Mosaic of S2 10m Subsets (B2, B3, B4) produces 300m output for incorrect footprint

Really having trouble understanding this one. We identified this issue as our gpt cli mosaic process (sill cannot get the gpt graph for this to work) was failing, so recreated directly in SNAP. Note, this does not occur with most of our other processes, but this date has been troublesome (and a few others, so likely the same underlying issue).

Using two adjacent S2 10m products that were previously subsetted with a gpt graph (not in SNAP, see the explanation after the image below), the resulting mosaic appears to be 300m and does not cover the entirety of the 2 subsets, nor the entirety of our AOI. Product 5 below is the mosaic (number reused from a previous mosaic task). Product 6 is is from our successful gpt graph subsetted version of S2A_MSIL1C_20220327T024551_N0400_R132_T50RNV_20220327T052037.SAFE, and product 7 is from S2A_MSIL1C_20220327T024551_N0400_R132_T50SNA_20220327T052037.SAFE

Furthermore, the two original S2 L1C products once subsetted to our AOI in SNAP using identical Geo Coordinates for bands B2, B3 and B4, cannot be displayed as RGB images, nor can they be mosaiced either.


When trying to display the RGB for either subset displays the error: “Referenced rasters are not of the same size”. When trying to mosaic, the error “Product ‘sourceProduct2’ contains bands of different sizes.” is displayed, which ought to be the case with both sourceProducts.

Product 1 properties:

Product 2 properties:

The problem might be that by default 0.05 degree is used as output resolution. You can change it to a higher resolution.

Marco. Where would the output resolution be specified? There is no parameter for this in the mosaicing options, nor in the individual bands of the resulting mosaic. When I inspect the band properties of the individual bands in the mosaic components, the Raster size is 5727 x 2422, however in the mosaiced image the Raster size is 12 x 6.

On the tab ‘Map Projection Definition’ you should see this:

Is this not shown?

My bad (eyes). Yes I see it. So changing it to 10 x 10 will give 10m resolution (and 20 x 20, 300 x 300, etc). Correct?

Which explains why this always works in our graph, as these values are discretely specified.

However, setting this in SNAP throws this error now for a Rayleigh corrected subset of two S2 tiles:


You made a common mistake. You have specified 20 degrees as pixel size. You can either set 0.00018 as degree or you can switch to e.g., UTM. Then you can enter the pixel size in meter.
The unit of the pixel size is always the unit of selected target CRS.

1 Like