Table 1.
Literature available on the status of Tonga’s coral reef ecosystem.
Fig 1.
Map of Tonga showing the locations of ecological survey sites in red.
Green represents land, black with grey outlines indicate villages, and blue shows coral reefs. Number of sites in Tongatapu = 60, Ha’apai = 143 and Vava’u = 172.
Table 2.
Summary of fish survey data sets available to the project.
Table 3.
Eleven socio-environmental variables included as potential influences on reef condition in Tonga.
Fig 2.
Patterns in benthic cover across the three main island groups of Tonga, arranged from south to north.
The Ha’apai group was split into southern, central and northern Ha’apai due to high latitudinal variation within the group. Values represent mean ± 95% confidence intervals. Letters denote significant groupings based on Tukey’s post hoc comparisons.
Fig 3.
Patterns of reef fish species richness, density and target biomass across Tonga’s three main island groups arranged from south to north.
The Ha’apai group was split into southern, central and northern Ha’apai due to high latitudinal variation within the group. Values represent mean ± 95% confidence intervals. Letters denote significant groupings based on Tukey’s post-hoc comparisons.
Fig 4.
Principal component ordination of the distribution of socio-environmental variables across Tonga’s island groups.
PCO was run on normalized data using Euclidean distances.
Fig 5.
Top left: Map of hard coral cover at sites sampled across Tonga. Light blue represents reef, green land, and black outlines villages. Each provincial capital is marked by a black star. Bottom left: Relative influence of the 11 predictor variables included in the Boosted Regression Tree. The dashed vertical line represents a reference point of relative influence that would be expected if all predictors were equally influential. Top right: Partial dependency plots with 95% confidence intervals for the most influential variables predicting hard coral cover. The plots show the effect of each predictor on the repsonse while all other variables were at their mean values. Relative influence of each predictor is reported in parentheses. Grey tick marks across the top of each plot indicate observed data points. Bottom right: Plots of the strongest pairwise interactions between influential variables. Contour lines indicate model predictions and points represent observed data. Units are as follows: SST−OCelcius, distance from provincial capital–km, rugosity– 1–5, reef density–km2, log wave energy–Joules per m2.
Fig 6.
Top left: Map of soft coral cover at sites sampled across Tonga. Light blue represents reef, green land, and black outlines villages. Each provincial capital is marked by a black star. Bottom left: Relative influence of the 11 predictor variables included in the Boosted Regression Tree. The dashed vertical line represents a reference point of relative influence that would be expected if all predictors were equally influential. Top right: Partial dependency plots with 95% confidence intervals for the most influential variables predicting soft coral cover. The plots show the effect of each predictor on the repsonse while all other variables were at their mean values. Relative influence of each predictor is reported in parentheses. Grey tick marks across the top of each plot indicate observed data points. Bottom right: Plots of the strongest pairwise interactions between influential variables. Contour lines indicate model predictions and points represent observed data. Units are as follows: distance from provincial capital–km, SST−OCelcius, log wave energy–Joules per m2.
Fig 7.
Top left: Map of CCA cover at sites sampled across Tonga. Light blue represents reef, green land, and black outlines villages. Each provincial capital is marked by a black star. Bottom left: Relative influence of the 11 predictor variables included in the Boosted Regression Tree. The dashed vertical line represents a reference point of relative influence that would be expected if all predictors were equally influential. Top right: Partial dependency plots with 95% confidence intervals for the most influential variables predicting CCA cover. The plots show the effect of each predictor on the repsonse while all other variables were at their mean values. Relative influence of each predictor is reported in parentheses. Grey tick marks across the top of each plot indicate observed data points. Bottom right: Plots of the strongest pairwise interactions between influential variables. Contour lines indicate model predictions and points represent observed data. Units are as follows: rugosity– 1:5, distance from provincial capital–km, reef density–km2, depth–meters, land area–km2, SST−OCelcius.
Fig 8.
Top left: Map of turf cover at sites sampled across Tonga. Note the color scale here is the inverse of other variables. Light blue represents reef, green land, and black outlines villages. Each provincial capital is marked by a black star. Bottom left: Relative influence of the 11 predictor variables included in the Boosted Regression Tree. The dashed vertical line represents a reference point of relative influence that would be expected if all predictors were equally influential. Top right: Partial dependency plots with 95% confidence intervals for the most influential variables predicting turf algae cover. The plots show the effect of each predictor on the repsonse while all other variables were at their mean values. Relative influence of each predictor is reported in parentheses. Grey tick marks across the top of each plot indicate observed data points. Bottom right: Plots of the strongest pairwise interactions between influential variables. Contour lines indicate model predictions and points represent observed data. Units are as follows: distance from provincial capital–km, rugosity– 1:5, SST−OCelcius, depth–meters, land area–km2.
Fig 9.
Top left: Map of reef fish species richness at sites sampled across Tonga. Light blue represents reef, green land, and black outlines villages. Each provincial capital is marked by a black star. Bottom left: Relative influence of the 12 predictor variables included in the Boosted Regression Tree. The dashed vertical line represents a reference point of relative influence that would be expected if all predictors were equally influential. Top right: Partial dependency plots with 95% confidence intervals for the most influential variables predicting reef fish species richness. The plots show the effect of each predictor on the repsonse while all other variables were at their mean values. Relative influence of each predictor is reported in parentheses. Grey tick marks across the top of each plot indicate observed data points. Bottom right: Plots of the strongest pairwise interactions between influential variables. Contour lines indicate model predictions and points represent observed data. Units are as follows: rugosity– 1:5, hard coral cover—%, distance from provincial capital–km.
Fig 10.
Top left: Map of reef fish density at sites sampled across Tonga. Light blue represents reef, green land, and black outlines villages. Each provincial capital is marked by a black star. Bottom left: Relative influence of the 12 predictor variables included in the Boosted Regression Tree. The dashed vertical line represents a reference point of relative influence that would be expected if all predictors were equally influential. Top right: Partial dependency plots with 95% confidence intervals for the most influential variables predicting reef fish density. The plots show the effect of each predictor on the repsonse while all other variables were at their mean values. Relative influence of each predictor is reported in parentheses. Grey tick marks across the top of each plot indicate observed data points. Bottom right: Plots of the strongest pairwise interactions between influential variables. Contour lines indicate model predictions and points represent observed data. Units are as follows: hard coral cover—%, slope– 1:5, rugosity– 1:5, reef density–km2.
Fig 11.
Top left: Map of target fish biomass at sites sampled across Tonga. Light blue represents reef, green land, and black outlines villages. Each provincial capital is marked by a black star. Bottom left: Relative influence of the 12 predictor variables included in the Boosted Regression Tree. The dashed vertical line represents a reference point of relative influence that would be expected if all predictors were equally influential. Top right: Partial dependency plots with 95% confidence intervals for the most influential variables target fish biomass. The plots show the effect of each predictor on the repsonse while all other variables were at their mean values. Relative influence of each predictor is reported in parentheses. Grey tick marks across the top of each plot indicate observed data points. Bottom right: Plots of the strongest pairwise interactions between influential variables. Contour lines indicate model predictions and points represent observed data. Units are as follows: rugosity– 1:5, distance from provincial capital–km, log wave energy–Joules per m2, hard coral cover—%, land area–km2, fishing pressure–weighted abundance of fishers scaled to 100.