Rational decision making by industries, agencies, and the public in anticipation of heavy precipitation, extreme temperature, snow storm, flood or other disruptive weather phenomena, requires information about the degree of certitude that the user can place in a weather forecast. It is vital, therefore, to advance the meteorologist's capability of quantifying forecast uncertainty to meet the society's rising expectations for reliable information.
The long-term goal of this research is to lay down a methodological foundation for the next generation of probabilistic forecasting systems. The specific objective is to develop and test (i) a set of statistical techniques for probabilistic forecasting of weather variates and (ii) a set of performance measures for verification of probabilistic forecasts. The basic technique, called Bayesian Processor of Output (BPO), processes output from a numerical weather prediction (NWP) model and optimally fuses it with climatic data in order to quantify uncertainty about a predictand. The BPO was developed and tested in Stage I of this project. The primary benchmark for evaluation of the BPO was the Model Output Statistics (MOS) technique used currently in operational forecasting; the BPO forecasts were better calibrated and more informative.
The objective of Stage II is to develop and test an extended technique. Called Bayesian Processor of Ensemble (BPE), it will process an ensemble of the NWP model output and optimally fuse it with climatic data in order (i) to obtain a probability distribution of each predictand, and (ii) to adjust the ensemble and thus to provide a well-calibrated assessment of the spatial, temporal, and inter-variate correlation within the approaching weather system. The BPE will harness recent advances in Bayesian statistical theory, multivariate distributions, estimation methods, and ensemble forecasting. It will be developed and tested in two versions, for (i) binary predictands (e.g., indicator of precipitation occurrence), and (ii) continuous predictands (e.g., precipitation amount conditional on precipitation occurrence, temperature, visibility, ceiling height, wind speed). The primary test will involve the production and verification of probabilistic quantitative precipitation forecasts (PQPFs) for up to 16 days ahead. The primary benchmark for evaluation of the BPE will be the frequentist technique used currently in operational forecasting.
The expected intellectual merit of this research will be an advancement of the Bayesian statistical theory and its applicability to complex forecasting problems, and an advance in the understanding of the stochastic properties of meteorological ensembles - a necessary step toward extracting more predictive information from the NWP models. The collaboration between the University of Virginia and the National Centers for Environmental Prediction, with provisions for exchanging data and expertise, testing the BPE, and transferring research results into operational use, will contribute to the aims of THORPEX - A Global Atmospheric Research Programme for the 21st Century under the aegis of the World Meteorological Organization.
The expected broader impacts of this new, state-of-the-art technique for ensemble forecasting will be improved forecasts of all major weather variates, and hence increased economic and societal benefits to the United States. In particular, reliable and informative PQPFs, produced by the BPE and suited to requirements of hydrologic models, will enable the production of probabilistic river stage forecasts, probabilistic flood forecasts, and flood warnings with explicitly stated detection probabilities.
Rational decision making by industries, agencies, and the public in anticipation of heavy precipitation, extreme temperature, snow storm, flood or other disruptive weather phenomenon, requires information about the degree of certitude that the user can place in a weather forecast. It is vital, therefore, to advance the meteorologist's capability of quantifying forecast uncertainty to meet society's rising expectations for reliable information. The long-term goal of this research is to lay down a methodological foundation for the next generation of probabilistic forecasting systems. The specific objective is to develop and test (i) a set of statistical techniques for probabilistic forecasting of weather variates and (ii) a set of performance measures for verification of probabilistic forecasts. The basic technique, called Bayesian Processor of Output (BPO), processes output from a numerical weather prediction (NWP) model and optimally fuses it with climatic data in order to quantify uncertainty about a predictand. The extended technique, called Bayesian Processor of Ensemble (BPE), processes an ensemble of the NWP model output and optimally fuses it with climatic data in order (i) to obtain a probability distribution of each predictand, and (ii) to adjust the ensemble and thus to provide a well-calibrated assessment of the spatial, temporal, and inter-variate correlation within the approaching weather system. The BPE harnesses recent advances in Bayesian statistical theory, multivariate distributions, estimation methods, and ensemble forecasting. It has two versions, for (i) binary predictands (e.g., indicator of precipitation occurrence), and (ii) continuous predictands (e.g., precipitation amount conditional on precipitation occurrence, temperature, visibility, ceiling height, wind speed). In a pilot study, probabilistic forecasts of surface temperature were produced for up to 15 days ahead. One property of these forecasts is this: As the lead time increases, the predictability of temperature by the NWP model decreases and eventually vanishes around the 15th day; the BPE automatically detects this trend and outputs a sequence of probability distributions which converge to the climatic probability distribution of temperature -- thereby characterizing smoothly the evolution of uncertainty between the current weather and seasonal climate (on day 15 and beyond). The expected intellectual merit of this research will be an advancement of the Bayesian statistical theory and its applicability to complex forecasting problems, and an advance in the understanding of the stochastic properties of meteorological ensembles -- a necessary step toward extracting more predictive information from the NWP models. The expected broader impacts of this new, state-of-the-art technique for ensemble forecasting could be improved forecasts of all major weather variates, and hence increased economic and societal benefits to the U.S. In particular, reliable and informative probabilistic precipitation amount forecasts, produced by the BPE and suited to requirements of hydrologic models, would enable production of probabilistic river stage forecasts, probabilistic flood forecasts, and flood warnings with explicitly stated detection probabilities -- the products that have been demanded by users since the Great Flood of 1993. For an extended summary of the nature and outcomes of this project, please see the article "From Theory to Application: Forward Thinking Forecasts", International Innovation, pages 10-12, June 2012, published and disseminated by Research Media Ltd., www.researchmedia.eu.
