HS4.2Hydrological forecasting: challenges in uncertainty estimation, data assimilation, post-processing and decision-making
|Convener: Robert Moore | Co-Conveners: Albrecht Weerts , Henrik Madsen|
This session will address methods for estimating predictive uncertainty, its reduction through data assimilation, improved model formulation and post-processing, and its use to support operational decision-making in flood warning and water management. Including uncertainty estimation in operational forecasting systems is becoming a more common practice. However, a significant research challenge and central interest of this session is the need to improve the specification of methods to estimate and reduce predictive uncertainty as well as its use in hydrological hazard mitigation and resource management. Providing uncertainty estimates for integrated catchment models involving forecasting models, either as a cascade or as alternative models, can prove particularly challenging and an issue of interest to the session.
Data assimilation or post-processing in real-time can provide important ways of reducing the uncertainty of hydrological forecasts. Methods that help update forecasts in real-time to reduce bias and increase accuracy, and case study demonstrations of their use, are of interest to this session. The models involved with the methods for predictive uncertainty, data assimilation, post-processing and decision-making may include catchment models, runoff routing models, groundwater models, coupled meteorological-hydrological models as well as combinations of these. Demonstrations of the reduction in predictive uncertainty through improved representation of model process and error are of interest to the session.
Whilst the value of a measure of forecast reliability is recognised by practitioners, realising the full benefit of predictive uncertainty in decision-making schemes is an issue that is actively being investigated in conjunction with researchers. Flood warning can be greatly improved in terms of better probability of detection and reduced false alarm rates. Of interest to the session are: (1) new performance measures relevant to probabilistic forecasting; (2) novel ways of communicating uncertainty through time-series and map displays in real-time, (3) ways of transforming predictive uncertainty into operational rules or decisions in flood warning and flood management. To raise awareness of the beneficial effects of operationally using predictive uncertainty, the session invites scientists and decision-makers to demonstrate, possibly through case studies, how it can be applied in practice to support decision-making.
Contributions are expected to address the following issues:
(i) Uncertainty propagation in meteorological-hydrological forecasting.
(ii) Measures of performance of probability-based hydrological forecasts
(iii) Assessment and use of predictive uncertainty for decision-making under uncertainty.
(iv) Improved ways of communicating forecast uncertainty
(v) Methods that allow use of ground-based hydrological data and remotely-sensed data in real-time forecasting, and account for their uncertainty.
(vi) Methods for preparing meteorological predictions as input to real-time hydrological probability forecasts.
(vii) Case studies of the above.