Numerical approximations can create chaos-related errors in blocking forecasts

The chaotic structure of two-dimensional atmospheric flow is illustrated. It is shown that certain errors in numerical approximations can prevent the correct prediction of chaotic processes. This is the case when the numerical approximations do not sufficiently allow air parcels to deviate from each other. The error mechanism is described with a case study and is proposed as one explanation for the errors observed when forecasting the development of blocking highs. It can explain why the errors in blocking highs are similarly found in different models from different centers, why they appear in medium-range forecasts but not in short-range forecasts, and why the error decreases only slowly with increasing resolution.