The Impact of Data Boundaries upon a Successive Corrections Objective Analysis of Limited-Area Datasets

Abstract

Successive corrections objective analysis techniques frequently are used to array data from limited area without consideration of how the absence of data beyond the boundaries of the network impacts the analysis in the interior of the grid. The problem of data boundaries is studied theoretically by extending the response theory for the Barnes objective analysis method to include boundary effects. The results from the theoretical studies are verified with objective analyses of analytical data. Several important points regarding the objective analysis of limited-area datasets are revealed through this study. Data boundaries impact the objective analysis by reducing the amplitudes of long waves and shifting the phones of short waves. Further, in comparison with the infinite plane response, it is found that truncation or the influence area by limited-area datasets and/or the phase shift of the original wave during the first pass amplified some of the resolvable short waves upon successive corrections to that first pass analysis. The distance that boundary effects intrude into the interior of the grid is inversely related to the weight function shape parameter. Attempts to reduce boundary impacts by producing a smooth analysis actually draw boundary effects father into the interior of the network. When analytical test were performed with realistic values for the weight function shape parameters, such as the GEMPAK default criteria, it was found that boundary effects intruded into the interior of the analysis domain a distance equal to the average separation between observations. This does not pose a problem for the analysis of large datasets bemuse sevens rows and columns of the grid can be discarded after the analysis. However, this option way not be possible for the analysis of limited-area datasets because there may not be enough observations. The results show that, in the analysis of limited-area datasets, the analyst should be prepared to accept that most (probably all) analyses will suffer from the impacts of the boundaries of the data field.