Scholarlink

Purpose of Review <p>We conducted a systematic review of 837 scientific articles published between 2016 and 2023 to examine how ALS pre-processing algorithms have been used in forestry, focusing on data acquired from manned platforms and on three pre-processing steps: ground filtering (GF), digital terrain model (DTM) interpolation, and individual tree crown delineation (ITCD). This review was combined with a benchmark of widely used algorithms, evaluated at three representative point densities (low: 1 pts/m<sup>2</sup>; medium: 8 pts/m<sup>2</sup>; and high: ~18.9–23.7 pts/m<sup>2</sup>) and three study areas spanning contrasting temperate forest environments in the United States. We asked how algorithm choices and study designs have evolved in recent forestry ALS literature, and which methods are more robust across point densities and forest contexts, including whether calibration meaningfully improves key forest parameters retrieval such as tree height and crown diameter.</p> Recent Findings <p>The review showed the dominance of area-based approaches, while revealing an increasing adoption of tree-level studies, higher point densities and repeated ALS acquisitions supporting multi-site and multi-temporal analyses. It also showed an increasing share of studies using publicly available ALS datasets. Informed by literature trends, the benchmark provided complementary evidence on algorithm behavior under operational conditions: Cloth Simulation Filter and Inverse Distance Weighting performed consistently well for GF and DTM interpolation, whereas no single ITCD algorithm clearly outperformed others. Input settings calibration had limited impact on tree height estimation but improved crown diameter retrieval, particularly at low density.</p> Summary <p>Based on our findings, no single ALS processing workflow can be recommended universally, as algorithm performance and calibration benefits depend on point density and forest context. Calibration showed limited effects on tree height estimation but improved crown diameter retrieval, particularly at low density. We therefore provide density-aware, context-specific recommendations to support reproducible ALS processing workflows in forestry studies.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Benchmarking Trends in Airborne LiDAR Pre-Processing Algorithms for Forestry Research

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月6日

查看全文
  • Aaron Cardenas-Martinez,
  • Emilia Guisado-Pintado,
  • Victor Rodriguez-Galiano

摘要

Purpose of Review

We conducted a systematic review of 837 scientific articles published between 2016 and 2023 to examine how ALS pre-processing algorithms have been used in forestry, focusing on data acquired from manned platforms and on three pre-processing steps: ground filtering (GF), digital terrain model (DTM) interpolation, and individual tree crown delineation (ITCD). This review was combined with a benchmark of widely used algorithms, evaluated at three representative point densities (low: 1 pts/m2; medium: 8 pts/m2; and high: ~18.9–23.7 pts/m2) and three study areas spanning contrasting temperate forest environments in the United States. We asked how algorithm choices and study designs have evolved in recent forestry ALS literature, and which methods are more robust across point densities and forest contexts, including whether calibration meaningfully improves key forest parameters retrieval such as tree height and crown diameter.

Recent Findings

The review showed the dominance of area-based approaches, while revealing an increasing adoption of tree-level studies, higher point densities and repeated ALS acquisitions supporting multi-site and multi-temporal analyses. It also showed an increasing share of studies using publicly available ALS datasets. Informed by literature trends, the benchmark provided complementary evidence on algorithm behavior under operational conditions: Cloth Simulation Filter and Inverse Distance Weighting performed consistently well for GF and DTM interpolation, whereas no single ITCD algorithm clearly outperformed others. Input settings calibration had limited impact on tree height estimation but improved crown diameter retrieval, particularly at low density.

Summary

Based on our findings, no single ALS processing workflow can be recommended universally, as algorithm performance and calibration benefits depend on point density and forest context. Calibration showed limited effects on tree height estimation but improved crown diameter retrieval, particularly at low density. We therefore provide density-aware, context-specific recommendations to support reproducible ALS processing workflows in forestry studies.