Consequences of land-cover misclassification in models of impervious surface
Photogrammetric Engineering and Remote Sensing journal article
By Gerard McMahon
Full Journal Article (11 pages, 233K)
Abstract
Model estimates of impervious area as a function of landcover
area may be biased and imprecise because of errors in
the land-cover classification. This investigation of the effects
of land-cover misclassification on impervious surface models
that use National Land Cover Data (NLCD) evaluates the
consequences of adjusting land-cover within a watershed to
reflect uncertainty assessment information. Model validation
results indicate that using error-matrix information to adjust
land-cover values used in impervious surface models does
not substantially improve impervious surface predictions.
Validation results indicate that the resolution of the landcover
data (Level I and Level II) is more important in
predicting impervious surface accurately than whether the
land-cover data have been adjusted using information in the
error matrix. Level I NLCD, adjusted for land-cover misclassification,
is preferable to the other land-cover options for use
in models of impervious surface. This result is tied to the
lower classification error rates for the Level I NLCD.
Citation:
McMahon, Gerard, 2007, Consequences of land-cover misclassification in models of impervious surface: Photogrammetric Engineering and Remote Sensing, v. 73, no. 12, p. 1343-1353.
For more information, contact
North Carolina Water Science Center
U.S. Geological Survey
3916 Sunset Ridge Road
Raleigh, North Carolina 27607
(919) 571-4000
E-mail
|