GSW

Tutorial 3

Mastering Supervised Land Use Land Cover (LULC) Classification with Google Earth Engine: A Step-by-Step Guide Using JavaScript API

Rifaat Abdalla (Department of Earth Sciences,College of Science,Sultan Qaboos University), Eltaib Saeed Ganawa (Faculty of Geographical and Environmental Sciences, University of Khartoum, Khartoum, Sudan), Mohammed Mahmoud Musa (Faculty of Computer Science and Information Technology, Alzaiem Alazhari University, Sudan), Anwarelsadat Eltayeb Elmahal (Faculty of Geographical and Environmental Sciences, University of Khartoum, Khartoum, Sudan)

Abstract : Google Earth Engine (GEE) is a powerful cloud-based platform designed for geospatial processing, providing unmatched capabilities for analyzing and visualizing vast amounts of satellite imagery and spatial data. Its ability to process data at scale makes it an essential tool for Land Use and Land Cover (LULC) classification, a critical task for accurately mapping land types and supporting informed decision-making in fields such as urban planning, resource management, and environmental conservation.

In this tutorial, participants will explore GEE’s JavaScript interface (GEE JS) to conduct supervised LULCC, showcasing its analytical strength in spatial analysis. Through a hands-on exercise, participants will not only gain practical experience but also enhance their understanding of key concepts in remote sensing, spatial analysis, and cloud computing. By working with real-world examples, they will learn to apply these skills effectively in diverse contexts.

Objectives: By the end of this tutorial, participants will:

  • Understand the principles of supervised classification in remote sensing, including the theoretical foundations and real-world applications.
  • Gain practical experience in using Google Earth Engine JavaScript (GEE JS) to import, process, and visualize satellite imagery for geospatial analysis.
  • Apply machine learning algorithms for training, validating, and refining Land Use and Land Cover (LULC) classifications within GEE.
  • Create thematic maps to represent various land cover types and learn how to export results for use in additional analyses or reports.

Target audience : This tutorial is suitable for:

  • Environmental scientists, practitioners, and researchers interested in LULC analysis.
  • Students and professionals in fields related to GIS, remote sensing, data science, and spatial data analysis.
  • Anyone with basic programming skills looking to learn more about GEE and LULC classification.

Tutorial Outline:

 1.  Introduction to Google Earth Engine (GEE)

  • Overview of GEE and its capabilities.
  • Introduction to the JavaScript code editor.
  • Accessing and managing Earth Engine assets.

 2. Accessing and preparing Satellite Imagery

  • Importing Landsat and Sentinel-2 imagery into GEE.
  • Preprocessing steps: cloud masking, region of interest (ROI) definition, and image composite creation.

 3.  Working with Training Data

  • Selecting land cover classes (e.g., water, vegetation, urban, barren land).
  • Creating regions of interest (ROI) for each land cover type using GEE’s interactive map.
  • Importing external shapefiles or creating custom polygons for training samples.

 4.  Supervised Classification Algorithms

  • Overview of supervised classification methods (e.g., Random Forest, SVM, CART).
  • Training the classifier with labeled data.
  • Classifying the image based on the trained model.

 5.  Validating and Evaluating the Classification

  • Splitting data into training and validation sets.
  • Accuracy assessment using confusion matrix and other statistical metrics.

 6. Visualizing and Exporting Results

  • Displaying the classified LULC map in GEE.
  • Exporting results to Google Drive or Earth Engine Asset for further use.

 7. Case Study: LULC Classification of a Specific Region

  • Application of the learned techniques to a real-world case study (e.g., a selected area in Sudan or another region).
  • Comparison of classification results with known data.

 8. Q&A and Troubleshooting

  • Application of the learned techniques to a real-world case study (e.g., a selected area in Sudan or another region).
  • Comparison of classification results with known data.

Technical requirements (software, hardware, etc.): 

  • Superior internet connection
  • Access to a Google Earth Engine account.
  • Basic understanding of remote sensing and geographic information systems (GIS).
  • Familiarity with the basic programming skills.