Top
Featured
Learn GIS & Remote Sensing for Free
Free
Crop LAI & Yield Sample Downloads
Crop LAI & Yield Calculators
Wildfire Risk (Sentinel-2) Sample Downloads
Hillshade & Elevation Tiles
Remote Sensing Background
Crop Agriculture
Other

Landsat 8 & Sentinel-2 Low Solar Elevation Imagery

Related Page: Sentinel-2 Lower Solar Elevation (~19⁰) Imagery Downloads

Landsat Scene Center Solar Elevation is 26.2⁰ (Sentinel-2 scene center solar elevation is nearly identical). Chavez (1996) stated further research for the COST Surface Refletance (SR) model is needed for solar elevations less than 35⁰. (IMAGERY DOWNLOADS & MORE INFORMATION ARE INCLUDED BELOW.)

12/18/2016: Central California, USA. Image extent (in kilometers) is 12.7 E-W x 14.6 N-S. Orange is download extent for Landsat 8 and Sentinel-2. Yellow segments are counties; extent is mostly in Madera County; the northwest portion is Merced County; the smallest portion in the north is Mariposa County. Background image is corresponding Sentinel-2 True Color imagery for 12/18/16. Surface is very similar between the Sentinel-2 12/18/16 date and Landsat 8 12/22/16 date, which can be seen when viewing the downloads.


Imagery downloads below include Landsat 8 & Sentinel-2 Top of Atmosphere (TOA) & SR converted with the Frequency 50 atmospheric correction method using relative scatter based on the red band (summed up on the Simplified Landsat 8 & Sentinel-2 SR Steps page) for blue, red, green, NIR, SWIR1, and SWIR2 bands, as well as corresponding Landsat 8 USGS Algorithm Surface Reflectance for the same bands (USGS SR is not a purely DOS method; as a result, surfaces will have different values if SR is subtracted from TOA). There is also a Sentinel-2 true color image than can be download for a larger area. Sentinel-2 imagery is for the same extent four days earlier. Very similar atmospheric conditions exist for both image dates - Landsat 8 and Sentinel-2 green band scatter were virtually identical for the two dates (Landsat 8 and Sentinel-2 green bands have virtually identical band center wavelengths and should have virtually identical scatter for the same atmospheric conditions). Surface reflectance methods are noted in Naming Convention below.

IMPORTANT : By comparing SR values in the downloads below, it can be seen that SR values are somewhat higher (and somewhat higher than they should be) at this low sun angle angle than at higher solar elevations (for this particular imagery, it seems that Sentinel-2 is modestly more accurate than Landsat 8 [though it is just on example]). To download higher solar elevation imagery with more accurate SR, access the following: 61.4⁰ (Tutorial downloads); 59.5⁰ (Rare L8-S2 Comparison downloads).

When comparing Landsat 8 F50 SR (converted with the Frequency 50-based method described here) to USGS Landsat 8 Algorithm SR (not a pure DOS method), F50 SR values are particularly high for the blue band (about .05 to .06 higher for vegetation pixels), but also to a lesser, but meaningful, extent for the green (about .02 to .03 higher for vegetation pixels) and red band (about .01 to .03) higher for vegetation pixels). For green vegetation, green SR can be slightly higher than blue but can also be lower than blue SR (particularly for healthy cropfields with low visible band SR) which is a clear incorrect relationship. A point on the histogram (for all visible bands) representing too many values would need to be selected as scatter in order to lower SR enough to match USGS Landsat 8 Algorithm SR (further providing evidence of problems visible bands have at low solar elevations).

When comparing Sentinel-2 8 F50 SR (converted with the Frequency 50-based method described here) to USGS Landsat 8 Algorithm SR (not a pure DOS method), F50 SR values are particularly high for the blue band (about .03 to .04 higher for vegetation pixels), but also to a lesser, but meaningful, extent for the green (about .01 to .02 higher for vegetation pixels) and red band (about .01 to .02 higher for vegetation pixels). For green vegetation, green SR can be fairly higher than blue but can also be even or very slightly lower than blue SR (particularly for healthy cropfields with low visible band SR) which is a clear incorrect relationship (though not a distinct as the Landsat image above). A point on the histogram (for all visible bands) representing too many values would need to be selected as scatter in order to lower SR enough to match USGS Landsat 8 Algorithm SR (further providing evidence of problems visible bands have at low solar elevations).

Accuracy at low elevation increases as wavelengths increase, so blue is the most erroneous - NIR AND SWIR WAVELENGTHS ARE ACCURATE AT THESE, AND EVEN LOWER, SOLAR ELEVATIONS.  

DOWNLOADS FOLLOW (SEE BELOW FOR NAMING CONVENTION):

* Downloads include blue, red, green, NIR, SWIR1, and SWIR2 bands for Landsat and Sentinel-2. *

Download naming convention is as follows (see extent below): L8 is Landsat 8; p42r34 is Landsat 8 path/row; 122216 is Landsat 8 imaging date of 12/22/2016; TOA is Top of Reflectance (surface reflectance is calculated from it); F50_B4RS_SR is the Frequency 50 minus .008 conversion to surface reflectance method described in the Landsat 8 Atmospheric Correction Tutorial and Sentinel-2 Atmospheric Correction Tutorial, where band 4 (red) is starting scatter; USGS SR is corresponding USGS Landsat 8 Algorithm surface reflectance; S2 is Sentinel-2A; 121816 is Sentinel-2A imaging date of 12/18/2016; T11SKB_20161218T184802_TCI.jp2 is the Sentinel-2 True Color background image (shown in graphic below) for the entire Sentinel-2 scene (much larger than the other image downloads; has the same file name as upon download from the Copernicus website). SR is the same as TOA (no scatter deducted) for SWIR bands.

Satellite imagery converted to surface reflectance for low solar elevation has not been assessed for accuracy as much as for higher elevations. Chavez (1996) stated further research for the COST model is needed for solar elevations less than 35⁰ (solar zenith angles greater than 55⁰). As is explained at the bottom of this page, you can order Landsat already converted to surface reflectance by the USGS (not a purely DOS method) and can download a toolkit to convert Sentinel-2 to surface reflectance. Surface reflectance for the download was calculated based on the QGIS tutorial method, the ArcGIS tutorial method would derived similar values; the only difference being that the starting scatter would be slightly different. Access the L8 & S2 Bands page on the top menu for a satellite comparison.

IF YOU PREFER, YOU CAN ORDER LANDSAT FROM THE USGS ALREADY CONVERTED TO SR; keep in mind that USGS Landsat 8 Algorithm SR is not purely a DOS method (subtracting USGS SR from USGS TOA will result in variable values). To order SR imagery, open Data Sources folder on the left and access EarthExplorer; in Data Sets, open Landsat then access Landsat Collection 1 Level-2 (On-Demand) (that is SR). Click the following links for USGS SR background information and Landsat Tier information (USGS offsite pages that open in new tabs). You need an account to order data (easy and free process). USGS SR is in an integer raster format; divide by 10,000 to convert to surface reflectance units (ratio of reflected to incoming radiation). The USGS will email you a link to download the data when it is ready. 

Sentinel-2 cannot be downloaded in SR format (which is considered Level-2A), but there is a free online toolbox that can be installed that can be used to convert to SR (Level-2A, also known as Bottom of Atmosphere reflectance). The following page can be used to access the toolbox: ESA Sentinel-2 Level 2A Toolbox (offsite ESA page; opens in new tab).

 

References

Chavez, P.S., Jr. 1996. Image-based atmospheric corrections–revisited and improved. Photogrammetric Engineering and Remote Sensing 62(9): pp.1025-1036.