Recent Progress in Estimating Geoposition Using Daylight

Abstract

In this paper we report on the results from a new experiment where we placed tags at three distinct depths on a fixed subsurface mooring. The accuracy of the daily geoposition estimates is substantially improved over our earlier work, with average positional errors of 30 km in longitude and 44 km in latitude. We attribute the result to improvements in both the tags and our software. From the resulting data, we were able to estimate a relationship between light and depth. We generated a simulated variable-depth light record, corrected the light values to a constant depth using the estimated relationship, and then applied our estimation algorithm to the depth-corrected values. This yielded an average absolute positional error of 42 km in longitude and 68 km in latitude, a significantly better result than previously reported. Finally, we provide a theoretical assessment of the expected differences that can be obtained by shielding the light sensor with a blue filter. Although this helps to linearize the relationship between light intensity and depth, the combination of the broad-band spectral sensitivity of the light sensor and the surface light spectrum means that a simple linear response should not be expected. Attempts to correct the light record for variations caused by changes in depth of a tagged animal are probably not substantially improved by filtering the light sensor. More attention needs to be given to the combined effect of the spectral sensitivity of the light sensor and the light spectrum present at twilight.