JRC Global Surface Water

Quick Win: Watching the Aral Sea Disappear

The Aral Sea was once the world's fourth-largest lake. Today, it's a shadow of itself, one of the most dramatic environmental catastrophes in modern history. Let's visualize its water occurrence to see the full scale of what happened.

// Visualize water occurrence for the Aral Sea
var aral = ee.Geometry.Point([59.0, 44.5]);

// Load the JRC Global Surface Water dataset
var gsw = ee.Image('JRC/GSW1_4/GlobalSurfaceWater');

// Select the occurrence band
var occurrence = gsw.select('occurrence');

// Visualize with a blue gradient palette
Map.centerObject(aral, 7);
Map.addLayer(occurrence, {
  min: 0,
  max: 100,
  palette: ['ffffff', 'ffbbbb', 'ff0000', '0000ff']
}, 'Water Occurrence (%)');

print('JRC GSW band names:', gsw.bandNames());

What's Inside the JRC Dataset?

Here's the cool part: the JRC team analyzed over 4.7 million Landsat satellite images spanning 1984 to the present. For every 30-meter pixel on Earth, they recorded when water was present and when it was absent. From that massive effort, they produced several derived layers, each capturing a different facet of water dynamics.

The dataset is stored as a single ee.Image with multiple bands. We don't need to filter dates or create composites because the JRC team already did that heavy lifting across the entire Landsat archive. Nice, right?

// Load the dataset and explore its bands
var gsw = ee.Image('JRC/GSW1_4/GlobalSurfaceWater');
print('Available bands:', gsw.bandNames());
// Bands: occurrence, change_abs, change_norm, seasonality,
//        recurrence, transition, max_extent

Reading the Occurrence Layer

The occurrence band is the most intuitive layer. Think of it like a report card for each pixel: "Out of all the cloud-free Landsat scenes we looked at, what percentage of the time was this pixel covered by water?"

Permanent lakes and rivers will score close to 100%. Seasonally flooded areas will have intermediate values. Areas that were once water but have since dried up? Low values.

// Map occurrence with descriptive classes
var occurrence = gsw.select('occurrence');

// Permanent water: occurrence > 75%
var permanent = occurrence.gte(75);

// Seasonal water: occurrence between 25% and 75%
var seasonal = occurrence.gte(25).and(occurrence.lt(75));

// Ephemeral water: occurrence between 1% and 25%
var ephemeral = occurrence.gte(1).and(occurrence.lt(25));

Map.addLayer(permanent.selfMask(), {palette: ['0000CC']}, 'Permanent Water');
Map.addLayer(seasonal.selfMask(), {palette: ['00CCFF']}, 'Seasonal Water');
Map.addLayer(ephemeral.selfMask(), {palette: ['FFAAAA']}, 'Ephemeral Water');

Where Has Water Come and Gone?

The transition band is like a before-and-after snapshot for every pixel. It classifies each pixel based on how its water status changed between the first and last observation periods. This is the fastest way to see where water has been permanently lost, where new water has appeared, and where seasonal patterns have shifted.

// Visualize water transitions
var transition = gsw.select('transition');

// Official JRC transition class palette
Map.addLayer(transition, {
  min: 0,
  max: 10,
  palette: [
    'ffffff', // 0: No change
    '0000ff', // 1: Permanent
    '22b14c', // 2: New permanent
    'd1102d', // 3: Lost permanent
    '99d9ea', // 4: Seasonal
    'b5e61d', // 5: New seasonal
    'e6a1aa', // 6: Lost seasonal
    'ff7f27', // 7: Seasonal to permanent
    'ffc90e', // 8: Permanent to seasonal
    '7f7f7f', // 9: Ephemeral permanent
    'c3c3c3'  // 10: Ephemeral seasonal
  ]
}, 'Water Transitions');

Here are the key transition classes to watch for:

• Permanent (1): Water that has been there throughout the entire period. Steady and reliable.
• New permanent (2): Areas that became permanently covered by water (think: new reservoirs).
• Lost permanent (3): Areas that were once permanent water but have dried up. This is the Aral Sea story.
• Seasonal (4): Areas that consistently show seasonal water patterns.
• Permanent to seasonal (8): Previously permanent water that now appears only part of the year. A warning sign.

The Rhythm of Water: Seasonality

The seasonality band tells us how many months per year a pixel is typically covered by water. A value of 12 means year-round water. A value of 3 means the pixel is wet for roughly one quarter of the year.

// Visualize water seasonality
var seasonality = gsw.select('seasonality');

Map.addLayer(seasonality, {
  min: 1,
  max: 12,
  palette: ['FFE4B5', 'FFA500', 'FF4500', '8B0000', '4169E1', '00008B']
}, 'Seasonality (months/year)');

print('Seasonality range:', seasonality.reduceRegion({
  reducer: ee.Reducer.minMax(),
  geometry: aral.buffer(200000),
  scale: 30,
  maxPixels: 1e9
}));

Seasonality maps are especially useful for understanding monsoon flooding, snowmelt cycles, and irrigation patterns. In arid regions, you'll see reservoirs with high seasonality values (water year-round) surrounded by areas with low values (wet only after seasonal rains). It's like reading the heartbeat of a landscape.

Then vs. Now: Tracking Water Loss

Here's where it gets interesting. The max_extent band records every pixel that was ever detected as water. By comparing this historical maximum with a current water mask, we can measure exactly how much water has been lost. Let's do this for the Aral Sea.

// Compare historical maximum with current water
var maxExtent = gsw.select('max_extent');
var currentWater = occurrence.gte(75); // Pixels currently permanent water

// Areas of water loss: was water historically but not currently
var waterLoss = maxExtent.eq(1).and(currentWater.eq(0));

Map.addLayer(maxExtent.selfMask(), {palette: ['AADDFF']}, 'Historical Maximum Extent');
Map.addLayer(currentWater.selfMask(), {palette: ['0000CC']}, 'Current Permanent Water');
Map.addLayer(waterLoss.selfMask(), {palette: ['FF0000']}, 'Water Loss');

// Calculate lost area in km²
var lostArea = waterLoss.multiply(ee.Image.pixelArea()).reduceRegion({
  reducer: ee.Reducer.sum(),
  geometry: aral.buffer(200000),
  scale: 30,
  maxPixels: 1e9
});
print('Water loss area (km²):', ee.Number(lostArea.get('max_extent')).divide(1e6));

Next Steps

• Lab 23: Water Mapping - Apply MNDWI and JRC GSW to map and compare water extent at Lake Mead.