GEE collection reductor

Question

I can't find out why either maxPixels and either bestEffort works to reduce and image collection like this below. I always get too many pixels.

// Load and filter the Sentinel-2 image collection.
var collection = ee.ImageCollection('COPERNICUS/S2_SR').select('B4', 'B8')
    .filterDate('2019-05-01', '2019-09-30');

var geom = ee.Geometry.Point(-122.08384, 37.42503).buffer(50000);

// Create and print the chart.
//print(ui.Chart.image.series(collection, geom, ee.Reducer.mean(), 20 ));
//print(ui.Chart.image.series(collection, geom, ee.Reducer.median(), 50 ));
//print(ui.Chart.image.series(collection, geom, ee.Reducer.min(), 50));

var mean = ee.ImageCollection('COPERNICUS/S2_SR').select('B4', 'B8')
  .mean()
  .reduceRegion({
    reducer: 'mean',
    geometry: ee.Geometry.Point([-122.27, 37.87]).buffer(1000),
    scale: 1,
    bestEffort: true,
    maxPixels: 1e9
});

print(mean);

print(ui.Chart.image.series(collection, geom, ee.Reducer.mean(), 20 ));

Do you have any ideas?

Kuik · Answer 1 · 2020-02-24T18:59:11.460

There are a couple of things why your task not completes. The task that actual fails is the presentation of the graph. The ui.Chart.series does not have the option to set a bestEffort or maxPixels, thus we first have to use reduceRegion, and then print a chart of the resulting featurecollection.

First, note that you do not filter your collection by region. As such, your are trying to print a chart of thousands of images which are not in your region of interest. Thus we first filter the collection by date and region.

// define geometry
var geom = ee.Geometry.Point(-122.08384, 37.42503).buffer(50000);
// Load and filter the Sentinel-2 image collection.
var collection = ee.ImageCollection('COPERNICUS/S2_SR').select('B4', 'B8')
    .filterDate('2019-05-01', '2019-09-30')
    // Don't forget filtering the image by region
    .filterBounds(geom)

Then note that we use the var collection in our following operations. We can get a feature for every mean B4 and B8 value by mapping over the image collection and applying reduceRegion on every image.

var meanVals = ee.FeatureCollection(collection.map(function(image){
  var mean = image.reduceRegion({ reducer: 'mean',
    geometry: geom,
    scale: 10, // native scale of the bands
    // bestEffort: true, // set bestEffort if run out of memory
    // tileScale: 4, // set higher to increase speed
    maxPixels: 1e15});
  return ee.Feature(null, {meanB4: mean.get('B4'), meanB8: mean.get('B8')
  }).set(image.toDictionary(image.propertyNames())); // copy the image properties
  }));

If we set the maxPixels high enough, we are able to print the chart. Note that you can increase the scale or tileScale and enable the bestEffort arguments to increase speed of the calculations. Make sure you read the documentation about this to understand the differences. Also, you have the possibility to export the table (to assets or drive). See the link below.

print(ui.Chart.feature.byFeature(meanVals, 'system:time_start', ['meanB4','meanB8']));

Additional note: You will see that you get multiple values per day. That is due to the data distribution of the ESA. See for example here or here to first aggregate images taken at a similar day.

link code

score 0 · Answer 2 · answered Feb 24 '20 at 13:27

0

You have set scale to 1. That means, that Earth Engine is calculating at a scale of 1 meter per pixel.

For Sentinel Data this resolution is way too high and is also the reason for your too many pixels error. Set scale: 30 and see if it runs. If it doesn't try maxPixels: 1e11.

That should definitely work then. If it still isn't working you can export the result without loss of accuracy.

answered Feb 24 '20 at 13:27

JonasV

3,754
8
24

Thanks TomazicM for the answer! I tried all of the mentioned methods, but neither of them worked unfortunately. If I set scale to 40 instead of 20 in the last line, it works but than I loose the high-resolution which I would like to keep. But with 155.178.686 points the graph does not show up. Even if I set the Maxpixel to 255.178.686 it does not work. What am I doing wrong? – Tamás Feb 24 '20 at 13:35
@Tamás You're not doing anything wrong. In the browser Earth Engine simply isn't going to calculate anything with that many pixels. The solution for your problem would be to calculate your data and then export it. When you're exporting the data Earth Engine is much more lenient with how many pixels you can reduce. – JonasV Feb 24 '20 at 14:05

GEE collection reductor

2 Answers2