ATSC 507 · Numerical Weather Prediction

Learning Goals

By the end of this course, students will be able to:

• justify the different assumptions that are often made to the governing dynamical equations of the atmosphere*

• convert dynamic and thermodynamic equations into finite-difference and spectral forms*

• anticipate errors associated with various finite-difference forms of eqs. of motion, and calculate their effects on forecast skill*

• explain why physical parameterizations are needed, and critically evaluate their value and limitations

• explain the importance of data assimilation in model initialization

• explain the roles of ensembles in reducing random forecast errors

• calculate different types of verification statistics for both deterministic and probabilistic forecasts

• anticipate the different types of numerical errors, and discuss their effect on predictability

• put into context the steps in an operational numerical forecast process

• explain the roles of statistical postprocessing in reducing systematic forecast errors

• recognize common operational models and their components, and be able to access their documentation

• run the most widely used NWP model in the world (currently, the WRF model)

• debate both sides of human vs. machine arguments for producing the best weather forecasts

Instructors

Roland Stull

Textbook

Thomas T. Warner, 2011: Numerical Weather and Climate Prediction.  Cambridge.  526 pp.  

    ISBN 978-0-521-51389-0.

Lecture Topics

Week 1:  Scientific basis for numerical weather prediction (NWP).
Weeks 2-3:  Numerical solutions to the equations.
Weeks 4-5:  Errors and effects of numerical approximations.
Week 6:  Spectral methods.
Week 7:  Overview of physical-process parameterizations.
Week 8:  Model initialization and data assimilation.
Week 9:  Ensemble methods.
Week 10:  Verification.
Week 11:  The numerical forecast process and statistical postprocessing.
Week 12:  Operational NWP models. 
Week 13 (not a full week):   Project discussions.

Labs

• Creating a simple NWP model from scratch.

• Learning how to run the WRF model.