Skip to main content
U.S. flag

An official website of the United States government

Estimation of a cover-type change matrix from error-prone data

Informally Refereed

Abstract

Coregistration and classification errors seriously compromise per-pixel estimates of land cover change. A more robust estimation of change is proposed in which adjacent pixels are grouped into 3x3 clusters and treated as a unit of observation. A complete change matrix is recovered in a two-step process. The diagonal elements of a change matrix are recovered from estimates of the temporal correlations of cover-type frequencies and an estimate of the odds-ratio of no change. Off-diagonal elements are recovered from least-squares solutions to a set of constrained linear equations. The proposed method produced less biased estimates on three of five sites when the average coregistration error was in excess of 0.3 to 0.7 pixels and on four of five sites if classification accuracy is below 0.9.

Parent Publication

Citation

Magnussen, Steen. 2009. Estimation of a cover-type change matrix from error-prone data. In: McRoberts, Ronald E.; Reams, Gregory A.; Van Deusen, Paul C.; McWilliams, William H., eds. Proceedings of the eighth annual forest inventory and analysis symposium; 2006 October 16-19; Monterey, CA. Gen. Tech. Report WO-79. Washington, DC: U.S. Department of Agriculture, Forest Service. 273-277.
https://www.fs.usda.gov/research/treesearch/17315