We sorted, identified, and enumerated Antarctic Silverfish larvae (n = 7086) collected in a 25-year time series (1993 – 2017; Fig. 2a) of plankton net tows (see Materials and Methods) as part of the Palmer Antarctica Long-Term Ecological Research Program (Palmer LTER) and archived in the Nunnally Ichthyology Collection at the Virginia Institute of Marine Science. Zero-inflated Generalized Linear Mixed-Effects model predictions show the abundance of larvae during this period is closely tied to sea surface temperature, with higher abundance in colder water (p < 0.001; Fig. 2b). Approximately 45% of Antarctic Silverfish larvae were encountered at sea surface temperatures ranging from - 2° to 0°C, and 95% of larvae were sampled in waters colder than 1.5°C. With 95% confidence, we predict no larvae will occur at sea surface temperatures of 1.7°C or greater (see Materials and Methods).
Based on the time series, at least two consecutive years of anomalously cold surface temperatures are necessary to produce peaks in larval abundance (Fig. 2a). Palmer LTER net tows are deployed to a depth of 120 meters, encompassing the majority of habitat occupied by Antarctic Silverfish larvae 19,37. Using CTD casts paired with net tows, we found Antarctic Silverfish abundance was significantly linked to colder temperatures averaged over the 120 m water column. However, sea surface temperature yielded optimal model performance (see Materials and Methods).
We next isolated the effects of sea ice dynamics and climatic variability on Antarctic Silverfish adults by lagging variables in the model 35,38. Model results indicate the timing of sea ice advance during austral autumn (March to May) controls larval abundance in the following year (p < 0.001; Fig. 2c). Model performance (see Materials and Methods) was considerably reduced when sea ice advance lagged two years or sea ice advance immediately prior to the cruise (no lag) were evaluated as alternative parameters. To our knowledge, this is the first statistically significant relationship reported between sea ice and the long-term abundance of any Antarctic fish.
There are relatively sparse data on the reproductive biology of Antarctic Silverfish, especially in the WAP region 14. Adult fish likely spawn during late austral winter to early spring (July to September), eggs develop for approximately four months and hatch in November and December (Fig. 3a). Larvae are then sampled annually by the Palmer LTER during January and February. However, there is evidence from the Ross Sea that Antarctic Silverfish are skip spawners 39. In austral autumn, adult fish migrate from their offshore pelagic habitat to coastal areas. Adults then improve their nutritional condition for a year before spawning the next season (Fig. 3a) 14,39.
We suggest that adult Antarctic Silverfish select their spawning area along the WAP based partially on the presence of sufficient sea ice cover during austral autumn (Fig. 2a, Fig. 3b, c). An early sea ice advance in late April or early May (days 110 to 130 in Fig. 2c) acts as a positive cue for migrating adults and increases spawning habitat (Fig. 3b). If advance is delayed, spawning habitat is reduced (Fig. 3c) and could cause adults to travel elsewhere or continue to postpone spawning. We predict a sea ice advance beginning on day 157 or earlier is necessary for spawning to successfully occur in this region (see Materials and Methods). Seasonal sea ice variability during the last 20 million years in the Southern Ocean has led to a high level of life-history plasticity among Antarctic Silverfish 3,13. However, given the sea ice dependence indicated by our model, it is likely that the acute reduction in sea ice during the 20th century has significantly decreased spawning in the northern WAP. Consequently, there has been a lower abundance of mature adults observed in the northern WAP for several decades 14,15,40.
Larval Antarctic Silverfish abundance was also significantly correlated with the ASL relative central pressure (RCP), an index of ASL strength, (p = 0.006; Fig. 4a) 23 and the latitudinal location of the ASL (p = 0.005; Fig. 4b). We considered ASL RCP and latitude averaged over summer (DJF), autumn (MAM), winter (JJA), and spring (SON) months. Lagged ASL strength and latitude averaged over autumn produced optimal model AIC (see Materials and Methods). Longitudinal location of the ASL was not included in this analysis because it exhibits a strong negative correlation with RCP during most of the year 41. These relationships further support our hypothesis that adult Antarctic Silverfish are selecting their spawning habitat during autumn, when sea ice is also beginning to advance (Fig. 3a). Autumn seasons with stronger (more negative) RCPs (Fig. 4a) and a poleward ASL location (Fig. 4b) were associated with diminished spawning success reflected by lower abundance of the larvae sampled in the following year. There was no correlation between larvae and lagged ENSO indices during any season. We observed a non-significant negative relationship between lagged Marshall SAM index and larval abundance during autumn (p = 0.10) and summer (p = 0.09); larval abundance was higher following years with a more negative SAM. However, including the SAM index for any season did not improve model performance and thus was not included in the final model (see Materials and Methods).
The positive relationship between the lagged latitude of the ASL and larval abundance is suggestive of CDW intrusions impacting adult fish. Increased intrusions of CDW onto the Amundsen continental shelf that occur during a poleward shift in the ASL (Fig. 3c) are well below the thermocline 29, with the CDW temperature maximum around 400 m 42. This warm layer of water, with temperatures approaching 2°C 43, expands as more CDW is transported onto the shelf 29. This water mass is well above the optimal thermal environment for Antarctic Silverfish, particularly as the physiological costs of spawning further impact their limited thermal tolerance 44. Adult Antarctic Silverfish, which occupy a depth range of 0 to 900 m 3, might therefore delay spawning or attempt to relocate when oceanic and sea ice conditions are unfavorable in the Bellingshausen Sea due to ASL strength and location (Fig. 3c). However, additional ocean modeling is required to verify specific causes of the correlation between ASL strength, location, and adult fish.
Although there are connections between the West Antarctic cryosphere inclusive of the WAP and ASL 24,25, there were no indications of collinearity between WAP ice advance, sea surface temperature, autumn ASL RCP, and the ASL latitudinal location in this analysis (see Materials and Methods). The positive correlation between ASL RCP and larval abundance is likely due to the warm northerly winds associated with strong (deep) RCPs reducing sea ice concentration in pathways that are not captured by the ice advance parameter.
Sea surface temperatures (< 20m) increased in the WAP region by approximately 2°C during the 20th century (1955 – 1998) 45. Although this abrupt warming has recently slowed 10, it is expected to intensify in coming decades 29,46. The altered near-surface winds associated with prolonged deepening and a poleward shift in the ASL, modulated by the SAM 47 and Pacific variability 48, contributes to past and future surface warming in the WAP region 23. Therefore, we posit that long-term strengthening of the ASL creates unfavorable water temperatures for Antarctic Silverfish larvae in the WAP region (Fig. 3c) 29,46.
In addition to sea surface temperature, larval Antarctic Silverfish abundance was positively correlated with chlorophyll concentration (p < 0.001; Fig. S1a), indicating bottom-up control in this food web. Phytoplankton are grazed by the early life stages of copepods (i.e., copepodites) which in turn are the primary prey of larval Antarctic Silverfish 3. While the focus of the present study is on modeling environmental variables, confounding influences of prey field dynamics due to a changing climate are also possible. The phytoplankton community shifts during warmer years with low sea ice 12,49, potentially altering copepod abundance and composition 50. In addition to the physiological cost of warmer waters on larval Antarctic Silverfish, climate change could cause their hatching period to become out of sync with peak abundances of their prey 11,19,51.
Higher abundance of larvae also occurred in areas of lower surface salinity (p < 0.001; Fig. S1b). Surface freshwater inputs from melting sea ice and glaciers lead to a more stratified surface layer and increased chlorophyll 52; such a stratified environment with ample food could be preferred by Antarctic Silverfish larvae. The relationship between salinity and larval abundance also points to the importance of sufficient sea ice cover. High-salinity brine is exported downwards in the water column during autumn sea ice growth and advance 52,53. Consequently, surface salinity is fresher following winters with high sea ice extent 53.
Slósarczyk 54 found Antarctic Silverfish larvae were more abundant in areas of high salinity in the WAP region (34.1 to 34.6) 11. However, their analysis used a single year of pelagic net tows, and the mean standard length (SL) of post-larval and juvenile Antarctic Silverfish they collected (75 mm) was over five times the mean SL of larvae used in our study (11.9 mm; Fig. S2). Osmoregulation functions change in fishes during their development 55; therefore, post-larval Antarctic Silverfish could possess a greater ability to tolerate more saline water 11. No experiments to date address the salinity thresholds of Antarctic Silverfish during any stage in their life history.