The National Center for Atmospheric Research is sponsored by the National Science Foundation.
Evaluation of lateral boundary errors in a limited-domain model
Version of Record online: 18 MAR 2010
1982 Blackwell Munksgaard
Volume 34, Issue 5, pages 409–428, October 1982
How to Cite
BAUMHEFNER, D. P. and PERKEY, D. J. (1982), Evaluation of lateral boundary errors in a limited-domain model. Tellus, 34: 409–428. doi: 10.1111/j.2153-3490.1982.tb01831.x
- Issue online: 18 MAR 2010
- Version of Record online: 18 MAR 2010
- Received July 14, 1981; in final form January 14, 1982
Computational errors that arise from the imposition of artificial lateral boundaries in numerical forecast models are evaluated, using the Perkey–Kreitzberg method of variable tendency averaging near the boundary in the NCAR second generation Limited Area Model. Boundary error is diagnosed for two sets of initial conditions by using an unbounded forecast as background truth. Several components of boundary error are identified and evaluated. These include the accuracy of the boundary specification, the formulation of the blending, the increased boundary diffusion, and the physical location of the boundary. The formulation is tested by comparing boundary errors generated by the Williamson–Browning method for the same initial conditions. Finally, the boundary errors are compared to the total forecast error to evaluate its significance.
Results indicate the Perkey–Kreitzberg method is stable for a wide variety of resolutions and situations. Significant errors propagate inward from the boundaries at speeds of 20–30° longitude per day. Inaccuracies in the data specified at the boundary can account for a major part of these errors, while increased diffusion near the boundary plays a lesser role in error generation. The two formulations produce similar error structures, but with less generation of spurious transients in the Williamson–Browning method. However this method proved to be unstable at high horizontal resolutions. In general, the boundary error is much less than the total forecast error.