Currently, I am busy planning a project that will use Landsat time series data. However, I need some advice on a few aspects.
Firstly, I will collect Landsat imagery from 2009 till 2020 April. The imagery data set will be comprised off multiple Landsat sensors (5 7 and 8). I can download these images as part of the Collection 1 Level-1 processed imagery, or Level-2 processed imagery. The main difference being that level-2 data is processed to be surface reflectance values (atmospherically corrected by NASA), whereas Level-1 data can be converted to Top of Atmosphere (TOA) reflectance values.
Question: For a time series analysis using Landsat imagery, which approach is better? Using Level-1 processed data and converting it to TOA reflectance (un-atmospherically corrected), or ordering Level-2 surface reflectance imagery and use as is. To complicate this further, I might include Sentinel data (either TOA or surface reflectance) as well.
Any opinion with scientific references are welcome.