Machine learning to improve assimilation of snow observations for (sub)seasonal hydrologic forecasts

点击翻译按钮获取中文摘要

Forecasts of streamflow in Canadian rivers between two weeks and two months ahead are of tremendous value to individuals, governments and industry. Aquanty Inc.would like to improve its capability to produce these forecasts, and a major challenge is obtaining accurate information about the quantity of water stored in the surface snowpack at the beginning of the forecast period. This proposal uses novel statistical and modelling techniques to produce a new optimal snow data product that can be used as an input for hydrologic forecasting. The new snow data product will be produced in real-time from an optimal combination of station snow surveys, remote sensing and reanalysis data, and will have great value to water managers. The impact of using this improved snow data to initialise hydrologic forecasts will be measured and reported in the scientific literature, where it will be of considerable interest to developers of similar forecast systems across the world.