Estimating Snow Mass and Peak River Flows for the Mackenzie River Basin Using GRACE Satellite Observations

• Published in: Remote Sensing
• Main Authors: Shusen Wang, Fuqun Zhou, Hazen A. J. Russell
• Format: Article in Journal/Newspaper
• Language: English
• Published: MDPI AG 2017
• Subjects: flood; GRACE satellites; snow; river flow; cold region; Mackenzie River basin; model; Science; Q; Canada; Mackenzie River; Snow River; Mackenzie river
• Online Access: https://doi.org/10.3390/rs9030256; https://doaj.org/article/d2deae57a1674c54aeda1e5d85354764

Flooding is projected to increase with climate change in many parts of the world. Floods in cold regions are commonly a result of snowmelt during the spring break-up. The peak river flow (Qpeak) for the Mackenzie River, located in northwest Canada, is modelled using the Gravity Recovery and Climate Experiment (GRACE) satellite observations. Compared with the observed Qpeak at a downstream hydrometric station, the model results have a correlation coefficient of 0.83 (p < 0.001) and a mean absolute error of 6.5% of the mean observed value of 28,400 m3·s−1 for the 12 study years (2003–2014). The results are compared with those for other basins to examine the difference in the major factors controlling the Qpeak. It was found that the temperature variations in the snowmelt season are the principal driver for the Qpeak in the Mackenzie River. In contrast, the variations in snow accumulation play a more important role in the Qpeak for warmer southern basins in Canada. The study provides a GRACE-based approach for basin-scale snow mass estimation, which is largely independent of in situ observations and eliminates the limitations and uncertainties with traditional snow measurements. Snow mass estimated from the GRACE data was about 20% higher than that from the Global Land Data Assimilation System (GLDAS) datasets. The model is relatively simple and only needs GRACE and temperature data for flood forecasting. It can be readily applied to other cold region basins, and could be particularly useful for regions with minimal data.