Glacier melt within alpine catchments provides a vital component of runoff that constitutes an important water resource for downstream populations. With future climate changes, it is expected that glacier volume change will be considerable in the coming decades, with associated implications for runoff. Estimation of future changes in glacier volume and catchment runoff is therefore essential for understanding future water resource implications in alpine environments.
This thesis focuses on glacier volume and runoff changes predicted using the statistical model GERM (Glacier Evolution and Runoff Model; Huss et al., 2008a) and has three novel aims. Firstly, to provide more robust assessments of the modelling uncertainty associated with predicted glacier and runoff changes from alpine catchments than previous studies, by challenging the model to reproduce historic changes in glacier volume and evolution over 120 year periods, and comparing predicted and measured runoff. Secondly, to use this assessment of uncertainty to contextualise and understand the precision of future (to 2100 AD) runoff projections for alpine catchments under a wide range of possible climate changes scenarios. Thirdly, to develop the model so that it can be applied to a debris-covered, downwasting glacier in the Himalaya. Two further novel aspects of this thesis are the development of a more systematic and robust calibration procedure for GERM, and the application of climate data downscaling techniques that are more sophisticated than have hitherto been applied in glacio-hydrological studies.
To achieve aim 1, GERM was used to forward model glacier volume and runoff for the Griesgletscher and Rhonegletscher catchments in the European Alps from 1884-2004. As a statistical model that requires catchment-specific calibration, GERM was first calibrated to each catchment using contemporary glacier volume and catchment runoff measurements (as is standard when using the model for future projections). Digital elevation models were then used to obtain the initial glacier geometry required to begin each model run, and each completed model run was subsequently used to estimate the accumulated uncertainty associated with the predicted glacier volume/runoff changes by comparing modelled with observed glacier volume/runoff change at the end of the simulation.
To achieve aim 2, future model runs (2010-2100) were conducted for the same two catchments and the glacier volume/runoff uncertainty calculated from model performance in the past (aim
1) applied to future projections. Future simulations were driven by a wide-range of climate inputs to allow quantification of the uncertainty associated with climate scenarios/models. The combination of these two sources of uncertainty (GERM and climate) provides future
II
projections with greater awareness and better quantification of uncertainties than previous studies.
Finally, to achieve aim 3, GERM was applied to the debris-covered Khumbu Glacier by adjusting the mass redistribution process of GERM (Δh-parameterisation) to reflect the downwasting behaviour of the debris-covered glacier tongue, based on observed thinning rates at Khumbu Glacier. Additionally, to account for the insulating effect of debris on ice, the modelled melt rate was reduced in proportion to debris thickness on a spatially distributed basis (i.e. debris thickness was not uniform) using observations of reduced melt at glaciers close to Khumbu.
Improvements to the calibration procedure used when applying GERM were made and applied throughout this thesis by developing an automated calibration which systematically adjusts the parameters, calculates a combined goodness-of-fit statistic that allows comparison to observations of both glacier volume and runoff, and selects the optimal parameter set. Improved downscaling methods were also used and applied to all future volume and runoff change projections made during this thesis. Specifically, state-of-the-art General Circulation Model simulations were dynamically-statistically downscaled using Regional Climate Model simulations and quantile mapping, and were used to drive future model runs at all three sites. Finally, the novel adjustments made to the mass redistribution process and the inclusion of reduced melt beneath debris indicate that GERM can now be applied to debris-covered glaciers. A recommendation for future research is that GERM is further tested on additional debris- covered glaciers and applied to additional catchments in the larger Everest region.
The results of the uncertainty analyses (aim 1) show that glacio-hydrological model uncertainty amounts to annual runoff errors of ±0.04 106m3yr-1 (±0.15 % yr-1), and glacier volume errors of
±0.16 % yr-1, over time periods of 120 years at Griesgletscher. At Rhonegletscher, the uncertainty assessment resulted in annual runoff errors of ±0.16 106m3yr-1 (±0.2 % yr-1) and glacier volume errors of ±0.13 % yr-1, over time periods of 120 years. Nonetheless, the key finding is that the main sources of future uncertainty relate to emissions scenarios and GCM-RCM (General Circulation Model - Regional Climate Model), combinations which lead to variations in predicted future runoff in 2100 of ±36 % at Griesgletscher and ±20 % at Rhonegletscher. The results of the future simulations (aims 2 and 3) indicate that all three glaciers that form the focus of this thesis will lose considerable volume. Specifically, by 2100, Griesgletscher is likely to have become an ice-free catchment (87-100 % ice loss); Rhonegletscher will have lost 70-90 % of ice; and Khumbu Glacier will have lost 61-92 % of ice. The results further show that mass losses will cause an initial increase in annual river discharge followed by a decline in discharge levels, such that annual discharge by 2100 will be considerably lower than present, with peak discharge at Griesgletscher occurring in 2020, at Rhonegletscher in 2075, and at Khumbu Glacier in 2045.
