Numerical Weather Prediction and Data Assimilation
Core concepts
Understand the governing equations discretised in NWP models, grid spacing, and resolution. Study parameterisations for radiation, convection, boundary layer, and microphysics. Learn data assimilation methods: 3D-Var, 4D-Var, ensemble Kalman filter, and hybrid approaches. Know the role of initial conditions, model spin-up, and predictability limits. Understand ensemble forecasting and the sources of forecast error.
Traps
Questions often confuse analysis with forecast and model resolution with grid spacing. Be careful with the difference between deterministic and ensemble output. Do not assume that higher resolution always improves forecasts. Distinguish systematic bias from random error.
Worked example
A 4D-Var assimilation uses observations over a time window to adjust the initial state. If a radiosonde is assimilated at 00 UTC, the analysis at 00 UTC and the forecast from it both reflect that observation, but the forecast beyond 24 hours may diverge due to error growth.