What Cities Look Like from Space

Your first urban map in 60 seconds

Let's jump right in. This script calculates NDBI from Landsat 8 and maps Houston's built-up areas. Bright pixels? Those are roads, buildings, and parking lots.

// Define Houston area
var houston = ee.Geometry.Point([-95.37, 29.76]);

// Load Landsat 8 SR, summer 2023
var l8 = ee.ImageCollection('LANDSAT/LC08/C02/T1_L2')
  .filterDate('2023-06-01', '2023-08-31')
  .filterBounds(houston)
  .filter(ee.Filter.lt('CLOUD_COVER', 20))
  .median();

// Apply scale factors
var scaled = l8.multiply(0.0000275).add(-0.2);

// Calculate NDBI = (SWIR1 - NIR) / (SWIR1 + NIR)
var ndbi = scaled.normalizedDifference(['SR_B6', 'SR_B5']).rename('NDBI');

// Visualize
Map.addLayer(ndbi, {min: -0.5, max: 0.5, palette: ['blue', 'white', 'red']}, 'NDBI');
Map.centerObject(houston, 11);
print('NDBI calculated for Houston');

Why are cities tricky to map?

Cities are a unique challenge for remote sensing. Unlike a nice uniform forest or a field of corn, urban areas are a wild mosaic of materials: concrete, asphalt, metal roofs, glass, patches of trees, and water bodies, all packed within a few hundred meters of each other.

So how do we make sense of all that? We have several tools in our toolkit:

• Spectral indices: NDBI highlights impervious surfaces; NDVI reveals urban vegetation; NDWI identifies water bodies. Used together, they separate the major urban land cover types.
• Nighttime lights: Artificial light emissions from cities give us an independent measure of urbanization that correlates with population density and economic output.
• Multi-temporal analysis: Comparing images from different years reveals where cities are growing, which neighborhoods are densifying, and where green space is disappearing.
• Global datasets: Pre-built products like the JRC Global Human Settlement Layer give us ready-to-use urban maps spanning decades.

How NDBI spots concrete from space

The Normalized Difference Built-up Index takes advantage of a neat spectral trick: built-up surfaces (concrete, asphalt) reflect more SWIR energy relative to NIR compared to vegetation. The formula is:

NDBI = (SWIR1 - NIR) / (SWIR1 + NIR)
// For Landsat 8: SWIR1 = SR_B6, NIR = SR_B5

NDBI values range from -1 to +1. Built-up areas typically produce positive values (0.0 to 0.3), while dense vegetation produces negative values. By applying a threshold, we can create a binary urban mask. Let's see it in action:

// Create a binary urban mask from NDBI
var houston = ee.Geometry.Point([-95.37, 29.76]);

var l8 = ee.ImageCollection('LANDSAT/LC08/C02/T1_L2')
  .filterDate('2023-06-01', '2023-08-31')
  .filterBounds(houston)
  .filter(ee.Filter.lt('CLOUD_COVER', 20))
  .median()
  .multiply(0.0000275).add(-0.2);

var ndbi = l8.normalizedDifference(['SR_B6', 'SR_B5']).rename('NDBI');

// Threshold: pixels with NDBI > 0 are classified as built-up
var urbanMask = ndbi.gt(0).selfMask();

Map.addLayer(urbanMask, {palette: ['red']}, 'Built-up Areas');
Map.centerObject(houston, 11);

A threshold of 0 is a solid starting point, but you may need to adjust it for your study area. Here's something to watch out for: arid regions with bare soil can produce false positives because dry soil also reflects strongly in SWIR. Don't worry, we'll fix that in the next section.

Cities that glow in the dark

Here's where it gets really interesting. The VIIRS Day/Night Band sensor captures visible light emitted at night from street lights, buildings, vehicles, and factories. This gives us a completely different view of urbanization compared to daytime spectral indices.

// Load VIIRS monthly nighttime lights
var ntl = ee.ImageCollection('NOAA/VIIRS/DNB/MONTHLY_V1/VCMSLCFG')
  .filterDate('2023-06-01', '2023-06-30')
  .first()
  .select('avg_rad');

// Visualize nighttime lights (radiance in nanoWatts/cm2/sr)
Map.addLayer(ntl, {
  min: 0, max: 60,
  palette: ['black', 'blue', 'purple', 'orange', 'yellow', 'white']
}, 'Nighttime Lights - June 2023');
Map.setCenter(-95.37, 29.76, 8);

Nighttime lights open the door to some fascinating urban analyses:

• Urban footprint delineation: Light emissions define the functional extent of a city, including suburban areas that barely show up in daytime imagery.
• Economic activity estimation: Nighttime light intensity correlates with GDP, which is especially powerful in regions where economic statistics are unreliable.
• Disaster assessment: Comparing pre-event and post-event nighttime lights reveals the extent of power outages after hurricanes, earthquakes, or conflict.

Three indices are better than one

No single index captures the full complexity of urban land cover. Here's the cool part: by combining NDVI, NDBI, and NDWI, we can build a simple land use classification that separates vegetation, built-up areas, and water.

// Combined urban land use classification
var houston = ee.Geometry.Point([-95.37, 29.76]);

var l8 = ee.ImageCollection('LANDSAT/LC08/C02/T1_L2')
  .filterDate('2023-06-01', '2023-08-31')
  .filterBounds(houston)
  .filter(ee.Filter.lt('CLOUD_COVER', 20))
  .median()
  .multiply(0.0000275).add(-0.2);

// Calculate three indices
var ndvi = l8.normalizedDifference(['SR_B5', 'SR_B4']).rename('NDVI');
var ndbi = l8.normalizedDifference(['SR_B6', 'SR_B5']).rename('NDBI');
var ndwi = l8.normalizedDifference(['SR_B3', 'SR_B5']).rename('NDWI');

// Simple classification using index thresholds
// Water: NDWI > 0; Vegetation: NDVI > 0.3; Built-up: NDBI > 0
var water = ndwi.gt(0);
var vegetation = ndvi.gt(0.3).and(water.not());
var builtUp = ndbi.gt(0).and(water.not()).and(vegetation.not());

// Combine into a single classification image
var classification = ee.Image(0)
  .where(water, 1)       // 1 = Water
  .where(vegetation, 2)  // 2 = Vegetation
  .where(builtUp, 3)     // 3 = Built-up
  .selfMask();

Map.addLayer(classification, {
  min: 1, max: 3,
  palette: ['blue', 'green', 'red']
}, 'Land Use Classification');
Map.centerObject(houston, 11);

This threshold-based approach is fast and interpretable. Think of it like sorting laundry: each index acts as a filter that catches a different material. For more nuanced classification, consider supervised classification methods (covered in earlier modules) that can distinguish commercial, residential, and industrial land uses.

Watching cities grow over time

This is a powerful insight: one of the most compelling applications of urban remote sensing is measuring how cities expand. By comparing NDBI or nighttime lights between two dates, we can map exactly where new development has popped up.

// Compare nighttime lights: 2014 vs 2023
var ntl2014 = ee.ImageCollection('NOAA/VIIRS/DNB/MONTHLY_V1/VCMSLCFG')
  .filterDate('2014-06-01', '2014-06-30')
  .first().select('avg_rad');

var ntl2023 = ee.ImageCollection('NOAA/VIIRS/DNB/MONTHLY_V1/VCMSLCFG')
  .filterDate('2023-06-01', '2023-06-30')
  .first().select('avg_rad');

// Calculate change
var ntlChange = ntl2023.subtract(ntl2014).rename('NTL_Change');

// Visualize: red = brighter (growth), blue = dimmer
Map.addLayer(ntlChange, {
  min: -10, max: 30,
  palette: ['blue', 'white', 'orange', 'red']
}, 'Nighttime Light Change (2014-2023)');
Map.setCenter(-95.37, 29.76, 9);

// Calculate urban area using pixel area
var roi = ee.Geometry.Rectangle([-95.8, 29.5, -95.0, 30.0]);
var urban2023 = ntl2023.gt(5).selfMask(); // Threshold for "urban"

var urbanArea = urban2023.multiply(ee.Image.pixelArea()).reduceRegion({
  reducer: ee.Reducer.sum(),
  geometry: roi,
  scale: 500,
  maxPixels: 1e9
});
print('Urban area (sq m):', urbanArea);

Next Steps

• Lab 26: Urban Analysis - Map and quantify urban expansion hands-on