Arctic charr is a partially migratory salmonid often undertaking annual marine migrations for feeding. As migration is an adaptive life-history trait, climate-driven environmental changes can alter selective pressures shaping movement behavior. To assess climate-related influences on migratory patterns and trophic niche use, we combined acoustic telemetry, otoliths, and stable isotope analysis to study populations in two Southwestern Greenland fjords differing in glacial runoff: Igaliku and Ikersuaq. Anadromous charr in Igaliku used the full fjord, traveling up to 21 km from their home river, while those in the more glacially influenced Ikersuaq remained nearshore with shorter movements (9km). Despite similar marine feeding duration (48 vs. 54 days), Igaliku fish exhibited lower marine feeding intensity. Arctic charr in Igaliku experienced warmer (8.2 ºC) conditions than those in Ikersuaq (5.8 ºC), with estuaries (Igaliku 8.2 ºC; Ikersuaq 7.6 ºC) consistently as warm or warmer than marine areas in both systems. Findings suggest Arctic charr is constrained by fjords with high glacial runoff and may shift to more open waters, thereby altering diet, growth rates and life-history strategies under future climate change scenarios.

Analyses of telemetry data

All statistical analyses were conducted using the R Program version 4.2.2 within the Rstudio version 2023.03.0 environment. Acoustic data were filtered following procedures described in Davidsen et al. (2024) to account for environmental effects that could have biased data interpretations. Signal collisions from simultaneously transmitting tags, or noise in the environment, can generate false detections that can either concatenate transmitter IDs and data or generate false IDs for tags that have not been deployed. Considering each receiver separately, a filter was applied to the tracking data in days when there were at least 1000 detections, or three or more detections of false transmitter IDs at the receiver. The automated filter excluded single detections on a receiver if separated by more than one hour from other detections for each of the tagged fish. The filter excluded 14,237 of 5,631,018 detections (0.25 %) from further analyses.

Since not all fish entered the freshwater lakes with installed receivers during the winter, an individual was considered to have completed a full marine migration if it was detected in the estuary (Ikersuaq: station 32-33; Igaliku: stations 10-12) during both the outward and inward migration to and from the fjords.

Marine residency was defined as the period from first detection in the estuary after leaving freshwater until the fish returned to the estuary again. Durations shorter than 10 days (Igaliku: n = 3; Ikersuaq: n = 6) were excluded from the analyses, as these individuals were only minimally influenced by fjord habitats. A generalized linear model (GLM) with a Gaussian error distribution and an identity link function was used to examine the relationship between marine residency duration and total body length, tagging location, and the timing of outward and inward migration to and from the fjord. Median population values were inputed for cases where marine residence duration could not be calculated. Time was measured as difftime in days from outmigration of the first fish to each of the individuals, and difftime in days from the first returning fish to each of the individuals relative time of sea entry. Continuous variables, including time variables, were scaled. Similarly, two GLM’s with Gaussian error distribution and an identity link function was used to examine correlations between 1) relative time of sea entry and total body length and tagging location and 2) between time of return, total body length and tagging location. The influences of fjord and habitat on the temperatures experienced (i.e. thermal habitat used) by individual fish during their movements were analyzed using a generalized additive mixed model (GAMM, Wood 2017; Pedersen et al. 2019). The model was fitted using the bam function from the mgcv package (Wood 2023), with temperature data recorded by the acoustic transmitters as the response variable for the period from 1 June to 1 August 2024. Prior to model fitting, the dataset was aggregated to daily mean temperature of each fish in each habitat. Here, a smooth term for time (formatted as a day of the year) within each fjord-habitat interaction was added as a thin plate regression spline smoother with 32 basis functions (k), and a random smooth term for day-of-year by fish ID with 5 basis functions. A parametric fjord-habitat interaction term (factors) was included. Temporal autocorrelation was accounted for by an AR1-structure for each individual fish.