IMPORTANT: YOU CAN NOW DOWNLOAD SENTINEL-2 IMAGERY FROM COPERNICUS THAT HAVE ALREADY BEEN CONVERTED TO SURFACE REFLECTANCE (THESE ARE LABLED "L2A"), THOUGH WE HAVE SEEN VALUES THAT SEEM VERY INACCURATE AND RECOMMEND USING THE TUTORIALS ON THIS WEBSITE ALONG WITH SENTINEL-2 1C IMAGERY (PARTICULARLY FOR BANDS SMALLER THAN NIR).
Simplified Landsat 8 & Sentinel-2 Conversion to Surface Reflectance Steps
(for Imagery from 2021 and Earlier)
Image-based atmospheric correction is a well established method of calculating surface reflectance (SR) that ultimately deducts atmospheric scatter reflectance from Top of Atmosphere (TOA) reflectance to retrieve surface reflectance (SR), and should be credited to Chavez (1988 & 1996). Converting Landsat or Sentinel-2 to SR is a quite simple process that can be completed for free (In SR units, .05=5%; .50=50%, 1.0=100%). If you do not have much experience with this process or GIS software, you may want to use a tutorial in the Landsat & Sentinel-2 Conversion to SR Tutorials drop-down menu above - the tutorials are much more detailed. If you have experience or do not want as many details, simply follow the steps below.
Download Landsat 8 & Sentinel-2 from the Free Satellite Imagery Page Accessed on the Top Menu
* If converting to SR for visible bands, try to use imagery with a solar elevation > 45⁰ (for mid-latitudes in the northern hemisphere, this is from about the middle of March to the middle of September for Landsat and Sentinel-2 [opposite in southern hemisphere). NIR and SWIR can be converted to SR at any solar elevation. See below for more information.
LANDSAT 8 CONVERSION TO SURFACE REFLECTANCE (Sentinel-2 is described after)
1) For Landsat 8, first establish atmospheric scatter reflectance for band 4 (red). This amount needs to be deducted from TOA reflectance to calculate surface reflectance (for the red band, the amount will likely be from .01 to .04).
If using ArcGIS or free QGIS (or any other GIS software that can produce a raster attribute table) you can use the Frequency 50 Attribute Table Method to establish the scatter DN (then deduct .008 to establish scatter reflectance, as is described later). If you choose to use the Lowest Valid Value method, you do not deduct .008. (IMPORTANT: We recommend using QGIS Version 3 [can be downloaded from this website here] because it can produce a raster attribute table for atmospheric correction, then applying the same Frequency 50 Attribute Table Method for starting scatter [same as you would do using ArcGIS]. To produce a raster attribute table in QGIS Version 3, open the Processing Toolbox, then open Raster Analysis, then open and run Raster Layer Unique Value Report [the Unique Value Report is and .html page, while the Unique Value Table is a .shp file that will open as a table in the QGIS Layers panel]). Landsat 8 imagery is downloaded in digital number (DNs) format. Surface reflectance is calculated by subtracting atmospheric scatter from Top of Atmosphere (TOA) reflectance. The Frequency 50 DN is converted to reflectance by applying following equation:
([DN x .00002] - 0.1) / Cosine of Solar Zenith
The cosine of the solar zenith can be calculated by inputting the Sun Elevation from the .MTL file into the Cosine of Solar Elevation Calculator on the top menu. DEDUCT .008 FROM THE REFLECTANCE AMOUNT FOR THE FINAL RED BAND SCATTER REFLECTANCE (so after the scatter deduction, the Frequency 50 value will now have a surface reflectance of .008, as opposed to zero). Input the final value into the Landsat 8 Relative Scatter Calculator on the top menu to calculate scatter for other bands.
2) For both ArcGIS and QGIS, convert the entire image scene (of a band that you want to convert to surface reflectance) to Top of Atmosphere (TOA) reflectance by applying the same equation as above with the Raster Calculator:
([DN x .00002] - 0.1) / Cosine of Solar Zenith
3) Deduct a scatter amount from Step 1 from the corresponding band TOA image in Step 2 with the Raster Calculator to calculate surface reflectance. YOU HAVE NOW CALCULATED LANDSAT 8 SURFACE REFLECTANCE.
SENTINEL-2 CONVERSION TO SURFACE REFLECTANCE STEPS
1) For Sentinel-2, first establish atmospheric scatter reflectance for band 4 (red).
If using ArcGIS or free QGIS (or any other GIS software that can produce a raster attribute table) use the Frequency 50 - .008 Attribute Table Method to establish scatter. Sentinel-2 is downloaded in Top of Atmosphere (TOA) integer format. Surface reflectance is calculated by subtracting atmospheric scatter reflectance (in reflectance units) from Top of Atmosphere (TOA) reflectance (in reflectance units).
The Frequency 50 DN is converted to reflectance units simply by dividing it by 10,000 (simpler than Landsat 8). DEDUCT .008 FROM THE REFLECTANCE AMOUNT FOR THE FINAL RED BAND SCATTER REFLECTANCE (so after the scatter deduction, the Frequency 50 value will now have a surface reflectance of .008, as opposed to zero). Input the final value into the Sentinel-2 Relative Scatter Calculator on the top menu to calculate scatter for other bands.
2) For both ArcGIS and QGIS, convert the entire image scene (of a band that you want to convert to surface reflectance) to Top of Atmosphere (TOA) reflectance by simply dividing it by 10,000 (same as above).
3) Deduct a scatter amount from Step 1 from the corresponding band TOA image in Step 2 with the Raster Calculator to calculate surface reflectance. YOU HAVE NOW CALCULATED SENTINEL-2 SURFACE REFLECTANCE.
* IMPORTANT: Retrieved Landsat 8 SR with the methods described on this website have only been tested here for accuracy for solar elevations greater than 50⁰ and less than 30⁰; we have found that visible band, particularly the blue band, SR values become much less accurate (too high) at solar elevations less than 30⁰ (because TOA becomes disproportionatly higher for shorter wavelengths at low solar elevations). We have not completed research for solar elevations from 30 to 50⁰. It is very likely that visible band retrieved SR, particularly the blue band, becomes increasingly less accurate (too high) as solar elevations decrease from about 45⁰ - NIR AND SWIR BANDS REMAIN ACCURATE AT VERY LOW SOLAR ELEVATIONS. Chavez (1996) stated further research for the COST model is needed for solar elevations less than 35⁰ (solar zenith angles greater than 55⁰). CLICK HERE FOR LOW SOLAR ELEVATION IMAGERY SR COMPARISONS & DOWNLOADS.
_______________________________________________________
Image-Based Atmospheric Correction Background & Current Recommended Methods
_______________________________________________________
RECOMMENDED Atmospheric Correction Background Information Pages (all on this website):
2) Chavez (1988) Original Landsat TM Base of Histogram DOS Scatter Reflectance Method
_______________________________________________________
RECOMMENDED Landsat 8 & Sentinel-2 Atmospheric ScatterEstablishment Methods for ArcGIS & QGIS
1) ArcGIS Users for Landsat 8: Lowest Valid Value or Frequency 50 Attribute Table Methods
2) ArcGIS Users for Sentinel-2: Frequency 50 Attribute Table Method
3) QGIS (free software) Users for Landsat 8: Bin 5 Histogram Method
4) QGIS (free software) Users for Sentinel-2: QGIS Base of Histogram Method
_______________________________________________________
SCATTER CALCULATORS & LOOKUP TABLES
Landsat 8 & Sentinel-2 Relative Scatter Calculator & Lookup Table
Landsat 4, 5, and 7 Relative Scatter Table
References
Chavez, P.S., Jr. 1996. Image-based atmospheric corrections–revisited and improved. Photogrammetric Engineering and Remote Sensing 62(9): pp.1025-1036.
Chavez, P.S., Jr. 1988. An improved dark-object subtraction technique for atmospheric scattering correction of multispectral data. Remote Sensing of Environment 24: pp.459-479.