UAV-LiDAR dataset of Pangandaran coastal tourism hotspots for tsunami and climate risk valuation and exposure mapping

Abstract

This article presents a high-resolution UAV–LiDAR dataset acquired over the main coastal tourism hotspots of Pangandaran, West Java, Indonesia (WGS84 / UTM Zone 49S). The survey was conducted using a DJI Matrice 300 RTK equipped with a CHCNAV AA450 LiDAR system at altitudes of 77–83 m AGL, following grid-based flight lines with 80% forward and 70% side overlap. The final point cloud, delivered in LAS format, exhibits a mean density of approximately 865 pts/m², with dominant values of 600–800 pts/m² across roads, roofs, and open terrain, and localized peaks exceeding 3,000 pts/m² in areas of flight-line overlap. Ground control was established using three static base stations, with 14 calibration control points and 8 independent validation check points. Accuracy assessment yields RMSE values of 0.072 m (Easting), 0.062 m (Northing), and 0.138 m (Elevation), with corresponding mean biases of 0.017 m, 0.017 m, and 0.044 m, confirming centimeter-level positional precision suitable for detailed coastal mapping. The dataset includes DSM and DTM derivatives, block-based tiles, metadata, and processing reports, supporting its use in tsunami exposure assessment, climate-risk valuation, urban coastal planning, and remote-sensing education. As one of the first openly accessible UAV–LiDAR datasets for an Indonesian coastal tourism hotspot, it provides a reproducible, high-density 3D resource for research, hazard analysis, and sustainable coastal development.

Specifications Table

Subject	Earth and Environmental Sciences – Geographical Information Systems; Earth-Surface Processes;
Specific subject area	High-resolution UAV–LiDAR point cloud dataset for coastal exposure mapping, tsunami hazard, and climate risk valuation
Type of data	LiDAR point clouds (.las); Camera Orbit POS (.txt); Ground Control Point data (.txt); POS Report (.pdf); Georeferenced image of point clouds overview (.tif); Shapefiles of AOI (.shp); Digital Surface Model (.tif); Digital Terrain Model (.tif);
Data collection	UAV platform: DJI Matrice 300 RTK; LiDAR sensor: CHCNAV AA450; Flight altitude: 77–83 m AGL; Overlap: 80% forward, 70% side; Horizontal mapping speed: 6.5 m/s; Ground control points (GCPs) with Leica GS18T RTK GNSS: 3 static base stations, 14 control points for calibration, 8 independent check points for validation; Processing and classification conducted using CHCNAV CoPre 2.7.7; Outputs georeferenced in WGS84 / UTM Zone 49S (EPSG:32749);
Data source location	Institution: Universitas Padjadjaran City/Town/Region: Pangandaran Coast, West Java, Indonesia Coordinate system: WGS84 / UTM Zone 49S Geographic extent (approx.): 7.691772°S – 7.705765°S, 108.649214°E – 108.665776°E
Data accessibility	Repository Name: Zenodo Direct URL to data: 10.5281/zenodo.17073404 DOI: 10.5281/zenodo.17073404
Related research article	None

1. Value of the Data

• •This dataset provides one of the first openly accessible UAV–LiDAR point cloud collections for a coastal tourism hotspot in Indonesia. It offers high-density 3D information that is typically unavailable for Pangandaran and other southern Java coastal regions. The dataset delivers a mean point density of approximately 865 pts/m², with dominant densities of 600–800 pts/m² across buildings, roads, and open terrain, and localized peaks exceeding 3000 pts/m² in areas where overlapping flight lines occur.
• •The dataset includes classified LAS point clouds, DSM and DTM products, processing reports, GCPs metadata, and georeferenced visualization raster, allowing users to directly use or further process the data for geomorphology, hazard modelling, and precision mapping. The availability of uncompressed LAS files preserves the exact classification flags, return information, and metadata generated by the LiDAR system, ensuring full reproducibility and scientific integrity.
• •This dataset is particularly valuable for tsunami exposure assessment, coastal infrastructure analysis, climate risk valuation, and evacuation planning, since it provides centimeter-level positional precision (RMSE = 0.072 m E, 0.062 m N, 0.138 m Z) derived from a well-distributed ground control network of 3 static base stations, 14 calibration control points, and 8 independent validation check points.
• •The dataset supports multiple research communities, including coastal engineers, risk modelers, geospatial analysts, and urban planners, enabling applications such as flood modelling, shoreline change analysis, asset-level risk assessment, and digital twin development for coastal cities.
• •The dataset is also valuable for education and training in remote sensing, photogrammetry, geomatics, and disaster risk reduction. Students and early-career researchers can practice working with a high-density, fully classified UAV–LiDAR dataset—including ground filtering, DSM/DTM derivation, accuracy assessment, and visualization—without the high cost typically associated with LiDAR acquisition [1].

2. Background

Coastal regions in Indonesia face increasing exposure to natural hazards, particularly sea-level rise and tsunamis, while also serving as key tourism destinations. Pangandaran, located on the southern coast of West Java, is one of the most visited beach areas and is recognised as highly vulnerable to tsunami events [2,3]. Despite this, high-resolution three-dimensional spatial data capturing the built environment and terrain in such coastal hotspots remains scarce. Traditional topographic datasets, such as national DEM products, provide regional coverage but lack the spatial resolution needed for high-resolution analysis of urban infrastructure, evacuation routes, and tourism facilities.

Recent advances in unmanned aerial vehicle (UAV) technologies and LiDAR sensors have enabled the capture of dense, accurate, and classified point clouds over complex urban and coastal environments [4,5]. The methodological framework applied in this dataset builds upon UAV-based remote sensing approaches for disaster risk mapping and urban coastal monitoring, integrating RTK/PPK positioning and ground control points to achieve centimetre-level accuracy. The survey was conducted on 24 June 2025 in Pangandaran, West Java, to provide openly accessible LiDAR data that can be reused in hazard, planning, and educational contexts.

3. Data Description

The dataset consists of high-resolution UAV–LiDAR point clouds and supporting metadata collected in Pangandaran, West Java, Indonesia, on 24 June 2025. All files are organized into three groups, each compressed in .rar format: pointclouds_X-X, metadata, and ancillary. These groups contain the primary LiDAR point clouds, supporting positional metadata, and supplementary visualization and vector data, ensuring complete documentation and the dataset's usability.

The first group, pointclouds_X-X.rar, contains the classified LiDAR point cloud files in .las format. The dataset is provided in LAS format rather than LAZ to preserve the original, uncompressed point-record structure generated by the CHCNAV processing workflow. Using LAS ensures full fidelity of classification flags, return information, and georeferencing metadata, supporting reproducibility and interoperability across LiDAR processing platforms. The entire survey area was partitioned into blocks measuring 260 × 260 m for optimal management of large datasets while maintaining accessibility and spatial consistency. Each compressed file corresponds to a specific block index (X-X), see Table 1 and as illustrated in Fig 1. In addition to the primary block coverage, each file contains an attached buffer area extending 10% beyond the block boundary to allow overlay between adjacent blocks and ensure seamless full coverage across the survey area. File sizes range from 111 MB to 1.7 GB, with the total size of the point cloud dataset approximately 16.6 GB.

Table 1.

UAV–LiDAR dataset description.

Project ID	Project Name	Collect time (min)	Flight altitude (m)
1	@@2025-06-24-002042	44.51	77.47
2	@@2025-06-24-031257	83.87	83.04
3	@@2025-06-24-052044	40.70	80.07

Fig 1.

Mapping result: block of project mapping division (top-left), point density within mapped area (top-right), surface elevation (bottom-left), point cloud in RGB with GCPs (bottom-right).

The second group, metadata, contains flight and positional information files generated during acquisition and processing. Two file types are included. The first, Camera_OrbitPos_X.txt, records the image position data captured by the onboard CHCNAV AA450 camera sensor, with the suffix X denoting the Project ID. These files indirectly represent sensor coverage and UAV trajectory information. The second, PosReport_X.pdf, summarizes the positional data for each Project ID in report format, providing an overview of the GNSS and trajectory solutions used during processing. Finally, the group also includes a Ground Control Point (GCP) data file (.txt) listing the surveyed coordinates of the ground control network used for calibration and accuracy validation. Together, these metadata files document the conversion of raw survey data into LAS format and serve as essential references for subsequent mapping and geospatial software processing.

The distribution of point density across the survey area, calculated using a 1 m² grid, is summarized in Fig 2. The histogram shows a multimodal structure with a dominant concentration of cells in the 600–800 pts/m² range, corresponding primarily to solid surfaces such as building roofs, roads, and open urban terrain where laser returns are strong and consistent. Lower-density bins (<100 pts/m²) occur mainly over water bodies and at block boundaries, where LiDAR pulses fail to generate reliable returns. Intermediate densities between 100–500 pts/m² occur in mixed contexts, including partial-occlusion zones and sparsely vegetated areas. Very high densities above 3,000 pts/m² are present only in localized regions where overlapping flight paths produce stacked multiple returns, typical of the non-repetitive Livox Avia scanner. Overall, the histogram demonstrates that the dataset provides exceptionally dense sampling, averaging 865 pts/m², with predictable variations tied to surface reflectance, canopy structure, and flight-line geometry.

Fig 2.

Histogram of point density (pts/m²) across the survey area derived from a 1 m² grid.

The third group, ancillary, contains additional materials to support data visualization and interpretation. This directory is subdivided into Image, Shapefile, and Screenshot folders. The Image folder includes georeferenced overview images of the LiDAR point clouds in .tif format, with corresponding support files. These raster products visualize the point clouds using different attributes such as classification, elevation, point density, RGB values, and return numbers (first, second, third). In addition to these point-cloud visualizations, this group includes example DSM and DTM rasters that serve as reference illustrations of the elevation products derived from the classified dataset. Representative examples of these images are shown in Fig 3. The Shapefile folder provides supplementary vector data in standard .shp format with associated support files, including block boundaries and orbit position distributions (see Fig 1). These shapefiles complement the metadata by documenting the spatial layout and acquisition geometry. The Screenshot folder contains a representative screenshot of the LiDAR point cloud, illustrating the dataset's visual quality.

Fig 3.

Mapping result: first return of LiDAR (top-left), second return of LiDAR (top-right), third return of LiDAR (bottom-left), classified point cloud (bottom-right).

Accuracy was evaluated using eight independent checkpoints that were not included in the calibration process. Using three static base stations and fourteen control points for LiDAR trajectory calibration (see Fig 1), the resulting positional accuracy of the classified point cloud was assessed in WGS84 / UTM Zone 49S. The mean absolute errors for easting, northing, and elevation are 0.0166 m, 0.0168 m, and 0.0441 m, respectively, while the corresponding RMSE values are 0.0724 m (E), 0.0623 m (N), and 0.1380 m (Z). The associated mean biases—0.0101 m, 0.0115 m, and 0.0292 m—indicate minimal systematic offset in both horizontal and vertical components. These values fall within the expected performance range for UAV–LiDAR systems operating in coastal urban environments and confirm the suitability of the dataset for high-resolution coastal mapping and geomorphological analysis.

4. Experimental Design, Materials, and Methods

4.1. Study area

The LiDAR dataset covers the main tourism area of Pangandaran, West Java, Indonesia (7.691772°S - 7.705765°S, 108.649214°E - 108.665776°E), one of the most popular coastal tourism destinations on Java's southern coast. The area has sandy beaches, dense rows of hotels and restaurants, and public facilities attracting over 3 million visitors annually [3,6,7]. Its geographical position also makes it a well-recognised tsunami hazard zone, which underlines the importance of generating high-resolution geospatial data for exposure and risk assessment. The Pangandaran dataset combines fine-scale terrain information with built-up infrastructure mapping to support hazard-related applications.

4.2. Selection of mapping area

The mapping area, covering approximately 94 hectares, included the West Beach and East Beach corridors and adjacent urban zones. These subareas were selected because they combine high tourism density with direct exposure to tsunamis and climate change risk. Criteria for inclusion followed approaches similar to those of other UAV–LiDAR campaigns, where accessibility for UAV flights and ground control deployment were critical. The design ensured the survey captured natural coastal features and dense built-up areas, which are often missing from national topographic products.

4.3. Equipment used

Data was acquired using a DJI Matrice 300 RTK UAV with the CHCNAV AlphaAir 450 (AA450) LiDAR system. The AlphaAir 450 integrates a Livox Avia laser scanner with a high-precision GNSS + IMU navigation system and an onboard 26 MP RGB camera, allowing simultaneous LiDAR and imaging data collection.

The laser scanner has a maximum range of 450 m. It supports multiple scanning rates of 240,000, 480,000, and 720,000 pulses per second, which makes it adaptable to different flight altitudes and point density requirements. The scanner has a 70.4° field of view. It can record up to three returns per emitted pulse, enabling the capture of multiple reflections from features such as vegetation canopy, building roofs, and the ground surface.

The navigation system comprises a complete GNSS receiver and a 500 Hz inertial measurement unit (IMU). Under standard operating conditions, the system achieves a maximum precision of approximately 10 cm horizontally and 5 cm vertically. These specifications ensure the dataset provides high-resolution and reliable spatial information suitable for detailed topographic and high-resolution mapping applications. The integrated 26 MP camera further enriches the dataset by enabling colorized point cloud generation and providing additional imagery for visual interpretation. During the survey campaign, a Leica GS18T RTK GNSS receiver was deployed as the local base station to ensure centimeter-level positional accuracy. The Leica receiver was connected via RTK to the AlphaAir 450, providing real-time corrections for UAV positioning. Flight missions were designed and executed using the DJI Pilot application, where parameters such as altitude, forward and side overlap, and flight speed were explicitly defined. This ensured consistent coverage of the survey area and reproducibility of the dataset.

4.4. UAV data collection

The survey was conducted on 24 June 2025 under clear skies, with visibility above 10 km and low wind speeds (below 3 m/s). Flight missions were designed with 80% forward and 70% side overlap, at altitudes around 80 m above ground level (AGL). Each sortie followed a grid-based flight path, systematically covering the designated hotspots. Three sorties were conducted, generating a cumulative dataset of around 17 GB.

4.5. Data processing

The raw trajectory data and LiDAR returns were processed using CHCNAV CoPro and CoPre software versions 2.6.0 and 2.7.7, respectively. The workflow included GNSS/IMU integration, RTK/PPK correction, and strip alignment for trajectory calibration. Trajectory calibration and strip alignment were performed in CHCNAV CoPre 2.7.7 using the multi-strip adjustment tools provided in the software. The alignment process produced mean strip-to-strip residuals of 3–5 cm, with maximum residual differences of 5–8 cm observed at intersections of overlapping flight lines. These values confirm that the trajectory solution is stable and internally consistent across all three mapping projects, providing a reliable geometric foundation for classification, surface modeling, and accuracy assessment.

Point clouds were then exported in LAS v1.4 format, followed by noise filtering to remove outliers and spurious points. Point cloud classification was performed using the default filtering parameters of CHCNAV CoPre 2.7.7. Noise points were removed using a statistical outlier filter (maximum allowed variance of 3 standard deviations within a 20 m neighborhood). Ground classification employed a curvature-based progressive filter with a 2 m base bin size, a minimum height departure threshold of 0.3 m for non-ground separation, an expected terrain slope of 7.5°, a maximum height delta of 50 m, and a maximum building-width constraint of 100 m to distinguish large planar elevated surfaces. Building classification used default planarity-based segmentation thresholds (planarity ≥ 0.8, minimum segment size 20–50 points, minimum building height 2 m), while vegetation classification was based on height above ground (>0.3 m), irregularity metrics, and return-number logic. Points not meeting any structural criteria were assigned to the unclassified category (see Fig 3 for the results of the point cloud classification).

Point cloud data from all projects was merged into a single dataset and then divided into blocks of regular grids, each not exceeding 2 GB, for easy access. Examples of a point cloud image as the final result, represented in color as height, are listed in Fig 4.

Fig 4.

Sample images of point clouds, visualized based on Elevation and RGB.

The Digital Surface Model (DSM) and Digital Terrain Model (DTM) derived from the classified UAV–LiDAR point cloud are presented in Fig. 5. The DSM captures the uppermost reflective surfaces across the Pangandaran coastal tourism hotspots, including building rooftops, vegetation canopy, and elevated man-made structures, reflecting the complex morphology of the densely built and tree-lined beachfront environment. In contrast, the DTM represents the underlying bare-earth terrain after the removal of above-ground features, providing a continuous elevation surface. Both products were generated from the classified LAS files using ground-filtering and surface interpolation routines in CHCNAV CoPre 2.7.7, and are included in the Zenodo repository together with the original point cloud data. Such DSM and DTM derivations are consistent with other recent UAV–LiDAR workflows used in high-precision terrain monitoring [8].

Fig 5.

Digital Surface Model (DSM) and Digital Terrain Model (DTM) derived from the LiDAR dataset.

Limitations

Despite the high overall point density and centimeter-level positional accuracy of the dataset, several localized limitations were identified that users should consider when applying the data for high-precision analysis. First, point density is not uniform across the study area. Solid surfaces such as roads, rooftops, and open terrain typically exhibit 600–800 pts/m², consistent with strong single-return behavior. In contrast, dense vegetation canopy produces markedly lower, less uniform densities, commonly ranging from 40–150 pts/m², due to pulse absorption and leaf scattering, which can reduce ground penetration. Over water surfaces and block boundaries, densities fall below 100 pts/m², reflecting the weak or absent returns due to water's low reflectance.

Conversely, extremely high densities exceeding 4,000–5,000 pts/m² occur in areas where overlapping flight lines intersect with tree canopies, producing stacked multiple returns characteristic of the non-repetitive Livox Avia scanner integrated in the CHCNAV AA450 system. These very high-density areas do not indicate measurement error but reflect sensor geometry and multi-return behavior.

Second, occlusion effects persist in narrow corridors (<5 m wide), between closely spaced buildings, and beneath dense vegetation. These conditions reduce sampling coverage and may lead to small gaps in building facades or incomplete terrain reconstruction at the micro-scale. Third, although the ground control network and RTK/PPK corrections provide strong georeferencing stability, minor systematic elevation bias and vertical RMSE remain, as expected for UAV-LiDAR surveys conducted in complex coastal urban environments.

Finally, the dataset covers only the central tourism hotspots of Pangandaran (∼94 ha) and does not include surrounding rural or natural coastal zones. Users requiring broader spatial context may need to integrate this dataset with satellite-based DEMs or other regional elevation products. These limitations reflect the inherent characteristics of UAV–LiDAR acquisition, flight geometry, surface reflectance variability, and the interaction of laser pulses with complex coastal vegetation and built structures.

Ethics Statement

Not applicable.

CRediT Author Statement

Mega L. Syamsuddin: Conceptualization, Resources, Writing, Supervision; Umar Abdurrahman: Conceptualization, Data Curation and Visualization, Writing; Ajeng R. Puspita: Data Curation, Writing - Review & Editing; Sunarto: Writing - Review & Editing; Qurnia W. Sari: Writing - Review & Editing; Fadli Syamsudin: Writing - Review & Editing; Indrawan F. Pratyaksa: Conceptualization, Data Curation and Validation; Iqbal M. Cipta: Conceptualization, Data Curation and Validation; Ivonne M. Radjawane: Resources, Writing - Review & Editing; Hansan Park: Resources, Supervision.

Data Availability

• ZenodoDrone LiDAR dataset of Pangandaran coastal tourism hotspots for tsunami and climate risk valuation and exposure mapping (Original data).