Stepping stones to isolation: Impacts of a changing climate on the connectivity of fragmented fish populations

Abstract

In the marine environment, understanding the biophysical mechanisms that drive variability in larval dispersal and population connectivity is essential for estimating the potential impacts of climate change on the resilience and genetic structure of populations. Species whose populations are small, isolated and discontinuous in distribution will differ fundamentally in their response and resilience to environmental stress, compared with species that are broadly distributed, abundant and frequently exchange conspecifics. Here, we use an individual-based modelling approach, combined with a population genetics projection model, to consider the impacts of a warming climate on the population connectivity of two contrasting Antarctic fish species, Notothenia rossii and Champsocephalus gunnari. Focussing on the Scotia Sea region, sea surface temperatures are predicted to increase significantly by the end of the 21st century, resulting in reduced planktonic duration and increased egg and larval mortality. With shorter planktonic durations, the results of our study predict reduced dispersal of both species across the Scotia Sea, from Antarctic Peninsula sites to islands in the north and east, and increased dispersal among neighbouring sites, such as around the Antarctic Peninsula. Increased mortality modified the magnitude of population connectivity but had little effect on the overall patterns. Whilst the predicted changes in connectivity had little impact on the projected regional population genetic structure of N. rossii, which remained broadly genetically homogeneous within distances of ~1,500 km, the genetic isolation of C. gunnari populations in the northern Scotia Sea was predicted to increase with rising sea temperatures. Our study highlights the potential for increased isolation of island populations in a warming world, with implications for the resilience of populations and their ability to adapt to ongoing environmental change, a matter of high relevance to fisheries and ecosystem-level management.

Keywords: Champsocephalus gunnari; Notothenia rossii; Scotia Sea; connectivity; individual‐based modelling; ocean warming; population genetics.

Figure 1

Map of the Scotia Sea region. Population boxes used in the connectivity analyses are outlined in red. Thin black lines indicate the mean positions of the Polar Front (PF) following Moore, Abbott, and Richman (1997), the Southern ACC Front (SACCF) and the southern boundary of the ACC (SB), both from Thorpe (2001). Bathymetry is from the GEBCO_2014 grid, version 20150318 (http://www.gebco.net)

Figure 2

Patterns of connectivity for Champsocephalus gunnari (a–f) and Notothenia rossii (g–k) for the present day (a, c, e, g, i, k) and under a climate change scenario of +2.5°C (b, d, f, h, j, l): mean (a, b, g, h), standard deviation (c, d, i, j) and frequency of nonzero connectivity (e, f, k, l). Population identifiers are Dallman Bay (DB), South Shetland Islands (SS), Joinville Island (JI), Elephant Island (EI), South Orkney Islands (SO), Shag Rocks (SR), South Georgia (SG) and South Sandwich Islands (SSa); see Figure 1 for locations. Note, the colour scales in (a–d) and (g–j) are transformed to allow visualization of the full range of values

Figure 3

Predicted changes to pairwise transport with increased sea temperatures (increased temperature scenario minus present day, expressed as the percentage of released particles on a transformed log scale to allow visualization of the full range of values) for Champsocephalus gunnari (a) and Notothenia rossii (b). Population identifiers are Dallman Bay (DB), South Shetland Islands (SS), Joinville Island (JI), Elephant Island (EI), South Orkney Islands (SO), Shag Rocks (SR), South Georgia (SG) and South Sandwich Islands (SSa); see Figure 1 for locations

Figure 4

Proportion of released particles that are exported, imported and retained at sites across the Scotia Sea in simulations of present‐day connectivity of Champsocephalus gunnari (a) and Notothenia rossii (b), and the change in these quantities for a rise in sea temperatures of 2.5°C (C. gunnari: c; N. rossii: d; increased temperature simulation minus present day). Site identifiers are Dallman Bay (DB), South Shetland Islands (SS), Joinville Island (JI), Elephant Island (EI), South Orkney Islands (SO), Shag Rocks (SR), South Georgia (SG) and South Sandwich Islands (SSa); see Figure 1 for locations. Note, y‐axis scales are not the same, and “retained” in (c) is scaled by a factor of 10⁻¹

Figure 5

Percentage change in connectivity due to increased sea temperatures against distance between sites, for Champsocephalus gunnari and Notothenia rossii

Figure 6

Projected genetic differentiation between populations (G” _ST) for Champsocephalus gunnari (a, c) and Notothenia rossii (b, d); present day above the diagonal, increased temperature scenario below the diagonal. Model projections were stopped after the same number of time steps as the present‐day simulations (a, b), or once the maximum projected genetic differentiation reached present‐day levels (c, d)

Figure 7

Discriminant analysis of principal components plots of the simulated genetic structure among populations for Champsocephalus gunnari (a, c) and Notothenia rossii (b, d); scatter plots of the first principal component illustrating the density of simulated individuals along the first discriminant function for present‐day connectivity simulations (a, b) and connectivity simulations under the climate change scenario (c, d). The | along the x axis represent simulated individuals, the different colours represent sites: Dallman Bay (DB), South Shetland Islands (SS), Joinville Island (JI), Elephant Island (EI), South Orkney Islands (SO), Shag Rocks (SR), South Georgia (SG) and South Sandwich Islands (SA). The number of principal components retained (1 in every case), and the cumulative variance explained are highlighted in black in the inset. Note: the climate change simulation resulted in no connectivity between SA and the other sites for C. gunnari, and hence, all 100 individuals are genetically monomorphic and there are no shared alleles between SA and the other sites, resulting in a single | in c