Skill and value of global seasonal streamflow forecasts

In our changing world, humans experience increasingly the negative consequences of floods and droughts. Seasonal forecasts with lead times of several months, and covering larger areas are necessary to increase global preparedness. This thesis explores the potential of global hydrological models in operational seasonal forecasting applications, assesses the skill and value of global seasonal streamflow forecasts and investigates possible ways to improve the current skill and value. To assess the prospect of applying a global hydrological model for seasonal forecasting, global terrestrial hydrology is simulated with the model PCR-GLOBWB. The model is forced with a meteorological dataset based on historical observations and model skill is assessed based on monthly discharges for twenty large rivers across the world. PCR-GLOBWB cannot forecast the historical hydrographs adequately for all basins but higher skills can be attained in forecasting the occurrence of monthly anomalies. The prospects for seasonal forecasting with PCR-GLOBWB or other comparable models are assessed to be positive. The simulated hydrological response depends on both the initial hydrological conditions and the meteorological forcing. Uncertainty in both inputs is evaluated by comparing ESP/revESP forecast ensembles with retrospective model simulations driven by meteorological observations. The results are analysed in the context of prevailing hydroclimatic conditions for larger rivers across the globe. The influence of the initial conditions and meteorological forcing on forecasting skill is found to vary considerably according to location, season and lead time. For arctic and snow fed rivers, forecasts of high flows may benefit from assimilation of snow and ice data. In some snow fed basins where the onset of melting is highly sensitive to temperature changes, forecast skill depends on better climate prediction. Groundwater and surface water states also strongly influence the skill in very large rivers. In monsoonal basins, the variability of the monsoon dominates forecasting skill, except for those where snow and ice contribute to streamflow. When the total skill is assessed in actual forecasting mode, actual seasonal meteorological forecasts are used as input into PCR-GLOBWB. The model is forced with S3 seasonal meteorological forecast ensembles from the ECMWF as well as with probabilistic meteorological ensembles obtained following the ESP procedure. Ensemble forecasts of monthly discharges for twenty large rivers of the world are produced with lead times of up to six months. Analysis of the results suggest that forecasting skill decreases with increasing lead time and varies considerably by region and season. The performance of ECMWF S3 forecasts is close to that of the ESP forecasts. In the current setup, the forecasting skill is limited and needs to be improved before forecasts can be adopted for water management applications. However, even with little added skill, forecasts may still be useful for end-users, allowing them to decide for themselves if they should take the risk of using the forecast information. The success of a hydrological forecasting system will ultimately be determined not only by its skill but also by its value to inform decision-making for water management. The interaction between skill and value is explored and possible ways to improve the value of seasonal hydrological forecasts on a global scale for water related applications are discussed with an emphasis on flood and drought mitigation. The ability of seasonal streamflow forecasting systems to predict the right category of an event months ahead is potentially valuable for many water-related applications. Seasonal hydrological forecasting on a global scale could be especially valuable for transboundary river basins as well as for developing regions of the world, where no effective local hydrological forecasting systems exist. The realization of the potential added value depends largely on the collaboration between forecast producers and users.