Google Earth Engine: Satellite Image Analysis in the Cloud

In short
- Google Earth Engine is a platform that analyzes satellite images directly in the cloud.
- Three ways of thinking, collection, filter and reducer, let you design most analyses.
- It is especially useful for large-scale time series analysis and for quick change detection at disaster sites.
Suppose a flood has occurred and you want to check the flooded extent. Getting satellite images of the area, downloading them to your computer, opening software and waiting for processing can take several hours on its own. A single image can be tens of gigabytes.
Google Earth Engine (GEE from here on) turns this flow around. Instead of bringing the images to your computer, you send your analysis commands to the server where the images are. Google's servers already store decades of satellite imagery, and the analysis runs right there. Only the result comes back to your screen, so local storage space and processing power matter very little. This post looks step by step at what that structure means in a time-critical field such as a disaster.
How is GEE different from ordinary GIS software?
Tools such as QGIS and ArcGIS download data to your computer and process it there. With GEE, you open code.earthengine.google.com in a browser, the editor opens right away, and when you run code the computation runs on Google's servers. Even when you work with hundreds of Landsat images, you can analyze them with your hard disk almost empty. One point to note: interface and menu names can change a little with updates, so it is a good idea to refer to the official documentation (developers.google.com/earth-engine) as well. For a disaster response agency that must quickly calculate the damaged area at a site, being able to start the analysis without a separate data download step is considered a major practical advantage.
Why do collections, filters and reducers matter?
When you first meet GEE, three unfamiliar terms keep appearing. A collection is a bundle of many satellite images. For example, all Sentinel-2 images form one collection. A filter is a sieve that picks only what you need from that bundle. You can set conditions such as a date range, a specific area, or a cloud percentage below a limit. A reducer is a calculation that summarizes the filtered images into one value or one image. Combining images from several dates by the median, or taking the minimum or maximum, belongs here. Once you are used to these three steps, turning a request such as "extract the NDWI change before and after the flood from a month of images" into code comes naturally.
What does GEE code look like?
You can use GEE in two ways: a JavaScript-based editor and the Python API (through libraries such as geemap). Below is the most basic skeleton in the JavaScript editor, which loads Sentinel-2 images over Korea, picks only the images with little cloud, and builds a median composite.
var collection = ee.ImageCollection('COPERNICUS/S2_SR_HARMONIZED')
.filterDate('2024-07-01', '2024-07-31')
.filterBounds(ee.Geometry.Point([127.5, 37.5]))
.filter(ee.Filter.lt('CLOUDY_PIXEL_PERCENTAGE', 20));
var composite = collection.median();
Map.centerObject(composite, 9);
Map.addLayer(composite, {bands: ['B4', 'B3', 'B2'], min: 0, max: 3000}, 'July median');
In this code, filterDate and filterBounds are the filters, and median() is the reducer. When you press the Run button at the top of the editor, the result appears in the map panel on the right. If you compare images from before and after wildfire damage with this structure, you can quickly visualize the extent of vegetation loss.
What kinds of work is GEE especially good at?
GEE shines in time series analysis and large-area processing. When you track yearly NDVI change over decades of imagery, or calculate land cover change over a whole country, work that takes days in a local environment finishes in much less time. On the other hand, for very detailed local-scale analysis, or when you must handle your own image data that GEE does not provide, QGIS or a Python environment may be more suitable. Some things you only learn by trying. You need to check whether images for the area and period you want to work with are included in the GEE catalog. In the editor, type the dataset name in the Search places and datasets box in the left panel and press the IMPORT button on the result card, and you can check the start date and spatial extent of that collection yourself. An agency that wants to analyze flood damage should first check in this way whether a SAR-based Sentinel-1 collection is in the catalog and whether it covers the dates it needs.
What points are easy to miss when using GEE?
GEE does not run computations immediately; it stacks up commands (lazy evaluation). No actual calculation happens at the moment you write the code. Computation starts only when an output command such as Map.addLayer or Export runs. If you do not know this, you may be confused when you try to print an intermediate result. Also, to download the analysis result as a file, you have to check commands such as Export.image.toDrive on the Tasks tab at the top of the editor and press the Run button before the export starts. If the file is large, it may be split and saved to Google Drive. If you must extract flood boundary polygons at a disaster site and share them as a GIS file, it is a good idea to include the time the Export step takes in your workflow in advance.
Step by step on screen
The figure below is a diagram that summarizes the order of the steps.

- Open GEE
Go to code.earthengine.google.com in a browser and sign in with a Google account - Check the dataset
Type 'Sentinel-2' in the Search places and datasets box in the left panel, click the result card and check the spatial and temporal extent - Load the collection
Click the IMPORT button on the dataset card; it is added as a variable in the Imports area at the top of the editor - Write the filter and reducer
Enter filterDate, filterBounds, filter (cloud condition) and median() code in the editor and click Run - Check the result
In the map panel on the right, turn layers on and off with the Layers toggle to look at the image; adjust the color range with the gear icon next to the layer name - Export the result
Click the Tasks tab at the top of the editor and press the Run button next to the Export entry to start saving to Google Drive
Menu names may differ slightly between program versions.
Prompts you can copy
A prompt for a GEE beginner who wants a code skeleton for a specific disaster type quickly
You are a geospatial analysis expert skilled in the Google Earth Engine JavaScript API. Write a GEE code skeleton that meets the conditions below. Disaster type: [choose one: flood / wildfire / landslide] Purpose of analysis: compare images from before and after the damage and extract the changed area Study area: [latitude and longitude, or place name] Analysis period: before the damage [date range], after the damage [date range] Satellite data to use: [choose one: Sentinel-1 SAR / Sentinel-2 optical] Required format: write in the order collection declaration, filter, reducer, map output, Export from the Tasks tab Explain the role of each line in a comment. Below the code, list separately in a numbered list the items that need checking. Do not use band names or dataset IDs that do not exist.
Replace the bracketed parts for your situation. Check the facts in any answer before you use it.
Try it yourself
- In the code.earthengine.google.com editor, type 'Sentinel-2' in Search places and datasets and check for yourself the period and region covered in the catalog.
- Paste the code example above into the editor and press Run to check whether a July median composite of the Korean Peninsula appears in the map panel.
- Change the date range in filterDate to the period around a wildfire, and compare how the result changes when you replace median() with min().
Common questions
Q. Do I need to be good at coding to use GEE?
A. For basic analysis, you can start with short code once you understand the collection, filter and reducer pattern. Even if you do not know JavaScript at all, you can get started by making small changes to the official tutorials and example code. If you are more comfortable in a Python environment, you can also use GEE with Python syntax through libraries such as geemap.
Q. Can I open results analyzed in GEE in QGIS?
A. If you run an export in GeoTIFF format from the Tasks tab, the file is saved to Google Drive, and you can load that file in QGIS through Layer > Add Layer > Add Raster Layer. Vector results are generally exported as Shapefile or GeoJSON.
Q. Can anyone use GEE for free?
A. Google runs separate programs for research, education and non-profit use, but you should check the specific terms of use and policies on the official site yourself. If you need commercial use or large-scale processing, a separate pricing plan may apply.
What comes next
Google Earth Engine has greatly lowered the barrier to working with satellite imagery. Without worrying about download time or storage space, once you learn the three ways of thinking, collection, filter and reducer, you can handle much of disaster analysis, from tracking flood extent to calculating wildfire damage area, in a single browser. Of course, GEE is not the best choice in every situation. Understanding the strengths and weaknesses of each tool and using it in the right place is the core of practical skill. A later post in this series covers how to connect images extracted from GEE to an artificial intelligence model, that is, the first step in using satellite data for training. What preprocessing is needed to feed satellite images into an AI model?
This is an English adaptation of a post in the Korean GeoAI basics series. The original is here, and more learning materials are on GeoAI for Disaster Safety.
This post was drafted with the help of AI tools and reviewed by the author. Tool features and data portals change over time, so check the official documentation when you practice.
Comments
Post a Comment