Uncertainty-based adaptive AXBT sampling with SPOTS

Naval operations continue to evolve toward Littoral Warfare as military action shifts to regional conflicts. To accomplish this evolution, new navigation, sensor, and data-analysis capabilities are needed to support operations in the highly variable and complicated near-shore waters of the littoral environment. Antisubmarine Warfare (ASW) is often conducted in shallow-water areas, where subsurface enemies pose a constant threat, and where knowledge of ocean thermal data is critical, but lacking. Planning operations in these harsh-environment areas is difficult because accurate predictions of sensor performance depend on detailed knowledge of the local conditions. Tactical mission planning is thus seldom optimal or efficient, often resulting in coverage gaps and increased risk. The Naval Air Systems Command has recently been exploring new environmental sonobuoy concepts to better characterize the littoral environment. Most designs contain a thermistor string, to measure ocean temperatures, and other environmental sensors. This type of sonobuoy, with a complex set of sensors, would be more expensive than a traditional AXBT but it could provide a more thorough littoral environment assessment. The increased cost implies the need for an Environmental Decision Aid to determine the minimum number and best locations for sensors to meet performance objectives. The work reported here concerns the development and evaluation of Sensor Placement for Optimal Temperature Sampling (SPOTS), which addresses these sampling requirements. The SPOTS process follows the following steps: 1) divide the area of interest into cells with varying volumes of water; 2) estimate the volume-weighted uncertainty of temperatures and the local anisotropic temperature covariance in each cell, based on current optimal interpolation nowcasts; 3) calculate the overall volume-weighted reduction in temperature uncertainty that would result from various sampling patterns; and 4) choose the pattern with the l- owest uncertainty. This uncertainty-based approach leads to sampling patterns that produce the highest accuracy temperature characteriz ations. SPOTS employs three innovations: 1) analysis of remotely-sensed data, confirmed with a numerical model, when needed; 2) adapting the covariance ellipse axes automatically to the predominant coastline features; and 3) using depth-weighted and volume-weighted uncertainty where the depth-dependent uncertainty and volume of water in a cell is considered in the optimization process. SPOTS uses an optimal interpolation technique that weights all input data by their uncertainties and provides uncertainty estimates for the output. That is a significant advantage over other interpolation schemes. Horizontal/vertical smoothing routines remove large discontinuities and produce the final "nowcast." As a result of these innovations, SPOTS sampling recommendations emphasize the upper water column, where most of the dynamic effects occur, and where acoustic variability is greatest. Data from several water-sampling flights in the Sea of Japan off the east coast of Korea were used to develop SPOTS. Approximately 44 AXBTs were dropped on a 15-min grid during each flight. Ten combinations of these AXBT measurements, ranging from three to all of the measurements, were assimilated into the Modular Ocean Data Assimilation System (MODAS). The climatology alone and climatology with assimilated satellite sea surface temperatures brought the number of cases to twelve. These were analyzed to determine the relationship between nowcast accuracy and the number (and placement) of assimilated in-situ measurements. The sub-sampled nowcast estimates were compared with the measured temperatures and reported as RMS temperature errors. The results show that: 1) a small number of well-placed measurements outperforms a larger number of gridded measurements; 2) a small number of poorly-placed measurements can significantly degrade a nowcast; and 3) approximately