South Dakota State University
A Validation and Update of Landsat Sensor Calibration Using Pseudo-invariant Sites and Dark Targets
Abstract
dc:description.abstractThe primary objective of this project was to consistently calibrate the entire Landsat series sensors to a common radiometric scale by establishing cross-calibration connections from the Landsat-8 OLI back to Landsat-1 MSS. First, the lifetime radiometric stability of all Landsat sensors was studied using bright-PICS, i.e. Sonoran Desert, and a new class of dark PICS, i.e. Crater Lake and Lake Tahoe. From the responses of these sensors, it was found that for dark targets there exist negative radiance values for green and NIR-1 bands of Landsat-1 and Landsat-2 MSS sensors indicating the need for re-calibration. Next, a consistent cross-calibration of all Landsat sensors was achieved using coincident/near-coincident scene pairs acquired during periods of spatial and temporal overlap of these sensors. For the whole archive calibration, initially all the MSS sensors (Landsat1-4) were mapped into Landsat-5 MSS and then Landsat-5 MSS was cross-calibrated to Landsat-5 TM, Landsat-5 TM was cross-calibrated to Landsat-7 ETM+, and finally Landsat-7 ETM+ was cross-calibrated to Landsat-8 OLI. The mismatching spectral filter responses for different sensors were addressed by deriving the scene specific spectral band adjustment factors prior to the cross-calibration. For determining the significance of the bias and gain terms derived from calibration, a t-test was performed with the null hypothesis being gain equals one and bias equals zero at the confidence interval of 99%. From the final calibration equation, it was found that for band 2 and 3 of Landsat-1 and -2 both bias and gain were significant. The effectiveness of these cross-calibration results were discussed by showing a significant improvement in the observed inconsistency of the absolute calibration of all Landsat sensors for both bright, as well as, dark targets. For the bright target, the uncertainty was reduced from 3.2% to 3.0% for the red band and from 4.5% to 3.6% for the NIR band. Similarly, for the dark targets, the uncertainties were reduced from 83% to less than 23% (high because of lower signal strength) for all the bands.
Degree
thesis:*- Name thesis:degree_name
- Master of Science (MS)
- Level thesis:degree_level
- Thesis - University Access Only
- Discipline thesis:degree_discipline
- Electrical Engineering and Computer Science
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Shrestha, Ashish
- Contributors dc:contributor
-
- Dennis Helder
Rights
dc:rights- Language dc:language
- en
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://openprairie.sdstate.edu/etd/1976