering becomes the most time-
consuming and expensive part of the process. Light Detection And Ranging (LiDAR), a recent advancement in RS
technologies, offers the potential of automated and expeditious data collection, and GIS provides a means to analyze
and process such data.
LiDAR and Its Characteristics
LiDAR is an active remote sensing technology that is used to collect topographic data (NOAA 2003). The data are
collected with aircraft-mounted lasers capable of recording elevation measurements at a rate of 5,000 to 50,000
pulses per second. The difference in time is measured from when a laser pulse is emitted from a sensor to when the
target objects in the path of the laser reflect back the pulse. Using the speed of light, these time measurements can be
converted into distance or range (Lim et al. 2001). The LiDAR instruments collect elevation data. To make these
data spatially relevant, the positions of the data points must be known. Thus, a high-precision global positioning
system (GPS) antenna, mounted on the aircraft, is used to determine the spatial positions of the data points. The end

4
product is accurate, geographically registered longitude, latitude, and elevation from the mean sea level (x, y, z)
positions for every data point (Baltsavias 1999). Latitude, longitude, and elevation are typically presented in a plane
co-ordinate system.
LiDAR is capable of providing both horizontal and vertical information at high spatial resolutions and vertical
accuracies. Airborne-based LiDAR data are accurate to +/-15 cm for vertical measurements and +/- 1.5 m (worst
case scenario) for horizontal distances (Flood 1999), although the system is marketed as having an accuracy of an
order of magnitude better in both directions (ALTM 2003). The extent of LiDAR point density is dependent on
flying height and system dependent factors such as platform velocity, sampling frequency, and field of view
(Axelsson 1999). The point density needs to be adjusted according to the application so that sufficient information is
harvested, while not collecting excessively detailed data. LiDAR technology has been used in many areas of
applications such as (a) generation for a variety of GIS/mapping related products, (b) forestry, (c) coastal
engineering, (d) flood plain mapping, (e) disaster response and damage assessment, and (f) urban modeling
(Airbornelasermapping 2003).
A current limitation of LiDAR is that it does not store any topological, shape, or size information of the
geographical features scanned. A LiDAR dataset is simply a collection of somewhat randomly distributed 3D points.
As it does not provide feature information, clever and efficient algorithms are required for feature identification
depending upon the desired application. Identification of tree hazards requires such an algorithm. LiDAR provides
only qualitative data, thus software capable of analysis is also required. GIS possesses such capabilities.
GIS and Its Characteristics
GIS can be defined as a computer-based tool set for collecting, storing, retrieving, transforming, and displaying
spatial data from a discipline-specific domain for a particular set of purposes (Burrough and McDonnell 1998).
Although computer aided design (CAD) software can perform similar functions, the true strength of GIS lies not its
ability to store and process data but in its spatial representation and spatial analysis (Rasdorf et al. 2000). Spatial
analysis involves examining the geographic patterns in data and observing relationships between geographical

Frequently asked questions

Q1. What are the contributions mentioned in the paper "Evacuation route selection based on tree-based hazards using lidar and gis" ?

This paper presents a novel methodology for automating the tree threat identification process by using airborne laser altimetry data and a Geographical Informat ion System ( GIS ). The proposed method harnesses the power of spatial analysis functionality provided by existing GIS software and high-quality, three-dimensional ( 3D ) data obtained from an airborne laser scanning system. This paper highlights the benefits related to using ( a ) height calculation of tall objects, ( b ) identification of hazardous objects, and ( c ) object identification from irregular 3D Light Detection And Ranging ( LiDAR ) point data over the currently employed manual methods. Th is methodology has the potential to be used for selecting the best possible evacuation routes based on tree hazards. 

Q2. What causes the blockage of major roads?

other strong winds, and ice storms cause tremendous damage across the United States (US) including flooding, wind-based destruction of property, and blockage of major road networks from fallen objects. 

Q3. What is the definition of a computer-based tool set for collecting, storing, retriev?

GIS can be defined as a computer-based tool set for collecting, storing, retriev ing, transforming, and displaying spatial data from a discipline-specific domain for a particular set of purposes (Burrough and McDonnell 1998). 

Q4. What is the way to achieve the high level of detail?

Airborne LiDAR systems are capable of producing extremely detailed informat ion of even small scanned areas (Baltsavias 1999), but to achieve a high level of detail requires substantial processing time. 

Q5. What is the cost of tree debris removal in North Carolina?

According to the North Carolina Division of Emergency Management, after hurricane Isabel, Bert ie County alone required tree debris removal from all roads in the county except three, totaling 780 km of road from which 52,865 cubic meter or 43,245.27 kg of debris was removed, at a cost of $1.6 million, over the course of 84 days (Canty 2004). 

Q6. What is the main objective of the aforementioned algorithms?

The main objective of all of the aforementioned algorithms is to extract features, especially those related to geometry, from the LiDAR points. 

Q7. What is the way to reduce the size of the dataset?

An effective measure to counter such resource intensiveness, while not losing the desired level of detail, is to reduce the dataset size. 

Q8. How is the distance between the LiDAR points and their adjoining road calculated?

After the height calculation is completed, the shortest distances between the LiDAR points (within the buffer zone) and their adjoining road are calculated by subtracting half the road distance from the centerline position location(Table 2). 

Q9. Why is there no generalized method to identify hazardous trees?

Because of budget restrictions, across the US there is no generalized, method to identify hazardous trees, despite the acknowledged expenses that they generate. 

Q10. How many data points were reduced in this example?

In this simple example, the number of data points was reduced by 83%, which decreased computation time by 67% compared to the time needed to process the non-buffered data.