The assimilation of different satellite and in situ products generally improves the hydrological model predictive skill. Most studies have focused on assimilating a single product at a time with the ensemble size subjectively chosen by the modeller. In this study, we used the European-scale Hydrological Predictions for the Environment hydrological model in the Umeälven catchment in northern Sweden with the stream discharge and local reservoir inflow as target variables to objectively choose an ensemble size that optimised model performance when the ensemble Kalman filter method is used. We further assessed the effect of assimilating different satellite products; namely, snow water equivalent, fractional snow cover, and actual and potential evapotranspiration, as well as in situ measurements of river discharge and local reservoir inflows. We finally investigated the combinations of those products that improved model predictions of the target variables and how the model performance varied through the year for those combinations. We found that an ensemble size of 50 was sufficient for all products except the reservoir inflow, which required 100 members and that in situ products outperform satellite products when assimilated. In particular, potential evapotranspiration alone or as combinations with other products did not generally improve predictions of our target variables. However, assimilating combinations of the snow products, discharge and local reservoir without evapotranspiration products improved the model performance.

Earth observations (EOs) are a valuable complement to in situ measurements in hydrology because they provide information in locations where direct measurements are unavailable or prohibitively expensive to make. Recent advances have enabled the assimilation of data sets of different physical variables into hydrological models to better estimate states and fluxes. Along with the meteorological forcings, the assimilated data exert controls on forecasts; it is therefore important to apportion the contributions from the forcing and assimilated data. Quality assessment is sometimes conducted in the pseudo‐reality, which neglects observations and model limitations. Here we introduce a diagnostic framework that accounts for observations to assess the sources of skill and infer the seasonal importance of assimilated and forcing data. We test the framework with a forcing data set from a downscaled Global Circulation Model and assimilate four EO and two in situ data sets to initialize the forecasts. We investigate the hydrological response and seasonal predictions over a Swedish snowmelt‐dominated catchment using the HYPE model over 2001–2015. The framework allows assessing the improvement in seasonal skill due to the different assimilated data and meteorological forcing. For the studied catchment, all EO and in situ data sets add information to the final forecast. The lead times during which data assimilation influences forecast skill also differ between data sets and seasons for example, assimilating snow water equivalent impacts the forecast for more than 20 weeks during winter. Lastly, assimilated data sets are generally more important to streamflow forecasting skill than meteorological forcing in the studied snow‐dominated catchment.

The assimilation of different satellite and in-situ products generally improves the hydrological model predictive skill. Most studies have focused on assimilating a single product at a time with the ensemble size subjectively chosen by the modeller. In this study, we use the European-scale Hydrological Predictions for the Environment hydrological model in the Umeälven catchment in northern Sweden with the stream discharge and local reservoir inflow as target variables to objectively choose an ensemble size that optimises model performance. We further assess the effect of assimilating different satellite products namely snow water equivalent, fractional snow cover, and actual and potential evapotranspiration; as well as in situ measurements of river discharge and local reservoir inflows. We finally investigate the combinations of those products that improve model predictions of the target variables and how the model performance varies through the year for those combinations. We found that an ensemble size of 50 was sufficient for all products except the reservoir inflow, which required 100 members and that in situ products outperform satellite products when assimilated. In particular, potential evapotranspiration alone or as combinations with other products did not generally improve predictions of our target variables. However, assimilating combinations of the snow products, discharge and local reservoir without ET products improves the model performance.

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 can further lead 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 forcings and the assimilated data. Here, we investigate the hydrological response and seasonal predictions over the snow-melt driven Umeälven catchment in northern Sweden. The HYPE hydrological model is driven by 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) technique. 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 measurements (discharge and reservoir inflow). We finally assess the impacts of the meteorological forcing data and the assimilated data on the quality of streamflow and reservoir inflow seasonal forecasting skill for the period 2001-2015. The results show that all assimilations generally improve the skill but the improvements vary 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 spring. We finally show that the assimilated datasets exert more control on the forecasting skill than the meteorological forcing data, highlighting the importance of initial hydrological conditions for this snow-dominated river system.

The potential skill of a seasonal hydrological forecasting system is well recognized to depend on the initial states, which among other things, can be improved through data assimilation and the meteorological forcing data used to drive the hydrological model. The operational skill of such a system will depend on the capacity of the model initialization and of the forcing data to capture key features in the catchment. For instance, studies at regional or continental scales have identified the baseflow index and the snowpack to be drivers of forecast skill in regions with high groundwater recharge and in high-latitude catchments, respectively. This study presents a seasonal forecasting system at a global scale using the global variant of the model HYPE and forced with the ECMWF SEAS5 meteorological data, with a focus on evaluating the forecast skill for key hydrological variables, e.g. discharge and snow water equivalent, and investigating the dependence of forecast skill on quantifiable physiographic and other catchment characteristics across the globe. Machine learning (ML) techniques are employed to identify the leading influencing factors linking forecast skills to catchment characteristics. Our goal is to demonstrate how forecast skill can be attributed to the properties of the catchments and how these links vary across the global hydrological gradient.