Accuracy Assessment of NOAA Coastal Change
Analysis Program 2006-2010 Land Cover and
Land Cover Change Data
John W. McCombs, Nathaniel D. Herold, Shan G. Burkhalter, and Christopher J. Robinson
Abstract
A new approach to locating accuracy assessment sample units
was used to quantify 2010 land cover accuracy, in addition to be-
ing able to make statements about 2006-2010 land cover change
mapping accuracy for National Oceanic and Atmospheric
Administration (
NOAA
) Coastal Change Analysis Program (
C-CAP
)
data. Three customized tiers of sampling strata were created,
as discussed, to meet these goals. Stratified random sampling
was employed in each stratum with a six out of nine pixel-ho-
mogeneity criteria (different from the final minimum mapping
unit) required for each sampling unit. Accuracy was assessed for
nine regions in the coastal United States with overall accuracy
ranging from 82.3 percent to 85.6 percent. Binary change was
mapped with 88.7 percent accuracy, with the largest error being
errors of commission (71.2 percent user accuracy). This sampling
design also allowed for the identification of 137 locations where
true change was not mapped, allowing for statements to be made
about missed change.
Introduction
Land cover and land cover change are of critical importance,
with implications for water quality (Kang
et al.,
2014; Lu and
Weng, 2006; Margriter
et al.,
2014), wildlife habitat (Lowe and
Peterson, 2014; Millette
et al.,
2014; Porter
et al.,
2015), forest
fragmentation (Civco
et al.,
2002; Nagendra
et al.,
2004), eco-
system health (Greene
et al.,
2014; Lowe and Peterson, 2014;
Nestlerode
et al.,
2014), human health (Cleckner and Allen,
2014; Liang and Gong, 2015; Raghavan
et al.,
2014), and
climate change (Galbraith
et al.,
2002; Hansen and Loveland,
2012; Morris
et al.,
2002).
In 2010, over 123 million people, or 39 percent of the na-
tion’s population, lived in coastal shoreline counties, repre-
senting less than 10 percent of the U.S. land area (excluding
Alaska) (U.S. Census Bureau, 2011). Population density with-
in this area is expected to increase to 14.26 persons per square
kilometer (37 persons per square mile) by 2020, while the ex-
pected increase for the entire U.S. is 4.25 persons per square
kilometer (11 persons per square mile) (Woods and Poole
Economics, 2011). Economic data for the U.S. indicate that
ocean- and Great Lakes-dependent businesses employed 2.8
million people, paid $107.5 billion in wages, and produced
$282.2 billion in goods and services in 2011. From 2010 to
2011, the ocean and Great Lakes economy gained 67,000 jobs,
an increase of 2.4 percent - twice the employment growth rate
of the U.S. economy as a whole. Real gross domestic product
grew by 2.7 percent, faster than the U.S. economy as a whole
(1.6 percent) (
NOAA
ENOW). Because of the importance of the
U.S. coastal zones, and the rate at which change is occurring
within it, the need for accurate and timely land cover and
land cover change data is vital.
Through its Coastal Change Analysis Program (
C-CAP
), the
National Oceanic and Atmospheric Administration (
NOAA
)
produces land cover for the coastal regions of the United
States.
C-CAP
inventories coastal intertidal areas, wetlands,
and adjacent uplands with a goal of monitoring these habitats
by updating the land cover every five years (Dobson
et al.,
1985;
NOAA
C-CAP
, 2015). Resulting data are then incorporat-
ed into, and serve as the coastal expression of, the United
States Geological Survey (
USGS
) National Land Cover Database
(
NLCD
).
The nationwide
C-CAP
baseline was developed from im-
agery acquired by the Landsat suite of satellites,
circa
2001,
using a standardized classification approach (unsupervised/
supervised classification, spatial modeling, hand edits, etc.).
Since that time, additional dates have been created for 1996,
2006, and 2010, in addition to limited geographies having
data for 1985 and 1992. The
C-CAP
approach to creating a new
date of land cover consists of identifying potential areas of
change between multiple dates of Landsat data (e.g., 2006 to
2010) through a variety of spectral change analyses, classify-
ing those areas of potential change in the new date, then over-
laying classified areas of change over remaining non-change
areas to create a wall-to-wall map for the new date. Over time,
as additional dates of land cover were created, the change de-
tection and mapping methods have improved, previous errors
have been addressed, and significant steps have been taken
to improve the overall quality of the map. The completion of
the 2010 data incorporated the largest improvements to date,
including improved impervious surface/developed land cover
from the
USGS
and improved wetland/upland distinction
through the development and application of a
NOAA
Office for
Coastal Management wetland potential layer.
A previous accuracy assessment of
C-CAP
land cover was
performed on the 2001 data set. This first assessment focused
on the single-date map accuracy and included no assessment
of the mapped change (1996 to 2001). Since that time, new
land cover classes have been added, the nation has expe-
rienced a considerable amount of land cover change, and
improvements have been made in detecting and mapping
change. For these reasons,
C-CAP
determined that an accuracy
assessment of the 2010 wall-to-wall land cover and mapped
2006-2010 change would be part of this update cycle.
John W. McCombs, Shan G. Burkhalter, and Christopher J.
Robinson are with The Baldwin Group at NOAA Office for
Coastal Management, 2234 South Hobson Avenue, Charleston,
SC 29405 (
).
Nathaniel D. Herold is with the NOAA Office for Coastal
Man-agement, 2234 South Hobson Avenue, Charleston, SC
29405.
Photogrammetric Engineering & Remote Sensing
Vol. 82, No. 9, September 2016, pp. 711–718.
0099-1112/16/711–718
© 2016 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.82.9.711
