RESULTS
The total MaxN over time increased from 56 (±9) in 2002 to 82 (±17) in 2018, an average increase of 46.4% (Chi = 0.003, df = 1, 209, p 0.003;\(R^{2}=0.038;\text{Fig.\ }1a).\) Similarly, species richness also increased over time from a mean of 10 (±1) to 18 (±0.8) (an 80% increase) (Chi = < 0.001, df = 1, 209, p < 0.001,\(R^{2}=0.338;\text{Fig.\ }1b\)).
The probability of occurrence of tropical fishes from all trophic guilds combined increased over time (Chi = < 0.001, df = 1, 1948, p < 0.001, \(R^{2}=0.051\)). When broken down into trophic guilds, three of the four guilds for tropical fishes showed a significant increase in probability of occurrence over time - herbivores, piscivores and invertivores - while planktivores showed no change (Fig 2a; SI T1). Temperate fish also increased overall (Chi = < 0.001, df = 1, 2598, p < 0.001, \(R^{2}=0.006\)) with a significant increase for piscivores and invertivores and no change for herbivores (Fig 2b, SI Table 1). There was, however, a significant decline in temperate planktivorous fishes (Fig. 2b; SI Table 1).
Out of 110 species, we identified 26 that individually showed significant changes in probability of occurrence over time (GLMM slope estimates and p-values in supplementary Table 2) (Fig.3). Most of these species (22 out of 26) increased in probability of occurrence through time, and this included both tropical and temperate affiliated species (Fig. 3). However, we also identified 4 temperate-associated species that significantly declined over time: the planktivore Atypicthys strigatus (Fig. 4; Chi < 0.001, slope estimate (Year) = -0.036; df = 1,48; \(R^{2}=0.100\)), the invertivoresOphthalmolepis lineolata (Fig. 3; Chi < 0.001, slope estimate (Year) = -0.025; df = 1,48; \(R^{2}=0.886\)) andPictilabrus laticlavius (Fig. 3; Chi < 0.001, slope estimate (Year) = -0.006; df = 1,48; \(R^{2}=0.522\)) and the piscivore Dinolestes leweni (Fig. 3; Chi < 0.001, slope estimate (Year) = -0.004; df = 1,48; \(R^{2}=0.794\)).
As a proxy for ecological niche space, body shape of all fish species present in the community was modelled using geometric morphometrics analysis. The complete list of Voronoi area size and distance to community weighted mean for each species are given in supplementary Table 2. Figure 4 shows a visual representation of the relative body shape analysis for each trophic guild. Species that showed a significant probability change (species labelled by picture) were randomly placed within their respective groups, with no obvious group outliers (Species not labelled).
Each species’ distance to the community weighted mean (CWM) and their Voronoi area derived from the morphometric model were tested as possible explanatory variables for change in probability of occurrence over time. Neither distance (\(F_{1,75}\)=0.102, p= 0.75) nor area (log(area):\(F_{1,\ 75}\)=0.50, p=0.482) in trait space were significant predictors of occurrence.