We use L2 atmospherically corrected products and are quite happy with their consistency.

At first we compared a few Sentinel-2 level 2 images to L1 products that we corrected ourselves and arrived at the conclusion that for our applications the "ready to go" L2A is good enough.

Same goes for Venus.

However, the Harmonized Landsat Sentinel-2 dataset was not up to par.

In any case where you are analysing images acquired by different sensors you have to take care of radiometric normalization first, and the "ready to go" products don't always do the job.

If you are interested - my team is finishing the development of transformation functions for Venus and Sentinel-2.

We use L2 atmospherically corrected products and are quite happy with their consistency.

At first we compared a few Sentinel-2 level 2 images to L1 products that we corrected ourselves and arrived at the conclusion that for our applications the "ready to go" L2A is good enough.

Same goes for Venus.

However, the Harmonized Landsat Sentinel-2 dataset was not up to par.

In any case where you are analysing images acquired by different sensors you have to take care of radiometric normalization first, and the "ready to go" products don't always do the job.

If you are interested - my team is finishing the development of transformation functions for Venus and Sentinel-2.

For Landsat and ASTER datasets, I prefer to use the datasets that needs to be corrected. There exists extensive, clear and concise documentation on the algorithms that may perform required corrections.

Also, several proprietary and open source software can handle these corrections satisfactorily.

The Quantum-GIS (QGIS) plugin " Semi-Automatic Classification Plugin for QGIS" can do atmos correction and more for Landsat and Sentinel: https://plus.google.com/u/1/communities/107833394986612468374?cfem=1

I prefer to do the pre-processing myself, then I know what has been done to the data and what can go wrong!

My sir

All band pixels give DN , for convertion its into reflection

At first radimetroc correction , atmospheric correction by ENVI

Second floating bands for NDVI etc.

Offer Rozenstein, Can we get 'ready to go' Level 2 Sentinel 2 which is atmospherically corrected from any source? I would be interested to know.

As you know, USGS provides both Level 2 Landsat data in addition to Analysis Ready Data (ARD only for CONUS). However, as you rightly pointed out, while doing time series analysis employing multiple sensors such as Sentinel-2 and Landsat 8 or even Landsat 8 and Landsat 5 TM/7 ETM+, the differences in spectral bands are not accounted for in Level 2 data. Therefore, there is a need for additional processing such as spectral statistical transformations, radiometric normalizations etc.. In this regard, I would be very interested in following what your team is developing.

Ittai Herrmann, as far as the original question is concerned, I also think Level 2 'ready to go' data works just fine as long as it does not involve multi-sensor temporal continuity applications. I would say, for Applied Earth Science students like me, getting bothered with image processing and calibration takes a lot of valuable time which could otherwise be utilized for making discussions/conclusions/decisions on the final products. That is the reason why USGS came up with ARD data in the first place. However, it is perfectly normal for young researchers and students, in addition to algorithm developing science teams to learn the science of image processing, calibration and analysis by manually performing the pre-processing steps from DN to Radiance and TOA Reflectance to Surface Relectance (BOA) etc. going from Level 0 to Level 1 to Level 2 and so on. I also agree with Sara Salehi 's point that the best way to learn is to do the pre-processing of raw data manually so that one has full control of the process instead of a black box output.

I would, in addition, like to point to some other issues with regard to Cloud Masking and correction. Cirrus Cloud bands (1.38µm) were only introduced in Landsat 8 and Sentinel 2 in addition to MODIS and VIIRS. Previous Landsat sensors such as TM, ETM+, MSS and other multi-spectral sensors did not have this Cirrus band, which makes it rather difficult to pin point cirrus clouds presence in historical satellite imagery without any coincident and contemporary spaceborne sensor with a cirrus band. USGS does not provide any operational product to correct Cirrus effects in Landsat yet. So the question that is still puzzling me is how do you deal with correcting Cirrus clouds and their effects that are present (or not?), for example, in Landsat TM, ETM+, MSS but that are not seen in the visible range of the spectrum?

I think that it depends on the aim and approach of the study. For instance, for time series analysis, I prefer to preprocess the data by myself. Then is possible to define the parameters of correction (e.g., atmospheric correction; haze/sky conditions) based on expert knowledge/ground truth data for my study area.

It depends of your study. The reason is that you can have the control or not with the corrections. If you do not have a lot of experience in remote sensing I consider you should use them (Ready to go).

For Landsat-8 sensor, USGS provides with pre-processed data after you send a request with scene details.

I have done pre-processing i.e. atmospheric correction on L1T products and then compared with the corrected products received from USGS and found both files to be exactly same.

Both ways are good.

it depends upon the satellite data

1. anyway radiometric correction done by the date center always. we no need do worry about the radiometric correction.

2. atmospheric correction will do when oceanographic application that type of sensors like OCM, MSS should need atmospheric correction

3. Geo-reference should do for all satellite image related to study specification and field specification