Lab 6 - Image Collections, Filtering & Sorting

Lab Instructions

1. Part 1 - Viewing an Image Collection

   Open up Google Earth Engine Code: https://code.earthengine.google.com/

   Copy and paste the following code into the center panel and click Run:

   /////
   // View an Image Collection
   /////
   
   // Import the Landsat 8 Raw Collection.
   var Landsat8 = ee.ImageCollection('LANDSAT/LC08/C02/T1');
   
   // Print the size of the Landsat 8 dataset.
   print('The size of the Landsat 8 image collection is:', Landsat8.size());
   
   // Try to print the image collection.
   // WARNING! Running the print code immediately below produces an error because 
   // the Console cannot print more than 5000 elements.
   print(Landsat8);
   
   // Add the Landsat 8 dataset to the map as a mosaic.
   Map.addLayer(Landsat8,
       {
           bands: ['B4', 'B3', 'B2'],
           min: 5000,
           max: 15000
       },
       'Landsat 8 Image Collection');

   Understanding the code:

   • ee.ImageCollection() - Loads an entire collection of images
   • .size() - Returns the number of images in the collection
   • When you add a collection to the map, Earth Engine mosaics all images together

   While the enormous image catalog is accessed, it could take a couple of minutes to see the result. You may note individual "scenes" being drawn, which equate to the way that the Landsat program partitions Earth into "paths" and "rows."

   

   About Rows and Paths: Images from Landsat are placed in a grid pattern. A row is a sequence of images taken at the same latitude, while a path is a series of images taken at the same longitude. Paths are numbered from west to east starting at the 180-degree longitude line. Rows are numbered from south to north starting at the equator.

   Notice the high amount of cloud cover and the "layered" look. This is because Earth Engine is drawing each of the images one on top of the other.

   

   Expected Console Output

   You should see a number greater than 1,000,000 and an error about too many elements to print.

   

2. Part 2 - Comment Out Code

   Edit your code to "comment out" the last two code commands using // at the beginning of each line. This will speed up your code in subsequent sections.

   

   Keyboard Shortcut

   Select multiple lines and press Ctrl + / (Windows) or Cmd + / (Mac) to toggle comments on/off quickly.

3. Part 3 - Filter by Date

   Now we will look at finding "the needle in the haystack." The basic approach to GEE is to add an image collection that brings in every image, then to filter the collection until you find the exact image you're interested in.

   filterDate allows us to narrow down the date range of the ImageCollection. Copy and paste:

   /////
   // Filter an Image Collection
   /////
   
   // Filter the collection by date.
   var LandsatWinter = Landsat8.filterDate('2020-12-01', '2021-03-01');
   
   Map.addLayer(LandsatWinter,
       {
           bands: ['B4', 'B3', 'B2'],
           min: 5000,
           max: 15000
       },
       'Winter Landsat 8');
   
   print('The size of the Winter Landsat 8 image collection is:', LandsatWinter.size());

   Understanding filterDate():

   • Takes two arguments: start date and end date
   • Format: 'YYYY-MM-DD' (year-month-day)
   • The end date is exclusive - images from that exact date are not included

   The result will load much faster because we filtered by date. Notice that before we had over a million images; after filtering, we have thousands.

   

   Comment out the Map.addLayer for winter imagery before moving on.

   

4. Part 4 - Filter by Location

   The filtering tool filterBounds is based on a location - for example, a point, polygon, or other geometry. Copy and paste:

   /////
   // Filter by location    
   /////
   
   // Create an Earth Engine Point object.
   var pointFL = ee.Geometry.Point([-82.32, 29.65]);
   
   // Filter the collection by location using the point.
   var LandsatFL = LandsatWinter.filterBounds(pointFL);
   Map.addLayer(LandsatFL,
       {
           bands: ['B4', 'B3', 'B2'],
           min: 5000,
           max: 15000
       },
       'FL Landsat 8');
   
   // Add the point to the map to see where it is.
   Map.addLayer(pointFL, {}, 'Point FL');
   
   print('The size of the Gainesville Winter Landsat 8 image collection is: ', LandsatFL.size());

   Understanding filterBounds():

   • Keeps only images that intersect with your geometry
   • Works with points, polygons, rectangles, or any geometry
   • Dramatically reduces the collection size by focusing on your study area

   Note that we add the pointFL to the map with an empty dictionary {} for the visParams argument - we are not specifying visualization parameters, using defaults instead.

   Run the code. Now only images that intersect our point have been selected. Note we only have 12 images.

   

   Try It

   Change the coordinates to your hometown and see how many images are available for that location during winter 2020-2021.

5. Part 5 - Select First Image

   The first function selects the first image in an ImageCollection. This allows us to place a single image on the screen. Because images are stored in time order, it will select the earliest image.

   

   Notice the image is very cloudy. Normally this is a problem for analysis - we prefer images without clouds. Each image in Landsat's collection has a cloud cover rating we can use.

6. Part 6 - Sort by Cloud Cover

   The basic approach is to sort the ImageCollection by cloud cover, then add the first image:

   // Sort the image collection by cloud cover
   var LandsatSort = LandsatFL.sort('CLOUD_COVER');
   
   // Select the first image in the sorted collection.
   var LandsatFirstSort = LandsatSort.first();
   
   // Display the first image in the sorted collection.
   Map.centerObject(LandsatFirstSort, 7);
   Map.addLayer(LandsatFirstSort,
       {
           bands: ['B4', 'B3', 'B2'],
           min: 5000,
           max: 15000
       },
       'Sort Landsat 8');

   Understanding the workflow:

   • .sort('CLOUD_COVER') - Orders images from lowest to highest cloud cover
   • .first() - After sorting, the first image is the one with least clouds
   • Map.centerObject() - Centers the map on your image at zoom level 7

   Now you have a single Landsat image with no cloud cover, filtered for Gainesville during the winter of 2020-2021!

   

   The Complete Filtering Pipeline

   1,000,000+ images → filterDate → ~50,000 images → filterBounds → 12 images → sort + first → 1 cloud-free image

   Full code link: GEE Code