Abstract
A series of observing system simulation experiments (OSSEs) are performed for the tropical Indian Ocean (±15° from the equator) using a simple analysis system. The analysis system projects an array of observations onto the dominant empirical orthogonal functions (EOFs) derived from an intermediate-resolution (2° × 0.5°) ocean circulation model. This system produces maps of the depth of the 20°C isotherm (D20), representing interannual variability, and the high-pass-filtered mixed layer depth (MLD), representing intraseasonal variability. The OSSEs are designed to assess the suitability of the proposed Indian Ocean surface mooring array for resolving intraseasonal to interannual variability. While the proposed array does a reasonable job of resolving the interannual time scales, it may not adequately resolve the intraseasonal time scales. A procedure is developed to rank the importance of observation locations by determining the observation array that best projects onto the EOFs used in the analysis system. OSSEs using an optimal array clearly outperform the OSSEs using the proposed array. The configuration of the optimal array is sensitive to the number of EOFs considered. The optimal array is also different for D20 and MLD, and depends on whether fixed observations are included that represent an idealized Argo array. Therefore, a relative frequency map of observation locations identified in 24 different OSSEs is compiled and a single, albeit less optimal, array that is referred to as a consolidated array is objectively determined. The consolidated array reflects the general features of the individual optimal arrays derived from all OSSEs. It is found that, in general, observations south of 8°S and off of the Indonesian coast are most important for resolving the interannual variability, while observations a few degrees south of the equator, and west of 75°E, and a few degrees north of the equator, and east of 75°E, are important for resolving the intraseasonal variability. In a series of OSSEs, the consolidated array is shown to outperform the proposed array for all configurations of the analysis system for both D20 and MLD.
Interannual variability in the tropical Indian Ocean
