Resampling: NDI calculation and Analytical download in EO Browser

[alessio.gentile]

Hello everyone,

I would like to have more info about:

• Sentinel-2 L2A bands resampling for the Normalised Difference Index (e.g., NDSI, NDVI) calculation within EO Browser.
• Resampling method used when a .TIFF is downloaded using the “Analytical” download in EO Browser.

Thank you for your help.

[chung.horng (Chung)]

Hi @alessio.gentile ,

The default setting for upsampling is [BICUBIC](https://en.wikipedia.org/wiki/Bicubic_interpolation) and for downsampling is [NEAREST](https://en.wikipedia.org/wiki/Nearest-neighbor_interpolation).

If the default setting does not fit your requirement, you could also change it in settings (Fig 1 & Fig 2).

Fig 1

Fig 2

[alessio.gentile]

Hi @chung.horng,

thank you very much for your reply, I have learned something new!

Just another info: if I calculate a NDI using Sentinel-2 L2A bands, for example using B3 (10 m resolution ) and B11 (20 m resolution ), what is the resolution (in meters) of the output? In other words, how does the built-in function that calculates the NDI in EO Browser manage calculations using bands with different resolution?

Thank you for your attention and availability.

[chung.horng (Chung)]

Hi @alessio.gentile ,

Depends on the zoom-level you could download it at different resolution (Fig 1). If you download NDSI in 10 m resolution, B11 is resampled to 10 m and vice versa.

Fig 1

[alessio.gentile]

Thank you @chung.horng,

the only issue is that, when I add an Area of Interest (AOI), I can only use the EPSG: 4326 (WGS84) in which I have not a projected resolution, but a resolution in degrees. In my case, the downloaded NDSI has lat.: 0.0001261 deg/px long.: 0.0001798 deg/px. In such case, how are the bands resampled for NDSI calculation?

Thanks for your patience.

[chung.horng (Chung)]

Hi @alessio.gentile ,

In this case it is resampled to what it states: lat.: 0.0001261 deg/px long.: 0.0001798 deg/px. You could convert it to meter using haversine formula but it is still an approximate.

[alessio.gentile]

Thank you very much @chung.horng,

In conclusion, the NDI function of EO-Browser resamples both bands according to the output resolution defined by the user (depending on zoom level or AOI) and, in case of upsampling, it uses the bicubic interpolation, while for downsampling it uses the nearest-neighbor interpolation. Is it correct?

[chung.horng (Chung)]

Yes, that’s correct.

If you would like to have more control on the request, I’d recommend using Requests Builder which allows you have full control of the request with a user-friendly GUI.