DATA SMOOTHING TECHNIQUES APPLIED TO PROTON MICROPROBE SCANS OF TELEOST HARD PARTS

We use a proton microprobe to examine the distribution of elements in otoliths and scales of teleost (bony) fish. The elements of principal interest are calcium and strontium in otoliths and calcium and fluorine in scales. Changes in the distribution of these elements across hard structures may allow inferences about the life histories of fish.

Otoliths and scales of interest are up to a centimeter in linear dimension and to reveal the structures of interest up to 200 sampling points are required in each dimension. The time needed to accumulate high X-ray counts at each sampling point can be large, particularly for strontium. To reduce microprobe usage we use data smoothing techniques to reveal changing patterns with modest X-ray count accumulations at individual data points.

In this paper we review performance for revealing pattern at modest levels of X-ray count accumulations of a selection of digital filters (moving average smoothers), running median filters, robust locally weighted regression filters and adaptive spline filters.