Mastering Earth Observation with AI & Google Earth Engine

Discover Google Earth Engine's cloud-based geospatial analysis with petabyte-scale satellite imagery from Landsat, Sentinel, and MODIS. Use JavaScript or Python APIs and integrate machine learning for large-scale analysis.

Access Google Earth Engine at code dot earth engine dot google dot com; create an account and explore maps. Draw ROI with lines or polygons; import data to Google Drive.

Master Google Earth Engine basics by printing messages, adding layers, and filtering datasets. Visualize Landsat and Sentinel-2 imagery, compute ndvi, and export results to Google Drive.

Compare ndvi changes between 2024 and 2020 using sentinel-2 data, mask clouds with qa60, then identify dense vegetation with ndvi >0.5 and reduce region stats.

Create a seven-class land use map in Google Earth Engine, using harmonized Sentinel-2 data, build training data, compute NDVI, NDWI, NDBI, NDSI, train random forest, and assess accuracy.

Build a map slider with a split panel to compare rgb imagery and a lulc map generated by random forest, with synchronized zoom via a linker and hidden map controls.

Train a 1D-CNN for land use land cover mapping with sentinel-2 data, using Google Earth Engine and Google Colab, achieving around 86% accuracy across seven classes.

Create a night-time light gif from an image collection over Nepal, overlay vectors and text annotations, and add hillshade for a compelling visualization.

Create daily rainfall animation maps from time-series precipitation data using the Google Earth Engine, with dynamic titles, color bars, and map overlays.

Analyze the 2025 Los Angeles forest fire using Landsat 9 imagery to compute NDVI and NBR changes, identify burned areas, and produce a bond area map with total area.

Explore a ResNet-like change detection approach using the Libor CDD dataset of 637 bi-temporal Google Earth Engine images (1024×1024 RGB) with ground-truth building changes.

Explore change detection with temporal imagery from Google Earth Engine. Crop 1024×1024 imagery to 256×256 tiles, train a residual unit model in PyTorch, and evaluate with accuracy and confusion matrix.

Analyze changes in Saudi Arabia's agricultural land with Google Earth Engine and Landsat imagery; compute NDVI with a 0.2 threshold to map vegetation and chart area over time.

Learn rapid flood mapping in Dubai 2024 using optical imagery from Landsat and Sentinel-2, and SAR data from Sentinel-1, with Google Earth Engine to map flood extent and assess impacts.

Generate a landslide susceptibility map, explain variable importance in a random forest, and illustrate histogram samples and ROC curve using 31 points within a cloud-based open-source methodology for higher accuracy.

Learn to generate landslide susceptibility maps in Google Earth Engine by compiling samples, building elevation, slope, aspect, hillshade, and flow accumulation maps, training a random forest, and visualizing ROC/histograms.

Generate a roc curve and a variable importance map for landslide susceptibility using a random forest model, explain variable importance, and visualize results with a user interface chart.