Font Size: a A A

Topographic correction of satellite images: Theory and application

Posted on:2003-03-01Degree:Ph.DType:Dissertation
University:Dartmouth CollegeCandidate:Reeder, David HenryFull Text:PDF
GTID:1460390011488595Subject:Geology
Abstract/Summary:
Steep terrain produces a topographic effect in digital satellite data that can make it difficult to distinguish changes in vegetation species or health from changes in illumination geometry. A multiple linear regression analysis of direct solar irradiance, slope, and elevation confirms that these topographic variables have significant leverage on Landsat TM brightness values from a deciduous/mixed forest in northern New Hampshire, USA. A number of standard correction algorithms are described, including the Regression model, Lambertian model, C-correction model, Minnaert model, and SCS model. A semi-empirical version of the SCS model is proposed and computer programs are described that apply these topographic correction algorithms directly from commercially available image analysis software. The semi-empirical methods of topographic normalization are found to be most effective at removing the correlation between brightness value and direct irradiance in deciduous/mixed forest regions, while entirely empirical and non-empirical methods prove insufficient. Successful methods reduce variability between brightness values on light and dark slopes and improve the accuracy and consistency of forest type classifications. A technique is proposed to estimate deciduous canopy percentages using Landsat data that are corrected for topographic effects. The technique produces estimates that are consistent from year to year under varying sun elevation angles and atmospheric conditions. Tests of the estimates' accuracy using modeling of forest canopies based on variable-area relascope surveys are inconclusive and do not confirm the accuracy of the technique.
Keywords/Search Tags:Topographic, Model, Correction, Forest
Related items