Introduction

Rapid warming is affecting Arctic and subarctic watersheds1,2, including across Interior Alaska, USA, and Yukon Territory, Canada, where changes in air temperature, precipitation, glacial dynamics, and permafrost interact to affect water and soil chemistry, streamflow, and river temperatures3,4,5,6. Environmental changes in watersheds that affect streamflow and river temperature7,8 have the potential to impact fish physiology, including metabolism9, and as such, fish populations are particularly influenced by climate change10,11,12. Migratory fishes are critical to the culture and food security of Alaska Native and First Nations Indigenous communities in the United States and Canada, respectively13, and essential to the structure and function of Arctic aquatic food webs14,15. Given the socioecological importance of Arctic fish and their sensitivity to ongoing rapid changes, more advanced assessments of the effects of climate change on fish populations are urgently needed16. Species responses to warming are often regionally dependent, with conflicting reports across Alaskan river basins; therefore, studies that include multiple basins and species may be particularly insightful. Likewise, basin-scale information on fish populations is critically needed to inform community subsistence adaptation plans and conservation measures17.

Increasing water temperatures can impact cold-water fish survival11,18. Observed increases in water temperature have been linked to decreased growth and abundance of cold-water species such as Arctic char (Salvelinus alpinus)19 and Arctic grayling (Thymallus arcticus)20 and increased growth, recruitment, and abundance of species with higher thermal ranges, such as Dolly Varden (Salvelinus malma)21 and threespine stickleback (Gasterosteus aculeatus)22. During the warmest summers, some Arctic rivers reach temperatures > 20\(\:^\circ\:\)C, a threshold known to be harmful to many cold-water species at different life stages23,24. These high temperatures vary at spatial scales smaller than watersheds and temporal scales shorter than seasons, and effects vary between species; thus, models that relate fish growth to river temperatures at daily and basin-wide scales are necessary to assess potential effects on fish25.

Chinook salmon (Oncorhynchus tshawytscha; also known as: king salmon in English; Taryaqvak in Yup’ik26; łuk-choo in Gwich’in27; and Iqalugruaq in Inupiaq28 and Dolly Varden (Salvelinus malma; also known as Iqalluyagaq and anerrluaq in Yup’ik26; and iqallugpik in Inupiaq28) are two fish species commonly harvested for subsistence and commercial use by Indigenous communities across Alaska and were chosen by community representatives during workshops held to guide this study29. In 2024, a seven-year moratorium on subsistence harvests of Chinook salmon was enacted across Interior Alaska and the Yukon Territories. The international action was taken following the closure of Chinook salmon subsistence fisheries in an attempt to help spawning Chinook salmon recover30. Historically, an estimated 51% of households across surveyed Alaskan communities used Chinook salmon (1980–2018) compared to 34% of households since 2019 (2020–2023)31. Dolly Varden are used in 27% of households across surveyed Alaskan communities since 1982, with some communities using the species in 70–100% of households31,32,33.

The two species have distinct growth temperature ranges for juveniles. Young-of-year (YOY) Chinook salmon are cold-water specialists with growth estimated to occur at temperatures between 7.3\(\:^\circ\:\)C and 14.6\(\:^\circ\:\)C34, whereas Dolly Varden may have a higher thermal growth range of 10\(\:^\circ\:\)C to 16\(\:^\circ\:\)C35. Both fish species inhabit large, complex, and remote river networks with varied life history patterns such as anadromy (Chinook salmon typically spend 1–2 years in freshwater from hatching before ocean migration while Dolly Varden typically spend 1–4 years in freshwater from hatching before ocean migration) and amphidromy (some Dolly Varden make recurrent returns to freshwater and brackish water throughout their life cycles) and freshwater residency for Dolly Varden36,37,38,39 across Interior Alaska. These complexities, even within the same populations, make juveniles difficult to study and contribute to limited observations of population change. Likewise, the effects of increasing river temperatures on these complex life history patterns may result in variable growth responses and population survival between species and individual populations. In recent years, Chinook salmon stocks in the transboundary Yukon River watershed have collapsed, with complete fishing closures enacted over large regions40. Yet, the status and trends of Dolly Varden populations across Interior Alaska are largely unknown despite their importance as a subsistence food39. Likewise, limited observational data exist for both species to disentangle their distribution, population status, and physiological resiliency to climate change41,42.

To overcome observational limitations and estimate future habitat suitability projections of YOY Chinook salmon and Dolly Varden, we used a model chain incorporating the effects of climate, landscape, and river hydrology on fish physiology through a bioenergetics model. Bioenergetics modeling leverages known relationships between fish physiology and various environmental factors. Bioenergetics describes individual-based growth processes using estimates of metabolic processes dependent on environmental conditions43. By simplifying complex eco-physiological processes into generalized mass-balance equations that describe energy allocation44, these models can simulate possible trends in individual-level growth under scenarios of environmental change43. Connecting bioenergetics models to recent advances in climate modeling at continental scales45,46 and hydrological modeling at the resolution of individual river reaches (i.e., 1–10 km)47 enables an exploration of historical, present, and future conditions of fish growth.

Few studies have provided a comprehensive assessment of possible climate change impacts on hydrology and temperatures of Alaskan rivers, and even fewer for multiple fish species across multiple watersheds in the Arctic-Yukon-Kuskokwim (AYK) region20. To explore habitat suitability changes and identify possible basin-level thermal refugia, we used outputs from a coupled land-atmospheric model specifically designed and calibrated for cold region processes and linked river models that simulated streamflow and river temperature to fish growth with a bioenergetics model. Using this modeling approach, we assessed historical conditions (1990–2021) and projected future mid-century (2034–2065) changes in growth outcomes of YOY Chinook salmon and Dolly Varden in seven river basins across the AYK region. During a project workshop, Indigenous community representatives identified relevant river basins and fish species through a participatory mapping exercise. The final seven basins and species were selected based on this identification and sufficient quantitative data for modelsimulations.

While fishes integrate the effects of warming over multiple seasons, habitats, and years, we focused on the YOY life stage because the first summer of life is critical to fish survival in winter and may be used as an indicator of smolt fitness14,48. Simulating growth outcomes for the first summer season can also inform assessments of life stage vulnerability to future thermal regimes and allow inferences on the effects of early growth on the survival of individuals into mature life stages. The seven river basins represent an area of nearly 250,000 km2, and each basin is diverse in size and ecosystem complexity (Fig. 1 and Fig. S1). Therefore, we assessed the relationships between basin-scale water temperature differences from the modeled historical to future 31-water-year periods and variables describing ecosystem complexity. A water year is a 12-month period from October 1 through the following September 30 and is named for the calendar year in which the period ends. For each fish species, river basin, and 31-water-year period, we simulated the first year of growth at daily resolutions for the coldest summer (June-September) season, the 31-water-year average summer season, and the warmest summer season. We used the growth results and differences between results for the historical and future simulations to examine habitat suitability in river basins and identify thermal refugia for Chinook salmon and Dolly Varden in the YOY life stage. Here we identify possible future scenarios for key Arctic fishes, species-specific thermal refugia, and vulnerabilities to regional and watershed-scale climate change impacts.

Results

Projected river temperatures

During the historical period (1990–2021), average temperatures of the seven rivers ranged between 8.77 °C and 11.68 °C (Fig. 1; Table 1). River basins were grouped into three categories: <10 °C (“coldest”), 10–11 °C (“intermediate”), and > 11 °C (“warmest”). In the historical period, three basins (total basin area: 36,314 km2) were characterized as “coldest”, three (202,654 km2) were characterized as “intermediate”, and one (5,777 km2) was characterized as “warmest” (Fig. 1a). On average, summer river temperatures are projected to warm by 1.26 °C from the historical to mid-century future scenario, with a cross-basin range of 0.95 °C to 1.58 °C (Table 1). In the future scenario, one basin remained in the coldest category, while all others either shifted into the next highest or remained in the highest temperature category (Fig. 1b). The degree of simulated warming was negatively related to the fraction of rivers characterized as first-order streams, or headwaters (r2 = 0.74), followed by changes in streamflow (r2 = 0.43) and latitude (r2 = 0.40). Warming was least correlated with projected air temperature changes (r2 = 0.15) (Fig. S1). The results suggest that topography, hydrology, and other physiographical landscape conditions control river temperature and the magnitude of expected warming.

Fig. 1
figure 1

River basin average summer temperatures increase from historical to future mid-century scenarios across the Arctic-Yukon-Kuskokwim (AYK) region. Lighter colors represent colder areas that could provide thermal refugia for cold-water fishes under warming scenarios. Maps of river basins and their respective average temperatures from (a) historical (1990–2021) and (b) future mid-century (2034–2065) pseudo-global warming (PGW) scenarios (Blaskey et al., 2024b). Maps were generated in Python version 3.12.2 ( https://docs.python.org/release/3.12.2/) using river network geopackages from MERIT Hydro49 and Cartographic Boundary shapefiles for the state of Alaska50.

Full size image
Table 1 River basins differ by basin size and degree of temperature change. River basin characteristics by basin area (km2), historical (1990–2021) and future (2034–2065) temperature classifications, and the difference in average summer temperature (future mid-century temperature- historical temperature). Temperature classifications are colored based on Fig. 1: <10 °C (“coldest”, yellow), 10–11 °C (“intermediate”, orange), and > 11 °C (“warmest”, red).
Full size table

In the historical period, daily water temperatures of all seven rivers remained below thresholds regarded as deleterious for YOY Chinook salmon and Dolly Varden, even in the warmest years. By mid-century, the warmest years are projected to bring unprecedented thermal conditions to many of these river basins; four of the seven rivers exceed the upper temperature (14.6 °C) for YOY Chinook salmon for an average of 8.75 days (9.4% of the summer). The coldest and average years by mid-century are likely influenced by the average and warmest years of the historical period in all river basins, and the average years marginally exceed peak growth ranges for YOY Chinook salmon (0.45 days, 0.49% of the summer). By comparison, even in the warmest mid-century years, daily temperatures in all rivers remain below the upper-temperature threshold for Dolly Varden growth (16 °C) (Fig. 2).

Fig. 2
figure 2

River basin temperature distributions shift upward from historical to future scenarios. Temperature distribution of individual river basins (a–g) based on model simulations from June to September and changes from historical (blue, 1990–2021) and future mid-century (pink, 2034–2065) pseudo-global warming (PGW) scenarios by given ‘Cold’, ‘Average’, and ‘Warm’ years. Colored bands represent historical and future average temperatures for each river basin, where yellow is < 10 °C, orange is 10–11 °C, and red is > 11 °C. Dashed black lines refer to the cited YOY Chinook salmon thermal range for growth34 and solid black lines refer to YOY Dolly Varden thermal range for growth35. Colored stars represent the average summer temperatures of the simulated ‘Cold’, ‘Average’, and ‘Warm’ years and the respective temperature classification of < 10 °C, 10–11 °C, and > 11 °C. Note that x-axes are not to scale, and the width of violin plots is based on the frequency of days at specific temperatures.

Full size image

Species-specific thermal preferences

While much of the discussion of climate change focuses on the effects of warming, summer water temperatures below growth thermal ranges for YOY Chinook salmon (7.3 °C) and Dolly Varden (10 °C) can also reduce growth (Figs. 3a and 4a). Historically, based on modeled summer water temperatures, the projected coldest, intermediate, and warmest rivers were below 14.6\(\:^\circ\:\)C for YOY Chinook salmon for 25, 17, and 3% of the summer, respectively. In the future, warmer summer temperatures resulted in fewer days below the growth temperature range for both species. By mid-century, on average, the coldest, intermediate, and warmest rivers were below YOY Chinook salmon lower threshold temperatures for 12, 4, and 0.6% of the summer season, respectively. Warming resulted in fewer days below growth temperatures, with the greatest change occurring during the coldest years (-9, -20, and − 12%, for the coldest, intermediate and warmest river classifications, respectively), while the warmest years marginally changed (-2, -3, and 0%, for the coldest, intermediate and warmest river classifications, respectively).

Historically, for YOY Dolly Varden, the projected coldest, intermediate, and warmest rivers were below growth temperatures for 70, 69, and 26% of the summer, respectively. By mid-century, the coldest, intermediate, and warmest rivers were below 10 °C for 48, 32, and 10% of the summer season, respectively. Similar to Chinook salmon, warming resulted in fewer days below growth temperatures, with the greatest change occurring during the coldest years (-27, -20, and − 22%, for the coldest, intermediate and warmest river classifications, respectively), while the warmest years marginally changed (-9, -11, and 0%, for the coldest, intermediate and warmest river classifications, respectively). Simulated warming resulted in fewer days with summer water temperatures below growth ranges for most years, basins, and for both species. In the future scenario, the percentage of the season within the 10–16 °C YOY Dolly Varden growth thermal range (+ 29.42 ± 6.81%) increased for all river basins. For YOY Chinook salmon, however, the increase in growth days in a warmer climate is counteracted by more days that exceed the growth range in the warmest years and the warmest river basins.

Fig. 3
figure 3

Young-of-the-year (YOY) growth in Chinook may be resilient to warming water temperatures in most, but not all, rivers. (a) Pie charts reflect the percent of the summer season temperatures above (purple), within (gray), or below (teal) the thermal range of Chinook salmon (7.3–14.6 °C)34. Percent differences in end-of-season weight from historical to future mid-century pseudo-global warming (PGW) scenarios for rivers classified as (b) < 10 °C, (c) 10–11 °C, and (d) > 11 °C in the mid-century. Boxplots show the interquartile range of growth potential (percent difference in weight) of simulated fish in a given year, including outliers. Percentage values above boxplots represent the mean growth potential of a given year type.

Full size image

Species-specific growth outcomes suggest habitat suitability

Individual YOY Chinook salmon growth increased in the mid-century warming scenario for the coldest and average years but declined in the warmest years when the thermal preference was exceeded (Fig. 3). In the coldest river basin, the coldest year produced the highest growth potential (Fig. 3; +23.54 ± 0.59%; P < 0.001). Likewise, across all rivers, fish sizes were significantly larger in the coldest years compared to the average years (Fig. 3). Simulated growth of YOY Chinook salmon in the intermediate river basins increased for the coldest (+ 18.89 ± 1.25%) and average years (Fig. 3c; +14.22 ± 0.32%) but decreased during the warmest years (-3.24 ± 0.10%). Growth potential showed greater variability during the coldest years (Fig. 3b-d). The type of year (i.e., the coldest, average, or warmest years of the respective historical and future periods) also significantly influenced the growth outcomes in all river basins. The warmest river basins produced larger individuals in the future compared to historical simulations, on average (Fig. 3d; +0.08 ± 0.24%) and during the cold years (+ 8.94 ± 0.41%), while growth declined during the warmest years (-3.46 ± 0.16%).

Fig. 4
figure 4

Young-of-the-year (YOY) Dolly Varden may be resilient to warming water temperatures. (a) Pie charts reflecting the simulation temperatures relative to the thermal range of Dolly Varden (10–16 °C)35. Percent differences in end-of-season weight from historical to future mid-century pseudo-global warming (PGW) scenarios for rivers classified as (b) < 10 °C, (c) 10–11 °C, and (d) > 11 °C in the mid-century. Boxplots show the interquartile range of growth potential (percent difference in weight) of simulated fish in a given year, including outliers. Percentage values above boxplots represent the mean growth potential of a given year type.

Full size image

YOY Dolly Varden have a higher growth thermal range and exhibited a larger growth potential than YOY Chinook salmon in simulated mid-century river basins across the AYK region (Fig. 4). Dolly Varden bioenergetics simulations produced the largest growth outcomes for average years, rather than the coldest or warmest years, in all river basins. However, for the coldest river basin, the type of year significantly influenced the degree of positive growth (Fig. 4b; P < 0.001), where the warmest years may result in less growth (+ 6.17 ± 0.14%) than the coldest years (+ 22.54 ± 1.63%). On average, the growth of individuals doubles in the future (+ 103.85 ± 5.94%). The growth potential variation in average years of the coldest river basin further demonstrates the disparate growth outcomes of the cold and warm years. Growth potential of individual Dolly Varden in intermediate river basins was highest under average conditions (Fig. 4c; +144.82 ± 5.80%), while the warm years produced slightly larger individuals (+ 68.39 ± 2.38%; P < 0.001) than the cold years (+ 43.16 ± 2.87%). Individuals in the warmest river basins demonstrated the largest increases in weight during average years (Fig. 4d; +59.20 ± 1.47%), while Dolly Varden had similar positive growth trajectories under the coldest (+ 43.43 ± 1.12%) and warmest years (+ 43.90 ± 2.56%; P = 0.98).

Regional species-specific growth rates and limitations

YOY Chinook salmon growth rates were highest around 11.7 °C across all river basins and years (historical and future mid-century) and began to decline or plateau above this temperature threshold in bioenergetics simulations (Fig. 5a). Growth rates increased steadily between 7 and 11 °C before plateauing near the upper temperatures examined (Fig. 5a). Meanwhile, consumption and respiration rates increased with higher temperatures, despite the growth rate-temperature relationship. This suggests that YOY Chinook salmon may need to consume and use more energy to maintain or improve growth rates at higher temperatures within the projected temperature range. YOY Dolly Varden maximum specific growth rates occurred near 14 °C across all river basins and years. Specific growth rate slowly increased at temperatures between 7 and 10 °C, and slopes were the highest at temperatures between 10 and 14 °C across tributaries (Fig. 5b). Further, consumption rates increased at warmer temperatures with the highest slopes between 7 and 9 °C and 12–14 °C. However, YOY Dolly Varden respiration rates decreased with warmer temperatures overall, exhibiting an initial increase in rates at temperatures less than 8.5 °C followed by a steady decline until plateauing at 14 °C (Fig. 5b).

The length of rivers above this 14 °C cross-tributary average was negligible. There were no river segments in the historical period above 14 °C, and by the future mid-century, the Koyukuk River may have 11 km or 0.1%, and the Porcupine River may have 96 km or 0.4% of the total river length above 14 °C. However, considering the threshold for Chinook salmon growth, historically the Koyukuk and Porcupine Rivers may have had 557 (2.3%) and 524 (3.1%) km of river, respectively, above 11.7 °C (Table 2). Further, by mid-century, the Chena (222 km, 19.1%) and Teedriinjik (57 km, 1.2%) Rivers may have some river segments above 11.7 °C, along with increasing river lengths for the Koyukuk (807 km, 4.8%) and Porcupine (2271 km, 9.3%) Rivers (Table 2).

Fig. 5
figure 5

Chinook salmon (left) and Dolly Varden (right) metabolic demands result in distinct growth outcomes, limiting growth potential in the warmest rivers for Chinook salmon. The inflection point in the average maximum specific growth rate at temperatures is lower for Chinook salmon than for Dolly Varden for all river basins. More river segments in the mid-century are projected to be warmer than temperatures required for optimal growth of YOY Chinook salmon based on modeled specific growth rates. Regression models for predicted average specific growth rate (J/g/d), respiration rate (J/g/d), consumption rate (J/g/d), and average river basin temperature (°C) across all years and river basins in the historical (1990–2021) and future mid-century (2034–2065) periods (n = 434) from bioenergetics simulations for (a) YOY Chinook salmon (maximum specific growth rate/red line = 11.7 °C) and (b) YOY Dolly Varden (maximum specific growth rate/red line = 14 °C). Note x-axes show different species-specific metabolic rate ranges. Shading represents ± 1 standard deviation of metabolic rates (J/g/d) at each 0.1 °C increment.

Full size image
Table 2 More rivers in the mid-century are projected to exceed the modeled Chinook salmon maximum growth rate temperature. Rivers examined in the current study and the length of river segments (km) above the cross-tributary average temperature of maximum specific growth rates of modeled YOY Chinook salmon (11.7 °C). Colored cells represent historical and future average temperatures for each river basin from Fig. 1, where yellow is < 10 °C, orange is 10–11 °C, and red is > 11 °C.
Full size table

Discussion

Rivers across the AYK region are projected to warm by 0.95–1.58 °C in the future (2034–2065) with ecological implications for different fish species, life stages, and life histories. This study showed that rivers most resilient to warming are those with more first-order headwater streams, those projected to increase in summer streamflow, and those at more northern latitudes. River basins with the highest projected warming (~ 1.5 °C) include the Aniak and Andreafsky Rivers, which are historically among the coldest AYK rivers examined. These rivers have the lowest percentages of first-order streams, projected reductions or negligible changes in streamflow, and are located at lower latitudes. Despite this high sensitivity, in the future, these rivers–– together with the Teedriinjik River—remain the three coldest rivers studied and may serve as possible thermal refugia for YOY Chinook salmon. The Chena River is projected to warm the least despite being the warmest historically out of the seven rivers examined, likely due to a higher percentage of first-order streams. Yet, the warming that occurs in the Chena, along with the Koyukuk, Porcupine, and Takotna Rivers, may be enough to limit the growth of Chinook salmon based on the frequency of days and river length in the upper-temperature range for individual growth. Our results illustrate the importance of considering topography, hydrology, river conditions, and local-scale climate when assessing thermal refugia for fish species.

Literature values of growth thermal preferences and growth temperatures of Chinook salmon and Dolly Varden suggest wide temperature ranges that may broadly define ecological niches based on other abiotic and biotic factors. While model simulations account for a range of growth trajectories, these simulations may also infer temperatures within the defined literature thermal ranges that may be physiologically relevant based on factors such as consumption and respiration that contribute to individual growth outcomes. For example, juvenile Chinook salmon are expected to increase in pre-smolt size in warmer conditions, provided that all other ecological conditions meet individual needs51,52. However, the longevity of this benefit is unknown due to limited knowledge regarding seasonal changes in prey biomass53,54, hydrological changes55, and stream network complexity51.

We show that YOY Chinook salmon growth may marginally increase as a result of warming in most river basins examined; however, the warmest years in rivers such as the Chena, Koyukuk, Porcupine, and Takotna will likely result in smaller Chinook salmon size at the end of summer due to increased metabolic demand at temperatures above their thermal optima. Reduced growth in the warmest years would likely result from an inability to compensate for increasing metabolic demand with a limited food supply, particularly when temperatures were above modeled peak growth temperatures of 11.7 °C. Feddern et al.56 also found that warming stream temperatures may not provide the large growth benefit expected for juveniles across the Yukon and Kuskokwim regions, possibly due to indirect effects of lower food quality or availability, predation, and higher streamflow. Considering reports of smaller returning adult Chinook salmon by many communities over recent years56,57, the mixed effects of warming stream temperatures during juvenile rearing combined with other factors such as ocean conditions and stream conditions during adult spawning migration may contribute to smaller adults and reduced spawning biomass56.

Historically, thermal regimes may have been unfavorable for higher growth rates of YOY Dolly Varden; however, future simulations suggest growth will increase across all river basins examined. The results are consistent with a recent study reporting larger sizes and abundance of Dolly Varden21. However, others expect localized declines for the species due to warming20, possibly making intermediate thermal regimes the best options for YOY Dolly Varden growth. Likewise, the lake trout (Salvelinus namaycush) profile used for Dolly Varden simulations may not fully represent Dolly Varden physiological relationships, particularly when examining the relationship between temperature and respiration rate. It is unknown whether Dolly Varden respiration rate can be represented with or without activity multipliers, in which case, the bull trout (Salvelinus confluentus) model from Fish Bioenergetics 4.0 may be more representative. Ultimately, testing life stage-specific responses to temperature for lesser-studied Dolly Varden will allow us to assess habitat suitability for this species more accurately. Further, obtaining more empirical observations in natural systems for consumption and growth responses under different environmental conditions for both species will allow for better parameterization of bioenergetics models and more relevant guidance for managers and communities alike58.

While we chose conservative models and temperature outputs at the river basin-scale to assess general trends in growth trajectories and possible vulnerabilities, we did not fully address impacts within smaller river reaches, extreme events, or outcomes likely to occur in mid-century. Extreme high or low stream temperature and streamflow events are known to negatively affect salmonid populations and increase mortality risk of certain life stages48,59,60. While we spatially aggregated river temperature simulations for historical47 and future61,62 periods to run the daily bioenergetics model at the basin-scale, running simulations at a finer scale would help narrow habitat suitability for early life stages and require tested bioenergetics parameters specifically for YOY. Future work could assess the likelihood of extreme river events by river segment, which could be used for identifying detailed thermal refugia, including upper headwater streams and deeper river segments. Likewise, fish habitat is changing with landscape processes, such as glacial retreat and riverbank erosion63,64, and the possibilities of future refugia for many species remain uncertain across such an expansive region.

Indigenous communities across Alaska and Canada have been faced with harvest declines and moratoriums specifically for Chinook salmon40. With an uncertain future for Chinook salmon populations across Interior Alaska and Yukon Territories, many questions remain about what can be done to manage species for resilience and to support Indigenous livelihoods. Early salmonid life stages are critical for recruitment in populations; therefore, prioritizing the conservation and restoration of current and possible future juvenile refuges in these freshwater systems may benefit fish species and human life under unprecedented ecosystem change. We have incorporated Indigenous guidance at all levels of the modeling scope from climate model decision-making, river usage, and fish species mapping for the fish bioenergetics modeling29.

Methods

Indigenous guidance of modeling decisions

Locations and species of interest were determined from engagement with representatives of Indigenous communities across Alaska and the Canadian Yukon Territories at the 2022 Arctic Rivers Summit29,65. The focal species and watershed-scale locations for regionally relevant bioenergetics modeling efforts were identified specifically from information shared during participatory mapping sessions, individual discussions, and available quantitative data across the Arctic-Yukon-Kuskokwim (AYK) region. Not all Alaska Native and Yukon First Nations communities were represented at the Summit; therefore, we consider the input for the modeling to represent the knowledge and priorities of a subset of communities that does not represent all Arctic Indigenous communities in the AYK region.

Species data collection

Data regarding species size, locations, dietary habits, and habitat preferences were obtained from various sources, including research articles, fishing reports, state and federal data repositories, gray literature reports, and individual and community knowledge. Species ecology was generalized among locations for model simulations across the AYK region. The timing of observations from reports and Indigenous Knowledge shared at the Arctic Rivers Summit was used to parameterize the timing of species and life stage presence within river systems (i.e., June to September) and prey consumption within model simulations using both quantitative and qualitative data. Research articles, data repositories, and gray literature reports contained quantitative information specifically on the timing of prey availability and fish stomach contents during the summer season (June to September). Individual and community-level Indigenous Knowledge contained qualitative information regarding observations of young-of-year presence during the summer season. Observed fish size records from agencies and institutions were used to parameterize the models from estimated sizes at the beginning (June 1st) and end (September 1st) of model runs to estimate growth outcomes for given years.

Young-of-year (YOY) Chinook salmon weight (g) and total length (mm) data for simulation parameterization were obtained for Aniak, Chena, the middle Yukon, and Porcupine River tributaries due to the larger number of observations available for these locations. Observational data from the Chena River Basin included both beginning season (May/June) and end-of-season (August/September) weights, therefore Chena River Chinook salmon size ranges were chosen for final model simulations (Fig. S2). Likewise, Chena River observational end-of-season size data represented intermediate size ranges across the AYK region and were statistically similar to data from both the Aniak and Porcupine Rivers (Fig. S2). Diet data from Falke et al.52 were used for all simulations because observations from this study demonstrate dietary variation between a regionally cold, warm, and average year that can be observed in bioenergetics simulations. According to Reiser and Bjornn34, the observed thermal preferences for the growth of YOY Chinook salmon populations across Alaska and Canada is estimated between 7.3 and 14.6 °C. Several literature sources have stated thermal preferences within or near this thermal range66,67, thus 7.3–14.6 °C was chosen because of the applicability to both observational daily river temperatures and modeled temperatures for river basins from this study.

YOY Dolly Varden are individuals < 70 mm at the beginning of the summer season (June-September)68. YOY Dolly Varden size data was estimated for the beginning of model simulations in June using growth rate estimates68and length-weight relationships from observed fish from the Kuskokwim River (W = 0.0000042*L3.172, W = weight (g), L = length (mm))69. End-of-season size data were provided for the Aniak and Salmon River systems of the Kuskokwim Basin from 2015 to 202169. The optimal temperature range for growth of YOY Dolly Varden has not been tested; however, studies on the optimal range for lake trout (Salvelinus namaycush) fry, a closely-related species, suggest temperatures from 10 to 16 °C70,71 while the British Columbia Ministry of Water, Land, and Air Protection35 suggests Dolly Varden juvenile rearing occurs in habitats between 8 and 16 °C. Other related species such as Arctic char (Salvelinus alpinus) and bull trout (Salvelinus confluentus) are cited to have juvenile optimal growth temperatures and thermal preferences in the range of 12–16 °C72, 10.9–11.6 °C73, 12.3–16 °C74, and 7–15 °C75. Early life stages of fishes are expected to vary in thermal tolerance based on different energetic requirements for specific tissue growth processes and developmental trajectories76,77. Therefore, For YOY Dolly Varden model simulations, we used a preferred growth temperature range of 10–16 °C reported for fry of related species (Arctic char and bull trout).

Bioenergetics parameters

Bioenergetics models have primarily been used to examine changes in growth and consumption based on equations describing relationships between environmental conditions (largely temperature) and organismal energy allocation to metabolism, waste, and growth processes43,78. Fish Bioenergetics 4.0 (FB4 v1.1.5)43 was used to estimate each species’ growth outcomes by weight (g) based on data collected from individuals across Alaska and Canadian populations. Fish Bioenergetics 4.0 (FB4) requires data concerning individual fish size, energy density, proportion of prey consumed, prey energy density, prey indigestibility, and either overall consumption rate estimates or end-weight to run model simulations with regional or local context. Data for most of these parameters were found across the region for both YOY Chinook salmon and Dolly Varden, while regional consumption rates were not found and were estimated using the bioenergetics model, which expresses consumption as a proportion (P) of a theoretical maximum consumption rate (Cmax) (Tables S1 and S2).

FB4 and other bioenergetics models also require species-specific estimates of constants related to laboratory or field-tested metabolic processes involved in respiration, consumption, energy assimilation, and excretion43. FB4 simulations were conducted using weight-dependent energy density (Model 2)43 and physiological parameters relevant to Chinook salmon using the FB4 Coho salmon profile79 rather than the FB4 Chinook salmon profile80. Recent literature52,81found that bioenergetic parameters from Stewart and Ibarra79 are more applicable for sub-yearling Chinook salmon across Alaska than more recent estimates from Plumb and Moffit80 because daily temperatures across many Alaskan river basins rarely exceed 18 °C, and daily temperature estimates used for our model simulations were below 18 °C61. While FB4 does not have a bioenergetics profile for Dolly Varden nor are there adequate literature references assessing components of Dolly Varden bioenergetics, there are profiles and sufficient data for two closely related species to Dolly Varden, lake trout and bull trout. The lake trout bioenergetics profile was ultimately used to estimate growth outcomes for Dolly Varden, given parameterization based on YOY Dolly Varden population size ranges (Fig. S2). Bioenergetics simulations for Dolly Varden using the FB4 profile of bull trout were also tested; however, growth outcomes were less similar to observational sizes of Dolly Varden than those using the lake trout model (Fig. S2).

The overall question of the current study addressed growth potential as a difference in historical and mid-century weight outcomes; therefore, estimates of the maximum proportion of consumption (P-value/ % Cmax) were required to run model simulations43. To estimate consumption data for running final growth simulations, observational temperatures from the Alaska Water Temperature database82 were run with size and diet information for the beginning of the season (May/June) for a given species, and set to an average observational fish weight (g) from relevant river systems with the most observational data available (Chena River YOY Chinook salmon52; Aniak River YOY Dolly Varden69. These consumption rates were averaged across observational temperature years to obtain a P-value that may represent observations within a given river basin. Observational temperatures from the AKTEMP database and estimated P-values were first used to validate growth estimates within the Chena River for Chinook salmon and the Aniak River for Dolly Varden. Bioenergetics simulation weights by September for years where both fish size and water temperature data were available for a given system to fish mass observations from those years (2014–2015, Chena River; 2015–2021, Aniak River; Table S1) were compared to masses by September using modeled river temperature data for the same years. Bioenergetics simulations are considered to be validated when weights fall within 25% of observations83 ( Fig. S2). Chinook salmon simulations using the Chinook salmon model and regional data slightly overestimated average weight (34% difference, Fig. S2); however, data used for parameterization is limited, and more data specifically on the early summer season (May/June) YOY Chinook salmon size may strengthen weight validation. Likewise, due to the generalization of sizes across basins, model estimations above 25% were acceptable when used for individual river systems. Dolly Varden simulations using the lake trout model and modeled temperatures were overestimated but showed the closest sizes to observed end-of-season weights in the Aniak River system compared to the bull trout model (27% difference versus 43%, Fig. S2). Following parameter validation, simulations were run using the observed temperatures from the AKTEMP database, standardized species-specific size ranges and diet data to estimate P-values for each river basin (Tables S2 and S3). The average P-value (proportion of maximum consumption) for each species and river basin was used for final growth simulations to obtain the end-of-season weight (g) (weight by September 1st), specific growth rates (J/g/d), consumption rates (J/g/d), and respiration rates (J/g/d) for historical and future mid-century scenarios. Fish Bioenergetics 4.0 (v1.1.7) is open access and available for download at http://fishbioenergetics.org/.

Model outputs

Model outputs for each simulation span estimates of respiration, consumption, egestion, excretion, and growth rates43. The current study focused on end-of-season weight (g) (the end of the model simulation on September 1st to estimate thermal effects on a cumulative growth metric and specific growth rate (J/g/d) to estimate daily growth. Both metrics can be used for interpreting the growth potential of early life stages and subsequent survival probability into the next age cohort. End-of-season weight was examined for cold, warm, and average years for each river basin and period. The specific growth rate was examined for all years across historical and future periods for each river basin. Estimates of the temperature at peak specific growth rate were examined using equations from a regression model (y = f(x)= β0 + β1 × 1 + β2 × 2 + β3 × 3+€) for each river basin (Fig. S3). Temperatures ranging from 3 to 16 °C were substituted into each equation at intervals of 0.1 °C to define a curve for each basin. The temperatures of the highest specific growth rates for each river basin from these curves were taken and averaged across all river basins to define a cross-tributary average. Outputs of consumption rate (J/g/d) and respiration rate (J/g/d) were also examined with specific growth rate (J/g/d) across all years for each river basin.

Water temperature data

The optimized river basin model (RBM)47 was used to simulate river temperatures for the open-water season (June-September) during the historical baseline (Water Year 1990–2021) and a future mid-century high pseudo-global warming period (Water Year 2034–2065). A chain of models, including the optimized Regional Arctic System Model (RASM), a coupled land-atmospheric model, and mizuRoute, a river routing model were used to force the RBM45,46,47. The future climate conditions were determined using hydroclimate change scenarios derived from an ensemble mean of Coupled Model Intercomparison Project Phase 6 Global Climate Models (CMIP6 GCM)84 under the Shared Socioeconomic Pathway 2-4.5 (SSP2-4.5)85 emission pathway for the mid-21st century. The high pseudo-global warming scenario is based on the mean monthly climate change perturbations (2034–2065–1991–2020) from the top quartile of SSP2-4.5 models based on Alaska regional annual air temperature change and late-winter early spring precipitation changes. These monthly perturbations were applied to the historical climate data to generate the future climate data46. Water temperature data used for bioenergetics simulations were aggregated to the basin level by taking the distance-weighted average of each river segment. Specific tributaries within the Yukon and Kuskokwim River watersheds were assessed, specifically the Aniak, Andreafsky, Chena, Koyukuk, Porcupine, Takotna, and Teedriinjik (formerly, Chandalar) Rivers (Fig. 1). In 2015, the Teedriinjik (shimmering river in the Gwich’in language; Ch’idriinjik translates to heart river) was re-designated its Gwich’in name from the Chandalar River in 2015 to “help revitalize Gwich’in culture and language” (Edward Alexander, former Second Chief of the Gwichyaa Zhee Gwich’in Tribal Government)86. River temperature data for all modeled river basins across Alaska are available at the Arctic Data Center ( https://doi.org/10.18739/A2125QB7M)61.

Observational data from the AKTEMP database were used to estimate bioenergetic parameters for model simulations before simulations were run with modeled water temperature. Similarly, water temperature data was used to validate the RBM47. While observational river temperature data inherently biases model simulations using the same temperature dataset, this protocol ensures the model uses the observational environmental data to produce historical model values near actual observed fish size and condition values (Fig. S2).

Model simulations were run for three year types (Cold, Average, and Warm) representing ‘Historical’ and high pseudo-global warming (‘mid-century’) climate scenario temperatures for each river basin. Cold and Warm years were chosen out of the 31-year periods based on adding accumulated thermal units (ATUs; Daily Water Temperature in °C − 0 °C = Daily TU) for the simulation period of interest (June 1st -September 1st ). Warm years for each river basin were years with the highest ATUs, while Cold years were those with the lowest ATUs. The average temperature profiles of each river basin were based on the daily average temperatures from June 1st to September 1st across the entire 31-water-year period. Pairwise differences were used to create temperature delineations of cold, average, and warm temperatures for river basins for further analysis of bioenergetics simulations. Rivers were grouped by average temperatures less than 10 °C (“coldest”), 10–11 °C (“intermediate”), and greater than 11 °C (“warmest”) based on an Analysis of Covariance (ANCOVA) model and post-hoc pairwise comparisons of the average temperatures from the simulation periods across historical and future scenarios.

Statistical analysis

Prior to analyses, temperature and bioenergetics data were tested for assumptions related to homogeneity of variance, heteroskedasticity, and collinearity. Temperature data met parametric assumptions for ANCOVA. ANCOVA was used to compare average historical and future mid-century summer temperatures (June 1st -September 1st, Fig. 1; Fig. S1) and temperature distributions among Cold, Average, and Warm year types for each tributary of interest (Table S2). Following significant ANCOVA results, Tukey’s HSD post-hoc comparisons were conducted to determine specific differences among locations, years, and scenarios. The changes in average basin summer temperature (future mid-century-historical, °C) for each river basin were examined using linear regression models for possible effects of ecosystem complexity such as basin area (km2), river basin latitude (degrees), percentage of first-order streams (%), elevation (m), streamflow difference (future-historical, %), and air temperature difference (future-historical, %) (Fig. S2). Modeled river temperature data were analyzed for each day (June 1st -September 1st, n = 93 per year) and year (n = 32 per scenario) within historical and future scenarios to determine the percentage of the season within, above, and below each species’ thermal optimal range (i.e., thermal availability) based on relevant literature values. ANCOVAs followed by Tukey’s HSD post-hoc comparisons were used to compare the historical and future theoretical thermal availability of each tributary for both Chinook salmon (Table S5) and Dolly Varden (Table S6) based on the aforementioned percentages of the season within optimal temperature ranges. Welch’s ANOVAs and Steel Dass post-hoc comparisons were conducted to assess differences in growth potential for each species between historical and future model simulations among different cold, average, and warm thermal regimes for each temperature classification of the coldest (10 °C), intermediate (10–11 °C), and warmest (> 11 °C) rivers (Tables S7 and S8).

Regression models for both species were used to analyze the relationship between the average temperature (°C) and metabolic rates (i.e., specific growth rate (SGR), consumption rate (CR), and respiration rate (RR); J/g/d) of individuals in each river basin for every year within both historical and future mid-century scenarios (n = 434 total years). Equations from each basin-specific regression were used to define a curve for the relationship between temperature and metabolic rate across a temperature range spanning 3 to 16 °C (encompassing values from the daily modeled temperature outputs for the seven river basins examined), at increments of every 0.1 °C (Tables S8-S10). Graphical outputs are shown for the temperature range of 7–14 °C because average temperature values for each year and river basin fell within this range. The maximum metabolic rates for each river basin were estimated by analyzing the temperatures at which the specific growth rate was the highest. The average of these temperature values for specific growth rates was taken as the cross-basin average for each species. All statistical analyses were conducted using R version 4.3.0 (R Core Team, 2023) and JMP Pro 17 (JMP Statistical Discovery LLC; Cary, North Carolina).