Remove duplicates from an Image Collection in Google Earth Engine

Question

I've been working with Sentinel-2 level 2A Image Collection in Google Earth Engine (GEE). I noted that on some dates, for the same pixel, there are 2 or more images. They were taken on the same day and the same hour (sometimes the difference is 1 second) though the reflectance values are different.

For instance, adding a point (pt) of coordinates 11.65, 37.34 and checking scenes of the 2021-10-02, it appears to be two images on the same day. Reflectance values are different for all bands. if you want to try, you just need to write the following code in GEE

var pt=ee.Geometry.Point(11.65, 37.34);
var start='2020-02-09';
var end='2020-02-11';
var sentinel_col=ee.ImageCollection("COPERNICUS/S2_SR")
                       .filterBounds(pt)
                       .filterDate(start,end);
print(sentinel_col)

Now my questions:

1. Why does this happen?
2. Is there a way to remove duplicates from the collection by means of GEE? Of course over a time span it occurs multiple times for different dates. I want to remove all the duplicates or maybe merging the scenes with the same date and getting averaged values for the bands.

Answer 1

They are not "duplicates". They are images with different Military Grid Reference System. It is referred in Sentinel 2 metadata as 'MGRS_TILE' property. In your case, you have two different MGRS TILEs: 32SQG and 33STB. So, following code is filtering by '32SQG' 'MGRS_TILE'.

var pt=ee.Geometry.Point(11.65, 37.34);
var start='2020-02-09';
var end='2020-02-11';
var sentinel_col=ee.ImageCollection("COPERNICUS/S2_SR")
                       .filterMetadata('MGRS_TILE', 'equals', '32SQG')
                       .filterBounds(pt)
                       .filterDate(start,end);
print(sentinel_col);

After running above code in GEE code editor, you can corroborate that is printed only one image in that Image Collection. By using '33STB' instead '32SQG', you can filter another Image Collection.

By using following code, it can be observed that each image, in its respective Image collection, corresponds to different monitoring areas.