Glacier retreat creating new Pacific salmon habitat in western North America

The study investigates how contemporary and future glacier retreat across the Pacific mountain ranges of western North America will affect freshwater habitat availability for Pacific salmon (Oncorhynchus spp.). While climate change has driven stressors such as ocean heat waves, low freshwater flows, and higher water temperatures that reduce habitat quality and productivity for many salmon populations, retreating glaciers simultaneously create new proglacial streams and lakes that could be colonized by salmon. Prior observations in places like Glacier Bay, Alaska, show rapid colonization and population growth of salmon following deglaciation. Yet, there had been no regional-scale projections of when and where new salmon-accessible freshwater habitats would emerge due to glacier loss. The purpose of this study is to quantify, map, and time the creation of new salmon-accessible streams and suitable spawning/rearing habitats under future climate scenarios, thereby informing forward-looking conservation and resource management decisions amid competing interests such as hydropower and mining. The research question centers on how much new salmon habitat will be created by glacier retreat by mid- and late-century, where it will occur, and how accessibility is constrained by stream gradients and watershed topography.

Regional colonization of recently deglaciated streams by salmon has been documented at the watershed scale, illustrating the potential for rapid use of new habitats after glacier retreat. Prior work has emphasized both beneficial and detrimental effects of glacier dynamics on salmon ecosystems, including influences on temperature, flow, turbidity, sediment supply, and channel morphology. Glacier change in western North America has been extensively studied, with rapid declines in glacier volume, thickness, and area linked to anthropogenic climate warming. Existing habitat models and intrinsic potential approaches relate stream gradient and channel type to salmon spawning and rearing suitability, with species-specific swimming and migration abilities constraining accessibility at higher gradients. However, before this study, no comprehensive, transboundary, regional projections existed that link modeled glacier retreat to the creation of new salmon-accessible stream networks and suitable spawning/rearing habitats across multiple sub-regions and climate scenarios.

Study region and data: The analysis spans glacierized watersheds within the Pacific salmon range from southeastern British Columbia to southern coastal Alaska (~623,000 km²). Glacier outlines (~43,963 glaciers; ~81,000 km²) were sourced from the Randolph Glacier Inventory (RGI v6.0), with outline epochs circa 2005–2014 in western Canada and 2008–2012 in Alaska. Digital Elevation Models (ASTER GDEM v2.0, ~30 m resolution) were used to derive synthetic stream networks across 18 sub-regions using GIS hydrology tools (ArcGIS 10.6, QGIS 2.18).

Synthetic stream network and accessibility: Present-day and future (post-retreat) stream networks were derived from DEMs, including reconstruction of subglacial terrain to represent streams that would be exposed as glaciers retreat. Stream segments were analyzed at ~500 m lengths to estimate slope/gradient from DEM elevations. Accessibility for adult salmon migration was determined using conservative stream gradient thresholds: primary threshold <10% gradient over >500 m for accessibility; an alternative, less conservative scenario used <15% gradient to reflect species with stronger swimming capabilities (e.g., Chinook) and to test sensitivity. Analyses focused on streams larger than first order (i.e., second order and above), acknowledging that salmon can use first-order streams but to maintain conservative estimates they were excluded.

Glacier retreat modeling and climate scenarios: Glacier retreat timing was modeled using a glacier model (GLOGM framework referenced) driven by ensembles of five Global Climate Models under Representative Concentration Pathways RCP4.5 and RCP8.5. The model provided time-evolving glacier extents and thickness, enabling determination of when subglacial valleys become ice-free and streams accessible to salmon under gradient thresholds. Projections were summarized for benchmark periods 2050 and 2100 using 10-year means.

Habitat suitability for spawning and rearing: Within the salmon-accessible networks, suitable habitat for spawning and juvenile rearing was identified using stream gradient criteria: conservative suitability at 0–2% gradient within accessible streams; a less conservative scenario used 0–4% gradient within streams accessible under the <15% threshold. Habitat quantities were computed by sub-region and time period.

Uncertainty and validation: Uncertainty was propagated from (i) GCM projections for each RCP, (ii) glacier ice thickness estimates and glacier model sensitivity (including ensemble spread), and (iii) stream segment length/elevation sampling for gradient estimates. Results are presented as ensemble means ± one standard deviation. Additional limitations include lack of comprehensive barrier datasets (e.g., waterfalls), potential DEM/segmentation artifacts, and species-specific variability in swimming performance. Regional examples and field knowledge informed gradient thresholds, with literature support for 10–20% migration constraints; the study adopted a conservative 10% threshold and an alternative 15% threshold for sensitivity.

• By 2100, glacier retreat is projected to create 6,146 ± 1,619 km of new streams accessible to Pacific salmon across the study region.
• Of these, 1,930 ± 569 km are projected to have gradients suitable for spawning and juvenile rearing under conservative criteria, representing gains of 0–27% among the 18 sub-regions.
• Temporal distribution: Approximately 23% of total potential salmon-accessible stream kilometers will be created by 2050, and about 63% by 2100, with substantial regional variation.
• Regional differences: Limited gains in the Skeena River watershed and the North Coast of British Columbia are projected mostly by 2050, whereas in the Gulf of Alaska only ~20% of new accessible stream kilometers will be created by 2050, with larger gains later due to the presence of large glaciers.
• Suitable habitat expansion under RCP4.5 by 2100 is largest in the Gulf of Alaska and Copper sub-regions, with 757 (+279) km and 408 (+105) km, respectively, of 0–2% gradient habitat within accessible networks.
• Sensitivity to thresholds: Using a less conservative accessibility threshold (<15% gradient) and broader suitability (0–4% gradient) yields ~65% more suitable habitat than the conservative estimates, with variability among sub-regions.
• Production implications: One km of suitable habitat can produce on the order of ~500–1,500 juvenile sockeye salmon; thus, hundreds to thousands of km of new habitat could translate into hundreds of thousands to millions of additional juveniles, depending on species and local conditions.

The findings address the central question by quantifying and mapping when and where glacier retreat will generate new freshwater habitats that are accessible and suitable for Pacific salmon. They demonstrate that, despite widespread warming-related stressors on salmon, deglaciation can create substantial new stream networks, particularly in regions with large, low-gradient glaciated valleys (e.g., Gulf of Alaska, Copper River). The temporal pattern indicates that most gains accrue later in the century, with earlier gains in regions where smaller or steeper glaciers retreat more quickly into accessible terrain.

The results are significant for conservation and management: identifying emerging habitats provides an opportunity for proactive protection, habitat connectivity planning, and consideration of competing land/water uses (e.g., hydropower, mining) before development forecloses salmon access. Downstream effects of glacier change complicate the picture: in southern/interior regions, diminishing meltwater may reduce summer flows and cool-water refugia, decreasing habitat quality; in southeastern Alaska, increased meltwater in coming decades may enhance cold-water availability but also increase turbidity, with mixed implications for salmon. The study underscores that watershed topography and species-specific swimming abilities will mediate realized habitat gains. Overall, new habitat creation via glacier retreat could partially offset some climate-related habitat losses elsewhere, but outcomes will vary regionally and by species.

This study provides the first regional-scale projection linking glacier retreat to the creation of new salmon-accessible streams and suitable spawning/rearing habitats across the Pacific mountain ranges of western North America. By 2100, thousands of kilometers of new accessible streams are expected, with substantial suitable habitat, especially in Alaska’s Gulf of Alaska and Copper sub-regions. These insights enable forward-looking management to safeguard migration pathways, prevent barriers, and balance conservation with development pressures in newly deglaciated landscapes. Future research should refine species-specific accessibility and habitat criteria, incorporate barrier mapping and dynamic channel evolution, improve DEM resolution and stream segmentation, couple hydrology and temperature/sediment models to assess habitat quality trajectories, and evaluate biological colonization rates and population responses under different climate and management scenarios.

• Gradient-based accessibility is a simplification; species-specific swimming ability and flow conditions vary and may allow or preclude passage beyond assumed thresholds.
• Lack of comprehensive datasets on migration barriers (e.g., waterfalls, culverts) between British Columbia and Alaska means some modeled accessible segments may be blocked in reality.
• Stream networks were derived from DEMs with ~30 m resolution and segmented at ~500 m; this can miss fine-scale features and introduce slope estimation errors.
• Focus on streams larger than first order yields conservative habitat totals; salmon use of first-order streams is underrepresented.
• Glacier retreat projections depend on GCMs, RCP scenarios, and glacier model assumptions (e.g., ice thickness, response times); ensemble uncertainty remains, and actual emissions trajectories will influence outcomes.
• Habitat suitability was inferred primarily from gradient; other determinants (temperature, flow regime, turbidity, sediment stability, channel morphology) were not explicitly modeled and may alter realized habitat quality and productivity.
• Some inconsistencies in available datasets and regional differences in data epochs for glacier outlines may affect temporal alignment.
• The study does not explicitly model estuarine/lake habitat formation that may also result from deglaciation.