267
© The Ecological Society of America www.frontiersinecology.org
H
uman populations are increasing exponentially in
coastal regions worldwide, rendering nearly 200
million people vulnerable to severe weather events and
sea-level rise (Small and Nicholls 2003). Many of the
world’s most densely populated coasts are fringed by pro-
tective tidal flats that stabilize coastlines, defend against
storm surges, and provide economic opportunities to
human communities (Healy et al. 2002; Nicholls et al.
2007). With exceptionally high biodiversity supported by
both terrestrial- and marine-derived nutrients, tidal flats
are among the world’s most productive ecosystems, acting
as nurseries for finfish and shellfish, and as habitat for
tens of millions of migratory birds (MA 2005). However,
recent reports of substantial reductions of sediment deliv-
ery from major rivers (Syvitski et al. 2005), sinking of
river deltas (Syvitski et al. 2009), coastal erosion
(Nicholls et al. 2007), and widespread degradation from
coastal development (Kirwan and Megonigal 2013) indi-
cate that coastal ecosystems are under extreme stress, yet
the changing status of tidal flats beyond local scales
remains largely unknown.
Such paucity of knowledge stems from the fact that
tidal flats are fully exposed only at low tide, which hin-
ders detection of change in their extent over large areas.
We developed a method to resolve this by mapping long-
term change of tidal ecosystems using publicly available
time-series satellite imagery. With additional reference to
historical topographic maps, here we report on more than
50 years of change in tidal flat extent for ~4000 km of
East Asia’s Yellow Sea coastline.
Tidal wetlands in the Yellow Sea are dominated by
tidal flats that, at up to 20-km wide, are among the most
extensive in the world (Healy et al. 2002), providing an
estimated $30 billion per year in ecosystem services
(MacKinnon et al. 2012) and buffering one of the most
densely populated coastal areas in the world from storms
and sea-level rise (Small and Nicholls 2003; Nicholls et
al. 2007). The Yellow Sea’s low-elevation coastal zone is
home to about 60 million people, and unprecedented
urban, industrial, and agricultural expansion in the
region has led to concern about coastal ecosystem
integrity and imperiled species conservation (CIESIN
2005; MacKinnon et al. 2012). With the Yellow Sea
coastal zone projected to be part of a 1800-km-long
urban corridor by 2030 (Seto et al. 2012), there is an
urgent need to understand the distribution and status of
remaining coastal ecosystems to allow the development
of complementary conservation and land-use planning
strategies.
We mapped the extent of tidal flats across the three
countries with a Yellow Sea coastline – China, North
Korea, and South Korea – at three time periods (mid-
1950s, early 1980s, and late 2000s). The analysis was
conducted on 80 Landsat Archive images in two clusters
(early 1980s and late 2000s) and 25 digitized 1:250 000
topographic maps (mid-1950s) across ~4000 km of
coastline between northern Jiangsu province, China
(34˚29' N, 119˚47' E) and eastern Busan province,
South Korea (35˚20' N, 129˚17' E; Figure 1). Here,
we report on areal changes of tidal flats over timescales
and spatial extents that have hitherto been impossible
to study.
RESEARCH COMMUNICATIONS RESEARCH COMMUNICATIONS
Tracking the rapid loss of tidal wetlands in
the Yellow Sea
Nicholas J Murray
1,2*
, Robert S Clemens
1
, Stuart R Phinn
3
, Hugh P Possingham
1,4
, and Richard A Fuller
1
In the Yellow Sea region of East Asia, tidal wetlands are the frontline ecosystem protecting a coastal popu-
lation of more than 60 million people from storms and sea-level rise. However, unprecedented coastal
development has led to growing concern about the status of these ecosystems. We developed a remote-sens-
ing method to assess change over ~4000 km of the Yellow Sea coastline and discovered extensive losses of
the region’s principal coastal ecosystem – tidal flats – associated with urban, industrial, and agricultural
land reclamations. Our analysis revealed that 28% of tidal flats existing in the 1980s had disappeared by
the late 2000s (1.2% annually). Moreover, reference to historical maps suggests that up to 65% of tidal flats
were lost over the past five decades. With the region forecast to be a global hotspot of urban expansion,
development of the Yellow Sea coastline should pursue a course that minimizes the loss of remaining
coastal ecosystems.
Front Ecol Environ 2014; 12(5): 267–272, doi:10.1890/130260 (published online 8 May 2014)
1
Australian Research Council Centre of Excellence for
Environmental Decisions, School of Biological Sciences, University
of Queensland, St Lucia, Australia
*
(murr.nick@gmail.com);
2
CSIRO Climate Adaptation Flagship and CSIRO Ecosystem
Sciences, Dutton Park, Australia;
3
Centre for Spatial En-
vironmental Research, School of Geography, Planning and
Environmental Management, University of Queensland, St Lucia,
Australia;
4
Imperial College London, Department of Life Sciences,
Silwood Park, Ascot, UK
Mapping coastal ecosystem loss NJ Murray et al.
268
www.frontiersinecology.org © The Ecological Society of America
n
Materials and methods
Satellite data and topographic maps
Using the Oregon State University China Seas tide
model (Egbert and Erofeeva 2002), we first estimated the
tidal elevation at the time of acquisition for all Landsat
Archive images available for the Yellow Sea coastal
region (5568 images). We visually reviewed all Landsat
Archive images that were acquired within the upper and
lower 10% of the tidal range and selected an image set
composed of images suitable for the subsequent remote-
sensing analysis (Murray et al. 2012). The final image set
comprised 32 ETM+ (Enhanced Thematic Mapper Plus),
12 Landsat TM (Thematic Mapper), and 36 Landsat
MSS (Multi-Spectral Scanner) satellite images across 20
Landsat 185-km × 170-km footprints (path-row tiles),
with images including areas with macro (>4 m), meso
(2–4 m), and micro (<2 m) tidal ranges (mean tide range
2.45 m) (WebTable 1). For each Landsat scene, we used
image differencing of classified land–water images to map
the area between the high- and low-tide waterline in
each time period, resulting in spatial datasets of tidal flats
present in the 1980s and 2000s (Murray et al. 2012).
To assess the accuracy of each tidal flat dataset, we adopted
a widely used accuracy assessment protocol termed an error
matrix (Congalton and Green 2008). An independent ana-
lyst was offered 240 randomly generated points over the
Figure 1. Change in tidal flats in the Yellow Sea between the 1950s and the 2000s, mapped at a 5-km grid resolution. Net change
between the two time periods is shown on a color ramp from blue (total gain) to red (total loss).
Net change per 5-km grid cell (%), 1950s–2000s
NJ Murray et al. Mapping coastal ecosystem loss
269
© The Ecological Society of America www.frontiersinecology.org
study area for each dataset and was
required to classify each point as
either tidal flat or other, using the
low-tide satellite image set for the
period in question (Murray et al.
2012). The analyst was able to use
all available bands of the Landsat
imagery to decide whether a point
was a tidal flat. The resulting
dataset of validation points, when
compared with our classification
points, allowed the error matrix for
each period to be populated
(WebTables 2 and 3). The accu-
racy assessment revealed >94%
overall classification accuracy for
mapping tidal flats with Landsat
Archive imagery.
For a historical baseline, we dig-
itized 25 coastal maps available
from the AMS L500 (China),
L541 (Manchuria), and L552
(Korea) series of 1:250 000 topo-
graphic maps (US Army Map
Service 1962). The collection was
produced between 1950 and 1962, and depicts tidal flats to
a smallest patch size of approximately 250 m × 250 m. All
of the maps contained reliability diagrams indicating their
origin, which included large-scale topographic maps, and
photogrammetric and hydrographic sources. We georefer-
enced the topographic maps against prominent topographi-
cal features in terrain-corrected Landsat imagery and delin-
eated the foreshore flat class using interactive digitization
methods (Hood 2004; Hughes et al. 2006). Given their
level of detail, the inclusion of reliability diagrams, informa-
tion on their source data, and the overlap with our Landsat-
derived datasets (Figure 2), we considered this dataset a
valuable historical baseline.
Change detection
To permit comparison across the three time periods (1950s,
1980s, 2000s), we resampled each dataset to 250-m spatial
resolution, which was larger than the smallest patch of tidal
flat depicted in the topographic maps, and then reprojected
each dataset to an Albers Equal Area projection. We also
accounted for the 2003 failure of the Scan Line Corrector
(SLC) aboard the Landsat 7 satellite, which resulted in data
gaps (striping) for approximately 26% of the tidal flat area
in the 2000s dataset. Areas of images lost due to the SLC
failure were overlaid onto the complete 1980s dataset,
allowing change between the two datasets to be calculated.
Thus, the 2000s extent of tidal flat (A
2
) is calculated as
A
2
= A
1
{1 – [(A
h
– A
p
)/A
h
]}
where A
1
is the 1980s extent of tidal flats, A
p
is the 2000s
dataset (with SLC-off data gaps), and A
h
is the artificially
striped A
1
dataset. Lastly, areas that could not be mapped in
one time period, primarily because of chronic cloud or ice
cover, were masked over all three datasets, resulting in final
coverage of 87.9% of the study area coastline (Figure 1).
We established the total area of each of the three tidal
flat datasets and calculated the net change over the study
region between each of the time periods (1950s–1980s,
1980s–2000s, 1950s–2000s; Table 1). We also calculated
the continuous rate of change (r, % yr
–1
) of tidal flats on
a per Landsat footprint basis between each of the time
periods as
r = [1/(t
2
– t
1
)] × ln(A
2
/A
1
)
where A
1
and A
2
are the areas of tidal flats in a Landsat
footprint at times t
1
and t
2
, respectively.
n
Results and discussion
Our analysis of the change in areal extent of tidal flats in
the Yellow Sea indicates that of the 545 000 ha present in
the 1980s, only 389 000 ha remained three decades later,
equating to a net loss of 28% at a mean rate of –1.2% yr
–1
(Table 1). Comparing the three countries in our analysis,
China lost more tidal flat area and at a faster rate (39.8%,
–1.8% yr
–1
) than South Korea (32.2%, –1.6% yr
–1
); in
North Korea, minor gains of tidal flats occurred (8.5%,
0.3% yr
–1
). Our area-related values underestimate the full
tidal flat extent in the Yellow Sea, because cloud cover,
ice cover, and lack of images acquired at suitable tide
heights precluded mapping 12.1% of the study area coast-
line (Figure 1). Nevertheless our data cover >4000 km of
the Yellow Sea coastline and reveal rapid and widespread
Figure 2. An example of tidal flat mapping results, showing the raw data (a–c) and
mapped tidal flats (d–f). The results reveal widespread loss of tidal flats from 1954 (left)
to 2010 (right). Satellite images (b) and (c) show that tidal flats present in 1981 (e)
were reclaimed for agricultural and industrial land by 2010 (f).
USGS
USGS
(a) (b) (c)
(d) (e) (f)
Mapping coastal ecosystem loss NJ Murray et al.
270
www.frontiersinecology.org © The Ecological Society of America
declines of tidal flats across the entire region (Figure 1;
Table 1).
According to historical topographic maps, tidal flats
occupied 1.12 million ha in the mid-1950s, equating to a
potential net loss of up to 65% over 50 years (Table 1).
Comparisons with historical mapping must of course be
interpreted cautiously, although we took care to match
the resolution of the three datasets and thoroughly inves-
tigated the reliability of the maps. Thus, our results sug-
gest that up to two-thirds of the tidal flats existing around
the Yellow Sea in the 1950s have since vanished, with
losses in China and South Korea accounting for most of
the decline (Figure 1; Table 1).
Losses of tidal flats were spatially pervasive, occurring
throughout heavily populated and rapidly developing
coastal areas (Figure 1). Tidal flats increased in extent in
a few isolated locations, such as the seaward edge of sev-
eral coastal embayments and at growing river deltas.
Much of the Yellow Sea coastline is under intense pres-
sure from land claims (commonly termed reclamation)
for agriculture, aquaculture, and industrial development
(Figures 2 and 3). For example, agricultural development
in Chungcheongnam-do Province, South Korea, caused
the loss of more than 7000 ha of tidal flats over the past
30 years (Figure 2). Two developments currently under-
way – Saemangeum, South Korea (40 100 ha) and the
Caofeidian port development in China (31 000 ha;
WebFigure 1) – are among the largest reclamation pro-
jects on Earth (CCICED 2010; MacKinnon
et al. 2012). Similarly, the conversion of tidal flats
to aquaculture ponds is widespread in the Yellow Sea and,
with Asia currently supplying 89% of global aquaculture
production (FAO 2012), further reclamation of tidal flats
will be required to meet increasing demand (Naylor et al.
2000). The impact of reclamation activity on tidal flats is
also reflected in our results for North Korea, where the
near absence of recent coastal development allowed
minor gains in tidal flat extent. This appears to result
from sediment deposition in the estuaries of the Yalu and
Chongchon rivers, perhaps owing to increased soil ero-
sion caused by abrupt land clearing that occurred in
North Korea during the 1990s (Stone 2012).
Tidal flats may be expected to shift seaward over the
long term in response to reclamation activities (Hood
2004; Kirwan and Megonigal 2013), but our data indicate
that this is not happening at a rate sufficient to compen-
sate for the loss, probably as a result of local compaction
and appropriation of tidal flat sediments for construction
purposes (MacKinnon et al. 2012). Studies of salt marsh
systems have shown that changes in sediment supply and
loss of coastal vegetation can lead to collapse of tidal wet-
lands, resulting in a runaway effect of tidal flat deepening
and bed erosion (Kirwan and
Megonigal 2013; Mariotti and
Fagherazzi 2013). Yellow Sea
tidal flats are highly dependent
on ongoing sediment supply
(Healy et al. 2002) and substan-
tial declines of sediment output
from major rivers in the region,
such as the 90% decline in sed-
iment flow from the Yellow
River during the 20th century
(Syvitski et al. 2009; Wang et
al. 2010), could be contributing
Figure 3. Tidal flat conversion to alternative land uses. (a) Aquaculture development encroaching approximately 250 m onto a tidal
flat (top of image) in Gomso Bay, South Korea. (b) Coastal reclamation for industrial land at an offshore island in South Korea,
noting the ships in port at the top left for scale.
Table 1. Tidal flat area and rates of change by country, 1950s–2000s
Estimated area of tidal flat Continuous rate of change
(ha) % change (% yr
–1
)
1950s– 1980s– 1950s– 1950s– 1980s– 1950s–
1950s 1980s 2000s 1980s 2000s 2000s 1980s 2000s 2000s
China 539 794 267 751 161 066 –50.4 –39.8 –70.2 –2.7 –1.8 –2.2
North Korea 231 813 99 333 107 765 –57.1 8.5 –53.5 –4.9 0.3 –1.6
South Korea 350 331 177 729 120 472 –49.3 –32.2 –65.6 –2.4 –1.6 –2.0
Yellow Sea 1 121 938 544 812 389 303 –51.4 –28.0 –65.3 –3.0 –1.2 –2.0
Notes: Area estimates should be considered minima for the Yellow Sea, because 12.1% of the coastline could not be
mapped owing to the presence of cloud or ice cover in satellite imagery obtained at suitable tide heights (Figure 1).
USGS
(a) (b)
USGS
NJ Murray et al. Mapping coastal ecosystem loss
to the broad-scale losses that we detected.
Consequently, although we consider coastal reclama-
tion to be an important driver of tidal flat losses in the
Yellow Sea, processes such as changes in sediment sup-
ply, loss of coastal vegetation associated with develop-
ment, erosion, redistribution of sediments due to storms,
and compaction and subsidence (sinking) caused by
extensive subsurface resource and groundwater extrac-
tion are also likely to be operating (Bartholdy and
Aagaard 2001; Syvitski et al. 2009; Nicholls and
Cazenave 2010; Higgins et al. 2013). These factors could
increase vulnerability of coastal communities and
coastal developments to storms and sea-level rise,
because land reclamations, intensive extractive activi-
ties, and sediment declines have been shown to lead to
relative sea-level rise that can be several orders of mag-
nitude greater than background levels of local and
global sea-level rise (Li et al. 2004; Cazenave and Le
Cozannet 2013; Higgins et al. 2013).
Although the magnitude of losses is alarming, our results
broadly agree with several other information sources on
coastal wetland loss in East Asia. For instance, the China
Council for International Cooperation on Environment
and Development reported that China has lost 57% of its
coastal wetlands since the 1950s, and that more than 1.3
million ha of coastal reclamation occurred between 1990
and 2008 (CCICED 2010). Other sources suggest that
51% of coastal wetlands in China were lost over the past
50 years (An et al. 2007), that more than one-third of
China’s tidal flats were reclaimed between 1950 and 1985
(Yu 1994), and that half of South Korea’s tidal wetlands
have been reclaimed in the past 50 years (Cho and Olsen
2003). With the implementation of a robust, repeatable
remote-sensing framework, our results provide the first
quantitative verification of widespread declines of tidal
flats in the Yellow Sea region. Globally, the status and dis-
tribution of tidal flats remain poorly understood, and with
about one-third of vegetated coastal ecosystems – includ-
ing mangroves, seagrass beds, and salt marshes – estimated
to have been lost in the past few decades (Mcleod et al.
2011), the total loss of tidal flats could be equally as high.
Our method for mapping tidal flats permits detailed mea-
surement of tidal habitats over thousands of kilometers,
and could provide a practical solution for establishing the
status of tidal wetlands for any large geographic region.
n
Conclusions
Our analysis indicates that tidal flats along the Yellow
Sea are declining at a rate comparable to many other at-
risk ecosystems, such as tropical forests (Achard et al.
2002), seagrass meadows (Waycott et al. 2009), and man-
groves (Giri et al. 2011). None of the drivers we identify
are unique to this region of the world. Degradation and
reclamation of coastal wetlands are worldwide phenom-
ena (MA 2005) and are likely to intensify, owing to the
increasing scarcity of land in coastal areas and the low
271
© The Ecological Society of America www.frontiersinecology.org
cost and rapid pace at which these areas can be devel-
oped (MacKinnon et al. 2012). Similarly, reduced sedi-
ment discharge, often associated with trapping of sedi-
ments in reservoirs, is associated with land loss at 26 of
the world’s major river deltas (Syvitski et al. 2009).
These factors, when combined with coastal subsidence
due to resource extraction and coastal development,
result in relative sea-level rise in coastal regions that is
far greater than the rate of global sea-level rise, poten-
tially leading to further loss of tidal flat ecosystems
(Cazenave and Le Cozannet 2013).
A combination of accelerating human population
growth along the world’s coastlines and impacts expected
from sea-level rise suggest that unless prompt action is
taken to protect remaining tidal wetlands, coastlines and
their associated ecosystem services will become increas-
ingly vulnerable in the 21st century. Major systems of
tidal flats protect 15 of the world’s 20 most flood-vulnera-
ble coastal cities (WebTable 4), and their maintenance
and protection offers an additional method for shielding
these communities from the impacts of storms and sea-
level rise (Arkema et al. 2013). Early warning signs from
the Yellow Sea suggest that the consequences of intertidal
ecosystem loss for coastal biodiversity may already be
apparent. Of the six migratory shorebird species that
depend solely on Yellow Sea tidal flats during migration,
the great knot (Calidris tenuirostris) and the far eastern
curlew (Numenius madagascariensis) have recently been
listed as globally threatened by the IUCN. In the Yellow
Sea region, where substantial urban expansion is forecast
in coastal areas, safeguarding ecosystem services provided
by tidal flats and ensuring protection of the region’s
coastal biodiversity will require coastal development
strategies that minimize ecosystem loss and protect
remaining coastal ecosystems.
n
Acknowledgements
We thank M Choi, R Ferrari, T Gill, J Hanson, D Melville,
C Roelfsema, and V Wingate for assistance with various
components of the project, and D Watkins, N Moores, and
the East Asian–Australasian Flyway Partnership for
regional expertise. This project was supported by an
Australian Research Council Linkage Grant
LP100200418, co-funded by the Queensland Department
of Environment and Resource Management; Com-
monwealth Department of the Environment; the
Queensland Wader Study Group; and the Port of Brisbane
Pty Ltd. Additional support was provided by Birds
Queensland and the CSIRO Climate Adaptation Flagship.
Landsat data are available from the US Geological Survey.
n
References
Achard F, Eva HD, Stibig H-J, et al. 2002. Determination of defor-
estation rates of the world’s humid tropical forests. Science 297:
999–1002.
An SQ, Li HB, Guan BH, et al. 2007. China’s natural wetlands:
