Back to search

South Dakota State University

Development of Automatic Cloud Mask Over PICS to Support Landsat Calibration

Abstract

dc:description.abstract

Clouds are visible and exist above the surface of the earth and cast their shadows in the direction of the sun-zenith and azimuthal angles. In satellite remote sensing, clouds obstruct the viewing of all radiometric spectral bands and are the source of many errors. The brightening effects of clouds and darkening effects of cloud shadows influence many data analysis techniques; these effects generate noise and also cause the satellite sensors to provide false target surface data. Therefore, the detection and masking of clouds is vital in the remote sensing field. Post-flight vicarious calibration requires cloud-free scenes from pseudo invariant calibration site (PICS) to retrieve the actual surface reflectance and calibrate the satellite sensors accurately. However, some cloud mask techniques developed previously are site specific and have problems detecting thin clouds and cloud shadows; some are global cloud cover algorithms and do not work properly with all sites. Furthermore, no cloud mask algorithm has yet been developed specifically for pseudo-invariant calibration sites (PICS). Thus, the overall goal of this thesis is to develop an automatic cloud mask algorithm over PICS to support Landsat calibration. This algorithm is developed based on the brightness temperature from the thermal band and radiometric reflectance profiles through the optical spectral bands of Libya-4 ETM+ scenes. There are a total of four different tests that work in a hierarchy to detect the clouds and their shadows accurately. The main effort has been to dynamically calculate an optimal scene-based threshold value for each and every test in order to detect all possible clouds and their shadows from cloudy Landsat scenes over PICS. The algorithm`s outstanding performance in detecting the clouds and their shadows over Libya-4 PICS scenes from Landsat-5, 7, and 8 has been visually and logically verified. Although this algorithm correctly identifies all of the cloudy sectors, an excess of cloud-free pixels, which have radiometric reflectance profiles similar to clouds and their shadow pixels, are also detected. The false detection rate and change in mean reflectance values are statistically analyzed after applying the algorithm to 140 cloud-free Libya-4, ETM+ scenes and have been found to be less than 5% and 1% respectively. The uncertainties of the temporal-trend of the top-of-atmospheric reflectance are less than 3% through all optical spectral bands of Libya-4 archived imagery from Landsat-5, 7, and 8 respectively. In addition, an empirical linear bidirectional reflectance distribution function (BRDF) correction model has been applied and the uncertainties have been reduced to less than 2.5% resulting in more stable top-of-atmospheric (TOA) reflectance trends. The performance of cloud masking using this algorithm is also analyzed with some additional PICS from the Saharan desert and North America, where all possible clouds and their shadows are detected accurately. The temporal trend uncertainties are found to be less than 3% in all spectral bands; except the SWIR1 and SWIR2 bands which have 4-5% uncertainties from the Sonora and the Algodunes Dunes of North America sites.

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
  • Koirala, Sudip
Contributors dc:contributor
  • Dennis Helder

Subjects

dc:subject × 1

Rights

dc:rights
Statement dc:rights
  • <p>In Copyright - Non-Commercial Use Permitted<br /><a href="http://rightsstatements.org/vocab/InC-NC/1.0/">http://rightsstatements.org/vocab/InC-NC/1.0/</a></p>
Language dc:language
en

Identifiers

dc:identifier.*
Repository record dc:identifier
https://openprairie.sdstate.edu/etd/1566

Chain of custody

source
Harvested from
South Dakota State University
Base URL
openprairie.sdstate.edu/do/oai/
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
citation

Koirala, Sudip. Development of Automatic Cloud Mask Over PICS to Support Landsat Calibration. Thesis - University Access Only thesis, https://openprairie.sdstate.edu/etd/1566