Evaluating the vertical accuracy of lidar and open source dem for oil palm plantation planning and design

Alfred Michael Sikab and Wong, Wilson Vun Chiong (2025) Evaluating the vertical accuracy of lidar and open source dem for oil palm plantation planning and design. Journal of Smart Farming and Food Security, 1 (1). 1 -15. ISSN 2180-1738

Text FULLTEXT.pdf Restricted to Registered users only Download (452kB) | Request a copy

Abstract

Accurate elevation data are essential for effective oil palm plantation planning, hydrological modeling, and environmental management particularly in tropical regions characterized by complex topography and dense vegetation. While LiDAR-derived Digital Elevation Models (DEMs) offer high vertical precision, their acquisition costs often hinder adoption in resource-limited settings. Open-source DEMs provide accessible alternatives but are frequently affected by vegetation interference and coarse resolution, leading to reduced vertical accuracy. This study proposes a hybrid correction framework that integrates a random forest (RF) machine learning algorithm and a geographically weighted regression (GWR) a spatially adaptive statistical method to enhance the vertical accuracy of open-source DEMs for terrain-sensitive applications. The study used377 high-precision Ground Control Points (GCPs) and LiDAR data to evaluate and correct six global DEMs: Advanced Land Observing Satellite (ALOS), TerraSAR-X add-on for Digital Elevation Measurement (TanDEM-X), Copernicus GLO-30, Shuttle Radar Topography Mission (SRTM), Forest And Buildings removed Copernicus DEM (FABDEM), Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) GDEM, and a high-resolution drone-derived DEM. RF was used to identify key topographic predictors, including aspect, slope, curvature, Topographic Position Index (TPI), and Terrain Ruggedness Index (TRI), while GWR applied spatially adaptive corrections to the RF residuals. The integrated RF–GWR model significantly improved the vertical accuracy across all DEMs. The post-correction R² values reached 0.914 for TanDEM-X, 0.910 for ALOS, and increased from 0.608 to 0.914for Copernicus, with the residual standard deviations reduced by up to 75% and near-zero mean bias. These results highlight the model’s ability to correct both systematic and spatially varying elevation errors. The framework presents a scalable and alternative to LiDAR for use in precision agriculture, flood risk modeling, and infrastructure planning. Future work should explore integration with deep learning to improve the temporal responsiveness and operational scalability. Keywords: Digital Elevation Models, Machine Learning, Random Forest, Geographically Weighted Regression, Vertical Accuracy.1. INTRODUCTION Accurate terrain data are fundamental to informed land-use planning, hydrological modeling, and precision agriculture, particularly in tropical environments characterized by dense vegetation, high rainfall, and complex topography. In the context of oil palm plantation development, reliable elevation models are critical for the design of efficient drainage systems, optimization of

Item Type:	Article
Keyword:	Digital Elevation Models, Machine Learning, Random Forest, Geographically Weighted Regression, Vertical Accuracy
Subjects:	H Social Sciences > HD Industries. Land use. Labor > HD28-9999 Industries. Land use. Labor > HD9000-9999 Special industries and trades > HD9000-9495 Agricultural industries T Technology > TN Mining engineering. Metallurgy > TN1-997 Mining engineering. Metallurgy > TN263-271 Mineral deposits. Metallic ore deposits. Prospecting
Department:	FACULTY > Faculty of Tropical Forestry
Depositing User:	JUNAINE JASNI -
Date Deposited:	13 Aug 2025 12:07
Last Modified:	13 Aug 2025 12:07
URI:	https://eprints.ums.edu.my/id/eprint/44856

Actions (login required)

View Item