Skip to main content

Lund University Publications

LUND UNIVERSITY LIBRARIES

Wall-to-Wall Mapping of Forest Canopy Height using ICESat-2 Data and Multi-source Remote Sensing Images in a Machine Learning Framework

Khankeshizadeh, Seyed Ehsan LU ; Jamali, Sadegh LU orcid ; Zaghian, Soheil ; Mauya, Ernest William ; Mohammadzadeh, Ali and Francis, Filbert LU (2025) 3rd International Conference on Machine Intelligence for GeoAnalytics and Remote Sensing, MIGARS 2025 In 2025 International Conference on Machine Intelligence for GeoAnalytics and Remote Sensing, MIGARS 2025
Abstract

Forest Canopy Height (FCH) is one of the key variables for understanding forest structure distribution and growth. Remotely sensed data such as the NASA Ice, Cloud and Land Elevation Satellite-2 (ICESat-2) ATL08 provides accurate FCH measurements; however, its point-based nature limits spatial continuity. This study addresses the challenge by generating a continuous FCH map over the West Usambara Mountain Forests in Tanzania through integration of ATL08 product with multi-source remote sensing inputs, including Sentinel-1, Sentinel-2, and SRTM DEM. The proposed methodology consists of four key stages: (1) preparation of a remote sensing feature cube, (2) preprocessing and noise removal of ATL08 segments, (3) spatial alignment and... (More)

Forest Canopy Height (FCH) is one of the key variables for understanding forest structure distribution and growth. Remotely sensed data such as the NASA Ice, Cloud and Land Elevation Satellite-2 (ICESat-2) ATL08 provides accurate FCH measurements; however, its point-based nature limits spatial continuity. This study addresses the challenge by generating a continuous FCH map over the West Usambara Mountain Forests in Tanzania through integration of ATL08 product with multi-source remote sensing inputs, including Sentinel-1, Sentinel-2, and SRTM DEM. The proposed methodology consists of four key stages: (1) preparation of a remote sensing feature cube, (2) preprocessing and noise removal of ATL08 segments, (3) spatial alignment and integration of ICESat-2 footprints with RS data patches, and (4) canopy height modeling using optimized machine learning regression, primarily Random Forest. Results showed that spectral indices and SAR features effectively predict canopy height, achieving an R2 of 0.29 and an RMSE of 8.64m.

(Less)
Please use this url to cite or link to this publication:
author
; ; ; ; and
organization
publishing date
type
Chapter in Book/Report/Conference proceeding
publication status
published
subject
keywords
Forest canopy height, ICESat-2, Machine learning, Remote sensing, Wall-to-Wall mapping
host publication
2025 International Conference on Machine Intelligence for GeoAnalytics and Remote Sensing, MIGARS 2025
series title
2025 International Conference on Machine Intelligence for GeoAnalytics and Remote Sensing, MIGARS 2025
publisher
IEEE - Institute of Electrical and Electronics Engineers Inc.
conference name
3rd International Conference on Machine Intelligence for GeoAnalytics and Remote Sensing, MIGARS 2025
conference location
Bucharest, Romania
conference dates
2025-09-02 - 2025-09-04
external identifiers
  • scopus:105025462397
ISBN
9798331579203
DOI
10.1109/MIGARS67156.2025.11232131
language
English
LU publication?
yes
id
32aaefb3-63f5-4b82-9863-5ec899da80e5
date added to LUP
2026-02-26 14:33:31
date last changed
2026-02-26 14:34:09
@inproceedings{32aaefb3-63f5-4b82-9863-5ec899da80e5,
  abstract     = {{<p>Forest Canopy Height (FCH) is one of the key variables for understanding forest structure distribution and growth. Remotely sensed data such as the NASA Ice, Cloud and Land Elevation Satellite-2 (ICESat-2) ATL08 provides accurate FCH measurements; however, its point-based nature limits spatial continuity. This study addresses the challenge by generating a continuous FCH map over the West Usambara Mountain Forests in Tanzania through integration of ATL08 product with multi-source remote sensing inputs, including Sentinel-1, Sentinel-2, and SRTM DEM. The proposed methodology consists of four key stages: (1) preparation of a remote sensing feature cube, (2) preprocessing and noise removal of ATL08 segments, (3) spatial alignment and integration of ICESat-2 footprints with RS data patches, and (4) canopy height modeling using optimized machine learning regression, primarily Random Forest. Results showed that spectral indices and SAR features effectively predict canopy height, achieving an R<sup>2</sup> of 0.29 and an RMSE of 8.64m.</p>}},
  author       = {{Khankeshizadeh, Seyed Ehsan and Jamali, Sadegh and Zaghian, Soheil and Mauya, Ernest William and Mohammadzadeh, Ali and Francis, Filbert}},
  booktitle    = {{2025 International Conference on Machine Intelligence for GeoAnalytics and Remote Sensing, MIGARS 2025}},
  isbn         = {{9798331579203}},
  keywords     = {{Forest canopy height; ICESat-2; Machine learning; Remote sensing; Wall-to-Wall mapping}},
  language     = {{eng}},
  publisher    = {{IEEE - Institute of Electrical and Electronics Engineers Inc.}},
  series       = {{2025 International Conference on Machine Intelligence for GeoAnalytics and Remote Sensing, MIGARS 2025}},
  title        = {{Wall-to-Wall Mapping of Forest Canopy Height using ICESat-2 Data and Multi-source Remote Sensing Images in a Machine Learning Framework}},
  url          = {{http://dx.doi.org/10.1109/MIGARS67156.2025.11232131}},
  doi          = {{10.1109/MIGARS67156.2025.11232131}},
  year         = {{2025}},
}