Wetland Mapping in the Upper Midwest United
States: An Object-Based Approach Integrating
Lidar and Imagery Data
Lian P. Rampi, Joseph F. Knight, and Keith C. Pelletier
Abstract
This study investigated the effectiveness of using high resolu-
tion data to map wetlands in three ecoregions in Minnesota.
High resolution data included multispectral leaf-off aerial
imagery and lidar elevation data. These data were integrated
using an Object-Based Image Analysis (
OBIA
) approach.
Results for each study area were compared against field and
image interpreted reference data using error matrices, accu-
racy estimates, and the kappa statistic. Producer’s and user’s
accuracies were in the range of 92 to 96 percent and 91 to
96 percent, respectively, and overall accuracies ranged from
96-98 percent for wetlands larger than 0.20 ha (0.5 acres). The
results of this study may allow for increased accuracy of map-
ping wetlands efforts over traditional remote sensing methods.
Introduction
Wetlands are naturally dynamic systems of important value
to the environment and society. The US Army Corps of
Engineers (
USACE
) in cooperation with the US Environmental
Protection Agency (
EPA
) have defined wetlands, incorporating
technical and policy considerations, as “…those areas that
are inundated or saturated by surface or ground water at a
frequency and duration to support and under normal cir-
cumstances do support, a prevalence of vegetation typically
adapted for life in saturated soil conditions” (Federal Register,
1980 and 1982). Wetlands can reduce some of the negative
effects of flooding and recharge groundwater by gradually re-
leasing flood water and snow melt. Wetlands offer habitat that
supports wildlife and fishing activities. Wetlands also provide
ecosystem services, including educational, aesthetic, and eco-
nomic opportunities. For example, intact freshwater marshes
in Canada have a total economic value of approximately 5,800
USD
per hectare compared to 2,400
USD
when those lands are
drained and used for agriculture (Millennium Ecosystem As-
sessment, 2005; Turner
et al
., 2000).
Due to wetland loss and degradation, many of the preced-
ing benefits have been reduced and are increasingly impacted.
About 215 million acres of wetlands existed in the United
States at the time of European settlement. However, by the mid-
1970s, only 99 million acres of the original wetlands remained.
Many of the lost wetlands were drained and are currently used
for agriculture, resource extraction, urbanization, and other
commercial purposes (Dahl and Johnson, 1991; Frayer
et al
.,
1983; Stedman and Dahl, 2008). Minnesota is not an excep-
tion to this large national wetland loss. Nearly half of Minne-
sota’s original wetlands were lost due to extensive agricultural
drainage and urban development. According to the Minnesota
Pollution Control Agency (
MPCA
) (2006), many original natural
wetlands were changed into local storm-water ponds to make
additional land available for urban development.
Currently in Minnesota only a few cities have updated wet-
land inventories. For the rest of Minnesota the only wetland in-
ventory available is the National Wetlands Inventory (
NWI
). The
Minnesota
NWI
maps were completed in the late 1980s using
aerial photos (some black and white) collected between 1979
and 1988 (
LMIC
, 2007). Several 7.5’ quadrangles in northwest-
ern Minnesota and a much larger area in northeastern Minneso-
ta were mapped based on 1970s 1:80 000 scale black-and-white
photos (
MPCA
, 2006). Changes in the landscape have occurred
which limit the use of the
NWI
maps due to the outdated data
and techniques used to create them. Thus, there is a need to
update wetland inventories with accurate boundaries and
improved delineation of smaller wetlands. An updated wetland
inventory would provide information that could be used to
make accurate decisions for the conservation, protection,
and restoration of wetlands. Although a Minnesota statewide
update is underway, it is a heavily image interpretation-based
project that is not expected to be completed until 2020. Thus,
more automated techniques may be useful in the near term.
A fast and effective method to identify accurate wetland
boundaries involves the use of remote sensing data and
techniques (Butera, 1983; Corcoran
et al
., 2011; Knight
et al
.,
2013). To the present time, the majority of wetland mapping
efforts using remote sensing data and techniques has been
focused on evaluating traditional pixel-methods with medium
to coarse resolution data. In many cases, the use of remote
sensing for wetland mapping has resulted in low accuracy
estimates, often due to mixed pixels and insufficient spectral
resolution (Grenier
et al
., 2007; Fournier
et al
., 2007; Lunetta
and Balogh, 1999; Ozesmi and Bauer, 2002).
Integration of
high resolution optical and elevation data has been shown
to reduce the mixed pixel problem (Frohn
et al
., 2009; Maxa
and Bolstad, 2009). Some studies have integrated optical
and elevation data to map wetlands using traditional pixel-
based methods. However, their accuracy results were low for
wetland classification due to the use of low to medium spatial
resolution data and pixel-based techniques (Baker
et al
., 2006;
Ozesmi and Bauer, 2002).
An object-based approach may be a better option to inte-
grate high resolution data and overcome some limitations,
including the mixed pixel problem and salt-and-pepper effect
of traditional pixel-based techniques (Myint
et al
., 2011; Zhou
The Department of Forest Resources, University of Minnesota, 1530
Cleveland Ave, N., Saint Paul, MN 55108 (
).
Photogrammetric Engineering & Remote Sensing
Vol. 80, No. 5, May 2014, pp. 439–449.
0099-1112/14/8005–439
© 2014 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.80.5.439
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
May 2014
439
