Integration of LiDAR and Pleiades Data in Identifying the Potential of Hazardous Trees

Integration of LiDAR and Pleiades Data in Identifying the Potential of Hazardous Trees

Suhaizad, L. S., Khalid, N.* and Tahar, K. N

Centre of Studies for Surveying Science and Geomatics, Faculty of Architecture, Planning and Surveying, Universiti Teknologi MARA, 40450 Shah Alam, Selangor, Malaysia

E-mail:nafisahkhalid@uitm.edu.my

*Corresponding Author

Abstract

Trees beautify the landscape and provide numerous benefits for human well-being. Without proper maintenance and mitigating, the hazardous trees with structural defects can be harmful to the local community. In practice, the hazardous trees can be assessed using visual tree assessment (VTA), expertise interview and questionnaire. The advancement in remote sensing technology provides users with alternative ways to identify the hazardous trees and is very efficient especially in large areas. The intent of this study is to identify the potential of the tree hazard in UiTM Shah Alam from the integration of LiDAR and Pleiades dataset. Tree health condition and tree height were derived from LiDAR and Pleiades dataset. Height of trees were extracted from a digital canopy height model (DCHM) and the condition of tree health were assessed based on the Normalized Difference Vegetation Index (NDVI) values. Both parameters were used to classify the individual into low, moderate and high-risk ratings. The findings demonstrated that the integration of both datasets successfully identified the potential of tree hazard at the study area. Altogether 108 of individual trees were detected. 7 and 67 trees were identified as high and moderate risk respectively. This study derived significant findings that can be used to support the maintenance of tree activities in the campus area and significant to private and government sectors such park management and local authorities.

1. Introduction

A tree is taken into account as hazardous if its defects will cause a failure and hit something as it falls. Not all tree hazards are obviously visible until an extreme weather event exposes their weakness. In other words, in order for a tree to be thought of as hazardous, it should have a structural weakness (Cahoon, 2020). Cracks, weak branch unions, branch decay, root damage, and root disease are common indicators of structural weakness that should be checked or observed on a constant schedule (National Tree Safety Group, 2011). Moreover, multiple and connected defects increase the potential of failure. Trees with unbalanced or asymmetrical crowns may have their weight distributed incorrectly throughout the stem. These trees are vulnerable to failure when combined with other flaws such as decay and root disease (Smiley, 2008).

Hazard-rating assessment is the evaluation of the individual tree that has the potential to fall and can cause damage to the surrounding area, especially during extreme weather events. The main objective of tree risk assessments is to identify defective trees in high-risk areas, assess the severity of the defects, and recommend corrective actions before the trees fail. Tree-risk assessments can help communities quantify the level of risk to public safety and prioritize implementation of corrective actions. The potential of tree hazard can be classified into three levels of hazard rating which are low, moderate and high (Chuon et al., 2011).

Based on the previous studies, there are some methods used in identifying hazardous trees, especially in the urban areas. According to Sreetheran et al., (2011), a form was used to record information such as tree species, tree structure, tree health, and hazard status for tree risk assessment and the street tree inventory. The tree height was measured with a hypsometer, and the crown diameter and diameter at breast height (DBH) were measured using tape. The hazardous trees also can be identified using the Delphi method (Maruthaveeran and Yaman, 2010). This is a procedure for obtaining the most credible consensus of opinion from a group of experts.

The process of visual tree assessment is an establish method in identifying the hazardous trees however it will be challenging for a large area. Therefore, advances in mapping technology are promising for providing tree inventories in a timely, flexible, accurate, and cost-effective manner (Ma et al., 2013). LiDAR is a remote sensing technology that can be used to extract comprehensive details of 3D structures. They can penetrate particularly dense canopies of vegetation and can be used to gain information on the 3D structure of trees and for trees that have been suppressed (Zhang et al., 2015).

Other than Pleiades image, there are past studies on SPOT-5 which have resolution of 10m with near infrared, red and green. 20m resolution of shortwave infrared band and with panchromatic 2.5m resolution. The study by (Meng et al., 2016) was mainly for mapping of forest health using remotely sensed image data. It states that this study is conducted because there is difficulty in assessing or monitoring individual tree health by using field observation. With the improvement of technology, remote sensing is the best way to conduct or to practice for this mapping activity.

2. Methodology

The overall methodology consists of 4 phases which is data acquisition, data processing and data analysis and map production as shown in Figure 1.

Figure 1 Flow of methodology

Data acquisition stage consists of LiDAR and Pleiades datasets. Data processing was divided into two phases, the first stage was for LiDAR data processing that involved rasterizing LiDAR data for delineating the individual tree polygons and the individual tree height were derived from digital canopy height model (DCHM). Both of the outputs are overlaid using Zonal Statistics to extract the tree height. Then, the condition of individual trees was derived from NDVI values. Derived parameters were overlaid to calculate the hazard rating of the individual trees. The map of the distribution of potential hazardous trees was produced.

2.1 Tree Height Estimation from LiDAR Data

The LiDAR data used for this study was obtained from Pejabat Pembangunan Infrastruktur & Infostruktur (PPII). The LiDAR data covered the study area of UiTM Shah Alam. Data obtained are in point cloud data and processed in ArcGIS software. The digital canopy height model (DCHM) was generated from the subtraction of Digital elevation model (DEM) from digital surface model (DSM). Individual tree heights were estimated from DCHM. Individual trees are delineated by rasterized the point cloud data and converted the raster data to polygon.

Zonal statistics is used to extract tree height. The tools extract summary statistics of the raster cells that fall within the polygon. In this study, maximum values were chosen since the highest tree tip or tree crown are the tree height. The polygon of trees is then converted into points to represent the tree locations.

2.2 Pleiades Data Processing

The Pleiades imagery used in this study was obtained from Malaysian Space Agency (MySA). This imagery consists of 4 bands which are blue, green, red and near infrared (NIR). In this study, Pleiades was used to assess the condition of the trees via NDVI values. The NDVI values are within -1 to 1 where 1 show tree with great condition. Raster calculator tool in ArcGIS was used to perform NDVI threshold. The health of the trees was classified using reclassify tools into four classes which are dead or inanimate range from -1 to 0, unhealthy trees range from 0 to 0.33, moderately healthy trees range from 0.33 to 0.66 and very healthy trees range from 0.66 to 1.

2.3 Identifying the Hazardous Tree

The hazardous trees were classified based on the hazard rating which mainly focus on the tree health and tree height as shown in Figure 2. The hazard rating was derived from the addition of tree height and tree health using add tools. According to USDA (2011), the classes of hazard rating are classified into three classes.

Figure 2 Tree hazard rating

Table 1 Results of DEM, DSM and CHM