Contents

(Search or Access Pages Above and Below)

Crop Agriculture

GIS Ag Maps Image-Based DOS Landsat 8 Surface Reflectance (SR) Results & Validation

(Values below are based on the Original Relative Scatter Table using the Lowest Valid Value Scatter Method)

The GIS Ag Maps dark-object subtraction (DOS) (concept originated in Chavez [1988]) Landsat 8 SR method is a simple and fast technique that establishes scatter by specific techniques (uses relative scatter) that are described and can be accessed through the Tutorial. The NDVI values in this article are based on the Original Relative Scatter Lookup Table; the current Revised Table is the same for the red band, while the NIR band has slightly lower scatter values; so the NDVI values would be slightly higher. The following article in Remote Sensing of Environment (Ke et al., 2015) examines and validates GIS Ag Maps DOS SR (page 303 [2.5.2]; labeled Landsat DOS on the graphic below) when applied to NDVI by comparing DOS-based Normalized Difference Vegetation Index (NDVI) values (as well as NDVI values from other reflectance methods and sensors) to in-situ NDVI measurements on a different continent (opens in new tab): Characteristics of Landsat 8 OLI-derived NDVI by comparison with multiple satellite sensors and in-situ observations.

GIS Ag Maps Lowest Valid Value Attribute Table Method is used for Landsat 8 DOS below (explained in article)

Landsat 8 surface reflectance based NDVI comparison results.

In the graphic above (Figure 10 from Ke et al. 2005), which can be accessed articles made accessible above, Landsat DOS (lighter orange) is GIS Ag Maps DOS SR method (as named and described in article), others are as follows: LED is an in-situ NDVI measurement based on light emitting diode instrumentation. Landsat TOA is Top of Atmosphere reflectance (not surface reflectance); Landsat 6SV is Second Simulation of a Satellite Signal in the Solar Spectrum Vector radiative transfer algorithm surface reflectance; Landsat FLAASH is Fast Line-of-sight Atmospheric Analysis of Hypercubes surface reflectance; MODIS is Moderate Resolution Imaging Spectroradiometer; GOCI is Geostationary Ocean Color Imager.

 

Comparison Between USGS & GIS Ag Maps Landsat 8 Surface Reflectance for Cloud - Free Areas

Related Pages: Landsat 8 SR Tutorial w/ Imagery Download    Relative Scatter Lookup Table 

The GIS Ag Maps DOS SR method uses continuous relative scatter which was suggested in Chavez (1988) and applied in Hollinger (2009). The table below shows a comparison of average surface reflectance and standard deviation for five areas (different dates and locations) that are shown below the table (abbreviation meanings are listed below table). Image-based methods to convert Landsat 8 digital numbers to surface reflectance were published on this website well before Landsat 8 surface reflectance was made available by the USGS. Surface reflectance retrieved by methods here based on a Continuous Relative Scatter Table are compared to USGS surface reflectance below - an adjustment is made to green scatter in the original scatter table in order to make green values closer to USGS Landsat 8 Algorithm surface reflectance (which is not solely DOS-based; abbreviated in table as DOS2 SR). DOS1 SR is the same as DOS2 SR, with the exception that DOS1 SR does not apply the modified green scatter (DOS1 SR is different than DOS SR because DOS1 SR uses TOA for NIR surface reflectance, while DOS SR applies relative scatter for NIR). Converting to surface reflectance independently can be useful if, for example, surface reflectance is not available for a particular reason or scene, or if surface reflectance is needed as soon as possible. See bottom of page for green scatter adjustment methodology.

Average of Green Mask Extent of Five Areas Below

  Landsat 8 USGS SR DOS1 SR DOS2 SR   TOA
  B2 (Blue) 0.0176 0.0207 0.0207   0.0944
  B3 (Green) 0.0420 0.0361 0.0420    0.0792
  B4 (Red) 0.0242 0.0240 0.0240   0.0501
  B5 (NIR)  0.4537 0.4525 0.4525   0.4525
  B6 (SWIR)  0.1774 0.1728 0.1728   0.1728
             
  Landsat 8 Stan. Dev. Stan. Dev. Stan. Dev.   Stan. Dev.
  B2 (Blue) 0.0026 0.0026 0.0026   0.0026
  B3 (Green) 0.0056 0.0047 0.0047   0.0047
  B4 (Red) 0.0049 0.0043 0.0043   0.0043
  B5 (NIR)  0.0540 0.0535 0.0535   0.0535
  B6 (SWIR)  0.0287 0.0273 0.0273   0.0273


Average of Full Extent of Five Areas Below

  Landsat 8 USGS SR DOS1 SR DOS2 SR   TOA
  B2 (Blue) 0.0280 0.0279 0.0279   0.1016
  B3 (Green) 0.0558 0.0467 0.0525    0.0898
  B4 (Red) 0.0442 0.0407 0.0407   0.0668
  B5 (NIR)  0.4051 0.4049 0.4049   0.4049
  B6 (SWIR)  0.2113 0.2051 0.2051   0.2051
             
  Landsat 8 Stan. Dev. Stan. Dev. Stan. Dev.   Stan. Dev.
  B2 (Blue) 0.0156 0.0120 0.0120   0.0120
  B3 (Green) 0.0206 0.0164 0.0164   0.0164
  B4 (Red) 0.0290 0.0247 0.0247   0.0247
  B5 (NIR)  0.0786 0.0771 0.0771   0.0771
  B6 (SWIR)  0.0600 0.0572 0.0572   0.0572

 

USGS SR is surface reflectance based on the USGS Landsat 8 Algorithm that can be ordered and downloaded for free. DOS1 SR uses the Relative Scatter Lookup Table to deduct blue and green scatter based on red starting scatter and uses TOA for NIR reflectance (DOS SR, included in Tutorial also applies relative scatter to the NIR band). DOS2 SR the same as DOS1 SR with adjusted green band scatter (Green 2) in the Relative Scatter Lookup Table. THE METHODOLOGY FOR THE ADJUSTED GREEN BAND VALUE IS SHOWN AT THE BOTTOM OF THIS PAGE. The Relative Scatter Lookup Table simplifies the conversion process for the visible bands and makes the process more reliable. TOA is top of the atmosphere reflectance (not surface reflectance) that can be calculated based on a published USGS equation. TOA reflectance is used for the NIR and SWIR bands as these values are shown to be closer to USGS values. "B" is band. Stan. Dev. is the average standard deviation of the five areas below. (DOS SR is not included here, but is included in the Tutorial; DOS SR is the same as DOS1 SR, except it also has relative scatter deducted from the NIR band.)

________

Individual Areas and Dates for Reflectance Comparison

(in order of clearest atmosphere based on Lowest Valid Value (also know as the DN100 Method) red band reflectance scatter; background image is LandsatLook for particular scene) 

 

Red Band Scatter (.011561)

August 4th, 2015; Northwest Ohio, USA (orange extent; 29.0 km x 13.8 km)

Landsat 8 DOS surface reflectance comparison extent

  Landsat 8 USGS SR DOS1 SR DOS2 SR   TOA
  B2 (Blue) 0.0306 0.0369 0.0369    0.0945
  B3 (Green) 0.0592 0.0566 0.0577    0.0859
  B4 (Red) 0.0503 0.0546 0.0546    0.0661
  B5 (NIR)  0.3662 0.3669 0.3669    0.3669
  B6 (SWIR)  0.2041 0.1973 0.1973    0.1973
             
  Landsat 8 Stan. Dev. Stan. Dev. Stan. Dev.   Stan. Dev.
  B2 (Blue) 0.0178 0.0145 0.0145   0.0145
  B3 (Green) 0.0223 0.0184 0.0184   0.0184
  B4 (Red) 0.0337 0.0296 0.0296   0.0296
  B5 (NIR)  0.0721 0.0711 0.0711   0.0711
  B6 (SWIR)  0.0560 0.0538 0.0538   0.0538

 

Green Mask (more than 99.99% of pixels within gray area have higher green reflectance than blue or red) 

Landsat 8 DOS surface reflectance comparison extent

  Landsat 8 USGS SR DOS1 SR DOS2 SR   TOA
  B2 (Blue) 0.0191 0.0284 0.0284   0.0861
  B3 (Green) 0.0451 0.0456 0.0466    0.0749
  B4 (Red) 0.0283 0.0357 0.0357   0.0472
  B5 (NIR)  0.4256 0.4254 0.4254    0.4254
  B6 (SWIR)  0.1717 0.1663 0.1663    0.1663
             
  Landsat 8 Stan. Dev. Stan. Dev. Stan. Dev.   Stan. Dev.
  B2 (Blue) 0.0037 0.0027 0.0027   0.0027
  B3 (Green) 0.0070 0.0055 0.0055   0.0055
  B4 (Red) 0.0066 0.0055 0.0055   0.0055
  B5 (NIR)  0.0455 0.0450 0.0450   0.0450
  B6 (SWIR)  0.0201 0.0192 0.0192   0.0192

 

Red Band Scatter (.022312)

August 18th, 2014; Easter Iowa, USA (orange extent; 117 km x 194 km)

Landsat 8 DOS surface reflectance comparison extent

  Landsat 8 USGS SR DOS1 SR DOS2 SR   TOA
  B2 (Blue) 0.0206 0.0172 0.0172    0.0867
  B3 (Green) 0.0435 0.0342 0.0388    0.0737
  B4 (Red) 0.0290 0.0255 0.0255    0.0478
  B5 (NIR)  0.4266 0.4259 0.4259    0.4259
  B6 (SWIR)  0.1852 0.1791 0.1791    0.1791
             
  Landsat 8 Stan. Dev. Stan. Dev. Stan. Dev.   Stan. Dev.
  B2 (Blue) 0.0120 0.0105 0.0105   0.0105
  B3 (Green) 0.0164 0.0139 0.0139   0.0139
  B4 (Red) 0.0201 0.0178 0.0178   0.0178
  B5 (NIR)  0.0897 0.0885 0.0885   0.0885
  B6 (SWIR)  0.0394 0.0378 0.0378   0.0378

 

Green Mask (more than 99.99% of pixels within gray area have higher green reflectance than blue or red)  

Landsat 8 DOS surface reflectance comparison extent

  Landsat 8 USGS SR DOS1 SR DOS2 SR   TOA
  B2 (Blue) 0.0159 0.0137 0.0137   0.0832
  B3 (Green) 0.0365 0.0285 0.0332    0.0681
  B4 (Red) 0.0203 0.0181 0.0181    0.0404
  B5 (NIR)  0.4693 0.4680 0.4680    0.4680
  B6 (SWIR)  0.1789 0.1730 0.1730    0.1730
             
  Landsat 8 Stan. Dev. Stan. Dev. Stan. Dev.   Stan. Dev.
  B2 (Blue) 0.0017 0.0026 0.0026   0.0026
  B3 (Green) 0.0038 0.0036 0.0036   0.0036
  B4 (Red) 0.0035 0.0034 0.0034   0.0034
  B5 (NIR)  0.0729 0.0723 0.0723   0.0723
  B6 (SWIR)  0.0317 0.0304 0.0304   0.0304

 

 

Red Band Scatter (.022550)

July 11th, 2013; Southern Illinois & Indiana (crosses border), USA (orange extent; 119.7 km x 35.7 km) 

Landsat 8 DOS surface reflectance comparison extent

  Landsat 8 USGS SR DOS1 SR DOS2 SR   TOA
  B2 (Blue) 0.0376 0.0353 0.0353   0.1051
  B3 (Green) 0.0710 0.0601 0.0647    0.0998
  B4 (Red) 0.0647 0.0599 0.0599   0.0824
  B5 (NIR)  0.3766 0.3777 0.3777    0.3777
  B6 (SWIR)  0.2451 0.2373 0.2373    0.2373
             
  Landsat 8 Stan. Dev. Stan. Dev. Stan. Dev.   Stan. Dev.
  B2 (Blue) 0.0241 0.0185 0.0185   0.0185
  B3 (Green) 0.0332 0.0268 0.0268   0.0268
  B4 (Red) 0.0472 0.0408 0.0408   0.0408
  B5 (NIR)  0.0821 0.0808 0.0808   0.0808
  B6 (SWIR)  0.0894 0.0856 0.0856   0.0856

 

Green Mask (more than 99.99% of pixels within gray area have higher green reflectance than blue or red)  

Landsat 8 DOS surface reflectance comparison extent

  Landsat 8 USGS SR DOS1 SR DOS2 SR   TOA
  B2 (Blue) 0.0165 0.0202 0.0202    0.0900
  B3 (Green) 0.0413 0.0368 0.0414    0.0765
  B4 (Red) 0.0235 0.0248 0.0248    0.0473
  B5 (NIR)  0.4405 0.4407 0.4407    0.4407
  B6 (SWIR)  0.1753 0.1705 0.1705    0.1705
             
  Landsat 8 Stan. Dev. Stan. Dev. Stan. Dev.   Stan. Dev.
  B2 (Blue) 0.0033 0.0037 0.0037   0.0037
  B3 (Green) 0.0074 0.0066 0.0066   0.0066
  B4 (Red) 0.0063 0.0057 0.0057   0.0057
  B5 (NIR)  0.0488 0.0484 0.0484   0.0484
  B6 (SWIR)  0.0304 0.0291 0.0291   0.0291

 

 

Red Band Scatter (.032949)

July 30th, 2014; East Central Illinois, USA (orange extent; 23.7 km x 15.0 km) 

Landsat 8 DOS surface reflectance comparison extent

  Landsat 8 USGS SR DOS1 SR DOS2 SR   TOA
  B2 (Blue) 0.0195 0.0221 0.0221    0.1033
  B3 (Green) 0.0433 0.0356 0.0437    0.0852
  B4 (Red) 0.0258 0.0232 0.0232    0.0562
  B5 (NIR)  0.4981 0.4959 0.4959    0.4959
  B6 (SWIR)  0.1838 0.1794 0.1794    0.1794
             
  Landsat 8 Stan. Dev. Stan. Dev. Stan. Dev.   Stan. Dev.
  B2 (Blue) 0.0102 0.0073 0.0073   0.0073
  B3 (Green) 0.0135 0.0102 0.0102   0.0102
  B4 (Red) 0.0173 0.0141 0.0141   0.0141
  B5 (NIR)  0.0806 0.0798 0.0798   0.0798
  B6 (SWIR)  0.0429 0.0406 0.0406   0.0406

 

Green Mask (more than 99.99% of pixels within gray area have higher green reflectance than blue or red) 

Landsat 8 DOS surface reflectance comparison extent

  Landsat 8 USGS SR DOS1 SR DOS2 SR   TOA
  B2 (Blue) 0.0160 0.0198 0.0198    0.1010
  B3 (Green) 0.0384 0.0320 0.0402    0.0817
  B4 (Red) 0.0195 0.0183 0.0183    0.0512
  B5 (NIR)  0.5224 0.5199 0.5199    0.5199
  B6 (SWIR)  0.1772 0.1732 0.1732    0.1732
             
  Landsat 8 Stan. Dev. Stan. Dev. Stan. Dev.   Stan. Dev.
  B2 (Blue) 0.0011 0.0011 0.0011   0.0011
  B3 (Green) 0.0034 0.0030 0.0030   0.0030
  B4 (Red) 0.0021 0.0018 0.0018   0.0018
  B5 (NIR)  0.0671 0.0670 0.0670   0.0670
  B6 (SWIR)  0.0389 0.0368 0.0368   0.0368

 

 

Red Band Scatter (.041205)

July 10th, 2014; North Central Iowa, USA (orange extent; 68.2km  x 37.1 km)

Landsat 8 DOS surface reflectance comparison extent

  Landsat 8 USGS SR DOS1 SR DOS2 SR   TOA
  B2 (Blue) 0.0320 0.0281 0.0281    0.1185
  B3 (Green) 0.0619 0.0468 0.0576    0.1043
  B4 (Red) 0.0510 0.0402 0.0402   0.0814
  B5 (NIR)  0.3579 0.3581 0.3581    0.3581
  B6 (SWIR)  0.2384 0.2322 0.2322    0.2322
             
  Landsat 8 Stan. Dev. Stan. Dev. Stan. Dev.   Stan. Dev.
  B2 (Blue) 0.0141 0.0094 0.0094   0.0094
  B3 (Green) 0.0175 0.0126 0.0126   0.0126
  B4 (Red) 0.0269 0.0213 0.0213   0.0213
  B5 (NIR)  0.0683 0.0655 0.0655   0.0655
  B6 (SWIR)  0.0724 0.0684 0.0684   0.0684

 

Green Mask (more than 99.99% of pixels within gray area have higher green reflectance than blue or red)  

Landsat 8 DOS surface reflectance comparison extent

  Landsat 8 USGS SR DOS1 SR DOS2 SR   TOA
  B2 (Blue) 0.0208 0.0213 0.0213    0.1116
  B3 (Green) 0.0486 0.0376 0.0485    0.0951
  B4 (Red) 0.0293 0.0233 0.0233    0.0645
  B5 (NIR)  0.4105 0.4086 0.4086    0.4086
  B6 (SWIR)  0.1839 0.1808 0.1808    0.1808
             
  Landsat 8 Stan. Dev. Stan. Dev. Stan. Dev.   Stan. Dev.
  B2 (Blue) 0.0034 0.0030 0.0030   0.0030
  B3 (Green) 0.0062 0.0048 0.0048   0.0048
  B4 (Red) 0.0060 0.0049 0.0049   0.0049
  B5 (NIR)  0.0357 0.0348 0.0348   0.0348
  B6 (SWIR)  0.0225 0.0212 0.0212   0.0212

 

Green Scatter Adjustment for DOS2 SR Methodology

A quite statistically significant correlation exists between the red band starting scatter amount (shown above) and the difference between USGS B2 (green) SR and DOS B2 SR (USGS - PM), as is shown below. In the plot: x-axis is red band scatter; y-axis is the different between USGS SR B2 and DOS SR B2 (USGS - DOS). Red band scatter was input into the linear regression equation below to calculate the amount that was added to DOS B2 to derive DOS2 B2.


Green Scatter Adjustment for DOS2 SR Methodology

 

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.

Hollinger, D. 2009. A GIS-based method to predict county corn yield based on retrieved Landsat reflectance variability in western Ohio. Papers of the Applied Geography Conference 32: pp.281-290.

Ke, Y., Im, J., Lee, J., Gong, H., and Y. Ryu. 2005. Characteristics of Landsat 8 OLI-derived NDVI by comparison with multiple satellite sensors and in-situ observations. Remote Sensing of Environment 164: pp. 298-313.