Return to Landsat Surface Reflectance Quality Assessment Overview
During Landsat 4-5 TM and Landsat 7 ETM+ Surface Reflectance generation, LEDAPS assesses the following quality conditions and expresses them using the bands described below. Specific attributes for each band described below are available in the Landsat 4-7 Surface Reflectance (LEDAPS) Product Guide.
LEDAPS quality conditions are expressed as either true or false, and are stored in a bit_packed file named sr_cloud_qa.
- Dark Dense Vegetation (DDV) – LEDAPS determined this pixel contains DDV, which is used to estimate aerosol optical thickness, a critical component to the atmospheric correction routine. The accuracy with which surface reflectance is calculated is dependent on how much DDV is available in a scene.
- Cloud – LEDAPS determined this pixel contains cloud.
- Cloud Shadow – LEDAPS determined this pixel contains cloud shadow.
- Snow – LEDAPS determined this pixel contains snow.
- Land/Water – LEDAPS determined this pixel is either land or water.
- Adjacent Cloud – LEDAPS determined this pixel is adjacent to a cloud pixel.
The table below describes the sr_cloud_qa band’s attributes. Bit pixel and pixel value attributes and interpretation tables are provided in the Landsat 4-7 Surface Reflectance (LEDAPS) Product Guide.
|Bit||Bit Value||Cumulative Sum||Attribute|
|0||1||1||Dark Dense Vegetation (DDV)|
|3||8||15||Adjacent to cloud|
Pixel QA – This alternative quality assessment algorithm, containing cloud, cloud confidence, cloud shadow, snow/ice and water, is generated by the CFMask algorithm and is likely to present more accurate results than its companion LEDAPS bands for cloud, cloud shadow, snow, and water identification. The Pixel QA band’s quality conditions are expressed as a confidence level or as either true or false.
The Radiometric Saturation band's quality conditions are expressed as either true or false and are stored in a bit-packed file, named radsat_qa. This band is a bit-packed representation of which sensor bands were saturated during data capture, yielding unusable data.