Baseline reef health surveys at Bangka Island (North Sulawesi, Indonesia) reveal new threats

Study area

Bangka Island belongs to a small archipelago located at the northern tip of Sulawesi, Indonesia (Fig. 1). These islands are exposed to the main current coming from the western Pacific Ocean and directed toward the Indian Ocean (Tomascik et al., 1997). The central area of the archipelago is shallow, while the outer sides drop rapidly to over 1,000 m depth. The islands are covered by lush vegetation and fringing reefs are alternated with mangroves and volcanic cliffs. The islands are home to some villages and few resorts, of which the oldest was built in 1987. Bangka, the largest of these islands is less than 48 km². It has a resident population of about 2,500 inhabitants (as of 2013), distributed throughout five main villages (Busabora, Libas, Kahuku and Lihunu, Sawang), four small resorts in the southeastern side (Murex, Blue Bay, Nomad, Mimpi Indah), and a private research station (the Coral Eye) which hosts researchers and occasional tourists. Threats faced by the reefs in this area include destructive fishing activities (e.g., blast fishing and poison fishing) (De Vantier & Turak, 2004) and mining which targets iron ore and other minerals. Such trends are similar to those faced by other reefs throughout Indonesia (e.g., Edinger et al., 2008; Caras & Pasternak, 2009; Lasut et al., 2010; Edinger, 2012; Reichelt-Brushett, 2012). At Bangka Island, the mine (managed by a foreign company) started in 2013 but has since closed in July 2015. The closure, sanctioned by the Supreme Court in Jakarta was a direct response to the opposition of residents and tourism operators.

Sampling design and survey method

Benthic assemblages, coral diseases and other signs of compromised health around Bangka Island (North Sulawesi, Indonesia; Fig. 1) were investigated at 10 randomly selected sites at three depths: 3, 6 and 9 m (below the mean lower low water). Surveys were conducted between the months of October and November 2013. Tide levels were calculated by WXTide, a free Windows tide prediction software (http://www.wxtide32.com), using the Manado subordinate station (based on Sungai Kutei reference station, Borneo, Indonesia) and locally calibrated using a depth logger (DST centi-TD from Star-Oddi, http://www.star-oddi.com). Six sites were characterised by fringing reefs, two sites by coral rims growing on volcanic cliffs and the other two by sparse coral heads on volcanic rockslides (see Table 1). Along fringing reefs, the three chosen depths roughly correspond to reef flat, reef crest and slope (i.e., front reef), which are standard subzones in coral reef monitoring studies (e.g., Hill & Wilkinson, 2004; Hoeksema, 2012). At each site and depth, five belt transects (10 × 2 m) were laid randomly along reef contours (Beeden et al., 2008). A gap of at least 5 m was left between each transect. Overall, 150 belt transects were analysed, covering a total area of 3,000 m². Scleractinian coral colonies larger than 20 cm in diameter were identified to genus level across all transects. Scleractinian taxonomy was mainly based on Veron (2000), with some exceptions taking into account the recent reclassification of certain coral species. For example, Acropora and Isopora genera were separated according to Wallace, Done & Muir (2012). Corals previously from the genus Favia were reclassified to the genus Dipsastraea (Budd et al., 2012). Phymastrea was used to indicate all Indo-Pacific species previously indicated as Montastrea in Veron (2000) (according to Budd et al. (2012)) and more recently reassigned to Astrea, Paramontastraea and Favites (Huang et al., 2014). Similarly we have used the functional group ’Fungia’ to indicate some subgenera recently elevated to genus level (e.g., Danafungia, Verrillofungia and Pleuractis; see Gittenberger, Reijnen & Hoeksema (2011)).

In order to reduce inconsistencies associated with disease identification (Lindop, Hind & Bythell, 2008), each visible sign of coral disease or other compromised health indicators were photographed, identified and assigned to one of the 21 categories defined according to the identification guides by Beeden et al. (2008) and Raymundo, Couch & Harvell (2008). The classification scheme and adopted category acronyms can be seen in Table 2.

Benthic sessile assemblages, bare rock, sand and coral rubble were quantified by using five photographic samples (50 × 50 cm), which were randomly located within each transect. Benthic organisms were assigned to 14 main groups: algae, encrusting-, massive-, erected- and boring- sponges, hydroids, anemones, soft- and hard- corals, gorgonians, giant clams, colonial-, solitary- and social- ascidians. Percent covers were estimated by superimposing a grid of 100 equal-sized cells, using the software PhotoQuad (Trygonis & Sini, 2012). The data of percent cover was averaged across transects. No hard corals were detected at two transects, both of which were at Pearl Garden.

Information on anthropogenic disturbance sources at each site was obtained by interviewing local people, including those working at the resorts, and from data collected in previous surveys carried out applying the Reef Check protocol (Ponti et al., 2012); freely available data at http://data.reefcheck.us/). Local human disturbances were grouped as mechanic (e.g., anchoring, boat strike, SCUBA diving, blast fishing), bio-chemical (e.g., boat engine leaking, village sewage) and fishing pressure. Storms are the primary natural disturbance in the area. Previous recent storms, such intense as to damage at least the shallow-water corals, were analysed in term of direction, occurrence date and wave exposure at each site. Intensities of human disturbances and wave exposure were classified into four ranked levels according to the maximum impacts recorded in the area (De Vantier & Turak, 2004): 0 = almost no human disturbance and/or well sheltered, 1 = low or occasional disturbance and/or sheltered, 2 = medium intensity, 3 = high and frequent disturbance and/or very exposed.

A research permit was granted prior to undertaking the survey work by the authority of Lihunu, Likupang eastern districts, Minahasa northern district government (Permit ID: 211/2019/SPP/IX-2013 issued at Lihunu on the 28th September 2013).

Data analyses

In the present study the abundances of each category of coral diseases and other compromised health indicators were expressed for each transect and analysed in terms of mean number of affected colonies per square metre of hard corals, based on the extension of hard corals in the transect.

Differences between sites (10 levels, random) and depths (three levels, fixed) were assessed by two-way crossed univariate and multivariate permutational analysis of variance (PERMANOVA, α = 0.05; Anderson & Robinson, 2001; Anderson & Ter Braak, 2003).

Univariate tests were performed on Euclidean distances calculated on untransformed data (Anderson & Robinson, 2001). Multivariate tests were based on Bray-Curtis similarity of square root transformed data (Clarke, 1993). Patterns of similarities in diseases and other signs of compromised health among sites and depths were tested on zero-adjusted Bray-Curtis similarity of square root transformed data. An adjustment was applied to solve the indetermination of the Bray-Curtis coefficient, which occurs when it is calculated for pairs of transects without signs of diseases. This was accomplished by adding a dummy variable with a value of 0.22361, that is the square root of the lowest non-zero value attainable (i.e., 0.05 affected colonies m${}_{\mathrm{hard corals}}^{-2}$) (Clarke, Somerfield & Chapman, 2006).

Multivariate similarity patterns were displayed by unconstrained ordination plots using the Principal Coordinate Analysis (PCoA, i.e., metric multidimensional scaling; Gower, 1966) based on the centroids of the similarity clouds at each site. Multivariate multiple regressions between similarity patterns and variables were performed by the DistLM procedure (marginal test; McArdle & Anderson, 2001) and significant correlations (P < 0.05) were graphically represented by correlation vectors, proportional to the Pearson’s correlation coefficients, superimposed on the PCoA plots. Correlations between abundances of diseases and other signs of compromised health were made compared with human and natural disturbances by the Spearman’s rank correlation coefficient (ρ). ρ assesses how well the relationship between two variables can be described using a monotonically increasing or decreasing function. Mean values were always reported together with their standard errors (SE). Statistical analysis was performed using PRIMER 6 with PERMANOVA +add-on package (Anderson, Gorley & Clarke, 2008). Spearman rank correlations and their tests were calculated by the computational language R (R Core Team, 2016).

Benthic community structure and coral cover

Benthic assemblages were very heterogeneous in the study area and the community structures significantly differed among sites and depths (Site × Depth: F_18,120 = 2.2216, P = 0.0001). Differences between sites were significantly related to the percent cover of bare rocks, sand, coral rubble, soft corals and encrusting sponges (Fig. 2). Percent cover of both total hard corals and coral rubble significantly varied among sites but not among depths (Coral rubble, Site: F_9,120 = 16.557, P = 0.0001; hard corals, Site: F_9,120= 2.0735, P = 0.0340). Total hard coral cover varied between 5.4 ± 1.7% at Coral Eye and 17.6 ± 3.4% at Busabora Kampung (Fig. 3A). Coral rubble cover varied from almost zero at Areng Kambing, a volcanic cliff, to 35.9 ± 5.0% at Coral Eye (Fig. 3B).

Overall, 42 genera of hard corals were found. The most abundant genera were Porites (mean cover 2.73 ± 0.35%, up to 27.2%) and Acropora (mean cover 1.75 ± 0.42%, up to 42.0%; Fig. 4). Local high percent cover of Tubastraea (up to 54.4%) and Montipora (up to 43.8%) was also observed. Hard coral assemblages were very heterogeneous and variable among sites and depths (Site × Depth: F_18,118 = 1.2787, P = 0.0044). Nevertheless, few hard coral genera significantly contributed to the observable differences among sites. Those that did, included: Seriatopora, Stylophora, Goniastrea, and Pocillopora (Fig. 5).

Occurrence of coral diseases and other signs of compromised health

20 different types of coral disease and other compromised health statuses were recorded on 598 scleractinian colonies from 35 of the 42 genera identified at the different sites (Table 3). The scleractinian genera that hosted the higher number of diseases and other signs of compromised health were Porites followed by Acropora, with 14 and 11 categories recorded respectively (Table 3).

Mean abundances and total number of occurrences of each category of coral diseases and other signs of compromised health are shown in Fig. 6. The most abundant type of compromised health recorded during this survey was coral bleaching (BL; Figs. 7A and 7B), followed by skeletal deformations caused by pyrgomatid barnacles (BA; Fig. 7C), damage caused by fish bites (FB), pigmentation response (PR) and galls caused by cryptochirid crabs (GA; Fig. 7D).

No significant differences were recorded for bleaching across sites or depths, with instances recorded across 69% of the transects (mean 1.21 ± 0.15 affected colonies m${}_{\mathrm{hard corals}}^{-2}$) (Fig. 8A, Table 4). Bleaching was shown to affect 26 hard coral genera (Table 3), with the most affected genera being Porites, Dipsastraea, Goniastrea, Platygyra, Seriatopora, Acropora and Pocillopora.

Skeletal deformations caused by pyrgomatid barnacles were found in 51% of transects with a mean abundance of 0.92 ± 0.20 affected colonies m${}_{\mathrm{hard corals}}^{-2}$. Their distribution was heterogeneous, showing significant differences of abundance among sites (Fig. 8B, Table 4). Pyrgomatid barnacles were found to occur in 16 genera (Table 3); mostly colonising massive forms of corals belonging to the Porites genus.

Fish bites were found in 27% of transects with no significant difference of abundance among sites and depths (0.36 ± 0.11 affected colonies m${}_{\mathrm{hard corals}}^{-2}$; Fig. 8C, Table 4). Again, these were more typically associated with massive Porites colonies and to a lesser extent Acropora and Pocillopora (Table 3).

Signs of pigmentation response (PR) i.e., tissue discoloration, often bordering specific lesions or scars, were found in 16% of transects with a mean of 0.24 ± 0.10 affected colonies m${}_{\mathrm{hard corals}}^{-2}$. Their distribution was very heterogeneous, with significant differences found in abundance amongst both sites and depths (Table 4). However PR was only associated with three genera; Acropora, Fungia group and Porites (Table 3). It should be noted that PR (as classified in this study) can be caused by a number of factors such as specific coral borers, competitors, algal abrasion, fish bites, breakages, etc. and is therefore a type of “inflammatory” response suggesting a compromised health state but not itself a sign of disease (Beeden et al., 2008). Galls (GA), caused by cryptochirid crabs were encountered only at six of the sites surveyed (out of 10). GA occurred on 15% of transects, with a mean of 0.21 ± 0.05 affected colonies m${}_{\mathrm{hard corals}}^{-2}$. Occasionally differences of GA abundances occurred between depths (Fig. 8D, Table 4). GA were observed more commonly with branching Seriatopora and to a lesser extent Stylophora (Table 3).

A number of unidentified signs of tissue loss (named as NI in this manuscript), were found in 20 of the coral genera (Table 3) and at all sites (25% of transects, 0.26 ± 0.08 affected colonies m${}_{\mathrm{hard corals}}^{-2}$), randomly distributed between site and depth (Table 4).

Amongst less abundant categories of coral diseases and other signs of compromised health, cases of ulcerative white spot (UWS) were the most prevalent (found at nine out of 10 of the sites surveyed). UWS occurred on 11% of transects, with 0.06 ± 0.02 affected colonies m${}_{\mathrm{hard corals}}^{-2}$. Eight genera showed signs of UWS (Table 3), with Porites being the most commonly affected and significant differences between sites (Table 4).

Signs of gastropod corallivory (GC; e.g., due to Drupella spp.) were found in five coral genera and were more abundant on Pocillopora. Sediment damage (SD) affected 12 genera but with relatively low prevalence at each site (Table 3). Focal (FBL) and non-focal bleaching (NFBL, e.g., patches, stripes) together with aggressive overgrowth by sponges and other invertebrates (AgO), e.g., the coral-killing sponges Terpios hoshinota and Chalinula nematifera, were found on 11, 12 and 12 coral genera respectively (Table 3), again at relatively low prevalence (Fig. 6).

Interestingly, trematodiasis (TR) was not found on any of the reefs surveyed and black band disease (BBD) was only found on one colony of Goniastrea, two colonies of Porites and two of Turbinaria. Similarly, skeletal eroding band (SEB) was found only on a colony of Ctenactis, a colony of Galaxea and two colonies of Porites. Only one colony of Acropora and one of Montipora showed signs of brown band disease (BrB) and atramentous necrosis (AtN) respectively.

Other occasional findings included one Acropora affected by crown-of-thorns starfish (Acanthaster planci; COTS), one Gardineroseris showing signs of algal overgrowth (AlO), and a few colonies of Gardineroseris, Platygyra, Porites and Turbinaria infested by acoelomorph flatworms (RW), more precisely Waminoa sp. (Table 3; Fig. 6).

Spatial distribution patterns of diseases and other signs of compromised health

Patterns of similarities in diseases and other signs of compromised health revealed significant differences among sites, but not among depths (Table 5; Fig. 9A). The categories that significantly contribute to the observed similarity pattern were skeletal deformations caused by pyrgomatid barnacles, occurrence of corals showing signs of white syndromes, fish bites, non-focal bleaching and gastropod corallivory. These signs of compromised health increased in abundance closer to the Coral Eye site. In contrast, galls caused by cryptochirid crabs increased towards Sipi and Pearl Garden sites (Fig. 9A).

The observed spatial pattern were not correlated with the distribution of the benthic assemblages; however, there was a weak correlation with the abundance of sand, total hard coral, coral rubble (which increased towards Coral Eye) and the amount of bare rock (that characterised the volcanic cliff and rockslide at Batu Gosoh, Tanjung Husi and Tanjung Husi 2; Fig. 9B).

Correlations between diseases and possible human and natural disturbances

According to information gathered from the local populace and previous surveys (Ponti et al., 2012), the most impacted sites in terms of mechanical disturbance, pollution and fishing, were Busabora Kampung (close to the homonymous village), Pearl Garden (located in the surroundings of a dismissed pearl farm), and Sipi (in front of the newly established metal mine) see Table 1. At Pearl Garden and Sipi, signs of recent blast fishing were evident. The most relevant recent storm happened a year earlier and hit the southeastern side of the island.

Few significant rank correlations between the abundances of diseases and other signs of compromised health were observed with regard to possible human and natural disturbances. Abundance of colonies affected by skeletal eroding bands increased with the intensity of bio-chemical disturbances (ρ = 0.6864, P = 0.0284) and decreased with distance from villages (ρ =  − 0.7647, P = 0.0100). Moreover, instances of damage caused by sedimentation together with observed instances of non-focal bleaching, decreased with regard to distance from the villages (respectively: ρ =  − 0.6933, P = 0.0262 and ρ = 0.7455, P = 0.0133), while skeletal deformations caused by pyrgomatid barnacles decreased with the intensity of bio-chemical disturbances (ρ =  − 0.6963, P = 0.0253).