Accelerating glacier volume loss on Juneau Icefield driven by hypsometry and melt-accelerating feedbacks

Mountain glaciers and icefields are major contributors to global sea-level rise, with Alaskan and Western Canadian glaciers providing some of the largest regional contributions and possessing large ice volumes. Plateau, top-heavy icefields with low-slope accumulation zones, such as in Alaska, are hypothesized to be especially vulnerable to small increases in equilibrium line altitude (ELA), to surface mass-balance (SMB)–elevation positive feedbacks as ice surfaces lower into warmer air, and to limited opportunities to retreat to higher elevations to re-equilibrate. Despite these theoretical expectations and regional importance, there is a paucity of long, multi-decadal empirical records to constrain how such icefields evolve and whether threshold or tipping-point behavior occurs. The study aims to quantify glacier change from the Little Ice Age (LIA; ~AD 1770) to 2020 across the Juneau Icefield to determine rates, patterns, and controls on icefield-wide shrinkage and thinning, and to diagnose the physical feedbacks driving recent acceleration.

Prior work shows: (1) global glacier mass loss substantially contributes to sea level rise (e.g., Zemp 2019; Slater 2021), with Alaska projected to remain a dominant regional contributor through 2100 under multiple emissions scenarios; (2) hypsometry strongly modulates mass-balance sensitivity in Alaska and NW Canada, with top-heavy/plateau icefields especially sensitive to ELA changes (McGrath et al. 2017); (3) SMB–elevation feedbacks can destabilize ice masses as surface lowering increases melt via lapse-rate warming (e.g., Zekollari et al. 2017; Sass et al. 2017); (4) models often project relatively smooth or delayed accelerations in mass loss for Juneau Icefield (e.g., PISM-based studies) with acceleration mainly after ~2070, yet observational constraints remain limited; (5) Juneau Icefield benefits from long-term mass balance records (since 1946) and documented changes at key glaciers (e.g., Lemon Creek, Taku), including rising ELAs and recent accelerations in thinning and retreat. Regional climate variability, especially the Pacific Decadal Oscillation and the Aleutian Low, modulates precipitation and temperature patterns, with a regime shift around 1976 increasing winter precipitation and warming, and strong warming trends since 2000 lengthening the melt season.

Study region: Juneau Icefield (Alaska/BC), temperate maritime icefield spanning 0–2300 m asl with extensive low-slope plateaus (~1200–1500 m asl). Baseline glacier inventory: RGI v6.0 outlines (census ~2005) edited using Landsat/Sentinel imagery to produce outlines for 2019, 2015, 2005, 1990, 1979, 1948, and reconstructed LIA (~1770). LIA extent reconstruction: manual extension of 2005 RGI outlines to LIA moraines, trimlines, and ice-scoured surfaces using published geomorphological mapping and historical data; applied morphostratigraphic principles and conservative selection of innermost LIA ridges. LIA ELAs estimated via AABR method (balance ratio = 1.88, maritime glaciers) to define ablation-area polygons for volume reconstruction. Photogrammetry and DEMs: Stereo aerial photographs from 1948 and AHAP 1979 processed in MicMac (tie points, relative orientations, orthomosaics, DEMs). Orthophotos registered to Sentinel-2; DEMs coregistered to Copernicus 30 m DEM; voids masked and filled via local mean elevation change. Additional DEMs: ArcticDEM v3 (2 m) as base topography; ASTER GDEM, ALOS PRISM DEM used where needed. DEMs of difference: generated for LIA–1948 (ablation areas only), 1948–1979, 1979–2000 (using SRTM), 2000–2010 and 2010–2020 (from Hugonnet et al. 2021 using ASTER/ArcticDEM). Volume change calculation: for each period, glacier outlines at the start date were divided into 50 m (or 10% elevation range) bands; mean dh computed per band (outliers removed via NMAD), scaled by band area, and summed per glacier; glaciers lacking coverage used icefield-wide average dh/dt for scaling. Uncertainty: DEM differencing uncertainty estimated using NMAD of off-glacier differences; additional interpolation uncertainty accounted for by an adapted formula combining NMAD, fraction interpolated, and on-glacier variability; total uncertainty combined glacier area and elevation-change uncertainties. Glacier area change: manual digitization from multi-sensor imagery; minimum mappable area ~0.001 km^2; annualized rates computed using acquisition dates; uncertainty in area via repeated digitization of 11 representative glaciers (seven rounds), regression to up-scale 95% CIs by size class. Snowlines: late-summer 2019 snowline mapped from cloud-free Sentinel-2 (Aug–Sep 2019), identifying the transition between snow and bare firn/ice. Disconnections: previously mapped glacier disconnections (often at icefalls) linked to time slices by overlay with updated outlines; elevation and slope at disconnections extracted from ASTER GDEM. Albedo: stack of 299 Landsat 5–9 scenes (1986–2023) in Google Earth Engine; clouds masked via USGS QA; broadband albedo computed from blue/red/NIR following established algorithms; mean albedo computed within 1990/2005/2019 glacier outlines and for plateau pixels >1500 m; seasonal means compiled; statistical comparisons conducted (e.g., t-tests). Climate datasets: NOAA station data (Juneau Airport) for temperature and precipitation; ERA5 reanalysis to document regional anomalies (1990–2005; 2015–2019 vs. 1950–1980).

• Juneau Icefield-wide acceleration: Area shrinkage rates increased sharply after 2005; 2015–2019 rates (0.96% a^-1; 38.47 km^2 a^-1) were ~5× those of 1979–1990 (0.18% a^-1; 8.33 km^2 a^-1) and ~7× those of 1948–1979.
• Volume loss rates: LIA–1979 remained relatively steady at ~0.65–1.01 km^3 a^-1; 1979–2010 increased to 3.08–3.72 km^3 a^-1; 2010–2020 doubled to 5.91 ± 0.80 km^3 a^-1. Mean dh/dt (all glaciers) rose to −0.74 ± 0.03 m a^-1 (2010–2020).
• Cumulative loss since LIA: 315.3 ± 237.5 km^3 (≈24.25% of 2017–18 reconstructed volume), with only 70.53% of glacier area remaining vs. the LIA; 108 glaciers disappeared by 2019; 47 new ice-contact proglacial lakes formed.
• Outlet glaciers: Accounted for 538.0 ± 32.4 km^2 of area loss since LIA; main outlet termini receded 4–5 km; severe thinning post-2010 with outlet glacier mean dh/dt ≈ −1.21 ± 0.21 m a^-1 (2010–2020).
• Thinning elevation extent: Thinning propagated upslope and across the plateau since 2005, reaching ~1500–1800 m asl (e.g., Meade plateau ~1550 m; Mendenhall ~1380 m; Tulsequah ~1490 m).
• Snowline/ELA rise: 2019 late-summer snowlines averaged 1612 m asl (SD 162 m; n=178); Taku mean snowline 1445 m (2019) vs. measured ELA 1528 m; Lemon Creek had no late-season snow (ELA 2023 m in 2019). ELAs at Lemon Creek rose from ~1038 m (1961–1970) to ~1499 m (2011–2020); Taku from ~912 m (1940–1950) to ~1159 m (2011–2020).
• Albedo decline: Mean albedo within 1990 outlines decreased from 0.81 ± 0.03 (1987–2009) to 0.67 ± 0.03 (2010–2023); plateau (>1500 m) from 0.92 ± 0.02 to 0.78 ± 0.04. 2018–2019 late-summer albedos exceptionally low (0.44–0.48).
• Fragmentation/disconnections: Increasing frequency post-2005; mean disconnection elevation ~1299 m asl (1362 m after 2005), aligning with plateau rim icefalls. Disconnections co-occur with strong local thinning (e.g., Thiel tongue up to 9 m a^-1; glacier-wide −3.33 m a^-1 in 2010–2020).
• Climate context: Since 1941, winter temperatures increased ~0.35 °C/decade; 2001–2020 winters were 2.07 °C warmer than 1941–1970; summer (2001–2020) ~0.97 °C warmer than 1941–1970. Post-1976 shift to a stronger Aleutian Low increased precipitation; post-2015 weakening noted.

Observed mass loss at Juneau Icefield has sharply accelerated in the early 21st century, contrary to some modeling studies that suggested more linear losses until mid/late century. The analysis attributes the acceleration to hypsometrically controlled, melt-accelerating feedbacks triggered once the ELA and late-summer snowlines intersect the low-slope plateau elevations (~1200–1500 m). This intersection rapidly reduces the accumulation area ratio, thins ice across the plateau, lowers regional and surface albedo (exposing darker firn/ice/rock and increasing absorption), and enhances SMB–elevation feedbacks as surfaces drop into warmer air. These processes are compounded by increasing glacier fragmentation at structurally vulnerable icefalls, which disrupts ice flux from accumulation basins to tongues, increasing stagnation and downwasting and further darkening through debris and rock exposure. Together, these feedbacks create a threshold response: modest additional warming (e.g., mean summer anomaly +0.55 °C in 2015–2019 vs. 1986–2005) corresponded with a disproportionate increase in area loss and thinning. Given thick plateau ice (>600 m in places), the SMB–elevation feedback can operate over large vertical ranges, impeding regrowth even under climate stabilization and potentially constituting a dynamic tipping point. The findings highlight the importance of hypsometry in modulating regional glacier responses and suggest that other plateau icefields in Alaska, Canada, Greenland, and Norway may experience similar accelerations once ELAs reach their plateaus.

This study provides a 250-year, icefield-wide reconstruction and quantification of glacier area, thickness, and volume change at Juneau Icefield, demonstrating a recent, marked acceleration in both area recession and volume loss, especially since 2005 and doubling in volumetric loss rates after 2010. The acceleration is driven by hypsometry-mediated feedbacks—ELA rising onto the plateau, reduced accumulation-area ratio, albedo darkening, SMB–elevation feedback, and enhanced fragmentation at icefalls—that together push the system towards threshold behavior and hinder potential regrowth (hysteresis). Given the prevalence of low-slope, top-heavy icefields regionally and globally, these processes likely generalize, implying higher near-future losses than linear projections suggest and significant implications for sea-level rise and water resources. The authors recommend that glacier model projections integrate longer historical observational constraints (LIA to present), explicitly represent hypsometric sensitivities and feedbacks, and evaluate the potential for threshold/tipping behaviors. Future work should broaden observations of ELAs, snowlines, albedo, and disconnections across other plateau icefields, and couple these datasets into model-data frameworks to improve predictive skill.

Volume change from the LIA is reconstructed only for ablation areas (due to limited geomorphic markers above ELA), so LIA–present changes represent minimum estimates; some mass loss above ELA is likely undetected. The LIA–1948 period has higher uncertainties due to DEM coverage gaps and interpolation (especially in Canada and accumulation areas), and potential positive elevation-change artifacts were masked, biasing toward conservative loss. Mass change (density conversion) was not calculated; volume changes are used as proxies. The 1948 extent is the best available estimate given mixed data sources (orthomosaics, topographic maps), but uncertainties remain in outlines and DEMs despite rigorous error assessment. Post-2000 volume-change estimates rely on published ASTER/ArcticDEM products and use outlines at fixed census dates, so some area-change effects within periods are not captured, making those rates conservative.