An assessment of the impact of climate change on the contribution of glaciers to the Upper Yukon River flows using GRACE data, spatial concentration functions, automated machine learning and hydrological modelling
Abstract
In the Upper Yukon Watershed (UYW, 20,000 km2), seasonal melt from glaciers contribute significantly to annual runoff and operation of the Whitehorse power plant. From a strategic planning point of view, this study aims to analyze the impact of climate change on the contribution of glaciers, which covers 5% of the UYW, to the annual runoff using hydrological modelling, GRACE data and machine learning algorithms. The spatial resolution of GRACE data remains too low to discriminate changes in glacier mass signal at the scale of the UYW. Thus, here we applied a spatial concentration function approach to build high resolution monthly time-series of glaciers mass changes over the UYW. To estimate glaciers mass changes, we decomposed four GRACE TWS solutions with different processing assumptions (i.e., CSR RL06 DDK5; CSR RL06 DDK8; GRGS RL05 and CSR RL06 MASCON) using monthly data from LSM GLDAS v2.1 (Rodell et al., 2004) and WGHM v2.2d (Müller Schmied et al., 2020). Spatial concentration functions were derived from four heterogeneous a priori of different resolutions and sources (Hugonnet et al., 2021; Larsen et al., 2015) and the leakage was subtracted by using glaciers over the Gulf Of Alaska (GOA). To analyze the accuracy of our assessments, we compared the trends resulting from the spatial concentration functions and the constrained forward approach (Doumbia et al., 2020) over the GOA and the Saint-Elias Mountains. To extend/reconstruct glacier mass change up to GRACE-FO (i.e. 2003-2020), we used the Automated Machine Learning (AML) H2O-AutoML (LeDell and Poirier, 2020). Then, glacier mass anomalies were used to calibrate the hybrid (i.e., degree-day/thermal energy balance) glacier melt model of HYDROTEL over UYW. For the period of 2003 to 2016, the trends in glaciers mass losses over the GOA and Saint-Elias Mountains varied from 41.61 to 53.43Gt/yr and 19.31 to 28.88Gt/yr, respectively. Our results compared well with the glaciers mass losses reported in other studies. The AML algorithms performed well with NSE values varying from 0.75 to 0.99; correlation coefficients from 0.93 to 0.99; P-bias from -2.4 to 4.8 and NMRSE from 0.8 to 49.6. The use of the glacier mass changes, in addition to stream flows, improved the calibration of HYDROTEL. To our knowledge, this study is the first to combine spatial concentrations with multiple GRACE data and heterogeneous a priori, AML algorithms and a glacier melt model to analyze the impact of climate change (based on CMIP6 data) on glacier melt contribution to stream flows. The latter results will be presented at the meeting.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2021
- Bibcode:
- 2021AGUFM.H25H1115D