Nonlinear Processes in Geophysics (2001) [IN PRESS]
State dependent model error makes a significant contribution to error in weather forecasting; in general, however, it is less well understood than error due to inaccurate observations of the current atmospheric state. Reliable estimates of model error are required to compare different models and to decide how resources would best be allocated to model development. This paper presents a technique for measuring state dependent model error and estimating shadow times (the time for which a model can stay close to the true state). Initial results comparing the T42 and T63 models with the TL159 model are presented, as well as an estimate of shadowing times for the operational model.
Last updated: 14 Feb 2001