Lab 8 - Band Arithmetic: NDVI

Lab Instructions

1. Part 1 - Set Up the Image

   Let's start by loading a Sentinel-2 image. Copy and paste this code into Earth Engine:

   /////
   // Band Arithmetic - NDVI with Sentinel-2
   /////
   
   // Define a point of interest (San Francisco area)
   var sfoPoint = ee.Geometry.Point(-122.3774, 37.6194);
   
   // Load and filter Sentinel-2 imagery
   var sfoImage = ee.ImageCollection('COPERNICUS/S2')
       .filterBounds(sfoPoint)
       .filterDate('2020-02-01', '2020-04-01')
       .first();
   
   // Display as false color composite
   Map.centerObject(sfoImage, 11);
   Map.addLayer(sfoImage, {
       bands: ['B8', 'B4', 'B3'],
       min: 0,
       max: 2000
   }, 'False Color');

   Understanding the code:

   • ee.Geometry.Point() - Creates a location to filter imagery around
   • filterBounds() - Only keeps images that overlap our point
   • filterDate() - Limits to a specific time period
   • .first() - Takes the first image from the filtered collection
   • The false color composite (NIR-Red-Green) makes vegetation appear red

   Sentinel-2 Band Reference: B8 = NIR, B4 = Red, B3 = Green, B2 = Blue

2. Part 2 - Calculate NDVI Manually (The Long Way)

   First, let's calculate NDVI step-by-step to understand exactly what's happening. This approach uses the basic arithmetic methods:

   // Extract the near-infrared and red bands
   var nir = sfoImage.select('B8');
   var red = sfoImage.select('B4');
   
   // Calculate numerator: NIR - Red
   var numerator = nir.subtract(red);
   
   // Calculate denominator: NIR + Red
   var denominator = nir.add(red);
   
   // Calculate NDVI: numerator / denominator
   var ndvi = numerator.divide(denominator);
   
   // Visualize with a vegetation palette
   var vegPalette = ['red', 'white', 'green'];
   Map.addLayer(ndvi, {
       min: -1,
       max: 1,
       palette: vegPalette
   }, 'NDVI (Manual)');

   Understanding the arithmetic methods:

   • .select('B8') - Extracts a single band as a new image
   • .subtract(red) - Subtracts red band values from NIR, pixel by pixel
   • .add(red) - Adds red band values to NIR, pixel by pixel
   • .divide(denominator) - Divides pixel by pixel

   Expected Result

   You should see an image where:

   • Green areas = healthy vegetation (NDVI close to 1)
   • White areas = bare soil/urban (NDVI near 0)
   • Red areas = water (NDVI negative)

   Try It

   Use the Inspector tool to click on different land cover types. Record the NDVI values you find for: forest, grass, bare ground, water, and urban areas.

3. Part 3 - Calculate NDVI with normalizedDifference (The Easy Way)

   Normalized differences are so common that Earth Engine provides a shortcut. The entire calculation can be done in one line:

   // Calculate NDVI using the built-in method
   var ndviND = sfoImage.normalizedDifference(['B8', 'B4']);
   
   Map.addLayer(ndviND, {
       min: -1,
       max: 1,
       palette: vegPalette
   }, 'NDVI (normalizedDifference)');

   Understanding normalizedDifference():

   • Takes an array of two band names: ['First', 'Second']
   • Computes: (First - Second) / (First + Second)
   • The result is identical to the manual method

   Band Order Matters!

   • For NDVI: NIR comes first, Red comes second
   • ['B8', 'B4'] ✅ Correct
   • ['B4', 'B8'] ❌ Wrong (inverts the result!)

4. Part 4 - Calculate NDWI (Water Index)

   The same technique works for other indices! The Normalized Difference Water Index (NDWI) measures vegetation water content using NIR and SWIR bands:

   

   // Calculate NDWI (Normalized Difference Water Index)
   var ndwi = sfoImage.normalizedDifference(['B8', 'B11']);
   
   var waterPalette = ['brown', 'white', 'blue'];
   Map.addLayer(ndwi, {
       min: -0.5,
       max: 0.5,
       palette: waterPalette
   }, 'NDWI');

   Interpreting NDWI:

   • High values (blue): High vegetation water content or water bodies
   • Low values (brown): Dry vegetation or bare soil

   Note: NDWI is sensitive to both open water AND vegetation moisture stress. It's commonly used for drought monitoring.

   Try It

   Compare NDVI and NDWI for the same areas. Do high NDVI areas always have high NDWI? What might cause differences?

5. Part 5 - Explore Other Indices

   Read these articles about additional spectral indices:

   • Remote Sensing Indices (Medium)
   • 4 Useful Remote Sensing Indices (Medium)

   Challenge: Calculate a New Index

   Choose a different index from the readings (e.g., NDBI for built-up areas, EVI for enhanced vegetation) and implement it in your code. Think about:

   • Which bands do you need?
   • What does the index measure?
   • What color palette would make sense?
   • What min/max values should you use?