Marine Pollution Bulletin 109 (2016) 734–743

Contents lists available at ScienceDirect

Marine Pollution Bulletin

journal homepage: www.elsevier.com/locate/marpolbul

Damage and recovery assessment of the Philippines' mangroves

following Super Typhoon Haiyan
Jordan Long a,⁎,2, Chandra Giri b, Jurgenne Primavera c, Mandar Trivedi d
a
InuTeq 1, Sioux Falls, SD 57198, USA
b
U.S. Geological Survey EROS Center, Sioux Falls, SD 57198, USA
c
Zoological Society of London, 48 Burgos Street, La Paz, 5000 Iloilo City, Philippines
d
Zoological Society of London, Regent's Park, London, NW1 4RY, United Kingdom

a r t i c l e i n f o a b s t r a c t

Article history: We quantified mangrove disturbance resulting from Super Typhoon Haiyan using a remote sensing approach.
Received 20 November 2015 Mangrove areas were mapped prior to Haiyan using 30 m Landsat imagery and a supervised decision-tree clas-
Received in revised form 15 June 2016 sification. A time sequence of 250 m eMODIS data was used to monitor mangrove condition prior to, and follow-
Accepted 22 June 2016
ing, Haiyan. Based on differences in eMODIS NDVI observations before and after the storm, we classified
Available online 7 July 2016
mangrove into three damage level categories: minimal, moderate, or severe. Mangrove damage in terms of ex-
Keywords:
tent and severity was greatest where Haiyan first made landfall on Eastern Samar and Western Samar provinces
Philippines mangrove and lessened westward corresponding with decreasing storm intensity as Haiyan tracked from east to west
Landsat across the Visayas region of the Philippines. However, within 18 months following Haiyan, mangrove areas clas-
eMODIS sified as severely, moderately, and minimally damaged decreased by 90%, 81%, and 57%, respectively, indicating
NDVI mangroves resilience to powerful typhoons.
Super Typhoon Haiyan © 2016 Elsevier Ltd. All rights reserved.
Mangrove disturbance

1. Introduction have resulted in extensive deforestation and degradation. Brackish-

water aquaculture, agriculture, salt production, infrastructure develop-
Mangroves are highly productive coastal ecosystems found within ment, and forest extraction are among the leading drivers of mangrove
the intertidal zones of the tropics and subtropics ranging from 35° N loss, worldwide (Food and Agriculture Organization of the United
to 40° S latitude (Giri et al., 2011b). Mangroves provide numerous eco- Nations, 2007; Primavera, 1995; Primavera, 2005a). Natural distur-
system goods and services that support human livelihoods and well- bances such as high water surge, tropical cyclones, tsunamis, and
being, and are critically important to nature and society (Alongi, 2002; wave action have also contributed to mangrove loss (Cornforth et al.,
Duke et al., 2007). Mangroves have traditionally been used by coastal 2013; Giri et al., 2008; Paling et al., 2008); however, mangroves are gen-
populations for construction materials, food, fuel, and medicine erally more resilient to natural perturbations than to human-induced
(Walters, 2005). In addition to providing suitable breeding and feeding disturbances (Jimenez et al., 1985).
grounds for numerous aquatic and avian species, undisturbed man- Typhoon Haiyan, known as “Super Typhoon Yolanda” in the Philip-
grove ecosystems provide coastal protection from erosion, tropical cy- pines, made landfall in Eastern Samar, Philippines, on November 8,
clones, and tsunamis (Alongi, 2008; Chmura et al., 2003; Lee et al., 2013. According to the Saffir-Simpson hurricane wind scale (SSHWS),
2014); are closely coupled to neighboring ecosystems (e.g., sea grasses Haiyan was rated as a category 5 hurricane with maximum sustained
and coral reefs) (Mumby et al., 2004); and store carbon (McLeod et winds of over 251 km/h and wind gusts greater than 300 km/h
al., 2011; Siikamäki et al., 2012; Twilley et al., 1992). (Fig. 1). The inordinately powerful typhoon damaged 1.1 million hous-
Despite the well-understood importance of mangroves, this biome ing structures, displaced 4.1 million people, and caused over 6000
has become among the most threatened in the past half century (Van deaths (National Disaster Risk Reduction and Management Council,
Lavieren et al., 2012). Globally, pressures from land use competition 2013; United States Agency for International Development, 2014). In
the aftermath of Typhoon Haiyan, the Philippines' Environment Secre-
tary announced plans for mangrove replanting and rehabilitation
⁎ Corresponding author. throughout the Philippines to provide bio-protection from future natu-
E-mail address: [email protected] (J. Long).
1
Contractor to the U.S. Geological Survey (USGS) Earth Resources Observation and
ral disasters similar to Haiyan (Department of Environment and Natural
Science (EROS) Center. Resources, 2013). Effective mangrove restoration requires precise infor-
2
Work performed under U.S. Geological Survey contract G13PC00028. mation on the geographic distribution as well as extent and severity of

http://dx.doi.org/10.1016/j.marpolbul.2016.06.080
0025-326X/© 2016 Elsevier Ltd. All rights reserved.
 J. Long et al. / Marine Pollution Bulletin 109 (2016) 734–743 735

Fig. 1. Typhoon Haiyan path and category in the Philippines (Hurrevac, 2014).

damage. Time-series remotely sensed measurements coupled with in was approximately 240,000 ha, with the greatest extent located in the
situ observations are needed to monitor mangrove damage and recov- provinces of Palawan, Sulu, and Siargao Island, Surigao del Norte
ery at a large spatial scale. Typically, studies of storm impacts on man- (Long et al., 2013). Mangrove biological diversity is relatively high
groves have used either remote sensing (Wang and D'Sa, 2009; Wang, with 35 true mangrove species (Food and Agriculture Organization of
2012) or re-surveys of forest plots (Kauffman and Cole, 2010; Smith et the United Nations, 2007); however, mangrove diversity is threatened
al., 2009); however, neither on their own provide a complete assess- and declining throughout the Philippines (Garcia et al., 2014; Richards
ment. Therefore, the aim of this study is to use remote sensing, coupled and Friess, 2016). The Philippines' total mangrove area decreased nearly
with ground observations, to map and quantify change in mangrove 11% from 1990 to 2010 (Long et al., 2013; Long and Giri, 2011). The
condition resulting from Typhoon Haiyan and monitor the subsequent principal threats are anthropogenic, with aquaculture and forest extrac-
temporal pattern of recovery. The data and information from this tion among the leading drivers of loss in recent decades (Primavera and
work are useful for developing rehabilitation and conservation strate- Esteban, 2008). Within the next century, mangrove loss and degrada-
gies and providing baseline information for future monitoring. tion is expected to continue as human populations increase and concen-
trate in coastal regions. The impacts of climate change, sea level rise, and
2. Materials and methods increased storm strength and frequency also contribute to expected
mangrove loss in the Philippines (Gilman et al., 2008).
2.1. Study area The Philippines has one of the highest disaster risk rankings in the
world due to its constant exposure to droughts, earthquakes, floods,
The Republic of the Philippines is an archipelago of 7107 islands lo- landslides, typhoons, and volcanic eruptions (UNU-EHS, 2014). Ty-
cated off the southeastern coast of Asia. With an extensive coastline phoons are the most frequently occurring natural disaster in the Philip-
comprising numerous low-wave energy intertidal bay areas and a trop- pines, historically impacting the greatest number of people and causing
ical climate, the Philippines provides ideal environmental conditions for the greatest amount of property damage (Bankoff, 2012). On average,
mangrove growth. In 2010, the total mangrove area of the Philippines 6.5 tropical typhoons make landfall in the Philippines per year (Joint
736 J. Long et al. / Marine Pollution Bulletin 109 (2016) 734–743

Typhoon Warning Center, 2008). The northern Luzon and central within 8 months prior to Haiyan's passing were selected from the U.S.
Visayas regions are the most frequently impacted during the wet season Geological Survey Earth Resources Observation and Science (USGS
from June to November, while typhoons impacting the southern Minda- EROS) Center archive through the Global Visualizing Viewer (GloVis)
nao region are less frequent. Mangroves in Luzon and Visayas make up (U.S. Geological Survey, 2016b) for 5 Landsat footprints (Table 1).
68% of the Philippines total mangrove area, and are exposed to recurrent Once Landsat imagery were visually selected, surface reflectance
disturbances such as high winds and tropical storms (Long et al., 2013). Landsat imagery were downloaded from the USGS EROS Science Pro-
Tropical cyclones potentially have long-term influences on the distribu- cessing Architecture (ESPA) on demand interface (U.S. Geological
tion and composition of mangroves. Wind effects can range from minor Survey, 2016d). Surface reflectance data provided via ESPA are generat-
defoliation to catastrophic uprooting of entire mangrove stands, while ed from the Landsat Ecosystem Disturbance Adaptive Processing Sys-
prolonged hydro-period and increased sediment deposition resulting tem (LEDAPS). The LEDAPS software applies the 6S radiative transfer
from storm surge can result in mortality from suffocation (Smith et al., model (Kotchenova et al., 2006) to Level-1 Landsat data products.
2009). Frequently disturbed mangroves are typically young succession- Landsat sensors have a relatively low temporal resolution with a
al forests and have a lower biomass than mature old growth mangroves, combined revisit time every 8 days, so regions of the world that are per-
but still provide important ecosystem goods and services (Twilley et al., sistently cloudy such as the Philippines, particularly during the rainy
1992). season (i.e., June through November), can be difficult to monitor with
single-date Landsat imagery. Therefore, several Landsat images were re-
2.2. Landsat mangrove classification quired per Landsat path/row footprint for complete mangrove classifi-
cation because of partial cloud cover. Landsat image pre-processing
Landsat 30 m resolution satellite imagery was used to map man- included stacking, masking cloudy pixels and cloud filling, and Normal-
grove areas in the storm-affected region prior to Typhoon Haiyan ized Difference Vegetation Index (NDVI) transformation (Fig. 2). Cloudy
(Fig. 2). Typhoon Haiyan transected 5 Landsat path/row footprints pixels were identified using the Fmask algorithm (Zhu and Woodcock,
(Fig. 3). Images from the Landsat 7 Enhanced Thematic Mapper Plus 2012) and filled using a Spectral Similarity Grouping (SSG) approach
(ETM+) and Landsat 8 Operational Land Imager (OLI) sensors captured with secondary Landsat imagery (Jin et al., 2013).

Fig. 2. Methods workflow for mapping impact of Typhoon Haiyan on the Philippines' mangrove forests.
 J. Long et al. / Marine Pollution Bulletin 109 (2016) 734–743 737

Fig. 3. Landsat path/row footprints over Typhoon Haiyan eye transect.

Mangrove areas were mapped using a supervised decision-tree clas- a 30 m Digital Elevation Model (DEM) derived from Shuttle Radar To-
sification applied to a multilayer data stack. The mapped region of inter- pography Mission (SRTM) (Farr et al., 2007), a Slope index generated
est included approximately 17.5% (42,098 ha) of the Philippines' total from the 30 m DEM, NDVI index, and a 30 m 2010 mangrove land
mangrove area (Fig. 3). Several studies have obtained high mangrove cover map consisting of three thematic classes: i) mangrove ii) water
land cover classification accuracy using a supervised decision-tree ap- and iii) terrestrial non-mangrove (Long et al., 2013). Band ratios, such
proach applied to Landsat imagery and vegetation indices (Giri et al., as NDVI, potentially increase spectral separability of mangrove from
2011a; Heumann, 2011b; Long et al., 2013; Zhang, 2011). For example, non-mangrove vegetation cover (Heumann, 2011a) and DEMs poten-
a 2010 national mangrove land cover map for the Philippines used a de- tially increase separability of low-lying mangrove from upland forest
cision-tree classification approach and has a reported overall accuracy (Jones et al., 2014; Lee et al., 2004). Following classification, we per-
of 93% (Long et al., 2013). Training data of known land cover samples formed a chi-squared test to determine which variables the decision-
for three target classes: i) mangrove ii) water and iii) terrestrial non- tree classifier used the most and found that the Landsat surface reflec-
mangrove, were collected from very high resolution imagery and tance bands were the highest used variables, followed by elevation
Landsat imagery. Very high resolution satellite data with a spatial reso- data layers, the 2010 mangrove land cover map, and NDVI (Table 2). It
lution of 5 m or less were downloaded from the USGS EROS Hazards is likely that NDVI values of mangrove were similar to other land
Data Distribution System (HDDS) Explorer (U.S. Geological Survey, cover types (e.g., upland forest and agriculture areas) and were ineffec-
2016c) and the National Geospatial Agency's (NGA) Web-based Access tive to the decision-tree classifier in this case study. Additionally, the
and Retrieval Portal (WARP) (National Geospatial-Intelligence Agency, thematic mangrove classification independent variable was not highly
2014). Very-high resolution data sensors included WorldView-2, utilized by the classifier likely because of differences in dependent
IKONOS-2, and QuickBird-2. The decision-tree ‘rule set’ was generated class labels and the independent class labels, most notably in the terres-
using See5 commercial software (Rulequest Research, 2011). Indepen- trial non-mangrove and water classes.
dent variables in the multilayer stack included Landsat spectral bands
1, 2, 3, 4, 5, and 7 for ETM+ sensor and 2, 3, 4, 5, 6, and 7 for OLI sensor,
2.3. Mangrove classification accuracy assessment

Table 1 An accuracy assessment was performed using a total of 250 random

Landsat imagery acquired for mangrove mapping analysis. points: 50 points for mangrove class, 100 points for terrestrial non-
Landsat path/row Satellite sensor Date acquired Table 2
P112 OLI 4/13/2013 Rank order of independent variables used by decision-tree classifier.
R52 OLI 4/29/2013
Rank Variable
P113 OLI 5/22/2013
R52 OLI 7/25/2013 1 ETM+ Band 1 OLI Band 2
OLI 8/10/2013 2 ETM+ Band 3 OLI Band 4
P114 ETM+ 4/3/2013 3 ETM+ Band 2 OLI Band 3
R52 OLI 9/2/2013 4 ETM+ Band 7 OLI Band 7
P115 ETM+ 8/16/2013 5 ETM+ Band 5 OLI Band 6
R52 OLI 7/7/2013 6 ETM+ Band 4 OLI Band 5
OLI 9/9/2013 7 SRTM DEM
P116 ETM+ 3/16/2013 8 DEM Slope
R52 ETM+ 4/1/2013 9 Mangrove 2010
ETM+ 5/19/2013 10 NDVI
738 J. Long et al. / Marine Pollution Bulletin 109 (2016) 734–743

Table 3
Error matrix for 2013 mangrove land cover classification.

Reference
Predicted
Mangrove Land Water Total User's %

Mangrove 45 1 0 46 97
Land 5 99 0 104 95
Water 0 0 100 100 100
Total 50 100 100
Producer's % 90 99 100
Overall 97

mangrove class, and 100 points for water class. The randomly generated Fig. 4. Phenology of mangrove area located in Palawan, Philippines (12°16′10.38″N, 119°
54′48.28″E). Monthly NDVI observations derived from 250 m 16-day MODIS NDVI
points were overlaid and compared with high resolution imagery from
products. NDVI remained between 7.5 and 9.5 during the 2-year observation period
Google Earth Pro™. High resolution imagery used in assessment were (2011–2012).
captured on 2/29/2012, 5/23/2012, 3/31/2013, 4/30/2013, 5/17/2013,
6/30/2013, and 8/26/2013. An error matrix was constructed to cross
tabulate the observed data with the reference data (Table 3) these pixels in a quality band (Jenkerson et al., 2010). We further pre-
(Congalton, 1991). The accuracy assessment indicates high classifica- processed the standard 10-day eMODIS data to generate NDVI time se-
tion accuracy, with an overall accuracy of 97%, a producer accuracy of ries composites prior to, and three time periods following, Typhoon
90% and a user accuracy of 97% for mangrove cover (Table 3). The ma- Haiyan: 6 months prior (April 15, 2013–November 4, 2013) (B); 1 to
jority of misclassification errors occurred where mangrove areas were 6 months after (November 21, 2013–May 26, 2014) (A1); 7 to
incorrectly mapped as terrestrial non-mangrove (i.e., land). 12 months after (June 5, 2014–November 6, 2014) (A2); and 12 to
18 months after (November 6, 2014 – May6, 2014) (A3). Our eMODIS
2.4. Time-series eMODIS pre-processing algorithm took raw 10-day eMODIS data and performed
data stacking, poor quality pixel removal based on the image quality
Expedited Moderate Resolution Imaging Spectroradiometer data provided, and calculated the median NDVI value per pixel to pro-
(eMODIS) data were used to monitor and quantify changes in mangrove duce a smoothed cloud-free NDVI composite for each of the three
condition resulting from Haiyan. The eMODIS collection is derived from time periods. Only data with known high quality indicated in the quality
250 m MODIS data (Jenkerson et al., 2010). MODIS provides surface re- layer were analyzed. Next, the 30 m mangrove map derived from
flectance data with a higher temporal resolution than Landsat, revisiting Landsat imagery was resampled to 250 m resolution using a nearest
once to twice per day, while Landsat combined sensor (ETM+ and OLI) neighbor resampling method and applied to mask areas where only
revisit time is every 8 days. The higher temporal resolution of MODIS is mangrove are located for all eMODIS time series composites (B, A1,
particularly advantageous in obtaining timely information and quality A2, and A3). The differences of B and A1, B and A2, and B and A3 NDVI
cloud-free observations in tropical regions like the Philippines, which composites were calculated to create NDVI change maps of mangrove
is not possible with Landsat data. Moderate resolution eMODIS data areas for the three time periods following Haiyan. Finally, we classified
were downloaded through EarthExplorer (U.S. Geological Survey, mangrove change into three damage level categories. Mangrove areas
2016a). The eMODIS datasets include acquisition, quality, and NDVI in- where NDVI decreased by 0.2 to 0.3, 0.3 to 0.5, and 0.5 or greater were
formation at 250 m spatial resolution. classified into Minimal (i.e., mostly intact, minor defoliation), Moderate
NDVI is one of the most widely applied proxies used to measure and (i.e., branches broken, tree trunks partially broken, but still standing and
monitor plant growth, vegetation cover, and biomass production and mostly defoliated), or Severe damage (i.e., trees uprooted or trunks bro-
has been extensively used for vegetation monitoring, crop yield assess- ken at base and completely defoliated) level categories, respectively.
ment, and drought detection (Ill et al., 1997; Peters et al., 2002;
Pettorelli, 2013; Tucker, 1979). NDVI exploits the contrasting character-
istics of the red and near-infrared (NIR) spectral bands and is calculated 2.5. Ground referencing
using the following formula:NDVI = (NIR − RED)/(NIR + RED). Values
for NDVI range from −1.0 to 1.0, where densely vegetated areas (e.g., Mangrove change maps were quality checked with independently
closed canopy tropical forest) generally yield high NDVI values (0.6 to collected field photos and field notes. Two field teams conducted field
0.9), sparsely vegetated areas (e.g., open shrub and grasslands) yield observations of mangrove damage from Haiyan in Eastern Samar and
moderate values (0.2 to 0.3), and non-vegetated surfaces (e.g., bare Leyte over a total of 7 days in January and March of 2014. With coordi-
rock, water, and snow) yield lower NDVI values (0.1 and lower) (U.S. nation by the Zoological Society of London (ZSL), the teams comprised
Geological Survey, 2015). Dense and healthy mangrove forests typically 17 participants from non-governmental organizations (NGOs), univer-
have NDVI values greater than 0.7 year round (Fig. 4). NDVI of man- sities, and the Philippine government. The locations surveyed included
grove forests remains high year round because they are largely ever- Quinapondan, Guiuan (Bagongbanwa Island), Salcedo (Maliwaliw Is-
green plants. Several studies have employed repeated measures of land and Abejao), General MacArthur, Hernani, and Lawaan in Eastern
NDVI to monitor mangrove vegetation response from varying distur- Samar, and Ormoc City, Palompon, Isabel, Merida, Carigara, Palo, and
bances (Giri et al., 2011a; Satyanarayana et al., 2011), but few have ap- Tacloban City in Leyte. Mangrove damage and recovery potential were
plied this approach to monitor mangrove disturbance from tropical assessed for 39 plots using two methods. First, counts of live seedlings,
cyclones (Wang, 2012). saplings, and trees were made in 10 × 10 m or 4 × 4 m plots. Second,
The eMODIS products available for our study region included 10-day the degree of defoliation and other damage of trees was classified as
composited data sets. NDVI can be sensitive to atmospheric scattering minimally damaged (i.e., intact), partially damaged (i.e., branches bro-
and absorption, as well as variations in the illumination and viewing ken, tree trunks partially broken but still standing and with or without
conditions (Kerekes, 1994); however, the eMODIS composite selection sprouts), or totally damaged (i.e., trees uprooted or trunks broken at
process minimizes these artifacts by filtering through input surface re- the base without sprouts). Mangrove damage maps produced from
flectance data and flagging poor quality pixels with negative values, this study were quality checked with ground observations by overlaying
clouds, snow cover, low view angles, or low sun angle and labeling georeferenced ground observations points over mangrove damage
 J. Long et al. / Marine Pollution Bulletin 109 (2016) 734–743 739

maps and comparing field note damage level classifications and The majority of severely and moderately damaged mangrove was
georeferenced photos with mapped damage level category. identified in Eastern Samar, Western Samar, and Leyte provinces, in-
cluding mangrove areas along Matarinao Bay north to Hernani and
3. Results from Guiuan west to Balangiga (Fig. 6). As Haiyan tracked east to west
it impacted mangroves in Leyte province, from Ormoc City west to
We found general patterns of mangrove damage with the greatest Isabel, and north to Tabango. Haiyan crossed the northern tip of Cebu,
decreases in NDVI values, indicating increasing severity of mangrove causing moderate damage to mangroves. Haiyan also caused moderate
disturbance, occurring in proximity to Typhoon Haiyan's eye transect. damage to mangroves in northern Iloilo, northern Capiz, and northern
According to Chen et al. (Chen et al., 2013), most tropical storms cause Aklan, although damage was minimal in extent because mangroves in
the greatest damage severity on the right and front side of the eye this region are highly fragmented from aquaculture development and
track, where wind and wave stress can be up to 25% greater than on mangrove area is relatively small. Haiyan's intensity lessened to a cate-
the left. Consistent with most tropical storms (Dahal et al., 2014), the gory 4 before passing over northern Palawan (Fig. 1); minimal and
greatest extent of moderately and severely damaged mangrove oc- moderate mangrove damage was detected here, but was negligible in
curred on the right side (i.e., north side of Haiyan transect) of the terms of area compared with mangroves in the Eastern Samar, Western
storm eye track and generally decreased with increasing distance from Samar, and Leyte provinces.
the eye path (Fig. 5); however, non-linear spatial patterns of mangrove Severely, moderately, and minimally damaged mangrove respec-
damage were observed in some areas. For example, mangroves in tively decreased by 90%, 81%, and 57% within 18 months following
Southwestern Matarinao Bay were less-severely damaged compared Haiyan (Fig. 7). Very high R-squared values near 1.0 indicate mangrove
with mangroves further north from the eye transect in Northern damage recovered near linearly for all damage level (Fig. 7). Based on
Matarinao Bay and near Hernani (Fig. 6). This non-linear damage likely linear extrapolation, if the remaining damaged mangrove continues to
resulted from variations in storm surge intensity, bathymetry and to- recover at constant rates, minimally damaged mangrove could poten-
pography, wind speed and direction, and type and condition of the tially recover completely within 12 months following time period A3
existing mangrove vegetation. Additionally, damaged mangrove extent and moderately and severely damaged mangrove could potentially re-
ranging from 0 to 10 km north of Haiyan was relatively low because cover completely within 6 months following time period A3; although,
mangrove extent is relatively low in this region (i.e. less than Wang (Wang, 2012) reported that mangrove recovery following a cate-
1500 ha.). Overall, mangrove damage was greatest in extent and sever- gory 5 hurricane disturbance in southern Florida did not recover linearly
ity 20 to 30 km north of Hiayan's eye path where Haiyan first made and recovery was greatest in the first year, slowed in the second year,
landfall in the eastern Philippines and lessened corresponding with de- and was minimal in the third year.
creasing storm intensity as Haiyan tracked from east to west (Fig. 5). A complete statistical validation was not performed on the man-
Total mangrove area initially impacted (i.e., mangrove experiencing grove damage maps because of inadequate reference data across the
NDVI decreases greater than or equal to 0.2 during time period A1) by study area; however, georeferenced field photos collected during inde-
Haiyan was 8568 ha, about 3.5% of the Philippines' total mangrove pendent field surveys following Haiyan indicate our results are consis-
area or 20.4% of the region of interest's total mangrove area. Nearly tent with on-the-ground mangrove condition observations (Primavera
870 ha of mangrove were severely damaged, 1820 ha were moderately et al., 2016). Fig. 8 (A) illustrates a mangrove area on Anahaw Island,
damaged, and 5900 ha were minimally damaged during this time peri- Eastern Samar, in Matarinao Bay where our change analysis indicated
od (Fig. 7). mangrove damage was minimal during time period A1. Ground obser-
vations showed mangrove cover remained intact and the field team
measured minimal observable defoliation or blowdown. Fig. 8 (B) illus-
trates a mangrove area on Tubabao Island, Guiuan, in the Leyte Gulf that
we classified as moderately damaged (NDVI decreased by 0.4) during
time period A1; consistent with field observations, Haiyan caused mod-
erate damage in this area. Fig. 8 (C) shows a mangrove area on Tubabao
Island, Guiuan, in the Leyte Gulf that we classified as severely damaged;
ground observation were consistent with damage maps indicating se-
vere damage to mangrove. Such field surveys found that the majority
severely damaged mangroves were plantations of Rhizophora while
most natural mangrove stands were recovering from minimal and mod-
erate damage (Primavera et al., 2016).

4. Discussion

Tropical storms are a frequently occurring phenomenon in the Phil-

ippines and greatly influence mangrove forest dynamics. Similar to
most mangroves worldwide, mangroves in the Philippines are mostly
located in remote and inaccessible regions, making it difficult, if not im-
possible, to conduct a comprehensive impact assessment from field sur-
veys alone. Remote sensing offers consistent, timely, and reliable
measurements that allow disturbance monitoring across large areas.
The results from this study indicate that a hybrid remote sensing ap-
proach using a combination of medium-resolution Landsat and time-se-
ries eMODIS data is effective for identifying storm damage severity and
recovery response of mangrove ecosystems. These findings are support-
ed by ground-referencing using field surveys at impacted sites. Our
analysis found that nearly 3.5% of the Philippines' total mangrove area
Fig. 5. Mangrove damage observed at 1 degree longitude intervals (A) and 10 km intervals was initially impacted by Haiyan. There was a high degree of spatial var-
north and south of Haiyan's eye path (B) during time period A1. iation in mangrove damage with damage found to be greatest in extent
740 J. Long et al. / Marine Pollution Bulletin 109 (2016) 734–743

Fig. 6. Mangrove damage across the Philippines. Mangrove damage was greatest in severity and extent in Eastern Samar where Haiyan made first landfall in the Philippines.

and severity in the Eastern Samar, Western Samar, and Leyte provinces any remote sensing methodology. These findings have implications for
where Haiyan initially made landfall. Of the total impacted mangrove government policy. The immediate response of the Philippine govern-
area, approximately 868 ha of mangrove experienced severe damage, ment to Haiyan was to allocate approximately 350 million Philippine
with Eastern Samar experiencing the majority of severe and moderate pesos (US $7.4 million dollars) to re-plant mangroves in thousands of
mangrove damage. hectares along hundreds of kilometers of coastline (Department of
Despite the initial damage caused to 3.5% of the Philippines' total Environment and Natural Resources, 2013). The present results suggest
mangrove area, the majority of damaged mangrove recovered within that mangrove recovery should be surveyed as late as 1.5 years
18 months following Super Typhoon Haiyan and, in the absence of fur- (18 months) following a category 5 typhoon to determine if mangrove
ther typhoons or other perturbations. These findings demonstrate the areas require rehabilitation. This study suggests that natural recovery
value of remote sensing as an ongoing monitoring tool to assess damage and regeneration would be a more economically and ecologically viable
and document recovery, and also indicate that some of the initial dam- strategy.
age was likely due to defoliation rather than mortality. Additionally, The high mangrove recovery rates we observed in a relatively short
ground referencing surveys provide an important validation of the amount of time (1.5 years) further demonstrate the resilience of man-
true status of mangroves post-typhoon and need to be integrated into groves to inordinately powerful tropical cyclones. These findings are
consistent with several studies affirming mangroves resilience and
high rate of recovery shortly following natural perturbations across
the globe (Alongi, 2008; Aung et al., 2013; Ayyappan et al., 2016;
Herbert et al., 1999; Wang, 2012). The resilience of the Philippines'
mangroves, and hence their ability to act as an important buffer for
coastal communities, could be enhanced through domestic policies
and measures to reduce recent rates of loss (i.e., 10.5% total mangrove
loss from 1990 to 2010) primarily resulting from anthropogenic activi-
ties (Long et al., 2013). Dating back to the 1970s, the Philippine green-
belt laws require mangrove buffer zones of 50–100 m facing open seas
and 20–50 m along riverbanks (e.g., Presidential Decree (PD) 705 of
1975, PD 953 of 1976, PD 1067, MNR Administrative Order (AO) 42 of
1986, DENR AO 76 of 1987), but these laws have been observed more
in the breach than in the compliance (Primavera, 2000). In the wake
of the 2004 tsunami, Primavera (Primavera, 2005b) called for their im-
plementation by planting mangrove greenbelts. Nearly a decade later
with the added urgency from the damage to life and property caused
by Typhoon Haiyan, this advocacy has resulted in the filing of the Na-
tional Coastal Greenbelt Bill in the Philippine Senate (National Coastal
Greenbelt Act of 2014, 2014), that mandates a 100 m wide band of man-
Fig. 7. Total mangrove area impacted by Typhoon Haiyan. Time periods after Haiyan: 1– grove vegetation. Remote sensing observations, as applied in this study,
6 months (A1), 7–12 month (A2), and 13–18 months (A3). can be a viable approach to documenting how effectively this legislation
 J. Long et al. / Marine Pollution Bulletin 109 (2016) 734–743 741

Fig. 8. (A) Mangrove area on Anahaw Island, Matarinao Bay in Eastern Samar where change analysis indicated minimal mangrove damage (NDVI decreased by 0.1 from B to A1 time
periods). Ground observation agreed with damage maps showing minimal damage to mangrove. (B) Mangrove area on Tubabao Island, Guiuan, in Leyte Gulf where change analysis
indicated moderate mangrove damage (NDVI decreased by 0.4 from B to A1 time periods). Consistent with field observations, Haiyan caused moderate damage (i.e., branches broken,
tree trunks partially broken but still standing and with shoots) in this location. (C) Mangrove area on Tubabao Island, Guiuan, in Leyte Gulf where change analysis indicated severe
mangrove damage (NDVI decreased by 0.6 from B to A1 time periods). Ground observation agreed with damage maps showing severe damage to mangrove. (Photos ZSL/Amado Blanco).

is implemented over time to restore coastal mangrove greenbelts in the The data and findings from this study are useful for developing man-
Philippines. grove rehabilitation and conservation strategies and provide baseline
It is vital that mangrove replanting activities are scientifically information for future monitoring. Long-term monitoring of mangrove
based to ensure the correct species are planted in the correct loca- areas impacted by Typhoon Haiyan using ground and remote sensing
tions and with the support of local communities to ensure ongoing observations is needed to fully understand the impacts on the
maintenance and protection (Primavera et al., 2012). Mangrove ecosystem.
replanting schemes are currently dominated by those resulting in
mono-type plantations (e.g., planting only Rhizophora) often with Author contributions
low survival rates, or impact on other important habitats such as
mudflats and seagrass beds (Alongi, 2002; Erftemeijer and Lewis, All listed authors contributed to study design. Authors Long and Giri
1999; Field, 1999; Walters, 2000). Further research is needed into performed analysis and Long prepared figures and wrote the first draft
the relative resilience of these mono-type species mangrove planta- of the manuscript. Giri, Primavera, and Trivedi revised the draft. Prima-
tions versus natural mixed species stands. Small-scale studies in vera and Filipino colleagues organized the field surveys, with sugges-
Panay have shown that severe damage from Typhoon Haiyan was tions from Trivedi. All authors read and approved the final version of
much higher in mono-type Rhizophora plantations than in diverse the manuscript.
natural mangrove forests (Primavera et al., 2016). Aung et al.
(2013) also observed higher mortality rates (N90%) among mono- Conflicts of Interest
type Rhizophora stands compared with non-Rhizophora stands
(b 20%) following a typhoon disturbance in Myanmar. The authors declare no conflicts of interest.
742 J. Long et al. / Marine Pollution Bulletin 109 (2016) 734–743

Acknowledgments Ill, E.W.R., Chappell, D.K., Baldwin, D.G., 1997. AVHRR imagery used to identify hurricane
damage in a forested wetland of Louisiana. Photogramm. Eng. Remote. Sens. 63,
293–297.
Any use of trade, firm, or product names is for descriptive purposes Jenkerson, C., Maiersperger, T., Schmidt, G., 2010. eMODIS: A User-friendly Data Source,
only and does not imply endorsement by the U.S. Government. Work per- in: Survey, U.S.G. (Ed.). Reston, Virginia, p. 10.
Jimenez, J.A., Lugo, A.E., Cintron, G., 1985. Tree mortality in mangrove forests. Biotropica
formed under U.S. Geological Survey contract G13PC00028. ZSL gratefully 17, 177–185.
acknowledge the efforts of the field team including representatives from Jin, S., Homer, C., Yang, L., Xian, G., Fry, J., Danielson, P., Townsend, P.A., 2013. Automated
Guiuan Development Foundation (GDFI), Environmental Leadership and cloud and shadow detection and filling using two-date Landsat imagery in the USA.
Int. J. Remote Sens. 34, 1540–1560.
Training Initiative (ELTI), Haribon Foundation, Tambuyog Development Center, Joint Typhoon Warning, 2008. Appendix B: Characteristics of Tropical Cyclones Af-
Center, Conservation International, University of the Philippines Diliman, fecting the Philippine Islands (Shoemaker 1991).
University of the Philippines Tacloban College, Ateneo University, La Salle Jones, T.G., Ratsimba, H.R., Ravaoarinorotsihoarana, L., Cripps, G., Bey, A., 2014. Ecological
variability and carbon stock estimates of mangrove ecosystems in northwestern
University, and the Department of Environment and Natural Resources
Madagascar. Forests 5, 177–205.
(DENR), Region 6. ZSL is also grateful to Christian Aid and the Darwin Ini- Kauffman, J.B., Cole, T., 2010. Micronesian mangrove forest structure and tree responses
tiative for funding the field surveys and Dr. Primavera. We appreciatively to a severe typhoon. Wetlands 30, 1077–1084.
acknowledge peer and anonymous reviewers. Kerekes, J.P., 1994. NDVI sensitivity to atmospheric water vapor as a function of spectral
bandwidth, Geoscience and Remote Sensing Symposium, 1994. IGARSS '94. Surface
and Atmospheric Remote Sensing: Technologies, Data Analysis and Interpretation.,
International. IEEE, pp. 1506–1508.
References Kotchenova, S.Y., Vermote, E.F., Matarrese, R., Klemm Jr., F.J., 2006. Validation of a vector
version of the 6S radiative transfer code for atmospheric correction of satellite data.
National Coastal Greenbelt Act of 2014 (Ed.), 2014. Senate Bill 2179, First ed. Part I: Path radiance. Applied optics Vol. 45, pp. 6762–6774.
Alongi, D.M., 2002. Present state and future of the world's mangrove forests. Environ. Lee, J., Shuai, Y., Zhu, Q., 2004. Using images combined with DEM in classifying forest veg-
Conserv. 29, 331–349. etations, Geoscience and Remote Sensing Symposium, 2004. IGARSS '04. Proceedings.
Alongi, D.M., 2008. Mangrove forests: resilience, protection from tsunamis, and responses 2004 IEEE International. IEEE, pp. 2362–2364.
to global climate change. Estuar. Coast. Shelf Sci. 76, 1–13. Lee, S.Y., Primavera, J.H., Dahdouh-Guebas, F., McKee, K., Bosire, J.O., Cannicci, S., Diele, K.,
Aung, T.T., Mochida, Y., Than, M.M., 2013. Prediction of recovery pathways of cyclone-dis- Fromard, F., Koedam, N., Marchand, C., Mendelssohn, I., Mukherjee, N., Record, S.,
turbed mangroves in the mega delta of Myanmar. For. Ecol. Manag. 293, 103–113. 2014. Ecological role and services of tropical mangrove ecosystems: a reassessment.
Ayyappan, N., Stephen, A., Muthusankar, G., Jeyakumar, S., Presena, J., 2016. Spatio-tem- Glob. Ecol. Biogeogr. 23, 726–743.
poral changes in fringe mangrove extent in Pondicherry, India after two phenomenal Long, J., Napton, D., Giri, C., Graesser, J., 2013. A mapping and monitoring assessment of
perturbations: tsunami and cyclone Thane. Trop. Ecol. 57, 361–368. the Philippines' mangrove forests from 1990 to 2010. J. Coast. Res. 260–271.
Bankoff, G., 2012. Cultures of Disaster: Society and Natural Hazard in the Philippines. Long, J.B., Giri, C., 2011. Mapping the Philippines' mangrove forests using Landsat imag-
Routledge. ery. Sensors 11, 2972–2981.
Chen, S.S., Zhao, W., Donelan, M.A., Tolman, H.L., 2013. Directional wind–wave coupling in McLeod, E., Chmura, G.L., Bouillon, S., Salm, R., Björk, M., Duarte, C.M., Lovelock, C.E.,
fully coupled atmosphere–wave–ocean models: results from CBLAST-Hurricane. Schlesinger, W.H., Silliman, B.R., 2011. A blueprint for blue carbon: toward an im-
J. Atmos. Sci. 70, 3198–3215. proved understanding of the role of vegetated coastal habitats in sequestering CO2.
Chmura, G.L., Anisfeld, S.C., Cahoon, D.R., Lynch, J.C., 2003. Global carbon sequestration in Front. Ecol. Environ. 9, 552–560.
tidal, saline wetland soils. Glob. Biogeochem. Cycles 17, 1111. Mumby, P.J., Edwards, A.J., Arias-González, J.E., Lindeman, K.C., Blackwell, P.G., Gall,
Congalton, R.G., 1991. A review of assessing the accuracy of classifications of remotely A., Gorczynska, M.I., Harborne, A.R., Pescod, C.L., Renken, H., 2004. Mangroves
sensed data. Remote Sens. Environ. 37, 35–46. enhance the biomass of coral reef fish communities in the Caribbean. Nature
Cornforth, W., Fatoyinbo, T., Freemantle, T., Pettorelli, N., 2013. Advanced Land Observing 427, 533–536.
Satellite Phased Array Type L-Band SAR (ALOS PALSAR) to inform the conservation of National Disaster Risk Reduction, Management Council, 2013. NDRRMC Update SitRep
mangroves: sundarbans as a case study. Remote Sens. 5, 224–237. No. 74 Effects of Typhoon “Yolanda” (Haiyan). http://www.ndrrmc.gov.ph/
Dahal, D., Liu, S., Oeding, J., 2014. The carbon cycle and hurricanes in the United States be- attachments/article/1125/updsitrep74.pdf.
tween 1900 and 2011. Sci. Rep. 4, 5197. National Geospatial-Intelligence Agency, 2014. Web-based Access and Retrieval Portal.
Department of Environment and Natural Resources, 2013. DENR sets aside P347M for https://warp.nga.mil.
coastal forest rehabilitation in Eastern Visayas. http://www.denr.gov.ph/news-and- Paling, E.I., Kobryn, H.T., Humphreys, G., 2008. Assessing the extent of mangrove change
features/latest-news/1604-denr-sets-aside-p347m-for-coastal-forest-rehabilitation- caused by Cyclone Vance in the eastern Exmouth Gulf, northwestern Australia.
in-eastern-visayas.html. Estuar. Coast. Shelf Sci. 77, 603–613.
Duke, N.C., Meynecke, J.-O., Dittmann, S., Ellison, A.M., Anger, K., Berger, U., Cannicci, S., Peters, A.J., Walter-Shea, E.A., Ji, L., Vina, A., Hayes, M., Svoboda, M.D., 2002. Drought mon-
Diele, K., Ewel, K.C., Field, C.D., 2007. A world without mangroves? Science 317, itoring with NDVI-based standardized vegetation index. Photogramm. Eng. Remote.
41–42. Sens. 68, 71–75.
Erftemeijer, P.L., Lewis, R.R., 1999. Planting Mangroves on Intertidal Mudflats: Habitat Pettorelli, N., 2013. The Normalized Difference Vegetation Index. Oxford University Press.
Restoration or Habitat Conversion, Proceedings of the ECOTONE VIII Seminar Enhanc- Primavera, J.H., 1995. Mangroves and Brackishwater Pond Culture in the Philippines, Asia-
ing Coastal Ecosystems Restoration for the 21st Century. Ranong, Thailand, Pacific Symposium on Mangrove Ecosystems. Springer, pp. 303–309.
pp. 23–28. Primavera, J.H., 2000. Development and conservation of Philippine mangroves: institu-
Farr, T.G., Rosen, P.A., Caro, E., Crippen, R., Duren, R., Hensley, S., Kobrick, M., Paller, M., tional issues. Ecol. Econ. 35, 91–106.
Rodriguez, E., Roth, L., 2007. The shuttle radar topography mission. Rev. Geophys. 45. Primavera, J.H., 2005a. Mangroves, fishponds, and the quest for sustainability. Science
Field, C., 1999. Rehabilitation of mangrove ecosystems: an overview. Mar. Pollut. Bull. 37, 310, 57–59.
383–392. Primavera, J.H., 2005b. Tsunamis and Mangrove Greenbelts The Philippine Star.
Food and Agriculture Organization of the United Nations, 2007. The World's Mangroves, Primavera, J.H., Cruz, M.D., Montilijiao, C., Consunji, H., Paz, M.D., Rollon, R., Maranan, K.,
1980–2005: A Thematic Study in the Framework of the Global Forest Resources As- Samson, M.C., Blanco, A., 2016. Preliminary Assessment of Post-Haiyan Mangrove
sessment 2005. Food and Agriculture Organization of the United Nations, Rome. Damage and Reocvery in Eastern Samar, Central Philippines (in Press).
Garcia, K.B., Malabrigo Jr., P.L., Gevaña, D.T., 2014. Philippines' mangrove ecosystem: sta- Primavera, J.H., Esteban, J., 2008. A review of mangrove rehabilitation in the Philippines:
tus, threats and conservation. In: Latiff, A., Hakeem, K.R., Ozturk, M. (Eds.), Faridah- successes, failures and future prospects. Wetl. Ecol. Manag. 16, 345–358.
Hanum, I. Mangrove Ecosystems of Asia, Springer New York, pp. 81–94. Primavera, J.H., Savaris, J., Bajoyo, B., Coching, J., Curnick, D., Golbeque, R., Henderin, A.G.J.,
Gilman, E.L., Ellison, J., Duke, N.C., Field, C., 2008. Threats to mangroves from climate Joven, R., Loma, R., Koldewey, H., 2012. Manual on Community-Based Mangrove Re-
change and adaptation options: A review. Aquat. Bot. 89, 237–250. habilitation, Mangrove Manual Series No. 1. Zoological Society of London, United
Giri, C., Long, J., Tieszen, L., 2011a. Mapping and monitoring Louisiana's mangroves in the Kingdom, p. 240.
aftermath of the 2010 Gulf of Mexico oil spill. J. Coast. Res. 27, 1059–1064. Richards, D.R., Friess, D.A., 2016. Rates and drivers of mangrove deforestation in Southeast
Giri, C., Ochieng, E., Tieszen, L.L., Zhu, Z., Singh, A., Loveland, T., Masek, J., Duke, N., 2011b. Asia, 2000–2012. Proc. Natl. Acad. Sci. 113, 344–349.
Status and distribution of mangrove forests of the world using earth observation sat- Rulequest Research, 2011. See5/c5.0. http://www.rulequest.com/see5-info.html.
ellite data. Glob. Ecol. Biogeogr. 20, 154–159. Satyanarayana, B., Mohamad, K.A., Idris, I.F., Husain, M.-L., Dahdouh-Guebas, F., 2011. As-
Giri, C., Zhu, Z., Tieszen, L., Singh, A., Gillette, S., Kelmelis, J., 2008. Mangrove forest distri- sessment of mangrove vegetation based on remote sensing and ground-truth mea-
butions and dynamics (1975–2005) of the tsunami-affected region of Asia†. surements at Tumpat, Kelantan Delta, East Coast of Peninsular Malaysia. Int.
J. Biogeogr. 35, 519–528. J. Remote Sens. 32, 1635–1650.
Herbert, D.A., Fownes, J.H., Vitousek, P.M., 1999. Hurricane damage to a Hawaiian forest: Siikamäki, J., Sanchirico, J.N., Jardine, S.L., 2012. Global economic potential for reducing
nutrient supply rate affects resistance and resilience. Ecology 80, 908–920. carbon dioxide emissions from mangrove loss. Proc. Natl. Acad. Sci. 109,
Heumann, B.W., 2011a. An object-based classification of mangroves using a hybrid deci- 14369–14374.
sion tree—support vector machine approach. Remote Sens. 3, 2440. Smith, T.J., Anderson, G.H., Balentine, K., Tiling, G., Ward, G.A., Whelan, K.R.T., 2009. Cu-
Heumann, B.W., 2011b. Satellite remote sensing of mangrove forests: Recent advances mulative impacts of hurricanes on Florida mangrove ecosystems: sediment deposi-
and future opportunities. Prog. Phys. Geogr. 35, 87–108. tion, storm surges and vegetation. Wetlands 29, 24–34.
Hurrevac, 2014. A storm tracking and decision support tool for government emergency Tucker, C.J., 1979. Red and photographic infrared linear combinations for monitoring veg-
managers, 1.4.2 ed. http://www.hurrevac.com. etation. Remote Sens. Environ. 8, 127–150.
 J. Long et al. / Marine Pollution Bulletin 109 (2016) 734–743 743

Twilley, R.R., Chen, R.H., Hargis, T., 1992. Carbon sinks in mangroves and their implica- Van Lavieren, H., Spalding, M., Alongi, D.M., Kainuma, M., Clüsener-Godt, M., Adeel, Z.,
tions to carbon budget of tropical coastal ecosystems. Water Air Soil Pollut. 64, 2012. Securing the future of mangroves.
265–288. Walters, B.B., 2000. Local mangrove planting in the Philippines: are fisherfolk and fish-
U.S. Geological Survey, 2015. Remote Sensing Phenology. http://phenology.cr.usgs.gov/ pond owners effective restorationists? Restor. Ecol. 8, 237–246.
ndvi_foundation.php. Walters, B.B., 2005. Patterns of local wood use and cutting of Philippine mangrove forests.
U.S. Geological Survey, 2016a. Earth Explorer. http://earthexplorer.usgs.gov. Econ. Bot. 59, 66–76.
U.S. Geological Survey, 2016b. Global Visualization Viewer (GloVis). http://glovis.usgs. Wang, F., D'Sa, E.J., 2009. Potential of MODIS EVI in identifying hurricane disturbance to
gov. coastal vegetation in the northern Gulf of Mexico. Remote Sens. 2, 1–18.
U.S. Geological Survey, 2016c. Hazards Data Distribution Systems Explorer. http:// Wang, Y., 2012. Detecting Vegetation Recovery Patterns After Hurricanes in South Florida
hddsexplorer.usgs.gov. Using NDVI Time Series.
U.S. Geological Survey, 2016d. LSDS Science Research and Development. http://espa.cr. Zhang, X.-H., 2011. Identification of Mangrove Using Decision Tree Method, Information
usgs.gov. and Computing (ICIC). 2011 Fourth International Conference on. IEEE, pp. 130–132.
United States Agency for International Development, 2014. Philippines - Typoon Yolanda/ Zhu, Z., Woodcock, C.E., 2012. Object-based cloud and cloud shadow detection in Landsat
Haiyan, FACT SHEET #22. FISCAL YEAR (FY). Vol. 2014, p. 7. imagery. Remote Sens. Environ. 118, 83–94.
UNU-EHS, 2014. World Risk Report. In: Lars Jeschonnek, M. (Ed.), United Nations Univer-
sity, Institute for Environment and Human Security.