Questions: Numerical Weather Prediction and Ensemble Forecasting

Question 1

What is the purpose of data assimilation in numerical weather prediction?

• A. To average the outputs of different forecast models
• B. To combine new observations with a short-range model forecast to produce the best estimate of the current state, which then starts the next forecast
• C. To remove chaotic behavior from the equations
• D. To convert model output into television graphics

Answer: (B)

Observations are scattered and imperfect, and the model needs a value at every grid point. Data assimilation blends observations with a recent short-range forecast, weighting each by its expected error, to build the starting state called the analysis.

Question 2

Why do ensemble forecasts start each run from a slightly different initial state?

Model answer: The true initial state is never known exactly, and in a chaotic atmosphere small differences grow over time. Starting many runs from slightly different but plausible states samples this uncertainty. If the runs stay close together the forecast is more reliable; if they spread apart, confidence is lower.

The spread of the ensemble is a forecast of the forecast's own uncertainty. This is the practical answer to the predictability limit that Edward Lorenz identified in the 1960s.

Question 3

A model with a grid spacing of 10 km can directly resolve individual thunderstorm updrafts that are about 1 km wide.

• T. True
• F. False

Answer: False

A feature must span several grid points to be represented well. A 1 km updraft is far smaller than a 10 km grid box, so its effects must be estimated with a convection parameterization.

Question 4

Lewis Fry Richardson's hand-calculated forecast, published in 1922, predicted a surface pressure change of about 145 hPa in six hours. What was the main lesson for later forecasters?

• A. The equations of the atmosphere are wrong
• B. Weather cannot be forecast by calculation
• C. The method was sound, but unbalanced initial data and numerical problems can wreck a forecast, so the starting state and numerical methods need careful treatment
• D. Only computers can solve the equations of motion

Answer: (C)

Richardson's approach is the basis of modern NWP. His huge error came mainly from imbalances in the initial data, which produced unrealistic fast waves. Later work on balanced initial states, stable numerical methods, and fast computers made his idea practical.