Google Earth Engine for Machine Learning & Change Detection

Explore Google Earth Engine via a web interface to access more than 40 years of satellite imagery for global geospatial analysis, with code editor and explorer for Python and JavaScript.

Explore how the ESA EO Browser enables cloud-based visualization and on-the-fly computation of spectral indices from Sentinel and Landsat data for land cover and vegetation monitoring.

Explore the Google Earth Engine code editor, a web-based development environment for writing JavaScript, visualizing maps, managing scripts and assets, sharing via links, and running unsupervised classification on Landsat data.

Explore the Google Earth Engine data catalog by function, including raster climate and atmosphere datasets, Landsat collections, high-resolution imagery, and geophysical layers for land use and change detection.

Learn JavaScript fundamentals for geospatial analysis in Google Earth Engine, covering variables, data types, arrays, dictionaries, objects, functions, and commenting to build reusable code.

Demonstrates how to declare variables in JavaScript for Google Earth Engine, covering numbers, strings, lists, and objects with semicolons and consistent quoting, and debugging variable scope errors in the console.

Map a Landsat eight collection to a specific time step and region, create a cloud-free Landsat eight composite, and explore visualization and input tools for region-specific analysis.

Import Landsat 8 surface reflectance data into Earth Engine, filter by region and date, compute a median image, and visualize it with adjustable visualization parameters on the map.

Learn to create a satellite image composite and calculate the normalized difference vegetation index (ndvi) in Google Earth Engine using bands, function operations, and expression-based workflows.

Learn to use JavaScript in Google Earth Engine to compute the normalized difference vegetation index from Landsat bands 5 and 4, and map the maximum NDVI over time.

Use Google Earth Engine to mosaic Sentinel-2 images for a Paris study area, filter by date and geometry, select bands, clip to the polygon, and export to Google Drive.

Explore exporting Google Earth Engine images to Google Drive or cloud storage, using region of interest buffers, bounding boxes, clip operations, and appropriate scale and projection settings.

Practice your skills in Google Earth Engine by applying JavaScript basics—change the area and time period, use Landsat images, and create and run your own image visualizations.

Explore the theory behind machine learning and image classification, including supervised and unsupervised algorithms, and connect these concepts to land use classification in Google Earth Engine.

Learn image classification in gis and remote sensing with machine learning, covering supervised and unsupervised approaches and accuracy assessment using confusion matrices to evaluate land cover maps.

Explore the basics of remote sensing for land use and land cover mapping, using spectral signatures and Landsat time-series to perform visual and automated change detection.

Learn how to convert Landsat satellite images into land use land cover maps using image classification, exploring spectral pattern recognition, training data, and supervised and unsupervised approaches.

Explore supervised and unsupervised image classification in remote sensing, using training areas and an interpretation key to map land cover, or clustering to reveal natural spectral groups.

Define mutually exclusive and exhaustive land use and land cover classes for supervised LULC classification. Collect representative training data, build spectral statistics, classify pixels, and assess accuracy to produce maps.

Discover image-based classification in Google Earth Engine Explorer using Landsat data and hand-drawn training points to train a random forest classifier and download results, including cloud and land cover classes.

Explore Landsat data characteristics, sensors, spectral bands, and resolution, including Landsat 8's 11 bands and level 1 and level 2 products for atmospheric correction and surface reflectance.

Learn to import Sentinel-2 data in Google Earth Engine, filter by Dubai geometry and 2020, and visualize true color and false color composites for land cover mapping with supervised classification.

Learn to visualize Landsat 8 imagery in Google Earth Engine by building true color and false color composites, selecting bands, applying stretches, and adjusting color palettes.

Master unsupervised image classification in Google Earth Engine by creating training data in Dubai, applying k-means clustering to obtain five or ten classes, and exporting results to Google Drive.

Explore supervised learning algorithms for image classification, from minimum distance to the mean and maximum likelihood to random forest and SVM, with parametric vs nonparametric methods and accuracy assessment.

Learn to perform supervised land-cover classification in Google Earth Engine using sentinel imagery and a random forest classifier, including training data creation, class merging, and model deployment.

Learn to perform supervised classification with CART (decision trees) in Google Earth Engine using Dubai satellite data, training and validation data, and evaluate overall, user, and producer accuracies.

Test accuracy assessment of a land-cover classification in Google Earth Engine using a validation dataset from the Dubai area. Compute confusion and error metrics and report the overall accuracy.

Assess accuracy of land use and land cover maps using confusion matrices, reference data, and error metrics like overall, user, and producer accuracy.

Explore change detection in remote sensing by comparing multi-temporal classifications and images, using image differencing, post-classification comparison, and time-series trend analysis to reveal land degradation and crop dynamics.

Learn to map burn severity using the normalized burn ratio index (nbr) in Google Earth Engine, comparing pre- and post-fire imagery to classify unburned to high severity and guide restoration.

Learn to perform simple change detection in Google Earth Engine using Landsat 8 data, computing a normalized burn ratio to map post-fire burn severity in Chile 2016.

Learn to compute precipitation anomalies from CHIRPS data for Madagascar by comparing a defined 2021 time frame to the long-term mean across years, and visualize the results in millimeters.

Select another area and use Google Earth Engine to calculate the normalized burn index before and after using Landsat 8 composites, then map burnt areas and discuss Australia’s 2019 fires.