Discrepancies in Sentinel-2 Satellite Data and Request for NDVI Workflow Implementation Guidance

[Dhruv-Mishra]

I am currently working with Sentinel-2 satellite data that I have obtained through the Google Earth Engine. However, I’ve noticed some discrepancies when comparing this data with the data retrieved from the Microsoft Planetary Computer. Despite ensuring that the parameters for region, date, time, and collection are identical in both datasets, there are significant differences in the values.

I’m seeking assistance to understand these discrepancies. Could these differences be due to variations in the data processing or retrieval methods between the Google Earth Engine and the Microsoft Planetary Computer?

In addition, I’m looking for guidance on implementing a workflow that fetches data from the Planetary Computer for a specific region, calculates the Normalized Difference Vegetation Index (NDVI), and applies a cloud mask. While I have some experience in Development, my exposure to remote sensing is relatively minimal. Therefore, any recommendations for a beginner-friendly approach would be greatly appreciated.

[TomAugspurger]

I'm not sure exactly what Earth Engine hosts, so it's a bit hard to give comparisons.

For the Planetary Computer, as mentioned in https://planetarycomputer.microsoft.com/dataset/sentinel-2-l2a we take the L1C data and use sen2cor and GDAL to produce the L2A Cloud Optimized GeoTIFFs we host.

A couple things you might want to check:

1. What baseline is being used? ESA changed some things, and different providers are doing different things. https://planetarycomputer.microsoft.com/dataset/sentinel-2-l2a#Baseline-Change has some background.
2. Make sure you're using the same units. https://docs.sentinel-hub.com/api/latest/data/sentinel-2-l2a/#units has some background on this.

[ghidalgo3]

Closed due to inactivity, feel free to reopen if you would like to continue this discussion.