Meltwater to Megawatts- A Masters Thesis: Modeling Dixon Glacier Peak Water for Renewable Energy Applications

CATEGORIES

Go Deeper

Built for an M.S. thesis at Alaska Pacific University, this project models Dixon Glacier on Alaska’s Kenai Peninsula to forecast when glacial meltwater will peak and how quickly it declines under different climate‑emission scenarios. This is important as the state is considering a $200-300 Million diversion project to add 60 MW of hydroelectric energy capacity to nearby Bradley Lake, an existing 120MW power facility. The research pipeline spans everything from raw SNOTEL weather‑station data to century‑scale ensemble projections. The human side contributed glaciology expertise, two summers of field‑stake measurements at three elevations, and 22 years of hand‑digitized Landsat snowlines. Claude Code (Opus) handled every layer of implementation — model architecture, numerical methods, data pipelines, calibration algorithms, and a complete thesis‑grade research log. Results: Ensemble projections showing peak meltwater around water year 2043 (SSP2‑4.5) to 2058 (SSP5‑8.5), this shows that there is likely enough water available to justify the expense of the diversion project. This model can also be used for other mountain glacier analyses.

Stack Used

Built With: - Claude Code (Opus 4.6 1M) - Python 3 + NumPy/SciPy/Pandas/Matplotlib - Numba JIT compilation for grid-scale physics loops - MCMC (adaptive Metropolis) + differential evolution for Bayesian calibration Model Stack: -Melt model inspiration from Hock (1999) DETIM Method 2 — distributed enhanced temperature-index melt model - Ice model inspiration from Huss (2010) delta-h parameterization — glacier retreat dynamics - 5m IfSAR DEM, 100m computational grid - RGI7 glacier outline, Farinotti (2019) ice thickness Data Sources: - NRCS SNOTEL — Nuka Glacier (primary) + 4 nearby stations for gap-fill (Middle Fork Bradley, McNeil Canyon, Anchor River Divide, Lower Kachemak Creek) - On-glacier AWS — Dixon Glacier ELA site (1078m), summer 2024–2025 field seasons - Hugonnet et al. (2021) — satellite geodetic mass balance 2000–2020 - Landsat archive — 22 years of digitized snowline positions (1999–2024) - NEX-GDDP-CMIP6 (AWS S3) — daily downscaled climate projections, 5 GCMs (ACCESS-CM2, EC-Earth3, MPI-ESM1-2-HR, MRI-ESM2-0, NorESM2-MM), SSP2-4.5 + SSP5-8.5