EG33: Land Use and Land Cover Mapping using Google Earth Engine: A Comparative Analysis of Machine Learning Algorithms

Salem Issa, Mubbashra Sultan

Research output: Contribution to conferencePaperpeer-review

Abstract

Land Use and Land Cover (LULC) mapping is of high significance for understanding and managing the changing environment. It provides valuable insights about distribution and composition of various land features, such as urban areas, vegetation, water bodies, and others. Accurate LULC mapping is applicable for multiple applications like urban planning, natural resource management, climate change monitoring, and sustainable development. Remote sensing plays a vital role in LULC mapping and monitoring. With the revolution in satellite remote sensing, better spectral and spatial resolution have also improved the mapping capabilities. Satellite missions like Landsat and Sentinel offer consistent data over long periods. This consistency allows for the comparison of LULC data across different time periods and helps in identifying long-term trends and changes. Google Earth Engine (GEE) is a powerful platform that enables access to a huge collection of satellite data and has significantly revolutionized the field of LULC mapping by providing data and tools for analysis. Abu Dhabi, a rapidly developing city, is facing the challenge of sustainable development. With urban expansion and population growth, effective land use and natural resource management is crucial to ensure sustainable development. This study utilizes GEE to create multitemporal LULC maps for three selected years i.e., 2001, 2014 and 2022. This paper attempts to compare the effectiveness and efficiency of two machine learning algorithms for LULC mapping. The maps demonstrate the increasing urbanization and vegetation growth, while there is a decline in desert areas. The LULC classification accuracy using the Random Forest (RF) algorithm ranges from 83.64% to 89.74% with Kappa values between 0.80 and 0.87. Comparatively, the Classification And Regression Trees (CART) algorithm achieves slightly lower accuracy. The study reveals the potential of GEE as an open-source platform for LULC mapping, with RF outperforming CART for the study area. Change analysis confirms the expansion of built-up areas and reduction of sand and sabkha. The findings emphasize the significance of GEE in supporting land management and planning decisions by providing reliable LULC information.

Original languageEnglish
Pages147-151
Number of pages5
DOIs
Publication statusPublished - 2024
Event7th International Conference on Engineering Geophysics, ICEG 2023 - Al Ain City, United Arab Emirates
Duration: Oct 16 2023Oct 19 2023

Conference

Conference7th International Conference on Engineering Geophysics, ICEG 2023
Country/TerritoryUnited Arab Emirates
CityAl Ain City
Period10/16/2310/19/23

ASJC Scopus subject areas

  • Geophysics

Fingerprint

Dive into the research topics of 'EG33: Land Use and Land Cover Mapping using Google Earth Engine: A Comparative Analysis of Machine Learning Algorithms'. Together they form a unique fingerprint.

Cite this