F. Anthony Eckel
ATM S 444
Covers the fundamental of chaos theory to help compare and contrast traditional, deterministic forecasting versus ensemble forecasting. Explores the various components of an ensemble prediction system. Introduces decision science to show how to apply probabilistic weather information in optimal decision making. Prerequisite: ATM S 370; either STAT/MATH 390 or Q SCI 381; AMATH 301. Offered: Sp.
People seek knowledge of the coming weather conditions in order to make good decisions. Today’s predominant weather support method of providing quality deterministic (single-value) forecasts is valuable, but is fundamentally limited by omission of forecast uncertainty information. Knowledge of forecast uncertainty (a measure of predictability) enables weighing the weather risks and potential consequences, and, when used properly, optimized decision making. An ensemble prediction system (EPS) can produce a comprehensive view of the forecast distribution (a range of possible future atmospheric states) and convey weather risk information. This capability is driving an evolution in weather support from a deterministic forecasting focus, based on single-solution numerical weather prediction (NWP) output, to a focus on multi-solution, or probabilistic forecasting. However, while operational EPS output has been available since 1993, it is not well utilized within the weather support process due to many factors including limited understanding in both the scientific and user communities. The goal of this course is to develop atmospheric scientists knowledgeable in EPS design and application who are then able to promote incorporation of the technology into the weather support process. The first half of the course covers fundamentals of chaos theory (as the scientific basis for ensemble forecasting) and explores the various components of an EPS. The second half introduces decision science, compares decision support using deterministic vs. probabilistic forecasts, and investigates aspects of communicating forecast uncertainty for optimal decision making.
Student learning goals
Be able to describe the behavior of a chaotic, dynamical system and what limits its predictability
•Know the function and interdependencies of the major EPS components w.r.t. simulating flow-dependent error growth o Accounting for initial condition uncertainty o Accounting for model uncertainty o Calibration of raw ensemble output o Communicating the forecast for optimal decision making o Verification and analysis of stochastic forecasts
Be able to critically evaluate the utility of an EPS
Understand the role of forecasts in creating value by providing decision support
Be able to analyze cost-loss decisions and understand more complex, dynamic decisions
Understand application/creation of stochastic predictions in the human forecast process
General method of instruction
Class assignments and grading