Application of global datasets and data assimilation to a distributed hydrological model in the Canadian Sub-Arctic

• Main Author: Casson, David R.
• Format: Other/Unknown Material
• Language: unknown
• Published: Delft : UNESCO-IHE Institute for Water Education; 2017
• Subjects: hydrology; hydrological modelling; snow hydrology; data assimilation; Arctic; Canada
• Online Access: https://doi.org/10.25831/8dm7-dm05; http://cdm21063.contentdm.oclc.org/cdm/ref/collection/masters2/id/105400

Hydrological modelling in the Canadian Sub-Arctic is hindered by sparse meteorological and snowpack data. Long term hydrological assessment and modelling tools available for operational water managers are similarly limited. This is true for the Snare Watershed in Northern Canada, which is a data sparse, snowmelt driven catchment providing inflow to a cascade of 4 hydropower stations. Operational water management in the Snare Hydro System could benefit from modelling tools that simulate river discharge, accumulation of winter snowpack and snowmelt. Given the remote location of the Snare Watershed, global re-analysis datasets are well suited to provide the required input meteorological data. Discharge gauges available in near-real time can be assimilated to improve model performance. This study explores the application of global re-analysis datasets from the European Union FP7 eartH2Observe project. Precipitation data from the Multi-Source Weighted-Ensemble Precipitation (MSWEP) product and temperature data from Watch Forcing Data applied to ERA-Interim data (WFDEI) provide the forcing data for a distributed version of the HBV-96 hydrological model. The GlobSnow-2 Snow Water Equivalent (SWE) product funded by the European Space Agency is evaluated for snowpack representation and operational utility. Model calibration and testing results demonstrate successful application of these datasets to the distributed WFLOW-HBV PCRaster model for discharge prediction, particularly for the MSWEP precipitation data. Data assimilation of river gauge measurements using an Ensemble Kalman Filter further improves discharge prediction and initial conditions for forecasting. The timing of both snowpack accumulation and melt is well-represented. However, the spatial representation of the snowpack is a persistent issue in deterministic and data assimilation model runs, which may be attributed to the structure of bias corrections in the MSWEP data. The GlobSnow-2 product is found to under-predict late season snowpack over the study ...