High-resolution weather models often use a technique called “nesting” to focus on specific regions. In this approach, a smaller inner model (called the nested domain) is placed inside a larger outer model (called the parent domain). There are two types of nesting: one-way nesting, where information only flows from the outer to the inner domain, and two-way nesting, where information flows both ways. One-way nesting can be done in an “offline” way, where the parent model is run in advance and boundary data are given to the inner model at regular time intervals (e.g., every hour). This is computationally efficient, but the infrequent boundary updates may introduce errors.

Small errors at the boundaries can grow over time and affect the accuracy of forecasts. Most past studies have focused on how uncertainties in the initial conditions limit predictability, but fewer have examined how boundary data quality and timing affect it.

In this study, we run an idealized weather model experiment to examine how the frequency of boundary updates affects the growth of errors in the inner domain. We use spectral analysis to understand how these errors evolve across different scales and how they affect predictability.