tübinger geowissenschaftliche arbeiten (tga) - TOBIAS-lib ...
tübinger geowissenschaftliche arbeiten (tga) - TOBIAS-lib ...
tübinger geowissenschaftliche arbeiten (tga) - TOBIAS-lib ...
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
18<br />
number of simulation runs and the modelled output parameters are compared with the<br />
input parameters. The correlation of input and modelled output data is statistically<br />
evaluated with the Kolmogorov-Smirnov test. The program provides a graphic display<br />
of the modelling results showing the best fit thermal evolution and statistically<br />
acceptable limits. The best fit thermal history may neither be geologically meaningful<br />
nor unique. The annealing model of Ketcham et al. (2000) was chosen with the<br />
measured chlorine data as input for the chemical composition. None of the modelled<br />
samples defined selected kinetic populations as all apatites were fluorapatites. Each<br />
sample was modelled in 50.000 simulation runs. The modelled time-temperature paths<br />
are shown in Appendix C, Fig. C3.<br />
2.13 Decomposition of zircon fission track grain age distributions<br />
Single fission track grain age spectra of sedimentary samples that are not thermally<br />
overprinted or reset, mirror the provenance ages of their source area. The<br />
decomposition of the mixed fission track grain age distribution of a sample yields age<br />
populations. These populations reflect low-temperature cooling phases during the<br />
geological history of the hinterland from which the sediment was eroded. The<br />
decomposition of the age spectra can be approximated statistically. The computer<br />
program BINOMFIT (Brandon 2002) was used, which is based on the binomial peakfitting<br />
method of Galbraith and Green (1990). The binomial peak-fitting method is<br />
based on the maximum likelihood method which determines the best-fit solution by<br />
directly comparing the distribution of the grain data to a predicted mixed binomial<br />
distribution (Brandon 2002). The entire grain age distribution is decomposed into a<br />
finite set of component binomial distributions, each of which is defined by a unique<br />
mean age, a relative standard deviation, and an estimated number of grains in the<br />
component distribution. The initial guesses for peak ages are generated from the<br />
probability density plot for the fission track grain age distribution. The quality of the fit<br />
for each solution is scored using the � 2 test (Brandon 2002). To test the number of<br />
underlying distributions and to define the minimum number of binomials needed, an Fratio<br />
test (Bevington 1969) is introduced to judge whether the reduction of � 2 , due to<br />
the addition of a new peak, is only by chance. When F is large the improvement in fit,<br />
associated with the additional peak is considered significant. The F distribution is used<br />
to assign a probability P(F), which is the probability that random variation alone could<br />
produce the observed F statistics. Brandon (2002) considers P(F) < ~5% to indicate<br />
that the improvement in fit is significant. Thus, the optimal number of significant peaks<br />
can be found by adding peaks until one gets a value of P(F) > ~5%. Results of fission<br />
track data and calculated age clusters using the computer program BINOMFIT are<br />
illustrated in Appendix C, Tab. C7 and the decomposed fission track age spectra in<br />
Appendix C, Fig. C5.