Forecasting intraspecific changes in distribution of a wide-ranging marine predator under climate change

Globally, marine animal distributions are shifting in response to a changing climate. These shifts are usually considered at the species level, but individuals are likely to differ in how they respond to the changing conditions. Here, we investigate how movement behaviour and, therefore, redistribution, would differ by sex and maturation class in a wide-ranging marine predator. We tracked 115 tiger sharks (Galeocerdo cuvier) from 2002 to 2020 and forecast class-specific distributions through to 2030, including environmental factors and predicted occurrence of potential prey. Generalised Linear and Additive Models revealed that water temperature change, particularly at higher latitudes, was the factor most associated with shark movements. Females dispersed southwards during periods of warming temperatures, and while juvenile females preferred a narrow thermal range between 22 and 23 °C, adult female and juvenile male presence was correlated with either lower (< 22 °C) or higher (> 23 °C) temperatures. During La Niña, sharks moved towards higher latitudes and used shallower isobaths. Inclusion of predicted distribution of their putative prey significantly improved projections of suitable habitats for all shark classes, compared to simpler models using temperature alone. Tiger shark range off the east coast of Australia is predicted to extend ~ 3.5° south towards the east coast of Tasmania, particularly for juvenile males. Our framework highlights the importance of combining long-term movement data with multi-factor habitat projections to identify heterogeneity within species when predicting consequences of climate change. Recognising intraspecific variability will improve conservation and management strategies and help anticipate broader ecosystem consequences of species redistribution due to ocean warming. Supplementary Information The online version contains supplementary material available at 10.1007/s00442-021-05075-7.

. Variation of mean (dark line) and standard deviation (light line) of the environmental variables as a function of the number of simulated random tracks. The final value of 45 random tracks (dashed red line) was chosen as it stabilized all environmental variables. Figure S3. Monthly distribution (a) of the Oceanic Niño Index (ONI) with horizontal bands representing the corresponding intensity of events. Geographical distribution within the Australian marine regions (b) of the mean sea surface temperature (SST) during peaked El Niño/La Niña months during the study period, with contour lines depicting the 1,000-m isobath used as the spatial resolution for the modelling approach.

Shark location processing
GPE3 software (Wildlife Computers) was used to generate raw tracks from light-level data of PSAT tags, and a Kalman Filter was applied to the corresponding processed geolocations and location errors. A correlated random-walk state-space model using a 5 m/s speed filter to avoid unrealistic swimming behaviour was applied to both PSAT and SPOT data, with location estimates reduced to standard sampling intervals of 24 hours using the foieGras R package (Jonsen et al. 2020). A bathymetric correction was performed with locations over land moved to the nearest in-water location within the corresponding location errors. False detections were identified and excluded from the acoustic tracking dataset.
These comprised either any acoustic detections of the tracked transmitters which occurred prior to the release of a tagged shark, or single/few detections which would correspond to biologically not plausible movements (e.g., hundreds to thousands of kilometres in one day/a couple of days).
One of the main differences between satellite and acoustic telemetry data is that in the latter the animal positions are restricted to the deployment locations of the acoustic receivers (Harcourt et al. 2019). To merge acoustic and satellite data and remove potential bias in the analysis from multiple daily detections of the same individual, the acoustic dataset was standardised to only include a single daily location for each tracked shark. These single daily acoustic locations were obtained using centre of activity calculations (Simpfendorfer et al. 2002). For double-tagged sharks, these processed acoustic locations were added to the satellite tracks. If the centre of activity location was placed within the error of a satellite position on a particular day, the acoustic position was used, and the satellite excluded as the latter are usually less accurate. If the acoustic location was outside of the satellite error, the midpoint of a straight line between the two positions was considered to representative that day's position.

Juvenile females
Sharks tagged in the Great Barrier Reef Marine Park tended to remain in the northern region though some individuals swum towards Papua New Guinea or the Solomon Islands, with the furthest movement a straight-line distance of 2,138 km from the Great Barrier Reef, October 2015 towards Tonga, February 2016 (Fig. 1a). Juvenile females tagged in the Temperate East Marine Region usually moved further, both into the northern region but also towards oceanic areas including to Papua New Guinea and New Caledonia (Fig. 1a).

Juvenile males
Juvenile males tagged in the Great Barrier Reef Marine Park remained north of the coast centroid, some moving towards Papua New Guinea (Fig. 1b). Juvenile males tagged in the Temperate East Marine Region generally travelled further, with one moving towards New Caledonia (Fig. 1b).

Adult females
Adult females tagged in the Great Barrier Reef Marine Park remained in the north (Fig.   1c). Most sharks remained within Australian coastal waters, but one individual moved into the Gulf of Carpentaria, and another travelled north towards Papua New Guinea (Fig. 1c).
Adult females tagged in the Temperate East Marine Region generally dispersed further than sharks tagged north of the coast centroid and moved both north and south into the South East Marine Region, using waters as far south as the Bass Strait ( Fig. 1c) between the months of February and April when SSTs ranged between 17.8 and 20.5°C. Table S1. Tagging and tracking metadata on the 115 tiger sharks analysed in the present study, including information on tagging date, location, sex, total length (TL; cm), types of transmitters deployed (Acoustic, SPOT and PSAT), total number of tracking days, maximum distances travelled away from tagging locations, and biological class.  Table S2. Details of the response and explanatory variables used in the tiger shark modelling approach. The coast centroid has been delimited at latitude 24.5°S, as it corresponds to the current boundary between the northern Great Barrier Reef Marine Park and the Coral Sea marine regions, and the southern Temperate East marine region (please see Figure 1 for further details). All explanatory variables were modelled as candidate effects interacting with the three significant biological classes (1 = juvenile females; 2 = juvenile males; 3 = adult females). Difference between present and past 6-day average local SST (°C), with 0.5° latitude x 0.5° longitude resolution. Sea Surface Temperature difference from the 22°C isotherm SST.22 Difference between present and past 6-day average local SST (°C) subtracted from the 22°C isotherm, with 0.5° latitude x 0.5° longitude resolution.
Derivative chlorophyll-a derChloro Difference between present and past 6-day average local surface concentration of chlorophyll-a (mg/m 3 ), with 0.5° latitude x 0.5° longitude resolution.

Region
Categorical variable representing the Marine Region (i.e. North or South Region) where a tiger shark track was located, based on the respective location latitude in relation to the 24°S latitude centroid.  Figure S4. Species-specific occurrences between January 2002 and December 2020, downloaded from the Ocean Biodiversity Information System for tiger shark potential prey species (sea turtles, snakes, crabs, birds, teleosts and elasmobranchs) found to be positively correlated with tiger shark presence (Table S8).