LANDSAT 8 & SENTINEL-2 STARTING SCATTER SELECTION METHODS FOR ARCGIS & QGIS

Landsat 8 ArcGIS & QGIS Scatter Method

Landsat 8 Attribute Table Frequency 50 and QGIS Histogram Bin 5 Scatter Methods

The base of the histogram (which is usually not the very lowest value) has been used to establish scatter (Chavez, 1988). Click here to view examples of Landsat TM histogram scatter selection from Chavez (keep in mind that Landsat 8 has a much longer statistical tail than Landsat TM; this can be seen in the graphics below). Chavez (1996) then calculated the base of the histogram scatter reflectance value to be .01 (as opposed to zero) by deducting .01 from it (this is one-percent dark object reflectance), because of the "fact that very few targets on Earth's surface are absolute black, so an assumed one-percent minimum reflectance is better than zero percent" (Chavez, 1996).

The Frequency 50 & Bin 5 scatter methods for ArcGIS and QGIS (QGIS software can be downloaded from this website here), respectively, are a systematic method for establishing a scatter (dark object) value which can essentially be used as the base of the histogram (is described below). The Frequency 50 & Bin 5 scatter amounts are usually not right at the base of the histogram, but are very close. An advantage of using Frequency 50 & Bin 5 scatter methods for Landsat 8 that similar surface reflectance will be calculated whether using the ArcGIS raster attribute table for the Frequency 50 value or QGIS histogram for the Bin 5 value (as explained in tutorial; QGIS cannot generate a raster attribute table). After the reflectance is calculated, we recommend .008 or .01 being deducted to calculate starting scatter for Landsat 8 band 4 (red) reflectance; relative scatter for other bands can then be computed with the Relative Scatter Calculator or Relative Scatter Table. The .008 deduction is based on research here with Landsat 8 data that shows the average and median difference between the Frequency 50 and Lowest Valid Value reflectance for 34 images throughout the year (10 of 12 months; included a significant amount of a variety of surfaces, including agriculture, mountain, desert, and water) were both .008. This is similar to the one-percent dark object scatter method from Chavez (1996), where the base of the Landsat TM histogram is established (by viewing the histogram) and .01 is deducted from the base of the histogram reflectance to established one-percent (.01) scatter reflectance for the dark object. The methods are described with examples in the Landsat 8 ArcGIS SR Tutorial and Landsat 8 QGIS SR Tutorial.

The Frequency 50 scatter value is visually deciphered from the ArcGIS attribute table and is the digital number (DN) where the frequency is 50 or the closest lower frequency to 50 (see examples below). The red band (Landsat 8, band 4) is recommended as the starting scatter. There may be a small (but negligible) amount of ambiguity when selecting a starting scatter amount with this method - very similar DNs result in virtually the same reflectance. A suitable value will be obvious in nearly all cases; if it is too difficult to select a starting red band value from the attribute table, select a value from the blue band (band 2). Relative scatter should then be applied as described in the tutorials.

The Bin 5 scatter value is from the QGIS histogram (processed as shown in tutorial [necessary to apply the Translate tool to convert zero values to no data]). The Bin 5 scatter method selects the lower value (left side of bin) from the lowest bin with at least a frequency of 5 (QGIS histogram do not represent all values, but are suitable for this application) that does not have a bin to the left of it (lower value bin) that has a frequency less than 5 (keeps value off statistical tail; see second example below).

ArcGIS Attribute Table Frequency 50 Value is 6002; QGIS Histogram Bin 5 Value is 6009.

ArcGIS Attribute Table Frequency 50 Value is 6327; QGIS Histogram Bin 5 Value is 6286.

ArcGIS Attribute Table Frequency 50 Value is 6114; QGIS Histogram Bin 5 Value is 6082.

ArcGIS Attribute Table Frequency 50 Value is 5970; QGIS Histogram Bin 5 Value is 5969.

ArcGIS Attribute Table Frequency 50 Value is 6019; QGIS Histogram Bin 5 Value is 6051.

ArcGIS Attribute Table Frequency 50 Value is 6116; QGIS Histogram Bin 5 Value is 6119.

ArcGIS Attribute Table Frequency 50 Value is 6704; QGIS Histogram Bin 5 Value is 6696.

ArcGIS Attribute Table Frequency 50 Value is 6214; QGIS Histogram Bin 5 Value is 6229.

ArcGIS Attribute Table Frequency 50 Value is 6004; QGIS Histogram Bin 5 Value is 6002.

Sentinel-2 ArcGIS & QGIS

Sentinel-2 Attribute Table Frequency 50 and Base of the Histogram Scatter Methods

WE RECOMMEND THE SENTINEL-2 FREQUENCY 50 VALUE SHOULD HAVE .008 OR .01 DEDUCTED FROM IT TO CALCULATE STARTING SCATTER. BASE OF THE HISTOGRAM REFLECTANCE SHOULD HAVE .01 DEDUCTED. SEE THE TUTORIALS FOR DETAILS.

The graphic below shows actual low portions of Sentinel-2 band 4 (red) attribute tables (the lower values to left represent exceptionally clear atmospheres). Frequency 50 values from left to right are: 179, 147, 145, 188, 236, 223, 295, 287, and 269. The Base of the Histogram could be deciphered, from left to right, as: 173, 143, 141, 181, 232, 215, 286, 278, and 249 (this is a somewhat subjective process, similar values are fine; values should not, however, be selected on a statistical tail). The second and third tables from the left represent extremely clear atmospheres.

Sentinel-2 QGIS Base of Histogram & Lowest Connected Value

Example 1: Base of Histogram, Lowest Connected Value, and Bin 5 Value

After processing the QGIS histogram properly (as shown in Course 2b), the histogram in Example 1 below is about as stretch as one will be on the low end. Below the following histogram image, is another image more zoomed in, as well as the corresponding ArcGIS histogram. The base of the histogram is subjective while the Lowest Connected Value is the lowest (left) side of the lowest bin that does not have a break in bins to the higher (right) side in the histogram. The Lowest Connected Value (LCV) in Example 1 is 287, which can be seen more clearly in the second histogram image. The base of the histogram is somewhat subjective and may be interpreted as a somewhat higher value than the LCV is this case, about 300. The Bin 5 Value (described above) is 312.

The Lowest Connected Value is 287. The Base of the Histogram is (about) 300. The Bin 5 Value is 312.

Example 2: Base of Histogram, Lowest Connected Value, and Bin 5 Value

This example shows the Base of Histogram, Lowest Connected Value, and Bin 5 values are all 295 (it is not very uncommon that these values are all the same).

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.