Refractivity Data Fusion

Description

Improved analyses of the refractivity field have been tested using a data fusion technique. The fusion algorithm modifies a model analysis profile of refractivity (base-point) at the time and location of an observed profile so as to more closely match the sounding by comparing a large number of randomly generated shifts in elevation and M units. The set of elevation shifts and M shifts which most closely matches the observed sounding is then distributed over a user-defined radius of influence R. A linear weighting is then applied such that the influence of the shifting decreases to zero at R. When the distributed and weighted field of shifts is applied to the model M field, a new three-dimensional model M field is produced. The data fusion algorithm has been validated using data from the VOCAR experiment using both single and dual base-points. Statistics documenting the performance of the data fusion approach were compiled

Results

Figure 1 shows a schematic diagram of the profile shifting procedure used in the data fusion algorithm. In the first set of tests, soundings and model profiles from the location of the Naval Postgraduate School (NPS) research ship were used to generate the field of elevation and M shifts (see Fig. 2 for locations). The field of shifts was then applied to the model M field and the impact was assessed at San Nicholas Island, Point Vicente, and San Clemente Island (Fig. 3). A statistical analysis of results over the entire VOCAR period is shown in Table 1. Note that the method yields a 26% improvement in duct base height estimates. The improvement in delta M is somewhat more modest. In the second set of test, soundings and model profiles from both the NPS ship and San Clemente Island were used to generate the field of shifts, assessed at San Nicholas Island and Pt. Vicente. The statistics were not significantly different from the single sounding results. We also assessed the sensitivity of the method to changes in the radius of influence and the vertical resolution of the computational grid to which the sounding and model data are interpolated.

Figure 1. Schematic of the shift imposed upon an original COAMPS refractivity profile by the data fusion technique. Red is the original COAMPS profile decreasing from 20 to 10 B-units over 300 m depth, blue is sounding decreasing from 0 to 10 B-units over a 200 m depth, and pink is modified COAMPS which is close to the observed sounding. The two green asterisks indicate a point on the original profile and the corresponding point on the shifted profile obtained by using a refractivity shift of DM=20 B-units and a height shift of Dh=65 m.

Figure 2. Map of the VOCAR domain in the Southern CA Bights. The NPS ship located at (33.5°N, 118.8°W) is the base point (B) from which the shifts are determined and three surrounding points (red circles) are locations for which the modified profiles are compared to soundings.

Figure 3. Refractivity profiles at three locations using shifts from the NPS ship. The black line is the sounding showing a higher and weaker duct than is predicted by COAMPS (blue line), and the red line is the modified COAMPS profile which is shifted to better match the observed sounding.

Table 1: Statistics from data fusion VOCAR test bed showing the improvement in

DBH and DS with the shifted COAMPS refractivity profiles.