Sentinel 2A MSIL2A

snap
#1

Dear ESA Forum members,

I am working with Sentinel-2 satellite imagery and have downloaded a Level-2A product in the MSI2A format. I am trying to determine whether the reflectance values in this product need to be transformed, and if so, how to do so.

The reason I am unsure whether a transformation is needed is because the reflectance values I am working with are reported in large values that are not in the range of 0 to 1. when I read the data using rasterio in Python, the values are reported as large numbers, such as 2000 and 6000.

I have read online that for Sentinel-2 , surface reflectance values need to be divided by 10,000(Quantification_value) to convert the values to the expected range of 0 to 1. However, I am not sure if this is applicable to the product I am working with, or if there are other factors that I need to consider because i have information that sentinel 2A MSIL2A is already corrected for atmospheric effects and is considered to be bottom-of-atmosphere (BOA) reflectance data but why reflectance values are not within the expected range of 0 to 1 . The metadata for this product indicates a Quantification_value of 10000. Should I divide the reflectance values by this factor?

file name that i have S2A_MSIL2A_20190720T100031_N0213_R122_T32SPB_20190720T161032.SAFE

I would appreciate any guidance or advice you can offer .

Best regards ,
Amani

#2

If you are working with SNAP, you do not need to do anything.

#3

I am using Python and the Rasterio library to read bands and obtain a Numpy array of reflectance values.

    with rasterio.open('T32SPB_20190720T100031_B02_10m.jp2', driver='JP2OpenJPEG') as band2:
        blue_band = band2.read(1)

I intend to perform all further analysis using programming language and integrate the results into a machine learning project, without the use of any external tools.
@abruescas

#4 
print(blue_band)

Screenshot from 2023-04-16 20-25-53
to take a look about the returned values

#5

Yes, you should do.
The values are stored as integer values because this is required by the JPEG2000 and the used wavelet transform compression. And that’s regardless of whether the values are BOA or TOA.

#6

thank you ! could you please provide me with a relevant reference that I could include in my study ?

#7

You can find the information in the Sentinel-2 Products Specification Document

On Page 326ff section 3.14.3.1 it is said
The conversion of the DN in BOA reflectance is applied as follows:
BOA = (DN - + BOA_ADD_OFFSET) / QUANTIFICATION_VALUE

#8

i didn’t find the BOA_ADD_OFFSET value in the metadata file

#9

This was introduced with some update. In particular with the migration to Processing Baseline 4.0.0.
Copernicus Sentinel-2: Major Products Upgrade Upcoming - Sentinel Online

For further questions regarding the data format please try asking the ESA support: eosupport@copernicus.esa.int

#10

Thanks for the quick response. , the date of the product that i’m working with is 2019/07 .

I’ll do , thanks

#11

May I please ask one more question?
If I open this image with Snap, will it provide me with the actual values of reflectance without requiring any further transformations or formula applications?

#12

Yes, in SNAP Desktop the converted reflectances will be shown.
If you would use snappy (snapy-python interface) it depends on which methods, you use.

#13

Your help is greatly appreciated, thank you so much !

#14

Dear Amanislh,
as already was mentioned you are using a quite old product with processing baseline N0213. You have only to divide the DN by the quantification value given in the metadata to get the reflectance of the BOA product. You have to account for the radiometric offset for newer products.

bdpg

#15

Thank you for your comment. Could you please provide me with a reference or source that supports the information you mentioned regarding the process of converting the digital number DN to physical reflectance for the specific product that i am working with? It seems that the references mentioned earlier are not applicable to my situation as they include formulas for newer products.
I found a reference that might be suitable for my research study. However, I’m not sure if it is the most appropriate one to use.
Sentinel-2-MSI-L2A-Product-Format-Specifications.pdf (908.2 KB)

May I ask one more question please?
For my study, I need to work with images that were acquired in 2019
and I was wondering if any of the images from this year were processed using the N0213 baseline? also , Are products that haven’t been updated still relevant and capable of producing accurate results, or can the type of processing used have an impact on their usefulness?

I would greatly appreciate any advice or guidance you can provide.
Best Regards ,
Amani

#16

@bdpg Thank you for your comment. Could you please provide me with a reference or source that supports the information you mentioned regarding the process of converting the digital number DN to physical reflectance for the specific product that i am working with? It seems that the references mentioned earlier are not applicable to my situation as they include formulas for newer products.
I found a reference that might be suitable for my research study. However, I’m not sure if it is the most appropriate one to use.
Sentinel-2-MSI-L2A-Product-Format-Specifications.pdf (908.2 KB)

May I ask one more question please?
For my study, I need to work with images that were acquired in 2019
and I was wondering if any of the images from this year were processed using the N0213 baseline? also , Are products that haven’t been updated still relevant and capable of producing accurate results, or can the type of processing used have an impact on their usefulness?

I would greatly appreciate any advice or guidance you can provide.
Best Regards ,
Amani