Application of High-spatial-resolution Satellite Images to Monitoring Coral Reef Habitat Changes at Weno Island Chuuk, Micronesia

1. Introduction

Marine ecosystems worldwide are undergoing significant changes on a range of scales caused by climate change, the direct and indirect effects of human activities (Andrefouet et al., 2003; Hochberg et al., 2003). In particular, many coral reefs have been threatened and degraded by disturbances to their local environments, including rising sea temperatures and increased anthropogenic stresses (Bruno and Selig,

2007; Scopelitis et al., 2009). Thus, the monitoring of changes in reef habitat is inevitable for rehabilitation and prevention of the reef environments.

Coral reefs in tropical or subtropical environments perform a number of key roles within these ecosystems, such as providing shelter from tropical storms, habitats for reef fish, and areas of elevated biodiversity, and supplying building materials (Mumby and Edwards, 2002). These reefs are known to be indicators of global climate change and have provided information that

Application of High-spatial-resolution Satellite Images to Monitoring Coral Reef Habitat Changes

at Weno Island Chuuk, Micronesia

Jong-Kuk Choi

·Joo-Hyung Ryu

·Jee-Eun Min

Abstract: We present quantitative estimations of changes in the areal extent of coral reef habitats at Weno Island, Micronesia, using high-spatial-resolution remote sensing images and field observations. Coral reef habitat maps were generated from Kompsat-2 satellite images for September 2008 and September 2010, yielding classifications with 78.6% and 72.4% accuracy, respectively, which is a relatively high level of agreement. The difference between the number of pixels occupied by each seabed type was calculated, revealing that the areal extent of living corals decreased by 8.2 percentage points between 2008 and 2010. This result is consistent with a comparison of the seabed types determined by field observations. This study can be used as a basis for remediation planning to diminish the impact of changes in coral reefs.

Key Words: Coral reefs; Kompsat-2 satellite image; Habitat mapping; Monitoring changes in habitat coverage;

Federated States of Micronesia, Weno Island Article

Received July 12, 2021; Revised July 26, 2021; Accepted August 3, 2021; Published online August 25, 2021

1)

Principal researcher, Korean Ocean Satellite Center, Korea Institute of Ocean Science and Technology

2)

Senior Researcher, Oceanic Research Division, Underwater Survey Technology 21

†

Corresponding Author: Joo-Hyung Ryu ([email protected])

This is an Open-Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License

(http://creativecommons.org/licenses/by-nc/3.0) which permits unrestricted non-commercial use, distribution, and reproduction in

any medium, provided the original work is properly cited.

is important for the monitoring of pollution and environmental change, especially in tropical or subtropical oceans (Hyun et al., 2008). This indicates that existing coral reef habitats are of significant scientific interest, and, as such, many studies have examined the management and preservation of coral reef communities with the aim of preventing reef degradation.

For a precise estimation of coral reef habitat, we need as much field data as possible; however, routine field observations in the tropical area are seriously restricted because of limitations in accessibility and possible dimensions. Remote sensing is an effective tool for the mapping of coral reef habitats by delineating the lateral dimensions of coral reef structures, with this technique successfully used by the coral reef research community in a number of areas (Andrefouet and Riegl, 2004; Riegl and Purkis, 2005). Hochberg et al. (2003) undertook in situ optical reflectance spectra measurements to develop remote sensing-based techniques for identifying and mapping reef-bottom-types in the Atlantic, Pacific, and Indian oceans. Andrefouet et al. (2003) examined the potential use of high-spatial-resolution IKONOS multispectral imagery for coral-reef habitat mapping.

High-spatial-resolution multispectral Quickbird images have also been used to monitor coral reefs and their physical environment (Mumby et al., 2004).

The above examples, in which high-spatial-resolution satellite images were successfully used to accurately map coral-reef habitats, indicate that this method could be used to monitor long-term changes in coral reef habitats—an important research task. Here, we present quantitative estimates of changes in coral reef areal extents and distributions using high-spatial-resolution satellite images combined with in situ measurements.

Although IKONOS and Quickbird images have allowed highly accurate mapping of coral reef habitats, these images are too expensive to be easily obtained.

Instead, we use high-spatial-resolution Korea Multi- Purpose SATellite-2 (Kompsat-2) images, a platform

that produces images with similar characteristics as IKONOS-derived data, but for only approximately 5%

of the price for Korean researchers. Kompsat-2 has the same band composition and similar spatial resolution with IKONOS or Quickbird, and has been successfully employed to the coastal area studies (Choi et al., 2011a;

Choi et al., 2011b). Habitat maps were generated from two Kompsat-2 images acquired during September 2008 and August 2010. Coral-reef habitat classification was undertaken using an object-based approach, with the resulting maps used to calculate coral-reef areal extents that allow the determination of changes in coral reef habitats between the acquisition times for each image.

2. Materials and Methods

1) General description of the study area The study area was the Chuuk Lagoon in the north- eastern part of Weno Island, in the Federated States of Micronesia (FSM) (Fig. 1(a)). The FSM consists of four states (Chuuk, Yap, Pohnpei, and Kosrae), with Chuuk (07°28’N, 151°53’E) containing Chuuk Lagoon, which is one of the largest lagoons in Micronesia, serving as the population and political centre of the state. Chuuk consists of five regions of partially submerged volcanic islands within a barrier reef containing numerous coral atolls and other small islands (Cole et al., 1999). This area has a tropical climate, with an annual average atmospheric temperature of 27°C, an annual sea surface temperature of 28-29°C, and annual average precipitation of 3000-10,000 mm.

Mangroves are present within the coastal area of each island within Chuuk atoll, with widespread coral reef development in areas of shallow waters separated by areas of seagrass, which is a typical vegetation structure for coastlines within tropical areas (Paik et al., 2007).

The study area is located on Weno Island (Fig. 1(b)),

an area with such a typical environment and with

fringing reefs. In situ measurements undertaken by Kim et al. (2013) indicate that surface waters around Weno Island have annual average temperatures of 28-30°C, with an annual average salinity of 33-35‰, indicative of low seasonal temperature and salinity variations.

The Chuuk region contains 615 known species of fish and 330 known species of corals (Houk and Leberer, 2008). Chuuk State comprises 14 outer island atolls, with reef areas that range in size from 0.4 to 4.6 km

(Edward, 2002). The overall condition of FSM coral reef ecosystems is generally good to excellent, with the majority of reefs in the low islands in excellent

condition. Houk et al. (2012) addressed that corals from inner reefs were larger but less diverse than those from outer reefs, in particular in Yap.

2) Field observations

Coral reef habitat mapping was undertaken on the north-eastern coastal region of Weno Island (Fig. 1(b)), using high-spatial-resolution remote sensing data combined with field observations. Two field campaigns were undertaken to identify seabed coral-reef habitat types and to determine the optical properties of each type. Field observations were taken from 42 locations

Fig. 1. (a) Map of FSM and Landsat ETM+ image of Chuuk Lagoon acquired on 23 November 2000. (b) Kompsat-2 image

acquired on 16 September 2008 showing Weno Island and (c) the study area overlaid with sampling locations

(red dots) visited between 28 August and 4 September 2008. (d) Kompsat-2 image of the study area acquired on

22 September 2010, overlaid with sampling locations (red dots) visited between 14 and 21 September 2011.

between 28 August and 4 September 2008 (Fig. 1(c)), and from 103 locations during 14-21 September 2011 (Fig. 1(d)). A handheld Garmin Oregon 500 global positioning system (GPS) receiver with a built-in three- axis compass was used for accurate positioning of sampling locations.

Each survey location was visited by a small research ship, with seabed habitat types being identified by snorkelling and underwater photography (Fig. 2(a), 2(c- g)). As shown in Fig. 2, the water was so clear that it was possible to see the seabed from the ship, with depths in the study area ranging from shallow areas

with 0.5 m water depths, to deeper areas at 1.8 m, with maximum water depth of 2.5 m at flood tide. Seven types of seabed coral-reef habitat were identified: coral, large seagrass, short seagrass, sand mixed with seagrass, sand, rubble, and rock; these are typical of the habitat characteristics of coastal areas at Weno Island. Remote- sensing-based classification of seabed coral-reef habitats usually employs five categories (coral, algae, seagrass, sand, and rubble) that can be subdivided or integrated according to the nature of the study area (Andrefouet et al. 2003). Fringing reef habitats in particular can be divided into six classes: sand, coral, rubble, seagrass,

Fig. 2. (a) Field observations of differing seabed types with snorkel and (b) in situ measurement of reflectance spectrum

using FieldSpec, with representative images of each seabed type, (c) coral, (d) rubble, (e) sand, (f) large seagrass,

and (g) short seagrass.

algae, and rocks (Roelfsema et al., 2010). In the present study area, the algae habitat was distributed intermittently, and seagrass habitats were subdivided into areas of large and short seagrass, primarily as these areas had significantly different reflectance characteristics. In addition, areas of mixed sand and seagrass were classified as a separate seabed habitat type. Although coral bleaching can also be used as a seabed classification type, this type of habitat was not identified in the study area. The radiometric characteristics of each habitat were measured using a FieldSpec Dual NVIR spectroradiometer (hereafter FieldSpec; Analytical Spectral Devices Inc.), with a spectral range of 350- 1050 nm (Fig. 2(b)).

3) Remote sensing data

Habitat maps of the study area were generated using two high-spatial-resolution images acquired by the Kompsat-2 satellite at 11:22 AM local time on 16 September 2008 (Fig. 1(c)) which was 2.5 h after low tide when the tide was at 0.1 m and at 10:30 AM local time on 22 September 2010 (Fig. 1(d)) which was 2.5h after low tide when the tide was at 0.2 m. The image obtained in September 2008 coincided with the timing of in situ measurements. The image obtained in September 2010 was the temporally closest of the available Kompsat-2 images to the in situ measurements undertaken during 2011. The spatial resolution of Kompsat-2 imagery is 4 m, with this platform having a Multi-Spectral Camera (MSC) payload that allows the acquisition of multispectral images in a 450-900 nm

spectral region. Each channel covers similar spectral regions to those used during IKONOS image acquisition, with detailed parameters provided in Table 1. The calibration and validation of the MSC to obtain standardised gain and offset values is ongoing;

consequently, the lack of standardised constant values means that accurate radiometric correction using typical atmospheric correction techniques is not possible (Lee, 2012). Thus, we used the digital number (DN) value instead of the optical reflectance in this study, because we could not acquire the calibration coefficient for the conversion of radiance (DN value) into optical reflectance for the Kompsat-2 data. In spite of this limitation, MSC images were used because of their high spatial and radiometric resolution (10 bits), and low cost. DN values have been used as spectral reflectance in many studies on the relationship between the spectral reflectance and the benthic environments in coastal areas (Choi et al., 2010; Rainey et al., 2003;

Ryu et al., 2004; Sorensen et al., 2006). The concentrations of chlorophyll-a and total suspended matter around the Weno Island were 0.11-0.49 mg/m

and 0.03-0.31 g/m

, respectively, showing a strong propensity for case-I water. Thus, although water column corrections for the Kompsat-2 satellite images were not possible, it was believed that the difference in relative reflectance by the bottom types would not have a significant impact on the classification. Geometric rectification for Kompsat-2 images was conducted using an image-to- image method with a rectified IKONOS image that was obtained on 5 December 2000.

Table 1. Band characteristics of the Kompsat-2 remote sensing platform

Satellite Band Wavelength (μm) Resolution (m)

Kompsat-2

(1) Blue 0.45-0.52 4

(2) Green 0.52-0.60 4

(3) Red 0.63-0.69 4

(4) Near-IR 0.76-0.90 4

(5) Panchromatic 0.50-0.90 1

Swath 15 km

4) Satellite-based habitat mapping

Coral reef habitat maps were generated using an object-based classification method that is known to be suitable for classification of high-spatial-resolution images, and that produces more accurate results than pixel-based methods (Blaschke and Strobl, 2001). This approach uses a classification process based on the contextual characteristics of the satellite images beyond the spectral characteristics of neighbouring pixels, such as texture, shape, and spatial relationships (Blaschke and Strobl, 2001). First, a satellite image that is composed of pixels is converted into spatially continuous and homogeneous segments called objects. Then, classification index which has been predefined by field data is assigned to each object (Conchedda et al., 2008; Desclee et al., 2006). This method has been successfully used to map surface sedimentary facies in tidal flat areas using IKONOS satellite images (Choi et al., 2010). The method was applied using Definiens® Developer 7 which is a commercial software developed as an object- based image analysis software especially for high resolution imagery (Walsh et al., 2008) by Definiens Imaging Co. During classification, we selected approximately 50% of in situ measurements of each seabed habitat type as a training dataset to map the

distribution of each seabed habitat type using Kompsat- 2 data, with the resultant habitat map validation undertaken using all in situ measurements. Thus, this approach used 23 field observation measurements of the 42 obtained during 2008, and 54 of the 103 obtained during 2011, as training datasets. We used Komsat-2 RGB (bands 3, 2, and 1) images for coral-reef habitat classification in the study area, as the high absorption of near-infrared (NIR) radiation by clear waters, as found in the study area, meant that negligible NIR reflectance passed through the water from the seabed (Choi et al., 2012).

3. Results and discussion

1) Coral reef habitat mapping

Fig. 3 shows 2008 (Fig. 3(a)) and 2010 (Fig. 3(b)) coral reef habitat maps for the study area. The area adjacent to the coast is dominated by large seagrass, with a progression through short seagrass, sand, coral, and rubble habitats towards the open sea. The density of the short seagrass vegetation meant that it could be clearly differentiated from areas of large seagrass vegetation. In addition, high-reflectance sandy areas

Fig. 3. Coral reef habitat maps of north-eastern Weno Island derived from (a) Kompsat-2 image acquired on 16

September 2008, and (b) Kompsat-2 image acquired on 22 September 2010.

were also easily distinguishable, but very few areas of mixed sand and seagrass were identified in the satellite images. Tables 2 and 3 show the error matrix generated from the 2008 and 2010 classifications, respectively.

Comparison with the in situ data (Table 2) indicates that areas of large seagrass were accurately classified by the object-based approach used here, with areas of coral and short seagrass also classified accurately. Of the 10 locations of mixed sand and seagrass, 3 were misclassified as large seagrass and of the 10 locations of sand, 3 were misclassified as mixed sand and seagrass. However, these misclassifications should be considered reasonable as the radiometric performance

of the satellite image means that areas of mixed sand and seagrass might be classified as sand- or seagrass- only areas. This misclassification is also visible in the 2010 classification map (Table 3), in which areas of large seagrass and sand habitats were classified relatively accurately, although other seabed types were classified with lower accuracy. In particular, approximately 18% of coral locations were misclassified as seagrass or areas of mixed of sand and seagrass, and approximately 23% of rubble locations were misclassified as areas of sand. These misclassifications are probably the result of these differing habitat types having similar remote sensing reflectance characteristics.

Table 2. Error matrix of Kompsat-2 derived coral-reef habitat classification using the September 2008 image Reference data (in-situ measurement)

Row total

User’s accuracy Coral Large (%)

Seagrass Short

Seagrass Sand +

seagrass Sand Rubble rock

Classification

Coral 8 0 0 0 0 0 0 8 100

Large Seagrass 0 4 1 3 0 0 0 8 50

Short Seagrass 0 0 4 0 0 0 0 4 100

Sand + seagrass 1 0 0 7 3 1 0 12 58

Sand 0 0 0 0 7 0 0 7 100

Rubble 0 0 0 0 0 2 0 2 100

rock 0 0 0 0 0 0 1 1 100

Column total 9 4 5 10 10 3 1 42

Producer’s accuracy (%) 89 100 80 70 70 67 100

Overall accuracy = 78.6%, Kappa coefficient = 0.74

Table 3. Error matrix of Kompsat-2 derived coral reef habitat classification using the September 2010 image Reference data (in-situ measurement)

Row total

User’s accuracy Coral Large (%)

Seagrass Short

Seagrass Sand +

seagrass Sand Rubble rock

Classification

Coral 14 0  0 0 0 3 0 17 82

Large Seagrass 0 12 2 1 0 1 0 16 75

Short Seagrass 1 1 4 2 0 0 0 8 50

Sand + seagrass 3 0 2 7 4 0 0 16 44

Sand 2 0 0 2 32 3 0 39 82

Rubble 2 0 0 0 0 6 0 8 75

rock 0 0 0 0 0 0 1 1 100

Column total 22 13 8 12 36 13 1 105

Producer’s accuracy (%) 64 92 50 58 89 46 100

Overall accuracy = 72.4%, Kappa coefficient = 0.65

Fig. 4 shows the spectral features of each seabed type acquired using the FieldSpec instrument. In the visible band (400-600 nm), sand and rubble have generally high reflectance values, whereas coral and seagrass have low overall reflectances, although all of them have similar spectral shapes having an absorption band around 700 nm. This indicates possible confusion between, and therefore potential misclassification of, areas of coral, short seagrass, large seagrass, rubble, and sand. If a hyperspectral remote sensed image were to be available for use in coral reef habitat classification, this confusion may decrease or be entirely eliminated, leading to more accurate remote-sensing-derived habitat maps. It is notable that areas of rubble have reflectance values that continuously increase, even in the NIR band (>700 nm) (Fig. 4). This feature is thought to be a product of bottom reflectance, as the rubble in the study area is predominantly located in relatively shallow water (<0.5 m depth, even at flood tide). Coral and large seagrass habitats have similar spectral patterns, in that their NIR band reflectance values show a gentle increase, possibly due to the proximity of canopy material to the water surface. This means that the seabed reflectance of each habitat in

clear and shallow-water environments should be accurately estimated to increase the accuracy of image- derived classification.

The overall accuracy of the remote sensing classification was 78.6% (Kappa value = 0.74) for the 2008 habitat map (Table 2) and 72.4% (Kappa value = 0.65) for the 2010 habitat map (Table 3). The relatively low accuracy for the 2010 habitat map might be due to one year difference between the field survey and the time for remotely sensed image acquisition. Nevertheless, the results demonstrate a relatively high level of agreement between remote sensing and field-based data, and indicating the reliability of this classification method.

This indicates that high-spatial-resolution remote sensing imagery, combined with in situ observation, is a useful tool that can effectively map coral-reef habitats.

2) Habitat change analysis

To quantitatively determine the change in coral reef habitat between 2008 and 2010, we calculated the number of pixels that were occupied by each seabed habitat type using remote-sensing-derived habitat maps, giving an estimated percentage of the areal extent of each habitat class (Table 4). The areal extent of rubble and sand habitats increased from 10.1% to 15.2%, and from 15.3% to 23.0%, respectively, with the extent of the coral habitat decreasing from 26.8%

to 18.6%. This suggests that parts of the coral habitat died, forming rubble or coral sand and indicating a change in habitat environment in the study area. These changes may have been caused by, for example, an increase in ocean current velocity or an increase in seawater temperature. Air temperature measurements in the Kwajalein station (http://tidesandcurrents.noaa.

gov), the nearest meteorological observation station from the study area, showed that average temperature increased about 0.18°C from 28.08°C in 2008 to 28.26°C in 2011. These changes need to be quantitatively investigated.

To examine the reliability of these coral reef Fig. 4. Mean reflectance spectrum values for each seabed

type acquired during field observations in the study

area.

habitat variations, we compared the field observations during 2008 and 2011 with the remote-sensing-based classifications outlined above. Fig. 5 shows the seabed types that were observed at identical locations in 2008 and 2011, with 5 of the 8 coral reef locations changing into areas of rubble or sand (sites 108, 144, 145, 177,

and 207), and the other 3 still being coral reef. These findings indicate the reliability of the remote-sensing- based coral reef habitat mapping. Two locations (148 and 213) changed from sand to coral reef habitats, in which the corals might have really increased, although this may reflect confusion during field observations, as Table 4. Comparative analysis of seabed type coverage between 2008 and 2010 habitat maps

2010 2008

Change

Pixels % Pixels %

Large Seagrass 9,971 25.7 8,989 23.3 2.4

Short Seagrass 1,399 3.6 2,795 7.2 -3.6

Sand + Seagrass 4,931 12.7 6,402 16.6 -3.9

Sand 8,932 23.0 5,906 15.3 7.8

Coral 7,197 18.6 10,345 26.8 -8.2

Rubble 5,878 15.2 3,884 10.1 5.1

Rock 376 1.0 209 0.5 0.4

unclassficied 100 0.3 89 0.2 0.0

Total 38,784 100 38,619 100 –

Fig. 5. Pictures showing (A) bottom type of “Sand+seagrass” obtained on 29 August 2008 and (B) bottom type of “Sand”

where most of seagrass had been gone obtained 17 September 2011. (C) In situ observation-based estimation

of changes in seabed type in identical locations between 2008 and 2011.

these locations were in areas of mixed coral reef and sand.

In this study, field observations were undertaken to qualitatively identify and determine the spectral characteristics of differing seabed types. However, for more accurate habitat mapping and change detection, quantitative estimation methods that allow the determination of seabed types are needed. One example of this would be measuring the percentage coverage of differing seabed types at each location using quadrats, although this approach would require appropriate criteria for coverage percentages that allow the consistent determination of seabed types. Although seabed spectral reflectance is a useful discriminant, the use of spatial relationships between habitat and water quality (e.g., total suspended matter, chlorophyll concentration, and dissolved organic matter) may also be an effective method for estimating the distribution of differing types of seabed habitat.

4. Conclusions

We used high-spatial-resolution remote sensing data combined with field observations to generate coral-reef habitat maps and to quantify the areal extent of coral reefs around Weno Island of the Federated States of Micronesia, in order to monitor changes in coral reef habitats in a tropical environment. An object-based classification method was successfully used to map the distribution of each seabed type in shallow-water areas of coral reef habitat. Our conclusions are as follows.

1) High-spatial-resolution Kompsat-2 remote sensing data can effectively classify coral reef habitats in tropical environments. The habitat maps were more than 70% accurate in terms of comparison with in situ observations. However, the simple use of RGB reflectance bands limits the accuracy of habitat type separations, primarily because a number of seabed types have similar remote sensing reflectance

characteristics. Use of hyperspectral imaging during future research may be able to overcome these limitations. Bottom reflectance can also affect the spectral reflectance of remote sensing images in clear and shallow water, which is an important variable in the accurate mapping of coral reef habitats.

2) A quantitative estimation of change in coral reef areal extents revealed that areas of coral in the study area decreased by as much as 8.2 percentage points from September 2008 to September 2010. Site-by-site comparisons between 2008 and 2011, from field observations, support this remote-sensing-based change analysis. This suggests that the study area has undergone changes in habitat environment that may have been caused by, for example, an increase in current velocity which could lead the death of coral in the reef boundary or an increase of seawater temperature. These variables need additional scientific research and effective remediation planning should be considered.

The present results could be helpful in determining priorities for policy implementations in terms of preventative or remediation measures in areas of coral reef decline in tropical or subtropical environments. We expect this study to further the development of novel classification and mapping methods for coral reef habitats that combine remote sensing data with in situ measurements, and to increase the awareness and use of Kompsat-2 satellite data.

Acknowledgements

This research was supported by the “Development of technology for constructing biological and environ - mental spatial information system of tidal flats through machine learning of remotely sensed visual data”

and the “Operation of Korea-South Pacific Ocean

Research Center (PO014006)” funded by the Korea

Institute of Ocean Science and Technology (KIOST).

References

Andrefouet, S., P. Kramer, D. Torres-Pulliza, K.E.

Joyce, E.J. Hochberg, R. Garza-Perez, P.J.

Mumby, B. Riegl, H. Yamano, W.H. White, M.

Zubia, J.C. Brock, S.R. Phinn, A. Naseer, B.G.

Hatcher, and F.E. Muller-Karger, 2003. Multi- site evaluation of IKONOS data for classification of tropical coral reef environments, Remote Sensing of Environment, 88: 128-143.

Andrefouet, S. and B. Riegl, 2004. Remote sensing: a key tool for interdisciplinary assessment of coral reef processes, Coral Reefs, 23: 1-4.

Blaschke, T. and J. Strobl, 2001. What’s wrong with pixels? Some recent development interfacing remote sensing and GIS, GeoBIT/GIS, 6: 12-17.

Bruno, J.F. and E.R. Selig, 2007. Regional Decline of Coral Cover in the Indo-Pacific: Timing, Extent, and Subregional Comparisons, Plos One, 2(8): e711.

Choi, J.K., J.H. Ryu, Y.K. Lee, H.R. Yoo, H.J. Woo, and C.H. Kim, 2010. Quantitative estimation of intertidal sediment characteristics using remote sensing and GIS, Estuarine Coastal and Shelf Science, 88: 125-134.

Choi, J.K., Y.J. Park, J.H. Ahn, H.S. Lim, J. Eom, and J.H. Ryu, 2012. GOCI, the world’s first geostationary ocean color observation satellite, for the monitoring of temporal variability in coastal water turbidity, Journal of Geophysical Research-Oceans, 117: C09004.

Choi, J.K., J. Eom, and J.H. Ryu, 2011a. Spatial relationships between surface sedimentary facies distribution and topography using remotely sensed data: Example from the Ganghwa tidal flat, Korea, Marine Geology, 280: 205-211.

Choi, J.K., H.J. Oh, B.J. Koo, S. Lee, and J.H. Ryu, 2011b. Macrobenthos habitat mapping in a tidal flat using remotely sensed data and a GIS-based

probabilistic model, Marine Pollution Bulletin, 62: 564-572.

Cole, T.G., K.C. Ewel, and N.N. Devoe, 1999. Structure of mangrove trees and forests in Micronesia, Forest Ecology and Management, 117: 95-109.

Conchedda, G., L. Durieux, and P. Mayaux, 2008. An object-based method for mapping and change analysis in mangrove ecosystems, Isprs Journal of Photogrammetry and Remote Sensing, 63:

578-589.

Desclee, B., P. Bogaert, and P. Defourny, 2006. Forest change detection by statistical object-based method, Remote Sensing of Environment, 102:

1-11.

Edward, A., 2002. Marine biodiversity of the Federated States of Micronesia, FSM National Biodiversity Strategy and Action Plan Project, Report the Global Environmental Facility, 2002: 1-20.

Hochberg, E.J., M.J. Atkinson, and S. Andrefouet, 2003. Spectral reflectance of coral reef bottom- types worldwide and implications for coral reef remote sensing, Remote Sensing of Environment, 85: 159-173.

Houk, P. and T. Leberer, 2008. Rapid Ecological Assessment of Chuuk, Hall, and Mortlock Islands, Chuuk State, Federated States of Micronesia, A Quantitative Assessment of Coral-Reef Assemblages and Coral Species Richness, Pacific Data Hub, Wellington, New Zealand, p. 53.

Houk, P., D. Benavente, and V. Fread, 2012.

Characterization and evaluation of coral reefs around Yap Proper, Federated States of Micronesia, Biodiversity and Conservation, 21:

2045-2059.

Hyun, S., H.S. Park, and S. Kim, 2008. Geochemical

elements in coral skeleton: potential proxies

for the earth climatic changes and oceanic

pollution, Journal of the Geological Society of

Korea, 44: 119-131 (in Korean with English

abstract).

Kim, T., Y.U. Choi, J.K. Choi, J.-K., M.S. Kwon, and H.S. Park, 2013. Comparison between in situ Survey and Satellite Imagery with Regard to Coastal Habitat Distribution Patterns in Weno, Micronesia, Ocean and Polar Research, 35:

395-405 (in Korean with English abstract).

Mumby, P.J. and A.J. Edwards, 2002. Mapping marine environments with IKONOS imagery: enhanced spatial resolution can deliver greater thematic accuracy, Remote Sensing of Environment, 82:

248-257.

Mumby, P.J., W. Skirving, A.E. Strong, J.T. Hardy, E.F.

LeDrew, E.J. Hochberg, R.P. Stumpf, and L.T.

David, 2004. Remote sensing of coral reefs and their physical environment, Marine Pollution Bulletin, 48: 219-228.

Paik, S.G., H.S. Park, R.S. Kang, H.S. Rho, and J.H.

Lee, 2007, Composition and configuration of tropical seagrass habitats in Chuuk Lagoon, FSM, The 2nd SPIRITS Workshop, Korea Ocean Research and Develoment Institute, Seoul, Korea, pp. 61-69.

Rainey, M.P., A.N. Tyler, D.J. Gilvear, R.G. Bryant, and P. McDonald, 2003. Mapping intertidal estuarine sediment grain size distributions through airborne remote sensing, Remote Sensing of Environment, 86: 480-490.

Riegl, B.M. and S.J. Purkis, 2005. Detection of shallow subtidal corals from IKONOS satellite and QTC View (50, 200 kHz) single-beam sonar data (Arabian Gulf; Dubai, UAE), Remote Sensing of Environment, 95: 96-114.

Roelfsema, C., S. Phinn, S. Jupiter, J. Comley, M.

Beger, and E. Paterson, 2010. The application of object based analysis of high spatial resolution imagery for mapping large coral reef systems in the West Pacific at geomorphic and benthic community spatial scales, Proc. of 2010 IEEE International Geoscience and Remote Sensing Symposium, Honolulu, HI, Jul. 25-30, pp. 4346-4349.

Ryu, J.H., Y.H. Na, J.S. Won, and R. Doerffer, 2004. A critical grain size for Landsat ETM+

investigations into intertidal sediments: a case study of the Gomso tidal flats, Korea, Estuarine Coastal and Shelf Science, 60: 491-502.

Scopelitis, J., S. Andrefouet, S. Phinn, P. Chabanet, O. Naim, C. Tourrand, and T. Done, 2009.

Changes of coral communities over 35 years:

Integrating in situ and remote-sensing data on Saint-Leu Reef (la Reunion, Indian Ocean), Estuarine Coastal and Shelf Science, 84: 342- 352.

Sorensen, T.H., J. Bartholdy, C. Christiansen, and J.B.T.

Pedersen, 2006. Intertidal surface type mapping in the Danish Wadden Sea, Marine Geology, 235: 87-99.

Walsh, S.J., A.L. McCleary, C.F. Mena, Y. Shao, J.P.

Tuttle, A. Gonzalez, and R. Atkinson, 2008.

QuickBird and Hyperion data analysis of an

invasive plant species in the Galapagos Islands

of Ecuador: Implications for control and land use

management, Remote Sensing of Environment,

112: 1927-1941.