چکیده:
Uncertainty associated with mixing models is often substantial, but has not yet been fully incorporated in models. The
objective of this study is to develop and apply a Bayesian-mixing model that estimates probability distributions of source
contributions to a mixture associated with multiple sources for assessing the uncertainty estimation in sediment
fingerprinting in Zidasht catchment, Iran. In view of this, 31 geochemical tracers were measured in 35 different sampling
sites of three sediment sources (rangelands, crop fields and stream banks) and 14 sediment samples from stream bed
deposition. Based upon statistical analysis, the best 20 composition subsets of tracers (e. g. 2, 3, 4 …21) were then
selected. Sediment source fingerprinting was used to explore the uncertainty in the contributions of sediment from the
three sources. The results showed that the main source of uncertainty was the number of tracers included in the model and
the higher number of tracer in the model the lower deviation in uncertainty. However, differences between the ranges of
uncertainty values from subset 5 to subset 21 of tracers are not statistically significant. In the study area, mean of relative
contributions associated with uncertainty from rangeland, crop field and stream bank sources (mean of subset 5 to 21)
were 0.526, 0.059, and 0.411 respectively. These results can be useful as a scientific basis of sediment management and
selecting the soil erosion control methods for decision makers of natural resources.
خلاصه ماشینی:
The objective of this study is to develop and apply a Bayesian-mixing model that estimates probability distributions of source contributions to a mixture associated with multiple sources for assessing the uncertainty estimation in sediment fingerprinting in Zidasht catchment, Iran.
Recent sediment source studies using mixing models have undertaken uncertainty analysis due to the spatial variability of source tracer properties to determine confidence limits of model estimates based on Monte-Carlo estimation approach (Krause et al.
2. Sampling and data collection Potential sediment sources were identified by observing the main land use types and soil erosion features within the study catchment and were dominated by three main groups; the rangelands, crop fields and stream banks.
In the first stage, the Kruskall– Wallis non-parametric test is used to identify those fingerprint properties (from 31 tracers) which were able to discriminate between the three calculating the probability distributions for the proportional contribution (fi) of each source i to the mixture in three stages: 1) determination of the prior probability distribution for model parameters, 2) construction of a likelihood function for the statistical model, and 3) derivation of the posterior probability distribution for the parameters using the Bayes rule to adjust the prior distribution based on the observed data (Bolstad, 2007).
Bayesian statistical methods quantify uncertainty by L f , fi i 1 1 fi i i q i i B , The model is based on the following constraints: 0 f i 1 ; the percentage source contributions must lie between 0 and 1; and n use a subset of different fingerprints to discriminate the three potential sediment sources in the study catchment.