DIGITAL IMAGE ANALYSIS AS A TOOL TO EXPLORE OSL OVERDISPERSION AND SEDIMENT IN THIN SECTION

Optically stimulated luminescence dating (OSL) estimates when sedimentary grains were last exposed to a bleaching event. In OSL, the variance, or overdispersion, is a measure of the distribution of the equivalent dose (De) values. Explanations for overdispersion include: partial or incomplete grain bleaching, grain migration resulting in mixed grain populations, and homogeneity in beta dosimetry resulting in variable De within a single grain population. Here we analyze OSL and related thin section images of soil developed on colluvium and alluvium from Missed Grouse Gulch in central PA to help explain De scatter and grain context. Preliminary results of RGB and greyscale analysis from thin sections show significant correlation between OSL overdispersion and observed red channel values (Mode, R2 = 0.81; StdDev, R2 = 0.51; Mean, R2 = 0.62). As overdispersion increases in the sample, Mode and Mean decrease while StdDev increases in the red channel. These values may act as visible spectrum proxies for Fe(III) concentration, associated with clay content which could offer clues about the origin of De scatter related to quartz grain dose rates. Results from greyscale image analysis suggest a relationship between the modality of the histograms and the degree of grain size heterogeneity within each thin section. Well sorted sediment matrix is associated with a single peak in the histogram while mixed grain assemblages are displayed as multi-modal distributions. This data can be used to delineate variability of grain size distribution, identify distinct grain populations, and understand the context of these populations to address questions concerning background dose rate and absorbed dose variability. When combined, these data suggest digital image analysis shows strong potential for characterization and analysis of in-situ sediment and could be useful in in explaining beta dosimetry factors. Further inquiries shall focus on image preprocessing and other visible spectrum measures to increase the reliability of reported values.