Relative efficacy of three approaches to mitigate Crown-of-Thorns Starfish outbreaks on Australia’s Great Barrier Reef

Does manual control reduce COTS densities?

We first asked whether manual control could control COTS densities at a site. To do this, we examined COTS control data from 52 sites with permanently marked coral monitoring (RHIS) sampling points, where repeated manual control of COTS took place from July 2013 to December 2017. These 52 sites were located at 21 reefs and were distributed across three different management zones (Fig. 1).

Over the 4.5 year period, individual sites were visited on average 15 ± 6.2 (s.d.) times (range 5–36), with the number of voyages to a site in a year ranging from 0 to 11 (mean 3.2 ± 2.3 s.d. voyages yr⁻¹). At the start of the Control Program, COTS densities at the 52 sites averaged 40 ± 54 s.d. individuals ha⁻¹ (range 0–237) and were above an ecologically sustainable density threshold of 3 ha⁻¹ at 45 sites. We use this threshold as a benchmark since coral growth is outpaced by predation by an ‘average’ COTS population at sites with low coral cover (estimated as 5 ha⁻¹ for just the three largest size categories)²⁷, and COTS fertilization (and thus reproductive) success increases substantially at densities of ≥ 3 ha⁻¹ due to Allee effects ²⁸. Manual control was effective in rapidly reducing COTS densities with the median density of COTS encountered being significantly lower on the second and subsequent voyages to a site than on the first voyage (Fig. 2; Friedman’s chi-squared = 9.31, df = 1, P = 0.0023; Table S1). This decline was initially rapid with one voyage sufficient to bring the median COTS density to below the ecologically sustainable threshold. The 75th percentile of sites reached this threshold following five culling voyages and fluctuated around the threshold until the number of sites in the analysis dropped to below 5 sites at 23 voyages and two and one site at voyages 27 and 29 respectively. Over the period of the study an average of 126 COTS ha⁻¹ (range 5–723) were removed from each site.

The second and subsequent voyages to a site resulted in additional COTS being culled indicating the need for repeated visits. This appears to be largely due to the fact that not all COTS present at a site are visible and available to be culled at any one time²⁹ and, to a lesser extent, to immigration into controlled sites from adjacent areas (see below). This result emphasizes the need for repeat voyages to a site in order to achieve sustained, reliable reductions to below the ecological threshold.

The impact of manual control was not consistent across the four COTS size categories (Fig. 3). The median densities of the largest size classes (> 15–25 cm, > 25–40 cm, and > 40 cm diameter) were significantly lower on the second and subsequent voyages than on the first voyage (Friedman’s chi-squared = 27, df = 1, P < 0.0000003) and reduced to levels below the ecological threshold (Fig. 3). While the densities of the smallest size class (< 15 cm diameter) also declined after the first voyage, this was to a smaller extent (Fig. 3, lower panel; Friedman’s chi-squared = 9.0, df = 1, P < 0.0027; Table S1). This might be due to smaller individuals being (1) harder to find, as reported by divers, (2) more likely to emerge in the absence of adults³⁰, or (3) more nocturnal and thus less exposed to culling³¹. These results suggest two things. First, manual control effectively targets the most damaging individuals. Because an individual’s coral consumption³² and fecundity^33,34, and therefore its contribution to population dynamics and the potential for irruptions²⁸, increases with its size, removing larger individuals from the population is important. The sooner these larger individuals are removed, the greater the reduction in coral loss at the site will be, and, the greater the reduction in the site’s contribution to downstream dynamics and impact will be. Second, the fact that, after just a small number of voyages, larger individuals had been removed from a site and that, thereafter, most individuals culled were from the smaller and harder to find size classes, points to generally low levels of immigration to sites post-control.

Combined, these findings indicate that strategic manual control at specific management locations removed disproportionate numbers of the larger, more fecund, and more damaging COTS, and was effective in keeping COTS densities below the ecologically sustainable level after five or more voyages to a site. This differential pattern of removal of COTS of different sizes suggests that rapid re-visitation in the initial phases of control is key to minimizing damage caused by these larger animals. Over the longer term, less frequent but regular visitation would be required to remove recruiting and immigrant individuals. Realizing these benefits will be most efficiently achieved by balancing re-visitation intervals to the minimum period that optimizes balance between the availability COTS and the economics of re-visitation. The fact that no COTS are available at a site at the end of a voyage but are available on subsequent voyages (an interval of ≥ 7 days) suggests that it is cycles in COTS behaviour, e.g. phases of active foraging and resting³¹, that is influencing their availability at short and long timeframes.

Manual control improves hard coral cover

The ultimate objective of COTS control is not to kill starfish but rather, to protect live hard coral. Consequently, an important measure of the effectiveness of a control program is the response of hard coral to control efforts. Our results indicate that manual control of COTS was effective in achieving this goal. At the start of the Control Program in July 2013, average hard coral cover at the 52 sites was 26.8% ± 12 s.d. During the 4.5 years period, average coral cover increased by 17.6% ± 85 s.d., with hard coral cover increasing at 25 sites (range: 0.35% to 305%) and decreasing at 27 sites (range: -85% to -3.2%). Specifically, hard coral cover in the last voyage that a site was visited was significantly and positively related to the number of control voyages to have visited that site previously (linear regression; R² = 0.17, F_{1, 50} = 10.08, P < 0.003, Table S2). This response in hard coral cover is not explained by sites with higher initial hard coral cover being visited more frequently by the Control Program (P = 0.73). In contrast, percent hard coral cover at fixed sites not receiving control (surveyed as part of the AIMS’ Long-Term Monitoring Program) decreased to “historical lows”³⁵ (see also Figure a) Benthic cover from fixed survey sites, Hard coral in³⁵) in the same period, with these declines being at least in part attributed to the current COTS population outbreaks³⁶. Interestingly, in reporting on inshore coral reef surveys, Thompson, et al.³⁷ noted that ongoing manual control of COTS at the Frankland Islands between January 2017 and March 2018 had contributed to mitigating their impacts on coral loss.

Not only was the final absolute hard coral cover related to the effort invested in manual control at a site but the proportional change in hard coral cover at a site relative to its initial cover across the 52 sites over the 4.5 years period was significantly and positively related to the number of voyages that visited these sites (linear regression: R² = 0.19, F_{1, 50} = 11.99, P < 0.0011; Fig. 4, Table S2). That is, coral cover was not just maintained but actually improved as the number of control voyages increased. These findings indicate that strategic manual control at specific management locations resulted in an average improvement in hard coral cover, with the proportional change in hard coral cover increasing with the number of control visits to a site.

Though the effect of repeated manual control on live hard coral cover was significant, the amount of variation in the change in hard coral cover relative to initial cover it explained, ~ 20%, could be considered relatively low. That it is of this order of magnitude, however, is not surprising as a multitude of factors, in addition to and independent of COTS predation, influence coral cover dynamics at a site on the GBR^15,38. De’ath, et al.¹⁵ estimated that coral predation by COTS was responsible for 42% of the 50% decline in coral cover across the GBR over a 27-year period. This provides us with an initial upper estimate of the magnitude of the effect we might expect from manual control. During this study, two mass bleaching events occurred in 2016 and 2017^16,39 with severe, though spatially uneven impacts on coral cover in the Cairns Sector of the GBR¹⁸. It is reasonable to expect that these additional, non-COTS related mortality factors would have had a significant influence on coral cover at the 52 sites during our 4.5 year study and would have limited the amount of variation available to be explained by manual control. Despite the operation of significant non-COTS drivers of hard coral cover during this study, the signal of the effect of manual control of COTS control persisted. This indicates that its effect is strong with manual control significantly improving outcomes for hard coral during a COTS population outbreak and multiple mass bleaching events.

Zoning influences initial COTS densities

The 52 sites at which manual control of COTS took place were located in three different management zones, namely in Marine National Park (i.e. ‘no-take’) zones where extractive use is prohibited (n = 26 sites); in Conservation Park (i.e. ‘limited-take’) zones where limited fishing (excluding gill netting and trawling) and collecting are permitted (n = 10); and in Habitat Protection (i.e. ‘take’) zones where fishing and other harvest activities are permitted with the exception of trawling (n = 17) (see “Methods—effect of marine protected areas” for more detail). Our initial analysis showed that at the start of the Control Program in July 2013, hard coral cover did not differ among differently zoned sites (Welch one-way test, F_2,23.78 = 0.99, P = 0.383). In contrast, zoning did influence the density of COTS culled on the first voyage to a site (Welch one-way test, F_2,29.6 = 5.95, P < 0.007, Fig. 5), with a higher COTS density in ‘take’ zones than in ‘limited-take’ or ‘no-take’ zones (Games Howell post-hoc comparisons, P = 0.008 and P = 0.008, respectively). In addition, the number of control voyages to a site was not independent of zoning (Welch one-way test, F_2,28.92 = 8.65, P = 0.0011), with sites located in ‘take’ zones tending to be visited less frequently than sites located in ‘limited-take’ or ‘no-take’ zones (Games Howell post-hoc comparisons, P = 0.012 and P = 0.001, respectively). Given this effect of zoning on the initial COTS densities encountered at a site (a dependent variable in our analyses) and on frequency of visitation to a site (an independent variable in our analyses), further analyses examining the combined effects of manual control and of zoning on changes in COTS densities and in hard coral cover was conducted.

Combined effects of manual control and zoning on hard coral cover

The combined effects of manual control and zoning on hard coral cover were examined for both the final coral cover at a site and for the proportional change in hard coral cover at a site (i.e. final hard coral cover as a proportion of the initial hard coral cover). First, absolute hard coral cover in the last voyage that a site was visited during the 4.5 year period was influenced by initial hard coral cover (β = 0.04, P = 0.005) and the number of voyages to visit the site (β = 0.08, P = 0.029), while the influence of increased protection through zoning reached only a trend (β = 0.45, P = 0.063) (fixed effects model: R² = 0.37, F_3,48 = 9.455, P < 0.0005; Table S3). Second, proportional change in hard coral cover over the 4.5 year period was not influenced by the zoning of a site (β = 0.06, P = 0.32), but did increase with the number of voyages to visit the site (β = 0.022, P = 0.015) (fixed effects model: R² = 0.21, F_2,50 = 6.49, P = 0.003; Table S3).

Zoning has been linked to a range of ecosystem benefits on the GBR^40,41,42, as well as to the potential of a reef to experience a COTS outbreak^43,44,45. Our results provide some support for this conclusion: sites zoned with greater protection, i.e. Marine National Parks and Conservation Park zones, had lower COTS densities than sites zoned Habitat Protection at the start of the manual control. Zoning showed a near-significant and positive effect on hard coral cover at the end of the study but was not a significant predictor of the proportional change in hard coral cover. In short, while zoning contributed to the initial conditions at a site, its contribution over the period of the Control Program was small relative to that of manual control. These results, and those of previous studies, suggest that current zoning arrangements act a means of moderating the impact of an active COTS outbreak, and that its role in a COTS Control Program will be as a complementary action used to support manual control or where manual control cannot be employed. The management utility of observed zoning effects on COTS populations is also compromised by the current lack of a mechanistic understanding of how such effects actually occur; whether directly through predation by targeted fish species⁴⁶ or indirectly through cascading trophic or behavioural effects, and whether the main effects are on the pelagic or settled phase or both^34,47. Determining the role of MPAs, and, in particular, the level of take of various fisheries on the GBR, on COTS densities and population outbreaks is a focus of current work. Finally, it should be remembered that current zoning arrangements were not designed with consideration of any influence of the spatial configuration of MPAs on COTS population outbreaks. Hence, effects of MPAs may well be much larger if designed with COTS outbreaks in mind, e.g. by protecting reefs that are identified as key nodes in COTS outbreak and spread processes^{48, 49}.

Efficacy of water quality improvement on COTS and hard coral cover

Finally, we compare our results against the relative efficacy of water quality improvement in controlling CoTS population outbreaks across the GBR and on reefs in the Cairns Sector up until December 2017. We first examine the scientific evidence for improvements in the quality of water flowing from the catchments adjacent to the GBR as a whole because water quality in the Cairns Sector is influenced by discharge from catchments well outside its boundaries⁵⁰. Progress towards improving GBR water quality, based on scientific monitoring and modelling published in peer-reviewed technical reports⁵¹, has been reported upon annually since 2011 (Table S4a,b, S5)^{52,53,54,55,56,57,58}. The most recent GBR Report Card 2019⁵⁸ reports that none of the main agricultural land uses (sugar cane, grazing, horticulture, grain) have achieved their 2018 target to manage 90% of land under best management practice (Table S4a). The reported improvements in agricultural best management practice systems were used to model estimates of the long-term annual river load reductions from 2009 to 2018⁵⁸. These estimates showed that the 2018 water quality targets for each of the key pollutants of concern, including those thought to influence COTS outbreaks (fine sediment, dissolved inorganic nitrogen, particulate nitrogen, particulate phosphorus) have also not been achieved (Table S4b). The marginal reductions in river pollutant loads are reflected in the lack of improvements, and in some cases further decline, in measured water quality along the Wet Tropics coast (adjacent to the Cairns Sector) since 2005 (see Sect. 5.2 and Fig. 5–31, 5,35, and 5–41 in⁵⁹). In particular, trends in chlorophyll-a concentrations have been relatively stable and are currently at, or slightly exceeding, the current water quality guideline values⁶⁰. Finally, since 2009 the overall score for water quality on the GBR given in the seven annual report cards (based on eReefs coupled hydrodynamic-biogeochemical model in the most recent report cards; see Supplementary Material Text S3), has fluctuated between ‘poor’ and ‘moderate’ (Table S6)^{52,53,54,55,56,57,58}. Combined, these measured and modelled results point to slow, if any, progress having been made in achieving GBR water quality improvement, including in the Cairns Sector^26,61.

The lack of meaningful improvement, and in some cases further decline in measured GBR water quality since the implementation of various Reef Plans starting in 2003^58,59, mean that the efficacy of water quality improvements in reducing COTS population outbreaks and their impacts on hard coral cover are likely to be negligible. This is borne out for the Cairns Sector and the GBR generally by the fact that the current COTS population outbreak began in 2010¹, well after implementation of water quality improvement programs commenced in 2003, and has since moved from further north, through the Cairns (this study) and adjacent Sectors, southward along the GBR^36,37. Not only did water quality improvement not prevent a COTS outbreak but hard coral cover in the study region showed no improvement after its implementation. The AIMS Long Term Monitoring Program reports that the trend in hard coral cover in the Cairns Sector at sites not receiving COTS control was a decline to “historical lows” (Figure a) Benthic cover from fixed survey sites, Hard coral ³⁵) in the period of this study. Moreover, predation by COTS had contributed to reductions in coral cover on some inshore reefs in the Wet Tropics region from 2012 to 2017³⁷.

These findings indicate there is little reason to expect that water quality improvement efforts have acted to suppress COTS population dynamics, or that any such influence would be detected at this point in time. The reductions in river pollutant loads (Table S4b) are sufficiently marginal to suggest that recent efforts to reduce land-based pollution (Table S4a) are unlikely to protect GBR ecosystems from declining water quality^26,62, including lowering phytoplankton biomass and associated recruitment of COTS larvae¹. This becomes particularly evident when compared with the magnitude of change in land use and management required to obtain substantial reductions in river pollutant loads to coastal receiving waters from international examples where measurable improvements in coastal water quality have been achieved⁶³. Hence, while water quality improvement may ultimately prove efficacious in influencing COTS population dynamics and outbreaks, it cannot yet be solely relied upon for COTS control in the GBR. Finally, further field observations linking larval abundance using eDNA⁶⁴ and environmental factors, combined with laboratory experiments on the effects of different nutrient and feeding regimes on juvenile COTS condition⁶⁵, are needed to elucidate the role of catchment-derived nutrients in driving COTS population outbreaks.

Source: Ecology - nature.com