Dataset for: Advancing projections of Crown-of-Thorns Starfish to support management interventions

Regular outbreaks of corallivorous Crown of Thorns Starfish (Acanthaster spp; CoTS) occur on the Great Barrier Reef (GBR) and are one of the leading drivers of coral mortality. Understanding the disparities between real-world observations and model predictions of CoTS densities is crucial for refining population modelling and developing effective control strategies. Using a spatially explicit ecosystem model of the Great Barrier Reef (ReefMod-GBR), we compare model predictions of CoTS densities to manta tow survey observations. We also incorporate a new zone-specific CoTS mortality rate to account for differences in predation of CoTS between fished and protected reefs. We found high congruence between predicted CoTS densities and observations: ~81% of categorical reef level CoTS densities were either the same density level or only differed by one level, however underpredictions increased as observed densities increased. The zone-specific CoTS mortality rate reduced severe underpredictions from 7.1% to 5.6%. Underpredictions are a key concern for reef managers as they indicate potential missing outbreaks where targeted culling efforts are necessary and may lead to an underestimation of the coral loss attributed to CoTS outbreaks. Reef protection status was an important driver of prediction accuracy, suggesting it plays a role in determining CoTS densities, emphasising the importance of further research on in situ CoTS mortality rates. The location of a reef inside or outside the “initiation box”, a speculative area of primary outbreaks on the GBR, was also important, with exact predictions more likely to occur outside the box. Accurately modelling initiation box dynamics is challenging owing to limitations of empirical data on CoTS outbreaks, but this highlights the need for focussed research on these dynamics to enhance overall predictive accuracy. Other spatial factors, such as region and shelf position, also contributed to the variance between observations and predictions, underscoring the importance of the spatial-temporal context of each observation. In conclusion, this study validates our CoTS population modelling efforts, showcasing a high congruence between CoTS density predictions and real-world observations. CoTS observations can help refine predictions and guide targeted control against CoTS populations and outbreaks, contributing to effective ecosystem management for long-term resilience of the GBR.

Data are mean CoTS density (per manta tow) observations and predictions for individual reefs and years on the Great Barrer Reef. Observations come from several sources (CCP, LTMP, FMP) and predictions come from a spatially explicit ecosystem model of the Great Barrier Reef (ReefMod-GBR). 

CoTS Control Innovation Program

README: Dataset for: Advancing projections of Crown-of-Thorns Starfish to support management interventions

https://doi.org/10.5061/dryad.31zcrjdtq

This dataset is for a paper that compares CoTS density observations to predictions for individual reefs on the Great Barrier Reef in individual years. CoTS manta tow observations derive from the CoTS Control Program (CP), Field Management Program (FMP), and the AIMS Long Term Monitoring Program (LTMP). Where multiple observations exist from the same observation source, they are averaged at the reef-level in that year. If multiple observations exist from different sources, they are kept separate. Predictions are generated by a spatially explicit ecosystem model of the Great Barrier Reef (ReefMod-GBR).

Description of the data and file structure

Observed and predicted CoTS densities are compared to determine prediction accuracy of the ecosystem model. Both observed and predicted CoTS per tow values were categorised as follows: 0 = None (Level 1); ≤ 0.1 = No Outbreak (Level 2); 0.11 - 0.22 = Potential Outbreak (Level 3); 0.22 - 1.0 = Established Outbreak (Level 4); 1.0 - 3.0 = Severe Outbreak (Level 5); > 3.0 = Extreme Outbreak (Level 6). The level of each observation was then compared to the level of each prediction and the difference calculated to determine prediction accuracy.

Different variables were extracted as potential predictors of CoTS prediction accuracy. As such, the dataset includes the following for each observation/prediction comparison:

1) RM_ID = an individual ID for each of the 3806 reefs that we model in our ecosystem model.

2) YEAR = year as an integer.

3) LAT and LON = the latitude and longitude of the reef.

4) GBRMPAID and GBR_NAME = identifiers for each reef from the Great Barrier Reef Marine Park Authority.

5) REGION: 1 = North, 2 = Central, 3 = South

6) SHELF: 1 = Inner, 2 = Middle, 3 = Outer

7) GZ = Whether a reef is protected (i.e., in a green zone = 1) or not (i.e., in a blue zone = 0).

8) IB = Whether a reef is inside the CoTS initiation box (1) or not (0).

9) GEOM_CH_KM2 = the area (km2) of coral habitat for that reef

10) OBS_SOURCE = the source of the CoTS observation (1 = LTMP, 2 = FMP, 3 = CP)

11) OBS_n = the number of observations that went into calculating the reef-level mean CoTS per tow for that year

12) OBS_COTS_CAT = observed CoTS density as a categorical level from GBRMPA

13) OBS_COTS = mean CoTS per tow, and OBS_COTS_1YB = mean COTS per tow at that reef in the preceding year

14) OBS_CC, OBS_CC_1YB = mean coral cover at that reef from that OBS_SOURCE, and OBS_CC_1YB = in the year preceding it.

15) PRED_COTS_CAT = predicted CoTS density as a categorical level from GBRMPA

16) PRED_COTS, PRED_COTS_1YB, PRED_COTS_2YB, PRED_COTS_3YB = predicted CoTS per tow densities at the current year, and in the one, two, and three years preceding it.

17) PRED_CC, PRED_CC_1YB, PRED_CC_2YB, PRED_CC_3YB = predicted coral cover (%) at the current year, and in the one, two, and three years preceding it.

18) PRED_ACRO, PRED_ACRO_1YB, PRED_ACRO_2YB, PRED_ACRO_3YB = predicted Acropora cover (%) at the current year, and in the one, two, and three years preceding it. Acropora is the preferred food of CoTS.

19) PRED_INSTR, PRED_INSTR_CUMU = incoming strength of CoTS larvae calculated as the CoTS larval input multipled by the size of the reef area. PRED_INSTR_CUMU is the sum of this value over the three years previous to the current year.

20) TOTAL_OBS = where data exist, the CoTS model ReefMod is forced with manta tow survey observations from the CP, FMP, and LTMP which override model predictions at individual reefs/years. This shows the total number of observations that have been used to force the CoTS predictions for this reef in all preceding years.

21) TIME_SINCE_OBS = similar to TOTAL_OBS, but gives the number of years since an observation last forced the CoTS prediction for this reef.

22) PRED_ERR_COTS = the difference in categorical CoTS density levels between observations and predictions.

23) PRED_ERR_CC = observed and predicted coral cover was categorised into AIMS coral cover categories: 0 = 0%, 1 = 0 - 10%, 2 = 10 - 30%, 3 = 30 - 50%, 4 = 50 - 75%, 5 = 75 - 100%. The difference in coral cover category between observations and predictions was then compared.

24) DIFF_CC = the difference in the % coral cover between the observations and predictions.

25) DIFF_CC_CUMU, DIFF_CC_CUMU_N = same as DIFF_CC except the cumulative % difference over the three preceding years, and the number of observations that went into calculating the %.