The Initialization of Numerical Models by a Dynamic-Initialization Technique

Abstract A dynamic-initialization technique is tested with three models of fluid flow. In this technique data are assimilated through the inclusion of terms in the forecast equations which force the model atmosphere toward the observations. Results indicate that accurate, dynamically balanced mass and momentum fields can be obtained from unbalanced, inexact first guesses. In midlatitudes for horizontal scales less than 2000 km, observations of the wind are more important than observations of mass in producing a successful initialization for the systems presented here. There is also evidence that forcing the model winds toward the divergent wind component may slow the initialization process.