SnowCast

SnowCast is an experimental data product that uses the Global Environmental Multiscale (GEM) model forecasts from Environment and Climate Change Canada (ECCC) to drive the Canadian Hydrological Model (CHM). Estimates of snowpack are provided over the a Bow River Basin, centered over Banff, Canada.

SnowCast is developed as part of Global Water Futures and the Centre for Hydrology, University of Saskatchewan. Thanks to collaboration with ECCC.

Two GEM forecasts are currently used:

DISCLAIMER: DO NOT USE FOR DECISION MAKING, EXPERIMENTAL!

HRDPS-CHM 2-day forecast

Snow water equivalent

Snow depth

Snow water equivalent change

Snow depth change

Current

GDPS-CHM 6-day forecast

Snow water equivalent

Snow depth

Snow water equivalent change

Snowdepth change

Simulation domain

The simulation domain covers the Upper Bow River watershed, above Calgary, Alberta. The domain is divided into 129,721 triangles using the Mesher algorithm (Marsh et al, in review), to better approximate complex terrain and variable vegetation. This has resulted in requiring only 4% of the total computational elements as the original raster, dramatically improving computational efficiency.

About

SnowCast is developed by Chris Marsh and Nic Wayand at the University of Saskatchewan. Further details can be found in Wayand et al. (in prep).

The Canadian Hydrological Model (CHM) is a modular, multi-physics, spatially distributed modelling framework designed for representing cold-regions hydrological processes. CHM uses existing high-quality open-source libraries and modern high-performance computing practices to provide a framework that allows for integration of a wide range of process representations, ranging from simple empirical relationships to physics-based, state-of-the-art algorithms. Modularity in structure and process representation allows for diagnosis of deficiencies in these aspects of the model. CHM also has sufficient flexibility in spatial representation and algorithm parameterisation to assess uncertainty in model structure, parameters, initial conditions, process representation, and spatial and temporal scales. By utilizing unstructured meshes, fewer than 1% of the computational elements of high-resolution structured (raster) grids are usually necessary. This preserves surface and sub-surface heterogeneity but results in fewer parameters and initial conditions.

SnowCast is licensed under the Creative Commons No Derivative 4.0 license. Full license details here.