The impact of assimilating Earth Observation and in situ data on seasonal hydrological predictions in a snow-dominated river system

Earth Observations (EO) have become popular in hydrology because they provide valuable information in locations where direct measurements are either unavailable or prohibitively expensive to make. Recent scientific advances have enabled the assimilation of EO’s into hydrological models to improve the estimation of initial states and fluxes which further leads to improved forecasting of different hydrometeorological variables. When assimilated, the data exert additional controls on the quality of the forecasts; it is hence important to apportion the effects according to model forcing and the assimilated data. Here, we investigate the impact of assimilating different EO and in itu data-sets individually and as combinations on the discharge and reservoir inflow estimations in the snow dominated Umeälven catchment in northern Sweden. We further assess the impact of the assimilations on seasonal predictions over the catchment. Six datasets are assimilated comprising of four EO products (fractional snow cover, snow water equivalent, and the actual and potential evapotranspiration) and two in situ datasets (discharge and reservoir inflow). For the latter investigation,  we drive the E-HYPE hydrological model with two meteorological forcings: (i) a down-scaled GCM product based on the bias-adjusted ECMWF SEAS5 seasonal forecasts, and (ii) historical meteorological data based on the Ensemble Streamflow Prediction (ESP) method. We finally assess the impacts of the meteorological forcing and the assimilated data on the streamflow and reservoir inflow seasonal forecasting skill for the period 2001-2015. We assessed the value of assimilating different data-sets and identified the datasets that can be meaningfully combined. We further show that all assimilations generally improve the forecasting skill but the improvement varies depending on the season and assimilated variable. The lead times until when the data assimilations influence the forecast quality are also different for different datasets and seasons; as an example, the impact from assimilating snow water equivalent persists for more than 20 weeks during the winter. We finally show that the assimilated datasets exert more control on the forecasting skill than the meteorological forcing, highlighting the importance of initial hydrological conditions for this snow-dominated river system.