Forgot your password?
Document Actions
Giri et al. 2008
Journal of Biogeography (J. Biogeogr.) (2008) 35, 519–528
ORIGINAL Mangrove forest distributions and
ARTICLE
dynamics (1975–2005) of the
tsunami-affected region of Asia
C. Giri1*, Z. Zhu2, L. L. Tieszen2, A. Singh3, S. Gillette4 and J. A. Kelmelis5
1
Science Application International Corporation ABSTRACT
(SAIC), contractor to U.S. Geological Survey
Aim We aimed to estimate the present extent of tsunami-affected mangrove
(USGS) Center for Earth Resources
Observation and Science (EROS), Sioux Falls,
forests and determine the rates and causes of deforestation from 1975 to 2005.
SD 57198, USA, 2USGS Center for Earth Location Our study region covers the tsunami-affected coastal areas of
Resources Observation and Science (EROS), Indonesia, Malaysia, Thailand, Burma (Myanmar), Bangladesh, India and Sri
Sioux Falls, SD 57198, USA, 3United Nations Lanka in Asia.
Environment Programme, Washington, DC
20006, USA, 4Colorado State University, Fort Methods We interpreted time-series Landsat data using a hybrid supervised
Collins, CO 80523, USA and 5US Geological and unsupervised classification approach. Landsat data were geometrically
Survey, VA and US Department of State, corrected to an accuracy of plus-or-minus half a pixel, an accuracy necessary
Washington, DC 20520, USA for change analysis. Each image was normalized for solar irradiance by converting
digital number values to the top-of-the atmosphere reflectance. Ground truth
data and existing maps and data bases were used to select training samples and
also for iterative labelling. We used a post-classification change detection
approach. Results were validated with the help of local experts and/or high-
resolution commercial satellite data.
Results The region lost 12% of its mangrove forests from 1975 to 2005, to a
present extent of c. 1,670,000 ha. Rates and causes of deforestation varied both
spatially and temporally. Annual deforestation was highest in Burma (c. 1%) and
lowest in Sri Lanka (0.1%). In contrast, mangrove forests in India and Bangladesh
remained unchanged or gained a small percentage. Net deforestation peaked at
137,000 ha during 1990–2000, increasing from 97,000 ha during 1975–90, and
declining to 14,000 ha during 2000–05. The major causes of deforestation were
agricultural expansion (81%), aquaculture (12%) and urban development (2%).
Main conclusions We assessed and monitored mangrove forests in the
tsunami-affected region of Asia using the historical archive of Landsat data. We
*Correspondence: Chandra Giri, Science also measured the rates of change and determined possible causes. The results of
Application International Corporation (SAIC), our study can be used to better understand the role of mangrove forests in saving
US Geological Survey (USGS) Center for Earth
lives and property from natural disasters such as the Indian Ocean tsunami, and
Resources Observation and Science (EROS),
Sioux Falls, SD 57198, USA. to identify possible areas for conservation, restoration and rehabilitation.
E-mail: cgiri@usgs.gov
Keywords
Work performed under USGS contract Change analysis, deforestation, image processing, Indian Ocean tsunami,
03CRCN0001. Landsat, mangrove forests.
and hurricanes. They also serve as breeding and nursing
INTRODUCTION
grounds for marine species and are sources of food, medicine,
Mangroves forests, distributed circumtropically in the inter- fuel and building materials for local communities. However,
tidal region between sea and land in the tropical and the forests have been declining at an alarming rate – perhaps
subtropical latitudes, provide important ecosystem goods even more rapidly than inland tropical forests (Aizpuru et al.,
and services. The forests help stabilize shorelines and reduce 2000) – and much of what remains is in degraded condition
the devastating impact of natural disasters, such as tsunamis (Valiela et al., 2001; Wilkie et al., 2003). Conversion of
ª 2007 The Authors www.blackwellpublishing.com/jbi 519
Journal compilation ª 2007 Blackwell Publishing Ltd doi:10.1111/j.1365-2699.2007.01806.x
C. Giri et al.
mangrove forests to aquaculture is on the rise in many rates and causes of change using multi-temporal satellite data
countries of the region without considering the fact that the and field observations. Our analysis sought to answer the
total economic value of intact mangrove forests is often higher following research questions: how much mangrove forest
than that of shrimp farming (Balmford et al., 2002). The remains; where are the remaining mangrove forests located;
remaining mangrove forests are under immense pressure from what is the rate of change; what are the main reasons for the
clearcutting, encroachment, hydrological alterations, chemical change?
spills, storms and climate change (Blasco et al., 2001; McKee,
2005).
Study area
The Indian Ocean tsunami of December 2004 and other
natural disasters have highlighted the importance of mangrove Our study area covers the coastal areas of Indonesia, Malaysia,
forests as a ‘bio-shield’ that protects vulnerable coastal Thailand, Burma (Myanmar), Bangladesh, India and Sri Lanka.
communities. Mangrove forests attenuated the Indian Ocean We chose this area for a number of reasons. First, this area was
tsunami waves and protected coastal communities in Indone- the most devastated during the Indian Ocean tsunami of
sia, Thailand, India and Sri Lanka (Danielsen et al., 2005; December 2004; as a result, many national governments and
IUCN, 2005; Kathiresan & Rajendran, 2005). In some areas, international organizations are now implementing ambitious
mangrove forests hit by the Indian Ocean tsunami suffered conservation and rehabilitation programmes. Second, the region
severe damage from breaking and uprooting. Recent findings contains approximately 10% of the total mangrove forests of the
suggest that the continued destruction and degradation of world, including the largest remaining contiguous mangrove
many mangrove forests throughout the tropics over the past forest in the world, the Sundarbans. Third, strong demographic
few decades has decreased the protective capacity of mangrove pressure and diverse climatic conditions in the region have
forest ecosystems and reduced their ability to rebound from created a mosaic of mangrove diversity that is changing
natural disasters (Dahdouh-Guebas et al., 2005; Nigel, 2005; constantly. Fourth, the region is the epicentre of mangrove
Weiner, 2005). However, accurate and reliable information on biodiversity and consists of many existing and planned national
the present extent of mangrove forests and the rate and causes parks, biosphere reserves and world heritage sties.
of deforestation in the tsunami-affected region of Asia has not The forest is under severe threat from both anthropogenic
been available (Adeel & Pomeroy, 2002; Danielsen et al., 2005; and natural forces. Anthropogenic threats include encroach-
UNEP, 2004). Such information is needed to better understand ment from expansion of agriculture (e.g. rice farming, coconut
the protective role of mangrove forests and to learn more and oil palm), aquaculture, urban development (e.g. resorts),
about deforestation dynamics, carbon fluxes, forest fragmen- mining, salt pan development and overexploitation of
tation and the provision of ecosystem goods and services. resources. Natural threats include erosion, sedimentation and
Remote sensing is an indispensable tool for assessing and sea level rise. Because of these threats, mangrove forest is the
monitoring mangrove forests, primarily because many man- most threatened habitat in the region. Only sporadic patches of
grove swamps are inaccessible or difficult to field survey. mangrove forests are left in India and Sri Lanka, and they have
Remote sensing provides synoptic coverage, and historical been depleted in Burma, Thailand, Malaysia and Indonesia.
satellite data dating back to the 1960s are available. Global In addition to deforestation, mangrove forests have been
mapping initiatives have failed to map the extent and rate of declining in biological diversity and economic value. Many
deforestation with sufficient detail because these studies have flora and fauna are vulnerable, near-threatened, threatened,
been based on satellite data with coarse spatial resolution endangered or critically endangered. Economic activities such
(1 km or coarser). For example, only extensive mangrove areas as extraction of timber and fuel wood, fishing and the
were mapped as part of the Global Land Cover 2000 survey collection of honey and other forest products have also
(Stibig et al., 2007). At local scales, several studies have used diminished.
moderate-resolution satellite data [e.g. Landsat, SPOT and the
India Remote Sensing Linear Imaging Self-scanning Sensor
DATA AND METHODOLOGY
(IRS LISS III)] to characterize and map mangrove forests
(Silapong & Blasco, 1992; Ramsey & Jansen, 1996; Blasco et al.,
Data acquisition
2001; Selvam et al., 2003; Ramasubramanian et al., 2006;
Vaiphasa et al., 2006). Synergistic use of optical and radar data We used Landsat GeoCover and recently acquired Enhanced
has been particularly useful in cloud-covered tropical man- Thematic Mapper Plus (ETM+) data, made available through
grove areas (Aschbacher et al., 1994; Giri & Delsol, 1995). the US Geological Survey (USGS) Center for Earth Resources
However, large areas of the tsunami-affected countries of Asia Observation and Science (EROS) (http://eros.usgs.gov). Geo-
(Indonesia, Malaysia, Thailand, Burma, Bangladesh, India and Cover is a collection of Landsat data with global coverage and
Sri Lanka) remain unmapped. As a result, the present extent of generally cloud-free images. Data were collected for three
mangrove forests, and the rate and causes of deforestation, are epochs: (1) the ‘1975’ imagery from 1973 to 1983, (2) the
unknown. ‘1990’ imagery from 1989 to 1993, and (3) the ‘2000’ imagery
We determined the extent and distribution of mangrove from 1997 to 2000. Detailed descriptions of GeoCover data
forests in the tsunami-affected countries and identified the can be found at https://zulu.ssc.nasa.gov/mrsid/. Additional
520 Journal of Biogeography 35, 519–528
ª 2007 The Authors. Journal compilation ª 2007 Blackwell Publishing Ltd
Mangrove forest distribution and dynamics
Multispectral Scanner (MSS) and Thematic Mapper (TM) data geometric correction error, noise arising from atmospheric
were acquired to supplement cloud-covered areas. The ‘2005’ effects, errors arising from changing illumination geometry
Enhanced Thematic Mapper Plus (ETM+) data collected from and instrument errors (Homer et al., 2004). Such errors are
2005 to 2006 were acquired from EROS. Approximately 216 likely to introduce biases and/or noise into mangrove forest
Landsat TM or ETM+ scenes each for 1990, 2000 or 2005, and classification and change analyses. Pre-processing is necessary
57 MSS scenes for 1975 were acquired for the study. Same-year to remove or minimize such errors.
and same-season data are best for this kind of study, but cloud- Landsat images acquired in the Universal Transverse Mer-
free images of the region were not available for all time cator (UTM) projection and coordinate system were
periods, prompting us to augment them with multi-season and re-registered to the Albers equal area projection. To improve
multi-year data. Twenty-four QuickBird scenes and eight the root mean square (RMS) error to ± 1/2 pixel, we used
IKONOS scenes, all collected in 2005 and 2006, were also additional ground control points (GCPs) collected from
acquired. 1:50,000-scale topographic maps. Images were resampled with
cubic convolution, which has a better spatial accuracy than the
commonly used nearest neighbour resampling technique
Field survey
(Shlien, 1979; Park & Schowengerdt, 1982).
We conducted a 4-week field survey in Malaysia, Thailand, Sri Removing or minimizing the presence of ‘noisy’ pixels is
Lanka and India during June and July 2006. A total of 182 also important. Noise in image data can be caused by several
calibration/validation points were collected. We used a Global factors: (1) differences in atmospheric scattering in the visible
Positioning System (GPS) to record the exact location of the bands, (2) differences in water and/or dust particles in the
survey site, and we took a photograph at every location. All the atmosphere, (3) temporal variations in the solar zenith and/or
photos were georeferenced using GPS-Photo Link (http:// azimuth angles, and (4) inconsistencies in sensor calibration
www.geospatialexperts.com/) and were used in tandem with for separate images (Homer et al., 2004). To reduce the noise
satellite imagery within Google Earth to verify mangrove caused by atmospheric effects and illumination geometry, we
locations and conditions. Data on presence or absence of applied the techniques developed by Homer et al. (2004) for
mangrove forests, and their condition, density, crown cover the US National Land Cover Database 2001. Each image was
and management regimes were collected during the field normalized for solar irradiance by converting digital number
survey. Density and crown cover were estimated visually. The values to the top-of-the atmosphere reflectance (Chander &
extent of damage caused by the Indian Ocean tsunami and any Markham, 2003). This conversion algorithm is ‘physically
recovery measures taken were also noted. The field data served based, automated, and does not introduce significant errors to
as training data, a portion of which served as independent the data’ (Huang & Townshend, 2003). A test of this technique
reference data for the verification of classification results. on mangrove areas in the Sundarbans (Giri et al., 2007)
We visited the national mapping agencies of the countries in showed it to be a reasonable pre-processing method for a large
the region to collect ancillary data such as forest classification data base covering several countries in Asia. Owing to the
maps, topographic maps and tsunami reports (Table 1). The unavailability of data on atmospheric conditions for the
team also visited local forestry departments to discuss various region, atmospheric correction was not performed. Cloud-
aspects of mangrove management. With their help, mangrove covered areas were replaced by additional Landsat scenes
and non-mangrove areas were delineated on hard-copy maps. obtained during the same period.
Local officials provided field data and information on mangrove
damage due to tsunami and afforestation/reforestation status.
Classification
They also guided the research team during the field visit.
Many image classification and change detection techniques
have been described in the literature (Singh, 1989; Civco et al.,
Pre-processing
2002). For change analysis, Civco et al. (2002) compared four
The use of multi-temporal satellite data (MSS, TM and ETM+) techniques – traditional post-classification, cross-correlation
at a subcontinent scale poses a number of challenges: analysis, neural networks and object-oriented classification
Table 1 Secondary data collected and used
Secondary data Country
during the study.
Mangrove forest maps prepared using Landsat data Thailand, India
Forest classification maps 1 : 250,000 using Sri Lanka, Malaysia, Indonesia
field inventory data
Land use/land cover maps at 1 : 1,000,000 scale Bangladesh, Malaysia
Topographic maps, 1 : 100,000 scale Sri Lanka, Thailand
QuickBird, IKONOS Selected areas in Thailand,
Bangladesh, India and Sri Lanka
Atlas of Mangrove Wetlands of India India
Journal of Biogeography 35, 519–528 521
ª 2007 The Authors. Journal compilation ª 2007 Blackwell Publishing Ltd
C. Giri et al.
Table 2 Class definitions. 1975–2005, 1990–2000, 1990–2005 and 2000–05. The change
areas were visually interpreted to identify the factors (e.g.
Classes Supervised classification class definitions
agriculture, aquaculture, urban development) responsible for
Mangrove Areas covered by both closed and open the change. Each of the change areas was checked individually
mangrove forests during the workshop by at least two experts to identify the
Non-mangrove Areas covered by croplands and other land uses causes of change. Published maps and high-resolution satellite
Barren lands Areas devoid of vegetation, e.g. sand dunes, data such as QuickBird and IKONOS were used for the
sediments or exposed soil purpose. A 3-pixel by 3-pixel window was used to identify the
Water bodies Areas of open water with no emergent vegetation, dominant land-cover types in the high-resolution satellite data.
e.g. channels and waterways Once the mangrove/non-mangrove areas were calculated for
each period, the rate of deforestation was calculated by using
– and concluded that each method has advantages and the following equation suggested by Puyravaud (2003):
disadvantages, and there is no single best way. Considering
1 A2
the large volume of data acquired from different time intervals, R¼ ln ð1Þ
t2 À t1 A1
we adopted the traditional post-classification approach.
For image classification, we used a hybrid supervised and where R is the rate of deforestation, A1 is the area at an initial
unsupervised classification approach because there was insuf- time t1 and A2 is the area at a later time t2. Maps and
ficient ground truth for a purely supervised classification. The photographs collected during the field visit, along with high-
images were not enhanced prior to unsupervised classification resolution IKONOS and QuickBird satellite data, were anal-
and the thermal band (band 6) was excluded from the ysed visually to discover the causes of deforestation. This
classification. Water bodies were mapped with a supervised involved visual analysis of change maps and reference data.
classification. We then used an ISODATA clustering algorithm
within ERDAS Imagine to generate 50 spectral clusters at the
Validation
99% convergence level. Through iterative labelling, mangrove
classes were identified and labelled with reference to field data Qualitative validation was performed with the help of local
and high-resolution QuickBird and IKONOS imagery, and experts and high-resolution satellite data such as QuickBird
then merged into a single mangrove category. Four land cover and IKONOS. We divided the entire area into 500 · 500 grids
classes were generated: mangrove, non-mangrove, barren lands and checked each grid visually to identify and correct ‘gross’
and water bodies (Table 2). Post-classification editing such as errors inherent in the classified maps. Quantitative accuracy
‘recoding’ was performed to remove obvious errors. Each assessment was not performed because sufficient ground truth
classified image was resampled to 50 m to be consistent with data for historical dates are not easily available. A workshop
MSS data. However, this resampling did not improve the was held at which local experts validated the change areas and
spatial details of MSS data. Finally, four classification images the causes of change. The experts provided ground truth data
were produced, one each for 1975, 1990, 2000 and 2005. and photographs for a number of sites throughout the study
area.
Change analysis
RESULTS AND DISCUSSION
We used a post-classification change analysis technique to
compare classification results from the four epochs of imagery. To meet the need for accurate and reliable information on the
This approach is probably the most common and intuitive present extent of mangrove forests of the region, our study
change detection method, because it provides ‘from–to’ change prepared a geospatial data base of mangrove distribution for
information. However, our approach may have three sources the year 2005 using Landsat data. The data base provides an
of uncertainty: (1) semantic differences in class definitions up-to-date and consistent overview of the extent and distri-
between maps, (2) positional errors, and (3) classification bution of mangrove forests with better spatial and thematic
errors. To minimize the semantic differences in class defini- details than previous data sets.
tions, we specified the same number of classes for all four We estimate that in 2005 there were approximately 1.67
dates. To minimize positional errors, additional GCPs were million ha of mangrove forest remaining in the region. This
selected and RMS error was reduced to ± 1/2 pixel as explained estimate is higher than previous estimates by Spalding et al.
in the pre-processing section. Post-classification editing (i.e. (1997), probably because it is based on moderate-resolution
recoding) using secondary data helped to correct classification Landsat data on which we were able to identify many small
errors. In doing so, problematic areas were identified visually (0.81 ha), previously unidentified, mangrove areas. However,
and using area of interest (AOI) in ERDAS Imagine followed we did not map mangrove patches smaller than the 1 ha,
by recoding. However, some positional and classification which was the minimum mapping unit (MMU) used for the
errors might still remain. analysis. We assumed that mangrove areas smaller than our
Change maps were generated by subtracting the classifica- MMU have no significance on the total mangrove area of the
tion maps between six periods: 1975–90, 1975–2000, region.
522 Journal of Biogeography 35, 519–528
ª 2007 The Authors. Journal compilation ª 2007 Blackwell Publishing Ltd
Mangrove forest distribution and dynamics
The largest percentage of the remaining mangrove forest mangrove forests along the Gulf of Thailand have disappeared
areas in our study area in 2005 is located in Burma in the last 30 years (Thampanya et al., 2006).
(551,361 ha; 33%), followed by Bangladesh (438,764 ha; This rate of deforestation was not uniform through space
27%), India (337,727 ha; 21%), Thailand (168,910 ha; 10%), and time. The annual rate of deforestation during 1975–2005
Malaysia (70,560 ha; 4%), Indonesia (68,194 ha; 4%) and Sri was highest (c. 1%) in Burma compared with Thailand
Lanka (10,379 ha; 1%). The largest expanse of mangrove ()0.73%), Indonesia ()0.33%), Malaysia ()0.2%) and Sri
forests is located in the Sundarbans (along the border between Lanka ()0.08%). In contrast, mangrove forests in Bangladesh
Bangladesh and India); the Ayeyarwady Delta, Rakhine, and (+0.14%) and India (+0.04%) remained essentially unchanged
Tahinthayi (Burma); Phang Nga and Krabi (Thailand); and or slightly expanded during this period. The increase in
Matang (Malaysia). mangrove area in India that we found is consistent with
The largest tracts of the remaining forests of the region are reports from the Forest Survey of India which stated that
in different states and conditions owing to different threats mangrove forest cover has increased or remained unchanged
and management interventions. Despite having the highest since 1995. However, almost all the mangrove areas in India
population density in the world in its immediate vicinity, the are severely degraded with reduced or negligible vegetation
areal extent of the mangrove forest in the Sundarbans has not cover (Wilkie & Fortuna, 2003). Bangladesh has started
changed significantly in the last 25 to 30 years (Giri et al., ambitious mangrove rehabilitation programmes, and man-
2007). A strong commitment from the governments of India grove forest areas have also increased by aggradation (Giri
and Bangladesh, in the form of various protection measures, et al., 2006). The reforestation programmes in both India and
such as forest reserves, wildlife sanctuaries, national parks and Bangladesh were initiated by the government and local
international designations, is responsible for keeping the communities.
Sundarbans mangrove forest relatively intact. This is an Net deforestation peaked at 137,000 ha (approximately 1%
excellent example of the coexistence of human, terrestrial and year)1) during 1990–2000, increasing from 97,000 ha
aquatic plant and animal life. In Phang Nga, mangrove forests (0.2% year)1) during 1975–90, and declining to 14,000 ha
are found all along the coast and on some larger islands, (0.06% year)1) during 2000–05. The main reason for the
including Ao Thalane and Ao Luk. Many areas in Phang Nga decline in the rate of deforestation is that the intensity of
and Krabi are protected under the conservation areas aquaculture expansion appears to have levelled off in all the
network. The Ao Phang Nga National Park covers an area countries except Burma and Indonesia. The highest rate of
of 4000 ha and represents the largest tract of remaining deforestation during 1976–90 occurred in Thailand (1.8%).
original primary mangrove forests of Thailand. All three The rate of deforestation in other countries was relatively low
mangrove areas in Burma are under immense pressure from during this period. However, during 1990–2000, the rate of
human exploitation. Forests that are not protected represent deforestation was highest in Burma (2.9%) and Malaysia
some of the most degraded or destroyed mangrove forests of (1.3%). Similarly, the deforestation rate during 2000–05 was
the region. Matang forest in Malaysia is intensely managed highest in Indonesia (0.75%), mainly because of the expan-
and is considered to be one of the best managed mangrove sion of aquaculture.
forests in the world. The forest, consisting mainly of At the local level, both deforestation and forest regeneration
Rhizophora apiculata, is the largest tract of mangrove forest occurred with varying intensities, with localized hotspots of
in Peninsular Malaysia (c. 40,000 ha). Approximately 80% of rapid change. We identified the major deforestation fronts that
the area is managed through a sustainable-yield production are located in the Ayeyarwady Delta, and Rakhine and
system with a 30-year rotation cycle. Smaller patches of Tahinthayi provinces of Burma; Sweetenham and Bagan in
mangrove forests are found in all seven countries with many Malaysia; Belawan, Pangkalanbrandan, and Langsa in Indone-
isolated patches in Sri Lanka and India. Many of these smaller sia; and Southern Krabi and Ranong in Thailand (Fig. 1).
patches are under immediate threat from human exploitation. Major reforestation and afforestation areas are located on the
Unfortunately, the majority of these forests are not protected south-eastern coast of Bangladesh, and in Pichavaram, Devi
under the existing protected areas network. In some cases, Mouth, and Godavari in India.
these small patches of forests are managed by local commu- The major causes of deforestation were agricultural expan-
nities. sion (81%), aquaculture (12%) and urban development (2%).
Time-series analysis of MSS, TM and ETM+ data has As expected, causes of deforestation also varied with space and
revealed a net loss of 12% of mangrove forests in the region time (Fig. 3 & Table 3). In Thailand, 41% (16,815 ha) of
from 1975 to 2005 (Fig. 1). The net deforestation resulted mangrove forests have been converted to aquaculture and
because the deforestation rate outpaced the afforestation/ another 2% (710 ha) have been converted to urban develop-
reforestation rate (Fig. 2). The overall net loss is lower than the ment. However, the largest factor was agricultural expansion
country average for Thailand, Malaysia and Indonesia, because (50%, 20,300 ha). Other factors, such as mining, are also
other parts of these countries are undergoing massive changes. responsible for deforestation in Thailand. These land conver-
For example, in Thailand, the Andaman coast experienced sions are particularly evident in Phuket, Ranong and southern
much less development pressure than the Gulf of Thailand Krabi. In Indonesia, 63% (20,960 ha) of the mangrove forests
(outside of our study area). Approximately 80–90% of have been cleared for shrimp ponds and another 32%
Journal of Biogeography 35, 519–528 523
ª 2007 The Authors. Journal compilation ª 2007 Blackwell Publishing Ltd
C. Giri et al.
Figure 1 Change in mangrove forest cover change from 1975 to 2005.
524 Journal of Biogeography 35, 519–528
ª 2007 The Authors. Journal compilation ª 2007 Blackwell Publishing Ltd
Mangrove forest distribution and dynamics
2,000
1,500
Area (in '000 ha)
1,000
500
0
–500
–1,000
1975–2005 1975–2000 1975–1990 1990–2000 1990–2005 2000–2005
No change (ha) 1,305,542 1,312,760 1,469,005 1,397,608 1,387,924 1,607,716
Deforestation (ha) –584,363 –577,884 –426,432 –398,695 –407,601 –50,212
Figure 2 Areal estimate of deforestation and
afforestation from 1975 to 2005. Afforestation (ha) 327,600 326,488 328,844 261,339 256,279 36,348
(10,625 ha) for agricultural lands. Conversion to urban destroyed or degraded by erosion and sedimentation (Barbier,
development accounted for approximately 4% (1420 ha). 2006); the delta has the fifth largest sedimentation in the
The large percentage of change in Malaysia is primarily world.
associated with the 30-year rotation cycle of clearcutting
mangrove forests in Matang Mangrove Reserve. The forests are
CONCLUSIONS
also being converted to urban areas.
The deforestation in Burma is due to the overexploitation of The results of our study will be useful in several ways. First, we
mangrove forests for fuel wood collection, charcoal production improved our understanding of the distribution of mangrove
and illegal logging, followed by encroachment for paddy forests in the region, and assessed the rates and causes of
cultivation. We estimated that 98% (293,035 ha) of mangrove deforestation. The geo-spatiotemporal data base generated by
deforestation in Burma during the period 1975–2005 was due this study provides an up-to-date, consistent and unbiased
to agricultural expansion (Fig. 4). During the same period, account of the extent, distribution and dynamics of mangrove
approximately 2% (6870 ha) of forests were converted to forests of the region, and has better spatial and temporal detail
aquaculture. In the Ayeyarwady Delta, forests are also being than the information previously available. Second, an unbiased
Figure 3 Spatial distribution of mangrove deforestation in Ayeyarwady Delta, Burma, during 1975–90, 1990–2000 and 2000–05.
Journal of Biogeography 35, 519–528 525
ª 2007 The Authors. Journal compilation ª 2007 Blackwell Publishing Ltd
C. Giri et al.
Table 3 Conversion of mangrove forest to other land use/cover categories.
Area converted [ha (%)]
To aquaculture To agriculture To urban To other
± Error ± Error ± Error ± Error
Country Area estimate Area estimate Area estimate Area estimate Total
Thailand 16,816 (41) 1009 (3) 20,296 (50) 1826 (5) 710 (2.1) 15 (2) 2745 (7) 220 (1) 40,567
Malaysia 1605 (7) 96 (0) 9605 (43) 864 (4) 4532 (20) 615 (14) 6557 (29) 525 (2) 22,299
Indonesia 20,956 (63) 1257 (4) 10,628 (32) 956 (3) 1420 (4) 60 (4) 0 (0) 0 (0) 33,004
Burma 6868 (2) 412 (0) 293,035 (98) 26,373 (9) 66 (0) 0 (0) 122 (0) 10 (0) 300,091
Bangladesh 1070 (11) 64 (1) 7193 (77) 647 (7) 0 (0) 0 (0) 1046 (11) 84 (1) 9309
India 7554 (22) 453 (1) 17,179 (50) 1546 (4) 168 (0) 1 (0) 9178 (27) 734 (2) 34,079
Sri Lanka 134 (1) 8 (0) 12,558 (92) 1130 (8) 26 (0) 0 (0) 901 (7) 72 (1) 13,619
Total 55,004 (12) 3300 (1) 370,495 (82) 33,344 (7) 6921 (2) 1437 (20) 20,549 (5) 1644 (0) 452,969
account of the status and trends of mangrove forest areas can sites are highly degraded by pollution and pesticides. Delin-
support region-wide decision-making on the distribution of eations of degraded mangrove forests are needed to promote
resources for the conservation and rehabilitation of mangrove regrowth and enrichment planting.
forests. Third, our regional analysis is a starting point from
which to assess the role of mangrove forests in saving lives and
property from natural disasters such as the Indian Ocean
tsunami of 2004.
Monitoring deforestation at a regional scale using moderate-
resolution satellite images over a long period of time requires
the processing of large volumes of data. We used simple but
efficient methods to analyse these data. This approach applied
semi-automated image analysis techniques to assess present
status and to monitor the rates and causes of change, and it
does so over a large area covering tsunami-affected regions of
Asia. Our analyses show the potential of producing consistent
and timely mangrove forest data bases of the region using the
historical archive of Landsat data.
Unlike many other areas in Asia, conversion to aquaculture
is not the major cause of mangrove deforestation in the region.
At the regional level, conversion to agriculture is the dominant Figure 4 Major causes of mangrove deforestation, by country.
factor, followed by conversion to aquaculture and urban
development. However, the degree to which these factors play
a role varies according to space and time. For example,
agriculture is the major factor in all the countries except in
Indonesia, where aquaculture is the dominant factor. In
Thailand, aquaculture accounts for 40% of mangrove defor-
estation, which is much higher than the regional average of
12%. Similarly, urban development is more dominant in
Malaysia.
Deforested and degraded mangrove areas can be rehabili-
tated and restored in some cases (Fig. 5). The identified
deforestation areas can be used to select potential rehabilita-
tion sites together with a matrix of criteria such as extent,
accessibility and socio-economic factors. Not all deforested
areas can be restored back to mangrove forests. For example,
urban areas are very unlikely to revert back to mangrove forest.
The majority of agricultural areas and some of the aquaculture
areas can be reforested. Other abandoned aquaculture areas are Figure 5 With some management intervention, this abandoned
very difficult to rehabilitate or regenerate, mainly because these shrimp pond can be rehabilitated (photo Chandra Giri).
526 Journal of Biogeography 35, 519–528
ª 2007 The Authors. Journal compilation ª 2007 Blackwell Publishing Ltd
Mangrove forest distribution and dynamics
ACKNOWLEDGEMENTS Bangladesh and India using multi-temporal satellite data
from 1973–2000. Estuarine, Coastal and Shelf Science, 73,
We would like to acknowledge four student interns for their
91–100.
help in classifying satellite images: Smita Srivastav, Supriti
Homer, C., Huang, C., Yang, L., Wylie, B. & Coan, M. (2004)
Shrestha, Arvind Pasula and Rene Siwe. We would also like to
Development of a 2001 national land cover database for the
acknowledge a number of people who helped us during the
United States. Photogrammetric Engineering and Remote
field data collection, including Anisara Pensuk (Thailand),
Sensing, 70, 829–840.
Rajendra Shrestha (Thailand), Professor Ong Jin Eong (Malay-
Huang, C. & Townshend, J.R.G. (2003) A stepwise regression
sia), Srimal Samansiri (Sri Lanka), Dolagovinda Prasad
tree for nonlinear approximation: applications to estimating
(India), A. K. Fazlul Hoque (Bangladesh) and Moe Myint
sub-pixel land cover. International Journal of Remote Sens-
(Burma).
ing, 24, 75–90.
IUCN (2005) Early observations of tsunami effects on mangroves
REFERENCES and coastal forests. The World Conservation Union (IUCN)
(available at: http://www.iucn.org/themes/wetlands/pdf/
Adeel, Z. & Pomeroy, R. (2002) Assessment and management
WaterWetlandsTsunami.pdf), Gland, Switzerland.
of mangrove ecosystems in developing countries. Trees, 16,
Kathiresan, K. & Rajendran, N. (2005) Mangrove ecosystems
235–238.
of the Indian Ocean region. Indian Journal of Marine
Aizpuru, M., Achard, F. & Blasco, F. (2000) Global assessment
Science, 34, 104–113.
of cover change of the mangrove forest using satellite imagery
McKee, K.L. (2005) Global change impacts on mangrove
at medium to high resolution. EEC research project no.
ecosystems (available at: http://www.nwrc.usgs.gov/factshts/
15017-1999-05 FIED ISP FR. Joint Research Center, Ispra.
2004-3125.pdf).
Aschbacher, A., Giri, C., Ofren, R., Tiangco, P.N., Suselo, T.B.,
Nigel, W. (2005) Tsunami insight to mangrove value. Current
Vibulsresth, S. & Charrupat, T. (1994) Tropical mangrove
Biology, 15, R73–R73.
vegetation mapping using advanced remote sensing and GIS
Park, S.K. & Schowengerdt, R.A. (1982) Image reconstruction
technology (main report). Asian Institute of Technology,
by parametric cubic convolution. Computer Vision, Graphics
National Research Council of Thailand, Royal Forest
and Image Processing, 23, 258–272.
Department, and UNEP-GRID, Bangkok.
Puyravaud, J.-P. (2003) Standardizing the calculation of the
Barbier, E.B. (2006) Natural barriers to natural disasters:
annual rate of deforestation. Forest Ecology and Manage-
re-planting mangrove after the tsunami. Frontiers in Ecology
ment, 177, 593–596.
and the Environment, 4, 124–131.
Ramasubramanian, R., Gnanappazham, L., Ravishankar, T. &
Blasco, F., Aizpuru, M. & Gers, C. (2001) Depletion of the
Navamuniyammal, M. (2006) Mangroves of Godavari –
mangroves of Continental Asia. Wetlands Ecology and
analysis through remote sensing approach. Wetlands Ecology
Management, 9, 245–256.
and Management, 14, 29–37.
Chander, G. & Markham, B. (2003) Revised Landsat-5 TM
Ramsey, E.W. & Jansen, J.R. (1996) Remote sensing of mangrove
radiometric calibration procedures and postcalibration
wetlands: relating canopy spectra to site-specific data. Photo-
dynamic ranges. IEEE Transactions on Geoscience and
grammetric Engineering and Remote Sensing, 62, 939–948.
Remote Sensing, 41, 2674–2677.
Selvam, V., Ravichandran, K.K., Gnanappazham, L. & Nava-
Civco, D.L., Hurd, J.D., Wilson, E.H., Song, M. & Zhang, Z.
muniyammal, M. (2003) Assessment of community based
(2002) A comparison of land use and land cover change
restoration of Pichavaram mangrove wetland using remote
detection methods. ASPRS-ACSM Annual Conference and
sensing data. Current Science, 85, 794–798.
FIG XXII Congress, 22–26 April, 2002. American Congress
Shlien, S. (1979) Geometric correction, registration and
on Surveying and Mapping (ACSM), MD, USA.
resampling of Landsat imagery. Canadian Journal of Remote
Dahdouh-Guebas, F., Jayatissa, L., Di Nito, D.J., Bosire, J., Lo
Sensing, 5, 75–89.
Seen, D. & Koedam, N. (2005) How effective were man-
Silapong, C. & Blasco, F. (1992) The application of geographic
groves as a defence against the recent tsunami? Current
information systems to mangrove forest management:
Biology, 15, R443–R447.
Khlung, Thailand. Asia Pacific Remote Sensing Journal, 5, 97–
Danielsen, F., Serensen, M.K., Olwig, M.F., Seklvam, V., Par-
104.
ish, F., Burgess, N.D., Hiraishi, T., Karunagaran, V.M.,
Singh, A. (1989) Review article: digital detection techniques
Rasmussen, M.S., Hansen, L.B., Quarto, A. & Suryadiputra,
using remotely-sensed data. International Journal of Remote
N. (2005) The Asian tsunami: a protective role for coastal
Sensing, 10, 989–1003.
vegetation. Science, 310, 643.
Spalding, M.D., Blasco, F. & Field, C.D. (eds) (1997) World
Giri, C. & Delsol, J.P. (1995) Mangrove forest cover mapping
mangrove atlas. The International Society for Mangrove
using remote sensing data in conjunction with GIS. Asian
Ecosystems, Okinawa.
and Pacific Remote Sensing Journal, 8, 13–26.
Stibig, H.-J., Belward, A.S., Roy, P.S., Rosalinea-Wasrin, U.,
Giri, C., Pengra, B., Zhu, Z., Singh, A. & Tieszen, L. (2007)
Agrawal, S., Joshi, P.K., Hildanus Beuchle, S., Fritz, S.,
Monitoring mangrove forest dynamics of the Sundarbans in
Journal of Biogeography 35, 519–528 527
ª 2007 The Authors. Journal compilation ª 2007 Blackwell Publishing Ltd
C. Giri et al.
Mubareka, S. & Giri, C. (2007) A land-cover map for South Wilkie, M.L., Fortuna, S. & Souksavat, O. (2003) FAO’s
and Southeast Asia derived from SPOT-VEGETATION database on mangrove area estimates. Forest Resources
data. Journal of Biogeography, 34, 625–637. Assessment Working Paper no. 62. Food and Agriculture
Thampanya, U., Vermaat, J.E., Sinsakul, S. & Panaptukkul, N. Organization of the United Nations (FAO), Rome.
(2006) Coastal erosion and mangrove progradation of
southern Thailand. Estuarine Coastal and Shelf Science, 68,
75–85.
UNEP (United Nations Environment Programme) (2004) BIOSKETCHES
Global environment outlook yearbook 2004. UNEP, Nairobi
(available at: http://www.unep.org/geo/yearbook/yb2004/ Chandra Giri is a principal scientist at Science Applications
007.htm). International Corporation (SAIC), contractor to the US
Vaiphasa, C., Skidmore, A.K. & Boer, W.F. (2006) A post- Geological Survey (USGS) Center for Earth Resources Obser-
classifier for mangrove mapping using ecological data. vation and Science (EROS). His work focuses on global and
ISPRS Journal of Photogrammetry & Remote Sensing, 61, continental-scale land use/land cover characterization and
1–10. mapping using remote sensing and Geographic Information
Valiela, I., Bowen, J.L. & York, J.K. (2001) Mangrove forests: Science (GIS).
one of the world’s threatened major tropical environments. Zhiliang Zhu is a research scientist with the US Geological
BioScience, 51, 807–815. Survey. His research focuses on mapping and modelling
Weiner, D. (2005) Tsunami and environment. Earth Island vegetation types and structure at different scales and assessing
Journal, 20, 44–46. spatial and temporal dynamics of ecosystems, including
Wilkie, M.L. & Fortuna, S. (2003) Status and trends of man- natural disturbances such as wildland fires.
grove area worldwide. Forest Resources Assessment Working
Paper no. 63. Food and Agriculture Organization of the
United Nations (FAO), Rome. Editor: David Bowman
528 Journal of Biogeography 35, 519–528
ª 2007 The Authors. Journal compilation ª 2007 Blackwell Publishing Ltd
ORIGINAL Mangrove forest distributions and
ARTICLE
dynamics (1975–2005) of the
tsunami-affected region of Asia
C. Giri1*, Z. Zhu2, L. L. Tieszen2, A. Singh3, S. Gillette4 and J. A. Kelmelis5
1
Science Application International Corporation ABSTRACT
(SAIC), contractor to U.S. Geological Survey
Aim We aimed to estimate the present extent of tsunami-affected mangrove
(USGS) Center for Earth Resources
Observation and Science (EROS), Sioux Falls,
forests and determine the rates and causes of deforestation from 1975 to 2005.
SD 57198, USA, 2USGS Center for Earth Location Our study region covers the tsunami-affected coastal areas of
Resources Observation and Science (EROS), Indonesia, Malaysia, Thailand, Burma (Myanmar), Bangladesh, India and Sri
Sioux Falls, SD 57198, USA, 3United Nations Lanka in Asia.
Environment Programme, Washington, DC
20006, USA, 4Colorado State University, Fort Methods We interpreted time-series Landsat data using a hybrid supervised
Collins, CO 80523, USA and 5US Geological and unsupervised classification approach. Landsat data were geometrically
Survey, VA and US Department of State, corrected to an accuracy of plus-or-minus half a pixel, an accuracy necessary
Washington, DC 20520, USA for change analysis. Each image was normalized for solar irradiance by converting
digital number values to the top-of-the atmosphere reflectance. Ground truth
data and existing maps and data bases were used to select training samples and
also for iterative labelling. We used a post-classification change detection
approach. Results were validated with the help of local experts and/or high-
resolution commercial satellite data.
Results The region lost 12% of its mangrove forests from 1975 to 2005, to a
present extent of c. 1,670,000 ha. Rates and causes of deforestation varied both
spatially and temporally. Annual deforestation was highest in Burma (c. 1%) and
lowest in Sri Lanka (0.1%). In contrast, mangrove forests in India and Bangladesh
remained unchanged or gained a small percentage. Net deforestation peaked at
137,000 ha during 1990–2000, increasing from 97,000 ha during 1975–90, and
declining to 14,000 ha during 2000–05. The major causes of deforestation were
agricultural expansion (81%), aquaculture (12%) and urban development (2%).
Main conclusions We assessed and monitored mangrove forests in the
tsunami-affected region of Asia using the historical archive of Landsat data. We
*Correspondence: Chandra Giri, Science also measured the rates of change and determined possible causes. The results of
Application International Corporation (SAIC), our study can be used to better understand the role of mangrove forests in saving
US Geological Survey (USGS) Center for Earth
lives and property from natural disasters such as the Indian Ocean tsunami, and
Resources Observation and Science (EROS),
Sioux Falls, SD 57198, USA. to identify possible areas for conservation, restoration and rehabilitation.
E-mail: cgiri@usgs.gov
Keywords
Work performed under USGS contract Change analysis, deforestation, image processing, Indian Ocean tsunami,
03CRCN0001. Landsat, mangrove forests.
and hurricanes. They also serve as breeding and nursing
INTRODUCTION
grounds for marine species and are sources of food, medicine,
Mangroves forests, distributed circumtropically in the inter- fuel and building materials for local communities. However,
tidal region between sea and land in the tropical and the forests have been declining at an alarming rate – perhaps
subtropical latitudes, provide important ecosystem goods even more rapidly than inland tropical forests (Aizpuru et al.,
and services. The forests help stabilize shorelines and reduce 2000) – and much of what remains is in degraded condition
the devastating impact of natural disasters, such as tsunamis (Valiela et al., 2001; Wilkie et al., 2003). Conversion of
ª 2007 The Authors www.blackwellpublishing.com/jbi 519
Journal compilation ª 2007 Blackwell Publishing Ltd doi:10.1111/j.1365-2699.2007.01806.x
C. Giri et al.
mangrove forests to aquaculture is on the rise in many rates and causes of change using multi-temporal satellite data
countries of the region without considering the fact that the and field observations. Our analysis sought to answer the
total economic value of intact mangrove forests is often higher following research questions: how much mangrove forest
than that of shrimp farming (Balmford et al., 2002). The remains; where are the remaining mangrove forests located;
remaining mangrove forests are under immense pressure from what is the rate of change; what are the main reasons for the
clearcutting, encroachment, hydrological alterations, chemical change?
spills, storms and climate change (Blasco et al., 2001; McKee,
2005).
Study area
The Indian Ocean tsunami of December 2004 and other
natural disasters have highlighted the importance of mangrove Our study area covers the coastal areas of Indonesia, Malaysia,
forests as a ‘bio-shield’ that protects vulnerable coastal Thailand, Burma (Myanmar), Bangladesh, India and Sri Lanka.
communities. Mangrove forests attenuated the Indian Ocean We chose this area for a number of reasons. First, this area was
tsunami waves and protected coastal communities in Indone- the most devastated during the Indian Ocean tsunami of
sia, Thailand, India and Sri Lanka (Danielsen et al., 2005; December 2004; as a result, many national governments and
IUCN, 2005; Kathiresan & Rajendran, 2005). In some areas, international organizations are now implementing ambitious
mangrove forests hit by the Indian Ocean tsunami suffered conservation and rehabilitation programmes. Second, the region
severe damage from breaking and uprooting. Recent findings contains approximately 10% of the total mangrove forests of the
suggest that the continued destruction and degradation of world, including the largest remaining contiguous mangrove
many mangrove forests throughout the tropics over the past forest in the world, the Sundarbans. Third, strong demographic
few decades has decreased the protective capacity of mangrove pressure and diverse climatic conditions in the region have
forest ecosystems and reduced their ability to rebound from created a mosaic of mangrove diversity that is changing
natural disasters (Dahdouh-Guebas et al., 2005; Nigel, 2005; constantly. Fourth, the region is the epicentre of mangrove
Weiner, 2005). However, accurate and reliable information on biodiversity and consists of many existing and planned national
the present extent of mangrove forests and the rate and causes parks, biosphere reserves and world heritage sties.
of deforestation in the tsunami-affected region of Asia has not The forest is under severe threat from both anthropogenic
been available (Adeel & Pomeroy, 2002; Danielsen et al., 2005; and natural forces. Anthropogenic threats include encroach-
UNEP, 2004). Such information is needed to better understand ment from expansion of agriculture (e.g. rice farming, coconut
the protective role of mangrove forests and to learn more and oil palm), aquaculture, urban development (e.g. resorts),
about deforestation dynamics, carbon fluxes, forest fragmen- mining, salt pan development and overexploitation of
tation and the provision of ecosystem goods and services. resources. Natural threats include erosion, sedimentation and
Remote sensing is an indispensable tool for assessing and sea level rise. Because of these threats, mangrove forest is the
monitoring mangrove forests, primarily because many man- most threatened habitat in the region. Only sporadic patches of
grove swamps are inaccessible or difficult to field survey. mangrove forests are left in India and Sri Lanka, and they have
Remote sensing provides synoptic coverage, and historical been depleted in Burma, Thailand, Malaysia and Indonesia.
satellite data dating back to the 1960s are available. Global In addition to deforestation, mangrove forests have been
mapping initiatives have failed to map the extent and rate of declining in biological diversity and economic value. Many
deforestation with sufficient detail because these studies have flora and fauna are vulnerable, near-threatened, threatened,
been based on satellite data with coarse spatial resolution endangered or critically endangered. Economic activities such
(1 km or coarser). For example, only extensive mangrove areas as extraction of timber and fuel wood, fishing and the
were mapped as part of the Global Land Cover 2000 survey collection of honey and other forest products have also
(Stibig et al., 2007). At local scales, several studies have used diminished.
moderate-resolution satellite data [e.g. Landsat, SPOT and the
India Remote Sensing Linear Imaging Self-scanning Sensor
DATA AND METHODOLOGY
(IRS LISS III)] to characterize and map mangrove forests
(Silapong & Blasco, 1992; Ramsey & Jansen, 1996; Blasco et al.,
Data acquisition
2001; Selvam et al., 2003; Ramasubramanian et al., 2006;
Vaiphasa et al., 2006). Synergistic use of optical and radar data We used Landsat GeoCover and recently acquired Enhanced
has been particularly useful in cloud-covered tropical man- Thematic Mapper Plus (ETM+) data, made available through
grove areas (Aschbacher et al., 1994; Giri & Delsol, 1995). the US Geological Survey (USGS) Center for Earth Resources
However, large areas of the tsunami-affected countries of Asia Observation and Science (EROS) (http://eros.usgs.gov). Geo-
(Indonesia, Malaysia, Thailand, Burma, Bangladesh, India and Cover is a collection of Landsat data with global coverage and
Sri Lanka) remain unmapped. As a result, the present extent of generally cloud-free images. Data were collected for three
mangrove forests, and the rate and causes of deforestation, are epochs: (1) the ‘1975’ imagery from 1973 to 1983, (2) the
unknown. ‘1990’ imagery from 1989 to 1993, and (3) the ‘2000’ imagery
We determined the extent and distribution of mangrove from 1997 to 2000. Detailed descriptions of GeoCover data
forests in the tsunami-affected countries and identified the can be found at https://zulu.ssc.nasa.gov/mrsid/. Additional
520 Journal of Biogeography 35, 519–528
ª 2007 The Authors. Journal compilation ª 2007 Blackwell Publishing Ltd
Mangrove forest distribution and dynamics
Multispectral Scanner (MSS) and Thematic Mapper (TM) data geometric correction error, noise arising from atmospheric
were acquired to supplement cloud-covered areas. The ‘2005’ effects, errors arising from changing illumination geometry
Enhanced Thematic Mapper Plus (ETM+) data collected from and instrument errors (Homer et al., 2004). Such errors are
2005 to 2006 were acquired from EROS. Approximately 216 likely to introduce biases and/or noise into mangrove forest
Landsat TM or ETM+ scenes each for 1990, 2000 or 2005, and classification and change analyses. Pre-processing is necessary
57 MSS scenes for 1975 were acquired for the study. Same-year to remove or minimize such errors.
and same-season data are best for this kind of study, but cloud- Landsat images acquired in the Universal Transverse Mer-
free images of the region were not available for all time cator (UTM) projection and coordinate system were
periods, prompting us to augment them with multi-season and re-registered to the Albers equal area projection. To improve
multi-year data. Twenty-four QuickBird scenes and eight the root mean square (RMS) error to ± 1/2 pixel, we used
IKONOS scenes, all collected in 2005 and 2006, were also additional ground control points (GCPs) collected from
acquired. 1:50,000-scale topographic maps. Images were resampled with
cubic convolution, which has a better spatial accuracy than the
commonly used nearest neighbour resampling technique
Field survey
(Shlien, 1979; Park & Schowengerdt, 1982).
We conducted a 4-week field survey in Malaysia, Thailand, Sri Removing or minimizing the presence of ‘noisy’ pixels is
Lanka and India during June and July 2006. A total of 182 also important. Noise in image data can be caused by several
calibration/validation points were collected. We used a Global factors: (1) differences in atmospheric scattering in the visible
Positioning System (GPS) to record the exact location of the bands, (2) differences in water and/or dust particles in the
survey site, and we took a photograph at every location. All the atmosphere, (3) temporal variations in the solar zenith and/or
photos were georeferenced using GPS-Photo Link (http:// azimuth angles, and (4) inconsistencies in sensor calibration
www.geospatialexperts.com/) and were used in tandem with for separate images (Homer et al., 2004). To reduce the noise
satellite imagery within Google Earth to verify mangrove caused by atmospheric effects and illumination geometry, we
locations and conditions. Data on presence or absence of applied the techniques developed by Homer et al. (2004) for
mangrove forests, and their condition, density, crown cover the US National Land Cover Database 2001. Each image was
and management regimes were collected during the field normalized for solar irradiance by converting digital number
survey. Density and crown cover were estimated visually. The values to the top-of-the atmosphere reflectance (Chander &
extent of damage caused by the Indian Ocean tsunami and any Markham, 2003). This conversion algorithm is ‘physically
recovery measures taken were also noted. The field data served based, automated, and does not introduce significant errors to
as training data, a portion of which served as independent the data’ (Huang & Townshend, 2003). A test of this technique
reference data for the verification of classification results. on mangrove areas in the Sundarbans (Giri et al., 2007)
We visited the national mapping agencies of the countries in showed it to be a reasonable pre-processing method for a large
the region to collect ancillary data such as forest classification data base covering several countries in Asia. Owing to the
maps, topographic maps and tsunami reports (Table 1). The unavailability of data on atmospheric conditions for the
team also visited local forestry departments to discuss various region, atmospheric correction was not performed. Cloud-
aspects of mangrove management. With their help, mangrove covered areas were replaced by additional Landsat scenes
and non-mangrove areas were delineated on hard-copy maps. obtained during the same period.
Local officials provided field data and information on mangrove
damage due to tsunami and afforestation/reforestation status.
Classification
They also guided the research team during the field visit.
Many image classification and change detection techniques
have been described in the literature (Singh, 1989; Civco et al.,
Pre-processing
2002). For change analysis, Civco et al. (2002) compared four
The use of multi-temporal satellite data (MSS, TM and ETM+) techniques – traditional post-classification, cross-correlation
at a subcontinent scale poses a number of challenges: analysis, neural networks and object-oriented classification
Table 1 Secondary data collected and used
Secondary data Country
during the study.
Mangrove forest maps prepared using Landsat data Thailand, India
Forest classification maps 1 : 250,000 using Sri Lanka, Malaysia, Indonesia
field inventory data
Land use/land cover maps at 1 : 1,000,000 scale Bangladesh, Malaysia
Topographic maps, 1 : 100,000 scale Sri Lanka, Thailand
QuickBird, IKONOS Selected areas in Thailand,
Bangladesh, India and Sri Lanka
Atlas of Mangrove Wetlands of India India
Journal of Biogeography 35, 519–528 521
ª 2007 The Authors. Journal compilation ª 2007 Blackwell Publishing Ltd
C. Giri et al.
Table 2 Class definitions. 1975–2005, 1990–2000, 1990–2005 and 2000–05. The change
areas were visually interpreted to identify the factors (e.g.
Classes Supervised classification class definitions
agriculture, aquaculture, urban development) responsible for
Mangrove Areas covered by both closed and open the change. Each of the change areas was checked individually
mangrove forests during the workshop by at least two experts to identify the
Non-mangrove Areas covered by croplands and other land uses causes of change. Published maps and high-resolution satellite
Barren lands Areas devoid of vegetation, e.g. sand dunes, data such as QuickBird and IKONOS were used for the
sediments or exposed soil purpose. A 3-pixel by 3-pixel window was used to identify the
Water bodies Areas of open water with no emergent vegetation, dominant land-cover types in the high-resolution satellite data.
e.g. channels and waterways Once the mangrove/non-mangrove areas were calculated for
each period, the rate of deforestation was calculated by using
– and concluded that each method has advantages and the following equation suggested by Puyravaud (2003):
disadvantages, and there is no single best way. Considering
1 A2
the large volume of data acquired from different time intervals, R¼ ln ð1Þ
t2 À t1 A1
we adopted the traditional post-classification approach.
For image classification, we used a hybrid supervised and where R is the rate of deforestation, A1 is the area at an initial
unsupervised classification approach because there was insuf- time t1 and A2 is the area at a later time t2. Maps and
ficient ground truth for a purely supervised classification. The photographs collected during the field visit, along with high-
images were not enhanced prior to unsupervised classification resolution IKONOS and QuickBird satellite data, were anal-
and the thermal band (band 6) was excluded from the ysed visually to discover the causes of deforestation. This
classification. Water bodies were mapped with a supervised involved visual analysis of change maps and reference data.
classification. We then used an ISODATA clustering algorithm
within ERDAS Imagine to generate 50 spectral clusters at the
Validation
99% convergence level. Through iterative labelling, mangrove
classes were identified and labelled with reference to field data Qualitative validation was performed with the help of local
and high-resolution QuickBird and IKONOS imagery, and experts and high-resolution satellite data such as QuickBird
then merged into a single mangrove category. Four land cover and IKONOS. We divided the entire area into 500 · 500 grids
classes were generated: mangrove, non-mangrove, barren lands and checked each grid visually to identify and correct ‘gross’
and water bodies (Table 2). Post-classification editing such as errors inherent in the classified maps. Quantitative accuracy
‘recoding’ was performed to remove obvious errors. Each assessment was not performed because sufficient ground truth
classified image was resampled to 50 m to be consistent with data for historical dates are not easily available. A workshop
MSS data. However, this resampling did not improve the was held at which local experts validated the change areas and
spatial details of MSS data. Finally, four classification images the causes of change. The experts provided ground truth data
were produced, one each for 1975, 1990, 2000 and 2005. and photographs for a number of sites throughout the study
area.
Change analysis
RESULTS AND DISCUSSION
We used a post-classification change analysis technique to
compare classification results from the four epochs of imagery. To meet the need for accurate and reliable information on the
This approach is probably the most common and intuitive present extent of mangrove forests of the region, our study
change detection method, because it provides ‘from–to’ change prepared a geospatial data base of mangrove distribution for
information. However, our approach may have three sources the year 2005 using Landsat data. The data base provides an
of uncertainty: (1) semantic differences in class definitions up-to-date and consistent overview of the extent and distri-
between maps, (2) positional errors, and (3) classification bution of mangrove forests with better spatial and thematic
errors. To minimize the semantic differences in class defini- details than previous data sets.
tions, we specified the same number of classes for all four We estimate that in 2005 there were approximately 1.67
dates. To minimize positional errors, additional GCPs were million ha of mangrove forest remaining in the region. This
selected and RMS error was reduced to ± 1/2 pixel as explained estimate is higher than previous estimates by Spalding et al.
in the pre-processing section. Post-classification editing (i.e. (1997), probably because it is based on moderate-resolution
recoding) using secondary data helped to correct classification Landsat data on which we were able to identify many small
errors. In doing so, problematic areas were identified visually (0.81 ha), previously unidentified, mangrove areas. However,
and using area of interest (AOI) in ERDAS Imagine followed we did not map mangrove patches smaller than the 1 ha,
by recoding. However, some positional and classification which was the minimum mapping unit (MMU) used for the
errors might still remain. analysis. We assumed that mangrove areas smaller than our
Change maps were generated by subtracting the classifica- MMU have no significance on the total mangrove area of the
tion maps between six periods: 1975–90, 1975–2000, region.
522 Journal of Biogeography 35, 519–528
ª 2007 The Authors. Journal compilation ª 2007 Blackwell Publishing Ltd
Mangrove forest distribution and dynamics
The largest percentage of the remaining mangrove forest mangrove forests along the Gulf of Thailand have disappeared
areas in our study area in 2005 is located in Burma in the last 30 years (Thampanya et al., 2006).
(551,361 ha; 33%), followed by Bangladesh (438,764 ha; This rate of deforestation was not uniform through space
27%), India (337,727 ha; 21%), Thailand (168,910 ha; 10%), and time. The annual rate of deforestation during 1975–2005
Malaysia (70,560 ha; 4%), Indonesia (68,194 ha; 4%) and Sri was highest (c. 1%) in Burma compared with Thailand
Lanka (10,379 ha; 1%). The largest expanse of mangrove ()0.73%), Indonesia ()0.33%), Malaysia ()0.2%) and Sri
forests is located in the Sundarbans (along the border between Lanka ()0.08%). In contrast, mangrove forests in Bangladesh
Bangladesh and India); the Ayeyarwady Delta, Rakhine, and (+0.14%) and India (+0.04%) remained essentially unchanged
Tahinthayi (Burma); Phang Nga and Krabi (Thailand); and or slightly expanded during this period. The increase in
Matang (Malaysia). mangrove area in India that we found is consistent with
The largest tracts of the remaining forests of the region are reports from the Forest Survey of India which stated that
in different states and conditions owing to different threats mangrove forest cover has increased or remained unchanged
and management interventions. Despite having the highest since 1995. However, almost all the mangrove areas in India
population density in the world in its immediate vicinity, the are severely degraded with reduced or negligible vegetation
areal extent of the mangrove forest in the Sundarbans has not cover (Wilkie & Fortuna, 2003). Bangladesh has started
changed significantly in the last 25 to 30 years (Giri et al., ambitious mangrove rehabilitation programmes, and man-
2007). A strong commitment from the governments of India grove forest areas have also increased by aggradation (Giri
and Bangladesh, in the form of various protection measures, et al., 2006). The reforestation programmes in both India and
such as forest reserves, wildlife sanctuaries, national parks and Bangladesh were initiated by the government and local
international designations, is responsible for keeping the communities.
Sundarbans mangrove forest relatively intact. This is an Net deforestation peaked at 137,000 ha (approximately 1%
excellent example of the coexistence of human, terrestrial and year)1) during 1990–2000, increasing from 97,000 ha
aquatic plant and animal life. In Phang Nga, mangrove forests (0.2% year)1) during 1975–90, and declining to 14,000 ha
are found all along the coast and on some larger islands, (0.06% year)1) during 2000–05. The main reason for the
including Ao Thalane and Ao Luk. Many areas in Phang Nga decline in the rate of deforestation is that the intensity of
and Krabi are protected under the conservation areas aquaculture expansion appears to have levelled off in all the
network. The Ao Phang Nga National Park covers an area countries except Burma and Indonesia. The highest rate of
of 4000 ha and represents the largest tract of remaining deforestation during 1976–90 occurred in Thailand (1.8%).
original primary mangrove forests of Thailand. All three The rate of deforestation in other countries was relatively low
mangrove areas in Burma are under immense pressure from during this period. However, during 1990–2000, the rate of
human exploitation. Forests that are not protected represent deforestation was highest in Burma (2.9%) and Malaysia
some of the most degraded or destroyed mangrove forests of (1.3%). Similarly, the deforestation rate during 2000–05 was
the region. Matang forest in Malaysia is intensely managed highest in Indonesia (0.75%), mainly because of the expan-
and is considered to be one of the best managed mangrove sion of aquaculture.
forests in the world. The forest, consisting mainly of At the local level, both deforestation and forest regeneration
Rhizophora apiculata, is the largest tract of mangrove forest occurred with varying intensities, with localized hotspots of
in Peninsular Malaysia (c. 40,000 ha). Approximately 80% of rapid change. We identified the major deforestation fronts that
the area is managed through a sustainable-yield production are located in the Ayeyarwady Delta, and Rakhine and
system with a 30-year rotation cycle. Smaller patches of Tahinthayi provinces of Burma; Sweetenham and Bagan in
mangrove forests are found in all seven countries with many Malaysia; Belawan, Pangkalanbrandan, and Langsa in Indone-
isolated patches in Sri Lanka and India. Many of these smaller sia; and Southern Krabi and Ranong in Thailand (Fig. 1).
patches are under immediate threat from human exploitation. Major reforestation and afforestation areas are located on the
Unfortunately, the majority of these forests are not protected south-eastern coast of Bangladesh, and in Pichavaram, Devi
under the existing protected areas network. In some cases, Mouth, and Godavari in India.
these small patches of forests are managed by local commu- The major causes of deforestation were agricultural expan-
nities. sion (81%), aquaculture (12%) and urban development (2%).
Time-series analysis of MSS, TM and ETM+ data has As expected, causes of deforestation also varied with space and
revealed a net loss of 12% of mangrove forests in the region time (Fig. 3 & Table 3). In Thailand, 41% (16,815 ha) of
from 1975 to 2005 (Fig. 1). The net deforestation resulted mangrove forests have been converted to aquaculture and
because the deforestation rate outpaced the afforestation/ another 2% (710 ha) have been converted to urban develop-
reforestation rate (Fig. 2). The overall net loss is lower than the ment. However, the largest factor was agricultural expansion
country average for Thailand, Malaysia and Indonesia, because (50%, 20,300 ha). Other factors, such as mining, are also
other parts of these countries are undergoing massive changes. responsible for deforestation in Thailand. These land conver-
For example, in Thailand, the Andaman coast experienced sions are particularly evident in Phuket, Ranong and southern
much less development pressure than the Gulf of Thailand Krabi. In Indonesia, 63% (20,960 ha) of the mangrove forests
(outside of our study area). Approximately 80–90% of have been cleared for shrimp ponds and another 32%
Journal of Biogeography 35, 519–528 523
ª 2007 The Authors. Journal compilation ª 2007 Blackwell Publishing Ltd
C. Giri et al.
Figure 1 Change in mangrove forest cover change from 1975 to 2005.
524 Journal of Biogeography 35, 519–528
ª 2007 The Authors. Journal compilation ª 2007 Blackwell Publishing Ltd
Mangrove forest distribution and dynamics
2,000
1,500
Area (in '000 ha)
1,000
500
0
–500
–1,000
1975–2005 1975–2000 1975–1990 1990–2000 1990–2005 2000–2005
No change (ha) 1,305,542 1,312,760 1,469,005 1,397,608 1,387,924 1,607,716
Deforestation (ha) –584,363 –577,884 –426,432 –398,695 –407,601 –50,212
Figure 2 Areal estimate of deforestation and
afforestation from 1975 to 2005. Afforestation (ha) 327,600 326,488 328,844 261,339 256,279 36,348
(10,625 ha) for agricultural lands. Conversion to urban destroyed or degraded by erosion and sedimentation (Barbier,
development accounted for approximately 4% (1420 ha). 2006); the delta has the fifth largest sedimentation in the
The large percentage of change in Malaysia is primarily world.
associated with the 30-year rotation cycle of clearcutting
mangrove forests in Matang Mangrove Reserve. The forests are
CONCLUSIONS
also being converted to urban areas.
The deforestation in Burma is due to the overexploitation of The results of our study will be useful in several ways. First, we
mangrove forests for fuel wood collection, charcoal production improved our understanding of the distribution of mangrove
and illegal logging, followed by encroachment for paddy forests in the region, and assessed the rates and causes of
cultivation. We estimated that 98% (293,035 ha) of mangrove deforestation. The geo-spatiotemporal data base generated by
deforestation in Burma during the period 1975–2005 was due this study provides an up-to-date, consistent and unbiased
to agricultural expansion (Fig. 4). During the same period, account of the extent, distribution and dynamics of mangrove
approximately 2% (6870 ha) of forests were converted to forests of the region, and has better spatial and temporal detail
aquaculture. In the Ayeyarwady Delta, forests are also being than the information previously available. Second, an unbiased
Figure 3 Spatial distribution of mangrove deforestation in Ayeyarwady Delta, Burma, during 1975–90, 1990–2000 and 2000–05.
Journal of Biogeography 35, 519–528 525
ª 2007 The Authors. Journal compilation ª 2007 Blackwell Publishing Ltd
C. Giri et al.
Table 3 Conversion of mangrove forest to other land use/cover categories.
Area converted [ha (%)]
To aquaculture To agriculture To urban To other
± Error ± Error ± Error ± Error
Country Area estimate Area estimate Area estimate Area estimate Total
Thailand 16,816 (41) 1009 (3) 20,296 (50) 1826 (5) 710 (2.1) 15 (2) 2745 (7) 220 (1) 40,567
Malaysia 1605 (7) 96 (0) 9605 (43) 864 (4) 4532 (20) 615 (14) 6557 (29) 525 (2) 22,299
Indonesia 20,956 (63) 1257 (4) 10,628 (32) 956 (3) 1420 (4) 60 (4) 0 (0) 0 (0) 33,004
Burma 6868 (2) 412 (0) 293,035 (98) 26,373 (9) 66 (0) 0 (0) 122 (0) 10 (0) 300,091
Bangladesh 1070 (11) 64 (1) 7193 (77) 647 (7) 0 (0) 0 (0) 1046 (11) 84 (1) 9309
India 7554 (22) 453 (1) 17,179 (50) 1546 (4) 168 (0) 1 (0) 9178 (27) 734 (2) 34,079
Sri Lanka 134 (1) 8 (0) 12,558 (92) 1130 (8) 26 (0) 0 (0) 901 (7) 72 (1) 13,619
Total 55,004 (12) 3300 (1) 370,495 (82) 33,344 (7) 6921 (2) 1437 (20) 20,549 (5) 1644 (0) 452,969
account of the status and trends of mangrove forest areas can sites are highly degraded by pollution and pesticides. Delin-
support region-wide decision-making on the distribution of eations of degraded mangrove forests are needed to promote
resources for the conservation and rehabilitation of mangrove regrowth and enrichment planting.
forests. Third, our regional analysis is a starting point from
which to assess the role of mangrove forests in saving lives and
property from natural disasters such as the Indian Ocean
tsunami of 2004.
Monitoring deforestation at a regional scale using moderate-
resolution satellite images over a long period of time requires
the processing of large volumes of data. We used simple but
efficient methods to analyse these data. This approach applied
semi-automated image analysis techniques to assess present
status and to monitor the rates and causes of change, and it
does so over a large area covering tsunami-affected regions of
Asia. Our analyses show the potential of producing consistent
and timely mangrove forest data bases of the region using the
historical archive of Landsat data.
Unlike many other areas in Asia, conversion to aquaculture
is not the major cause of mangrove deforestation in the region.
At the regional level, conversion to agriculture is the dominant Figure 4 Major causes of mangrove deforestation, by country.
factor, followed by conversion to aquaculture and urban
development. However, the degree to which these factors play
a role varies according to space and time. For example,
agriculture is the major factor in all the countries except in
Indonesia, where aquaculture is the dominant factor. In
Thailand, aquaculture accounts for 40% of mangrove defor-
estation, which is much higher than the regional average of
12%. Similarly, urban development is more dominant in
Malaysia.
Deforested and degraded mangrove areas can be rehabili-
tated and restored in some cases (Fig. 5). The identified
deforestation areas can be used to select potential rehabilita-
tion sites together with a matrix of criteria such as extent,
accessibility and socio-economic factors. Not all deforested
areas can be restored back to mangrove forests. For example,
urban areas are very unlikely to revert back to mangrove forest.
The majority of agricultural areas and some of the aquaculture
areas can be reforested. Other abandoned aquaculture areas are Figure 5 With some management intervention, this abandoned
very difficult to rehabilitate or regenerate, mainly because these shrimp pond can be rehabilitated (photo Chandra Giri).
526 Journal of Biogeography 35, 519–528
ª 2007 The Authors. Journal compilation ª 2007 Blackwell Publishing Ltd
Mangrove forest distribution and dynamics
ACKNOWLEDGEMENTS Bangladesh and India using multi-temporal satellite data
from 1973–2000. Estuarine, Coastal and Shelf Science, 73,
We would like to acknowledge four student interns for their
91–100.
help in classifying satellite images: Smita Srivastav, Supriti
Homer, C., Huang, C., Yang, L., Wylie, B. & Coan, M. (2004)
Shrestha, Arvind Pasula and Rene Siwe. We would also like to
Development of a 2001 national land cover database for the
acknowledge a number of people who helped us during the
United States. Photogrammetric Engineering and Remote
field data collection, including Anisara Pensuk (Thailand),
Sensing, 70, 829–840.
Rajendra Shrestha (Thailand), Professor Ong Jin Eong (Malay-
Huang, C. & Townshend, J.R.G. (2003) A stepwise regression
sia), Srimal Samansiri (Sri Lanka), Dolagovinda Prasad
tree for nonlinear approximation: applications to estimating
(India), A. K. Fazlul Hoque (Bangladesh) and Moe Myint
sub-pixel land cover. International Journal of Remote Sens-
(Burma).
ing, 24, 75–90.
IUCN (2005) Early observations of tsunami effects on mangroves
REFERENCES and coastal forests. The World Conservation Union (IUCN)
(available at: http://www.iucn.org/themes/wetlands/pdf/
Adeel, Z. & Pomeroy, R. (2002) Assessment and management
WaterWetlandsTsunami.pdf), Gland, Switzerland.
of mangrove ecosystems in developing countries. Trees, 16,
Kathiresan, K. & Rajendran, N. (2005) Mangrove ecosystems
235–238.
of the Indian Ocean region. Indian Journal of Marine
Aizpuru, M., Achard, F. & Blasco, F. (2000) Global assessment
Science, 34, 104–113.
of cover change of the mangrove forest using satellite imagery
McKee, K.L. (2005) Global change impacts on mangrove
at medium to high resolution. EEC research project no.
ecosystems (available at: http://www.nwrc.usgs.gov/factshts/
15017-1999-05 FIED ISP FR. Joint Research Center, Ispra.
2004-3125.pdf).
Aschbacher, A., Giri, C., Ofren, R., Tiangco, P.N., Suselo, T.B.,
Nigel, W. (2005) Tsunami insight to mangrove value. Current
Vibulsresth, S. & Charrupat, T. (1994) Tropical mangrove
Biology, 15, R73–R73.
vegetation mapping using advanced remote sensing and GIS
Park, S.K. & Schowengerdt, R.A. (1982) Image reconstruction
technology (main report). Asian Institute of Technology,
by parametric cubic convolution. Computer Vision, Graphics
National Research Council of Thailand, Royal Forest
and Image Processing, 23, 258–272.
Department, and UNEP-GRID, Bangkok.
Puyravaud, J.-P. (2003) Standardizing the calculation of the
Barbier, E.B. (2006) Natural barriers to natural disasters:
annual rate of deforestation. Forest Ecology and Manage-
re-planting mangrove after the tsunami. Frontiers in Ecology
ment, 177, 593–596.
and the Environment, 4, 124–131.
Ramasubramanian, R., Gnanappazham, L., Ravishankar, T. &
Blasco, F., Aizpuru, M. & Gers, C. (2001) Depletion of the
Navamuniyammal, M. (2006) Mangroves of Godavari –
mangroves of Continental Asia. Wetlands Ecology and
analysis through remote sensing approach. Wetlands Ecology
Management, 9, 245–256.
and Management, 14, 29–37.
Chander, G. & Markham, B. (2003) Revised Landsat-5 TM
Ramsey, E.W. & Jansen, J.R. (1996) Remote sensing of mangrove
radiometric calibration procedures and postcalibration
wetlands: relating canopy spectra to site-specific data. Photo-
dynamic ranges. IEEE Transactions on Geoscience and
grammetric Engineering and Remote Sensing, 62, 939–948.
Remote Sensing, 41, 2674–2677.
Selvam, V., Ravichandran, K.K., Gnanappazham, L. & Nava-
Civco, D.L., Hurd, J.D., Wilson, E.H., Song, M. & Zhang, Z.
muniyammal, M. (2003) Assessment of community based
(2002) A comparison of land use and land cover change
restoration of Pichavaram mangrove wetland using remote
detection methods. ASPRS-ACSM Annual Conference and
sensing data. Current Science, 85, 794–798.
FIG XXII Congress, 22–26 April, 2002. American Congress
Shlien, S. (1979) Geometric correction, registration and
on Surveying and Mapping (ACSM), MD, USA.
resampling of Landsat imagery. Canadian Journal of Remote
Dahdouh-Guebas, F., Jayatissa, L., Di Nito, D.J., Bosire, J., Lo
Sensing, 5, 75–89.
Seen, D. & Koedam, N. (2005) How effective were man-
Silapong, C. & Blasco, F. (1992) The application of geographic
groves as a defence against the recent tsunami? Current
information systems to mangrove forest management:
Biology, 15, R443–R447.
Khlung, Thailand. Asia Pacific Remote Sensing Journal, 5, 97–
Danielsen, F., Serensen, M.K., Olwig, M.F., Seklvam, V., Par-
104.
ish, F., Burgess, N.D., Hiraishi, T., Karunagaran, V.M.,
Singh, A. (1989) Review article: digital detection techniques
Rasmussen, M.S., Hansen, L.B., Quarto, A. & Suryadiputra,
using remotely-sensed data. International Journal of Remote
N. (2005) The Asian tsunami: a protective role for coastal
Sensing, 10, 989–1003.
vegetation. Science, 310, 643.
Spalding, M.D., Blasco, F. & Field, C.D. (eds) (1997) World
Giri, C. & Delsol, J.P. (1995) Mangrove forest cover mapping
mangrove atlas. The International Society for Mangrove
using remote sensing data in conjunction with GIS. Asian
Ecosystems, Okinawa.
and Pacific Remote Sensing Journal, 8, 13–26.
Stibig, H.-J., Belward, A.S., Roy, P.S., Rosalinea-Wasrin, U.,
Giri, C., Pengra, B., Zhu, Z., Singh, A. & Tieszen, L. (2007)
Agrawal, S., Joshi, P.K., Hildanus Beuchle, S., Fritz, S.,
Monitoring mangrove forest dynamics of the Sundarbans in
Journal of Biogeography 35, 519–528 527
ª 2007 The Authors. Journal compilation ª 2007 Blackwell Publishing Ltd
C. Giri et al.
Mubareka, S. & Giri, C. (2007) A land-cover map for South Wilkie, M.L., Fortuna, S. & Souksavat, O. (2003) FAO’s
and Southeast Asia derived from SPOT-VEGETATION database on mangrove area estimates. Forest Resources
data. Journal of Biogeography, 34, 625–637. Assessment Working Paper no. 62. Food and Agriculture
Thampanya, U., Vermaat, J.E., Sinsakul, S. & Panaptukkul, N. Organization of the United Nations (FAO), Rome.
(2006) Coastal erosion and mangrove progradation of
southern Thailand. Estuarine Coastal and Shelf Science, 68,
75–85.
UNEP (United Nations Environment Programme) (2004) BIOSKETCHES
Global environment outlook yearbook 2004. UNEP, Nairobi
(available at: http://www.unep.org/geo/yearbook/yb2004/ Chandra Giri is a principal scientist at Science Applications
007.htm). International Corporation (SAIC), contractor to the US
Vaiphasa, C., Skidmore, A.K. & Boer, W.F. (2006) A post- Geological Survey (USGS) Center for Earth Resources Obser-
classifier for mangrove mapping using ecological data. vation and Science (EROS). His work focuses on global and
ISPRS Journal of Photogrammetry & Remote Sensing, 61, continental-scale land use/land cover characterization and
1–10. mapping using remote sensing and Geographic Information
Valiela, I., Bowen, J.L. & York, J.K. (2001) Mangrove forests: Science (GIS).
one of the world’s threatened major tropical environments. Zhiliang Zhu is a research scientist with the US Geological
BioScience, 51, 807–815. Survey. His research focuses on mapping and modelling
Weiner, D. (2005) Tsunami and environment. Earth Island vegetation types and structure at different scales and assessing
Journal, 20, 44–46. spatial and temporal dynamics of ecosystems, including
Wilkie, M.L. & Fortuna, S. (2003) Status and trends of man- natural disturbances such as wildland fires.
grove area worldwide. Forest Resources Assessment Working
Paper no. 63. Food and Agriculture Organization of the
United Nations (FAO), Rome. Editor: David Bowman
528 Journal of Biogeography 35, 519–528
ª 2007 The Authors. Journal compilation ª 2007 Blackwell Publishing Ltd