SIMULATION OF THE GAMMA-RADIATION FIELD IN LUMPY ENVIRONMENTS

Many of the environments used for ESR and luminescence dating are inho mogeneous over the distances up to 0.3 m which contribute to gamma dos es, and the consequent variability in gamma dose usually cannot be asc ertained on a sample by sample basis. In this paper we examine the cha racteristics of such ''lumpy'' environments by simulation. Random envi ronments were generated in which spherical source-free calcite inclusi ons were embedded in a source-rich sediment matrix. Using dose point k ernels based on build-up factors, the gamma dose rate at a large numbe r of random points in the sediment was then calculated. Various values for the sphere radius and volume fraction of the inclusions were test ed. The gamma dose rate frequency distributions exhibit negative skewn ess, with the low outliers arising from points close to one or more in clusions. The variance of the distribution and the significance of the negative tail increase with the inclusion radius. Lognormal and gamma distributions were found to fit these histograms quite well. By sampl ing from the dose rate distribution, it was possible to repeatedly gen erate two hypothetical samples representing the positions of contempor aneous datable objects (and thus their accumulated doses) and of dose meters. Various statistical estimators were trialled to establish the optimum approach to extracting a site age from given sets of accumulat ed dose and gamma dose rate values. For ease of analysis, it was assum ed that the gamma dose is the total dose, but the techniques can be ge neralised to realistic dose environments also. Most estimators produce d values accurate to a few percent for typical environment parameters, and the one with the least variance was based on a two-stage maximum likelihood approach to age estimation. In the absence of such a sophis ticated approach, culling of low valued outliers for both accumulated dose and dose rate leads to improved site age estimation. (C) 1997 Els evier Science Ltd.