Shoreline kinematics in Libreville (Gabon) derived from very high spatial resolution (VHR) imagery using machine and deep learning techniques
Abstract
Artificial Intelligence (AI), remote Sensing (RS) and observations and field data offer a number of advantages for detection and automatic extraction of coastline evolution. However, for many countries in the intertropical zone, such as Gabon, accurate and up-to-date mapping systems and spatial statistics are lacking or incomplete. In this context, RS using Very High Spatial (VHR) resolution satellite imagery and algorithms of AI, Machine Learning (ML) and Deep Learning (DL) associated with field data represents a major asset. The primary objective on this work is to contribute to the understanding of the shoreline dynamics of Libreville, the political and administrative capital of Gabon, through a diachronic analysis covering the period from 2003 to 2024. To this end, a panchromatic aerial image from 2003, two Pleiades images (a stereo image from 2022 and tri-stereo image for 2024), as well, in situ shoreline surveys conducted simultaneously with the Pleiades image acquisition, were utilized. Several AI models have been implemented: ML (Pixel-Based Image Analysis (PBIA) and Object-Based Image Analysis (OBIA) with Random Forest (RF) and Support Vector Machine (SVM) algorithms) and DL (Convolutional Neural Network (CNN) processing by pixel-based approach and combined with OBIA). The RF-PBIA and CNN-OBIA approaches yielded the best classification performances (over 80%), enabling the extraction of shorelines from 2022 (Pleiades stereo) and 2024 (Pleiades tri-stereo). The analysis of the End Point Rate index revealed that between 2003 and 2022, the coastline exhibited a general trend of accretion (0.87 m/year). In contrast, the period from 2022 to 2024 recorded an average annual rate of -0.32 m/year.
Related
- A Deep‐Learning Model Cascade for Automating Large‐Scale Shoreline Mapping From Satellite Imagery
- High resolution spatial documentation of Renaissance church interiors through multiple non-invasive survey techniques
- Digital Meaningful Learning Module For Two-Dimensional Kinematics To Enhance Critical Thinking Skills In High School
- Advanced machine learning techniques for daily streamflow forecasting: a case study of the Brahmaputra River
- Spatial distribution, origins, and driving mechanisms of groundwater nitrate contamination in a rural-agricultural-mining catchment: Insight from PMF, machine learning and multi-isotopic techniques
BibTeX
@article{manombambadinga2026shoreline,
title = {{Shoreline kinematics in Libreville (Gabon) derived from very high spatial resolution (VHR) imagery using machine and deep learning techniques}},
author = {Nina Manomba-Mbadinga and Simona Niculescu},
journal = {Applied Computing and Geosciences},
year = {2026},
month = sep,
doi = {10.1016/j.acags.2026.100407},
url = {https://hal.science/hal-05752486v1},
note = {CC BY-NC-ND 4.0},
}
RIS
TY - JOUR
TI - Shoreline kinematics in Libreville (Gabon) derived from very high spatial resolution (VHR) imagery using machine and deep learning techniques
AU - Nina Manomba-Mbadinga
AU - Simona Niculescu
PY - 2026
DA - 2026/09/16
JO - Applied Computing and Geosciences
DO - 10.1016/j.acags.2026.100407
UR - https://hal.science/hal-05752486v1
AB - Artificial Intelligence (AI), remote Sensing (RS) and observations and field data offer a number of advantages for detection and automatic extraction of coastline evolution. However, for many countries in the intertropical zone, such as Gabon, accurate and up-to-date mapping systems and spatial statistics are lacking or incomplete. In this context, RS using Very High Spatial (VHR) resolution satellite imagery and algorithms of AI, Machine Learning (ML) and Deep Learning (DL) associated with field data represents a major asset. The primary objective on this work is to contribute to the understanding of the shoreline dynamics of Libreville, the political and administrative capital of Gabon, through a diachronic analysis covering the period from 2003 to 2024. To this end, a panchromatic aerial image from 2003, two Pleiades images (a stereo image from 2022 and tri-stereo image for 2024), as well, in situ shoreline surveys conducted simultaneously with the Pleiades image acquisition, were utilized. Several AI models have been implemented: ML (Pixel-Based Image Analysis (PBIA) and Object-Based Image Analysis (OBIA) with Random Forest (RF) and Support Vector Machine (SVM) algorithms) and DL (Convolutional Neural Network (CNN) processing by pixel-based approach and combined with OBIA). The RF-PBIA and CNN-OBIA approaches yielded the best classification performances (over 80%), enabling the extraction of shorelines from 2022 (Pleiades stereo) and 2024 (Pleiades tri-stereo). The analysis of the End Point Rate index revealed that between 2003 and 2022, the coastline exhibited a general trend of accretion (0.87 m/year). In contrast, the period from 2022 to 2024 recorded an average annual rate of -0.32 m/year.
SN - 2590-1974
ER -