Problems with the models

• Models represent a "simplified" atmosphere - not every real process in atmosphere can be resolved in the models
• Many are not global in coverage
• Initial atmospheric state is not well-known - want a dense, global network of observations
• Have many data-parse regions, particularly over the oceans
• The data may also have errors in it
• The model equations compute quantities at grid points. Currently, grid spacing ranges from 30-50 km apart. Any phenomena smaller in size that grid spacing will not be resolved in models (e.g., thunderstorm) -->>
• small-scale terrain features will not be handled properly
• models can not resolve boundary layer very well.
• The atmosphere is fundamentally chaotic - small differences in the model initial conditions can produce radically different results later in time
• Each model can produce different predictions.... which do you believe????

QUESTION FOR THOUGHT:

Explain how the phrase "sensitive dependence on initial conditions" relates to the final outcome of a computer-based weather forecast.