Image Operations

Your first win: load, select, clip, and add NDVI

Let's get a quick win. This script loads a single Landsat 8 image over San Francisco, selects the visible bands, clips it to a county boundary, and adds a computed NDVI band. All four core operations in under 20 lines.

// Load a specific Landsat 8 image by ID
var image = ee.Image('LANDSAT/LC08/C02/T1_L2/LC08_044034_20230715');

// Apply scale factors for surface reflectance
var scaled = image.select('SR_B.*').multiply(0.0000275).add(-0.2);

// Select RGB bands for display
var rgb = scaled.select(['SR_B4', 'SR_B3', 'SR_B2']);

// Clip to San Francisco County
var sfCounty = ee.FeatureCollection('TIGER/2018/Counties')
  .filter(ee.Filter.eq('NAME', 'San Francisco')).first().geometry();
var clipped = rgb.clip(sfCounty);

// Compute and add NDVI
var ndvi = scaled.normalizedDifference(['SR_B5', 'SR_B4']).rename('NDVI');
var withNDVI = clipped.addBands(ndvi.clip(sfCounty));

// Display results
Map.centerObject(sfCounty, 12);
Map.addLayer(clipped, {min: 0, max: 0.3}, 'RGB');
Map.addLayer(withNDVI.select('NDVI'), {min: 0, max: 0.8, palette: ['brown', 'yellow', 'green']}, 'NDVI');

Peeking inside an image

Before we process anything, let's peek inside. GEE images carry rich metadata and band information that tells us exactly what data is available.

var image = ee.Image('LANDSAT/LC08/C02/T1_L2/LC08_044034_20230715');

// What bands does this image have?
print('Band names:', image.bandNames());

// What properties (metadata) are available?
print('Property names:', image.propertyNames());

// Read specific properties
print('Cloud cover:', image.get('CLOUD_COVER'));
print('Acquisition date:', image.date().format('YYYY-MM-dd'));
print('Sensor ID:', image.get('SENSOR_ID'));

// Check the projection and scale of a band
print('B4 projection:', image.select('SR_B4').projection());
print('B4 scale (meters):', image.select('SR_B4').projection().nominalScale());

Picking just the bands you need

Most images contain way more bands than we need. The .select() method pulls out only the ones we want, keeping scripts clean and fast.

var image = ee.Image('LANDSAT/LC08/C02/T1_L2/LC08_044034_20230715');

// Select by exact name (most common)
var rgb = image.select(['SR_B4', 'SR_B3', 'SR_B2']);
print('RGB band count:', rgb.bandNames().size()); // 3

// Select by pattern (regular expression)
var allSR = image.select('SR_B.*');
print('All SR bands:', allSR.bandNames()); // SR_B1 through SR_B7

// Select by index (0-based)
var firstTwo = image.select([0, 1]);
print('First two bands:', firstTwo.bandNames());

// Select and rename in one step
var renamed = image.select(['SR_B5', 'SR_B4'], ['NIR', 'RED']);
print('Renamed bands:', renamed.bandNames()); // NIR, RED

Best practice: Select by name, not by index. Band order can change between datasets, but names are consistent within a collection.

Cutting to your study area

The .clip() method masks an image to a geometry or feature boundary. Pixels outside become transparent. It is great for clean map displays and focusing on our area of interest.

// Clip to a simple geometry
var roi = ee.Geometry.Rectangle([-122.5, 37.6, -122.3, 37.8]);
var clipped = image.clip(roi);

// Clip to a feature from a FeatureCollection
var state = ee.FeatureCollection('TIGER/2018/States')
  .filter(ee.Filter.eq('NAME', 'California')).first();
var clippedToState = image.clip(state.geometry());

Map.addLayer(clippedToState.select(['SR_B4', 'SR_B3', 'SR_B2'])
  .multiply(0.0000275).add(-0.2),
  {min: 0, max: 0.3}, 'California');

Performance note: Clipping is computationally expensive on large or complex geometries. For analysis (reducers, exports), prefer .filterBounds() on the collection and specify the geometry in reduceRegion() instead of clipping first. Save .clip() for display purposes.

Stitching scenes together with mosaic

When our study area spans multiple satellite scenes, .mosaic() stitches them into a single seamless image. Here is the catch: the last image in the collection "wins" where scenes overlap, so sort order matters.

// Create a mosaic from multiple Landsat scenes
var collection = ee.ImageCollection('LANDSAT/LC08/C02/T1_L2')
  .filterDate('2023-07-01', '2023-07-31')
  .filterBounds(ee.Geometry.Point([-122.4, 37.7]))
  .filter(ee.Filter.lt('CLOUD_COVER', 15));

// Sort by cloud cover so cleanest image is on top
var mosaic = collection.sort('CLOUD_COVER').mosaic();

// Apply scale factors
var scaled = mosaic.select('SR_B.*').multiply(0.0000275).add(-0.2);

Map.setCenter(-122.4, 37.7, 8);
Map.addLayer(scaled, {bands: ['SR_B4', 'SR_B3', 'SR_B2'], min: 0, max: 0.3}, 'Mosaic');

For a statistically robust composite (better cloud removal), use .median() or .mean() instead of .mosaic(). Mosaicking is best when you want speed or need to preserve the original pixel values from a specific scene.

Stacking on new layers with addBands

The .addBands() method attaches new layers to an existing image. This is how we include computed indices (NDVI, NDWI), elevation data, or any derived layer alongside the original bands. Think of it like adding a new sheet to a stack of transparencies.

var image = ee.Image('LANDSAT/LC08/C02/T1_L2/LC08_044034_20230715');
var scaled = image.select('SR_B.*').multiply(0.0000275).add(-0.2);

// Compute NDVI and NDWI
var ndvi = scaled.normalizedDifference(['SR_B5', 'SR_B4']).rename('NDVI');
var ndwi = scaled.normalizedDifference(['SR_B3', 'SR_B5']).rename('NDWI');

// Add both to the original image
var enriched = scaled.addBands(ndvi).addBands(ndwi);
print('Enriched band names:', enriched.bandNames());
// SR_B1, SR_B2, ..., SR_B7, NDVI, NDWI

// Add elevation from a different source
var dem = ee.Image('USGS/SRTMGL1_003').rename('elevation');
var withElevation = enriched.addBands(dem);
print('With elevation:', withElevation.bandNames());

By default, .addBands() keeps all bands. If the new band has the same name as an existing one, you can use .addBands(newBand, null, true) to overwrite the existing band.

Reading the fine print: image metadata

Metadata properties tell us when, where, and how an image was captured. Think of it as the EXIF data on your phone photos, but for satellites. This information is essential for quality control, filtering, and documenting our analyses.

var image = ee.Image('LANDSAT/LC08/C02/T1_L2/LC08_044034_20230715');

// Key metadata fields for Landsat 8
print('Date acquired:', image.get('DATE_ACQUIRED'));
print('Cloud cover (%):', image.get('CLOUD_COVER'));
print('Cloud cover (land):', image.get('CLOUD_COVER_LAND'));
print('Sun elevation:', image.get('SUN_ELEVATION'));
print('WRS Path:', image.get('WRS_PATH'));
print('WRS Row:', image.get('WRS_ROW'));

// The system timestamp (milliseconds since epoch)
print('System time:', image.get('system:time_start'));
print('Formatted date:', image.date().format('YYYY-MM-dd'));

// Image footprint geometry
var footprint = image.geometry();
Map.addLayer(footprint, {color: 'blue'}, 'Image footprint');
Map.centerObject(footprint, 7);

Use .getInfo() sparingly. It forces a synchronous server call, which blocks your script. For display purposes, print() is asynchronous and much safer.

Next Steps

• Band Arithmetic - perform mathematical operations between bands.
• NDVI - compute and interpret the Normalized Difference Vegetation Index.
• Cloud Masking - remove cloud-contaminated pixels before analysis.
• Image Collections - filter, sort, and composite multi-image stacks.