Marine Pollution Bulletin xxx (xxxx) xxx–xxx Contents lists available at ScienceDirect Marine Pollution Bulletin journal homepage: www.elsevier.com/locate/marpolbul Change detection of Bunaken Island coral reefs using 15 years of very high resolution satellite images: A kaleidoscope of habitat trajectories ⁎ Eghbert Elvan Ampoua,b,c, , Sylvain Ouillonc, Corina Iovana, Serge Andréfouëta a UMR9220 ENTROPIE, IRD, Université de la Réunion, CNRS, B.P.A5, 98848 Noumea, New Caledonia b Institute for Marine Research and Observation, Ministry of Marine Affairs and Fisheries, SEACORM/INDESO center, Jl. Baru Perancak, Negara-Jembrana, Bali 82251, Indonesia c Laboratoire d'Etudes en Géophysique et Océanographie Spatiales (LEGOS), Université de Toulouse, IRD, CNRS, CNES, UPS, 14 avenue Edouard-Belin, 31400 Toulouse, France ARTICLE INFO ABSTRACT Keywords: In Bunaken Island (Indonesia), a time-series of very high resolution (2–4 m) satellite imagery was used to draw Indonesia the long-term dynamics of shallow reef flat habitats from 2001 to 2015. Lack of historical georeferenced ground- INDESO truth data oriented the analysis towards a scenario-approach based on the monitoring of selected un- Remote sensing ambiguously-changing habitat polygons characterized in situ in 2014 and 2015. Eight representative scenarios Reef flat (coral colonization, coral loss, coral stability, and sand colonization by seagrass) were identified. All occurred Marine habitat simultaneously in close vicinity, precluding the identification of a single general cause of changes that could Resilience have affected the whole reef. Likely, very fine differences in reef topography, exposure to wind/wave and sea level variations were responsible for the variety of trajectories. While trajectories of reef habitats is a way to measure resilience and coral recovery, here, the 15-year time-series was too short to be able to conclude on the resilience of Bunaken reefs. 1. Introduction management in light of disturbances and climate change, monitoring coral reef communities at a decadal temporal scale is necessary to un- Coral reefs are highly biodiverse and productive tropical ecosystems derstand and project their dynamics since managers also need to be that provide numerous services to human populations. Coral reefs are cognizant of normal successional changes. also highly vulnerable to numerous natural and anthropogenic threats. Unfortunately, despite numerous in situ monitoring programs now The general consensus is that reefs worldwide have significantly de- reaching two decades or more of data acquisition on habitat and graded in the past decades due to pollution, overfishing, coastal de- benthic community dynamics and trajectories, how different reefs in velopments and reclamations, physical destruction due to hurricanes, different locations can cope with natural, direct human impacts (e.g. global warming and coral bleaching, and ocean acidification which is a fishing) and climate change stressors remains difficult to predict. The new threat triggered by increased concentrations of atmospheric main reason is the very high heterogeneity of situations in terms of greenhouse gases. While reefs have always been exposed to natural biological communities, physical environment, isolation and con- disturbances, the accumulation of natural and human-induced stressors nectivity with other reefs (and other ecosystems and watersheds), may now impair their natural capacity to recover, even in remote un- stressors, and finally management levels. The question of whether any populated places (Pandolfi et al., 2003). Coral reef communities re- given reef will have the capacity to recover to their pre-disturbances spond with different levels of sensitivity to these stressors reflecting state (engineering resilience, sensu Nyström et al., 2008), or evolve to a both the normal environment and the site-specific history (Done, 1992; functionally different type of reef (ecological resilience) still remain Wakeford et al., 2008; Scopélitis et al., 2009, 2010, 2011; Perry and unsolvable considering the data presently available. In this context, we Smithers, 2011). Monitoring programs attempt to document changes in suggest that it is necessary to investigate the history of the reef and the spatial distribution and composition of coral communities, and to understand its spatio-temporal heterogeneity prior to try to predict the understand dynamics of degradation and recovery at different spatial long-term response of a reef to disturbances (Scopélitis et al., 2011). and temporal scales, from the global (e.g. Wilkinson, 2008) to the local The challenge of reconstructing reefs and habitat trajectories sy- scale (e.g. Done et al., 2007). Temporally, to provide a basis for reef noptically across its entire mosaic of habitats is that, in most cases, very ⁎ Corresponding author at: UMR9220 ENTROPIE, IRD, Université de la Réunion, CNRS, B.P.A5, 98848 Noumea, New Caledonia. E-mail address: [email protected] (E.E. Ampou). http://dx.doi.org/10.1016/j.marpolbul.2017.10.067 Received 31 October 2016; Received in revised form 21 October 2017; Accepted 23 October 2017 0025-326X/ © 2017 Elsevier Ltd. All rights reserved. Please cite this article as: Ampou, E.E., Marine Pollution Bulletin (2017), http://dx.doi.org/10.1016/j.marpolbul.2017.10.067 E.E. Ampou et al. Marine Pollution Bulletin xxx (xxxx) xxx–xxx Table 1 Selected publications on change detection of reef habitats across nearly two decades or more using high (Landsat: 30 m), and very high (others: 1 to 4 m) spatial resolution satellite sensors and aerial photographs. Reference Location Type of images Objective Period Method Yamano et al. (2000) Kabira Reef, Japan Aerial photographs Habitat changes 1973–1994 Photo-interpretation Lewis (2002) Barbados Aerial photographs Structural changes 1950–1991 Photo-interpretation Palandro et al. (2003) Carysfort Reef, Florida Aerial photograph, IKONOS Coral mortality 1981, 1992, 2000 Supervised classification Houk and van Woesik Saipan Aerial photograph, IKONOS Habitat changes 1940, 2004 Historical map, present time photo- (2008) Northern Mariana interpretation and classification Islands Palandro et al. (2008) Florida Keys, USA Landsat TM, ETM+ Habitat changes 1984–2002 Supervised classification Scopélitis et al. (2009) Saint Leu-Reef, La Aerial photographs, Quickbird Habitat changes 1973–2006 Photo-interpretation Réunion Scopélitis et al. (2011) Heron Island, Australia Aerial photographs, Quickbird Coral growth 1972–2007 Classification Andréfouët et al. (2013) Toliara, Madagascar Aerial photographs, Quickbird, Loss of coral 1962–2011 Photo-interpretation Landsat habitats El-Askary et al. (2014) Hurghada, Egypt Landsat TM, ETM+, OLI Habitat changes 1987, 2013 Supervised classification Nurdin et al. (2015) Spermonde, Indonesia Landsat TM, ETM+, OLI Coral loss 1972–2013 Unsupervised classification few, if any, georeferenced quantitative information describe what the site in Indonesia, ii) the period of availability of multispectral very high reefs and their habitats looked like in the past, even only 30 years ago. resolution images. In doing so, we aim to verify how the coral com- To complement infrequent and localized in situ observations, time- munities of Bunaken Island have changed in 15 years, and, if they had series of very high resolution (1–4 m) black-and-white and color aerial degraded at some point, do they appear to recover? photographs combined with recent multispectral remotely sensed sa- tellite images often can provide the only source of information. Using 2. Material and methods these data pose a number of challenges, including: data mining, pro- cessing consistently a series of heteroclite images, possible lack of un- 2.1. Study site ambiguous invariant features to help the calibration and rectification of different images, and absence of ground-truth historical data. Few Bunaken Island (BI) is located by 1.62379°N, 124.76114°E, off studies have attempted to reconstruct the trajectory of reef and habitats Manado city in North Sulawesi (Fig. 1). Bunaken Island is one of the across multiple decades (Table 1). In La Reunion Island, Scopélitis et al. first Indonesian managed site. It is part of the Bunaken National Park (2009) showed the recovery of reef flat coral communities after hurri- (BNP) which was established in October 1991 and encompasses 5 is- cane and coral bleaching across a 35-year period, suggesting en- lands, and is the main visited location in the park. BI is surrounded by a gineering resilience. Conversely, on southwest Madagascar, Andréfouët simple fringing reef system, comprising reef flats, several small en- et al. (2013) showed a steady decrease in the past 40 years of coral closed lagoons and forereefs. The total extent of this ensemble covers cover on reef flats due to destructive fishing. These studies have focused 5.42 km2. on shallow reef flats (< 5 m deep), where remote sensing is the most The forereefs include spectacular walls that have brought the dive informative. No studies on deep (5–25 m) forereef benthic changes have industry to BI since more than three decades. The reef flats themselves been done, or even can be considered feasible considering the intrinsic are seldom visited by tourists, except one location dedicated to snor- limitation of remote sensing in deep water (Maritorena et al., 1994; keling activities. However, reef flats are systematically combed by the Hedley et al., 2012). Often the images of the ocean side of reef suffer resident population at low tide to harvest crustaceans, mollusks and