Filtering Landsat images base on cloud cover over region of interest

Question

My goal is to create an image collection that has 100% free cloud over a small region of interest, for example a lake.

This script filters landsat 8 images based on location and cloud cover:

var nocloudimages =  landsat8.filterBounds(ROI)
                     .filter(ee.Filter.lt('CLOUD_COVER', value))
                     .sort('system:time_start', true)

However, as we know, the 'CLOUD_COVER' accounted for the whole landsat 8 image percentage of cloud, not a particular Region of Interest (ROI) cloud cover.

Is there a method to achieve that?

Answer 1

It's going to be something like this, but you'll need to play with the threshold (10 in this example) to meet your needs. Watch out for ROIs that overlap a scene's footprint, but do not contain any valid pixels. Also watch out for ROIs that are very large or span multiple WRS cells.

var ic = ee.ImageCollection("LANDSAT/LC08/C01/T1_RT_TOA");

// A polygon representing the roi.
var geometry = ee.Geometry.Polygon(
        [[[-121.85897778617999, 37.70881514186375],
          [-121.83975284337708, 37.76202899390253],
          [-121.94137041100134, 37.759857750255144]]]);

var c = ic.filterBounds(geometry);

var withCloudiness = c.map(function(image) {
  var cloud = ee.Algorithms.Landsat.simpleCloudScore(image).select('cloud');
  var cloudiness = cloud.reduceRegion({
    reducer: 'mean', 
    geometry: geometry, 
    scale: 30,
  });
  return image.set(cloudiness);
});

var filteredCollection = withCloudiness.filter(ee.Filter.lt('cloud', 10));
print(filteredCollection);

Answer 2

There is standard example in the code editor that is pretty close to what you want. https://code.earthengine.google.com/1be28850c6c7880d8fcd5f1e0a808986

// SimpleCloudScore, an example of computing a cloud-free composite with L8
// by selecting the least-cloudy pixel from the collection.

// A mapping from a common name to the sensor-specific bands.
var LC8_BANDS = ['B2',   'B3',    'B4',  'B5',  'B6',    'B7',    'B10'];
var STD_NAMES = ['blue', 'green', 'red', 'nir', 'swir1', 'swir2', 'temp'];

// Compute a cloud score.  This expects the input image to have the common
// band names: ["red", "blue", etc], so it can work across sensors.
var cloudScore = function(img) {
  // A helper to apply an expression and linearly rescale the output.
  var rescale = function(img, exp, thresholds) {
    return img.expression(exp, {img: img})
        .subtract(thresholds[0]).divide(thresholds[1] - thresholds[0]);
  };

  // Compute several indicators of cloudyness and take the minimum of them.
  var score = ee.Image(1.0);
  // Clouds are reasonably bright in the blue band.
  score = score.min(rescale(img, 'img.blue', [0.1, 0.3]));

  // Clouds are reasonably bright in all visible bands.
  score = score.min(rescale(img, 'img.red + img.green + img.blue', [0.2, 0.8]));

  // Clouds are reasonably bright in all infrared bands.
  score = score.min(
      rescale(img, 'img.nir + img.swir1 + img.swir2', [0.3, 0.8]));

  // Clouds are reasonably cool in temperature.
  score = score.min(rescale(img, 'img.temp', [300, 290]));

  // However, clouds are not snow.
  var ndsi = img.normalizedDifference(['green', 'swir1']);
  return score.min(rescale(ndsi, 'img', [0.8, 0.6]));
};

// Filter the TOA collection to a time-range and add the cloudscore band.
var collection = ee.ImageCollection('LC8_L1T_TOA')
    .filterDate('2013-05-01', '2013-07-01')
    .map(function(img) {
      // Invert the cloudscore so 1 is least cloudy, and rename the band.
      var score = cloudScore(img.select(LC8_BANDS, STD_NAMES));
      score = ee.Image(1).subtract(score).select([0], ['cloudscore']);
      return img.addBands(score);
    });

// Define visualization parameters for a true color image.
var vizParams = {'bands': ['B4', 'B3', 'B2'], 'max': 0.4, 'gamma': 1.6};
Map.setCenter(-120.24487, 37.52280, 8);
Map.addLayer(collection.qualityMosaic('cloudscore'), vizParams);

Answer 3

I've created the following function to add the cloud+shadow+snow percentage for a particular region as a new image property so you can filter the image collection based on that property. The 'region' in the function refers to your area of interest or aoi.

    // Function to add the cloud, shadow and snow percentge as a new property for each image. It uses 'QA_PIXEL' band
function add_cloudCover(img){
  // Bit 3 - Cloud
  // Bit 4 - Cloud Shadow
  // Bit 5 - Snow
  var cloudsBitMask = 1 << 3;
  var cloudShadowBitMask = 1 << 4;
  var snowBitMask = 1 << 5;
  var qa = img.select('QA_PIXEL').clip(region);
  var unmasked = qa.gt(0)
  var unmasked_area = unmasked.multiply(ee.Image.pixelArea());
  var mask = qa.bitwiseAnd(cloudShadowBitMask).eq(0)
    .and(qa.bitwiseAnd(cloudsBitMask).eq(0)
    .and(qa.bitwiseAnd(snowBitMask).eq(0)));
  var masked_area = mask.multiply(ee.Image.pixelArea());
  var u_area = unmasked_area.reduceRegion({
  reducer: ee.Reducer.sum(),
  geometry: region.geometry(),
  scale: 30,
  maxPixels: 1e10
  });
  var area = masked_area.reduceRegion({
  reducer: ee.Reducer.sum(),
  geometry: region.geometry(),
  scale: 30,
  maxPixels: 1e10
  });
  var landCover = ee.Number(area.get('QA_PIXEL')).multiply(100).divide(ee.Number(u_area.get('QA_PIXEL')));
  var cloudCover = ee.Number(100).subtract(landCover).round();
  return img.set('CLOUD_SHADOW_SNOW_COVER_REGION', cloudCover);
}