Fifty years of land use and land cover mapping in the United Arab Emirates: a machine learning approach using Landsat satellite data

This study analyses the spatiotemporal distribution of land use and land cover (LULC) in the United Arab Emirates (UAE) over the past 50 years (1972–2021) using 72 multi-temporal Landsat satellite images. Three machine learning (ML) classifiers, Classification and Regression Tree (CART), Support Vec...

Full description

Saved in:
Bibliographic Details
Main Authors: Mubbashra Sultan, Salem Issa, Basam Dahy, Nazmi Saleous, Mabrouk Sami
Format: Article
Language:English
Published: Frontiers Media S.A. 2024-12-01
Series:Frontiers in Earth Science
Subjects:
Online Access:https://www.frontiersin.org/articles/10.3389/feart.2024.1510510/full
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1846128062419173376
author Mubbashra Sultan
Salem Issa
Basam Dahy
Nazmi Saleous
Mabrouk Sami
author_facet Mubbashra Sultan
Salem Issa
Basam Dahy
Nazmi Saleous
Mabrouk Sami
author_sort Mubbashra Sultan
collection DOAJ
description This study analyses the spatiotemporal distribution of land use and land cover (LULC) in the United Arab Emirates (UAE) over the past 50 years (1972–2021) using 72 multi-temporal Landsat satellite images. Three machine learning (ML) classifiers, Classification and Regression Tree (CART), Support Vector Machine (SVM) and Random Forest (RF), were tested, with RF finally chosen for its higher performance. Spectral, spatial, topographic, and object aspect attributes were extracted and used as input for the RF algorithm to enhance the classification accuracy. A dataset comprising 46,146 polygons representing four LULC classes was created, with 80% allocated for training and 20% for testing, ensuring robust model validation. The algorithm was trained to develop a machine learning model that classified the data into four LULC classes namely: built areas, vegetation, water, and desert and mountainous regions, producing eight thematic maps for the years 1972, 1986, 1992, 1997, 2002, 2013, 2017, and 2021. The results reveal the dominance of desert and mountainous regions, with their coverage gradually declining from over 97% in 1972 to nearly 91% in 2021. In contrast, built areas grew from less than 1% to nearly 6%, while vegetation cover increased from 0.71% to 2.85%. Water bodies have exhibited periodic fluctuations between 0.4% and 0.35%. These changes are attributed to extensive urbanization, agricultural expansion, forest plantation programs, land reclamation, and megaprojects. Accuracy assessment of the classified maps showed high overall accuracy, ranging from 85.11% to 98.4%. The study provides a unique long-term analysis of the UAE over 50 years, capturing key developments from the 1970s oil boom through subsequent megaprojects at the onset of the new millennium, leading to reduced reliance on oil. These findings underscore the role of machine learning and geospatial technologies in monitoring LULC distribution in challenging environments, and the results serve as a vital tool for policymakers to manage land resources, urban planning, and environmental conservation.
format Article
id doaj-art-cb4a5e06eced4853b57e86a74afab4b7
institution Kabale University
issn 2296-6463
language English
publishDate 2024-12-01
publisher Frontiers Media S.A.
record_format Article
series Frontiers in Earth Science
spelling doaj-art-cb4a5e06eced4853b57e86a74afab4b72024-12-11T06:44:59ZengFrontiers Media S.A.Frontiers in Earth Science2296-64632024-12-011210.3389/feart.2024.15105101510510Fifty years of land use and land cover mapping in the United Arab Emirates: a machine learning approach using Landsat satellite dataMubbashra Sultan0Salem Issa1Basam Dahy2Nazmi Saleous3Mabrouk Sami4Department of Geosciences, College of Science, United Arab Emirates University, Al Ain, Abu Dhabi, United Arab EmiratesDepartment of Geosciences, College of Science, United Arab Emirates University, Al Ain, Abu Dhabi, United Arab EmiratesDivision of Engineering, New York University (NYU Abu Dhabi), Abu Dhabi, United Arab EmiratesDepartment of Geography and Urban Sustainability, College of Humanities and Social Sciences, United Arab Emirates University, Al Ain, Abu Dhabi, United Arab EmiratesDepartment of Geosciences, College of Science, United Arab Emirates University, Al Ain, Abu Dhabi, United Arab EmiratesThis study analyses the spatiotemporal distribution of land use and land cover (LULC) in the United Arab Emirates (UAE) over the past 50 years (1972–2021) using 72 multi-temporal Landsat satellite images. Three machine learning (ML) classifiers, Classification and Regression Tree (CART), Support Vector Machine (SVM) and Random Forest (RF), were tested, with RF finally chosen for its higher performance. Spectral, spatial, topographic, and object aspect attributes were extracted and used as input for the RF algorithm to enhance the classification accuracy. A dataset comprising 46,146 polygons representing four LULC classes was created, with 80% allocated for training and 20% for testing, ensuring robust model validation. The algorithm was trained to develop a machine learning model that classified the data into four LULC classes namely: built areas, vegetation, water, and desert and mountainous regions, producing eight thematic maps for the years 1972, 1986, 1992, 1997, 2002, 2013, 2017, and 2021. The results reveal the dominance of desert and mountainous regions, with their coverage gradually declining from over 97% in 1972 to nearly 91% in 2021. In contrast, built areas grew from less than 1% to nearly 6%, while vegetation cover increased from 0.71% to 2.85%. Water bodies have exhibited periodic fluctuations between 0.4% and 0.35%. These changes are attributed to extensive urbanization, agricultural expansion, forest plantation programs, land reclamation, and megaprojects. Accuracy assessment of the classified maps showed high overall accuracy, ranging from 85.11% to 98.4%. The study provides a unique long-term analysis of the UAE over 50 years, capturing key developments from the 1970s oil boom through subsequent megaprojects at the onset of the new millennium, leading to reduced reliance on oil. These findings underscore the role of machine learning and geospatial technologies in monitoring LULC distribution in challenging environments, and the results serve as a vital tool for policymakers to manage land resources, urban planning, and environmental conservation.https://www.frontiersin.org/articles/10.3389/feart.2024.1510510/fullremote sensingmultispectralclassificationrandom forestaccuracy assessmentarid environment
spellingShingle Mubbashra Sultan
Salem Issa
Basam Dahy
Nazmi Saleous
Mabrouk Sami
Fifty years of land use and land cover mapping in the United Arab Emirates: a machine learning approach using Landsat satellite data
Frontiers in Earth Science
remote sensing
multispectral
classification
random forest
accuracy assessment
arid environment
title Fifty years of land use and land cover mapping in the United Arab Emirates: a machine learning approach using Landsat satellite data
title_full Fifty years of land use and land cover mapping in the United Arab Emirates: a machine learning approach using Landsat satellite data
title_fullStr Fifty years of land use and land cover mapping in the United Arab Emirates: a machine learning approach using Landsat satellite data
title_full_unstemmed Fifty years of land use and land cover mapping in the United Arab Emirates: a machine learning approach using Landsat satellite data
title_short Fifty years of land use and land cover mapping in the United Arab Emirates: a machine learning approach using Landsat satellite data
title_sort fifty years of land use and land cover mapping in the united arab emirates a machine learning approach using landsat satellite data
topic remote sensing
multispectral
classification
random forest
accuracy assessment
arid environment
url https://www.frontiersin.org/articles/10.3389/feart.2024.1510510/full
work_keys_str_mv AT mubbashrasultan fiftyyearsoflanduseandlandcovermappingintheunitedarabemiratesamachinelearningapproachusinglandsatsatellitedata
AT salemissa fiftyyearsoflanduseandlandcovermappingintheunitedarabemiratesamachinelearningapproachusinglandsatsatellitedata
AT basamdahy fiftyyearsoflanduseandlandcovermappingintheunitedarabemiratesamachinelearningapproachusinglandsatsatellitedata
AT nazmisaleous fiftyyearsoflanduseandlandcovermappingintheunitedarabemiratesamachinelearningapproachusinglandsatsatellitedata
AT mabrouksami fiftyyearsoflanduseandlandcovermappingintheunitedarabemiratesamachinelearningapproachusinglandsatsatellitedata