Application of the affinity propagation clustering algorithm based on grain-size distribution curve to discrimination of sedimentary environment——A case study in Baiyangdian area
-
Graphical Abstract
-
Abstract
In the paper, 22 sedimentary sections are selected from the Baiyangdian region as research carriers. The affinity propagation (AP) clustering algorithm is adopted to cluster the similar grain size distribution of sediment into clusters. The results are then compared with the characteristics of grain size distribution which are known in typical sedimentary environments. The grain size distribution patterns in different sedimentary environments in Baiyangdian area are concluded and the feasibility of application of the affinity propagation clustering algorithm based on sediment grain characteristics to environmental interpretation discussed. The results suggests that the AP clustering algorithm can gather the sediments formed under the same or similar dynamical conditions into groups, and dig out the sedimentary dynamical information contained in the grain size data; The distribution curves of grain size for all samples are subdivided into 11 clusters. Considering the change in dynamic conditions in the same sedimentary environment, the 11 cluster curves are further categorized into 4 sets of curves, which could be compared with the grain size curves from known environment. 4 sedimentary facies i.e. the lacustrine-swamp facies, lacustrine facies, fluvial facies and alluvial facies are recognized. Among them, in terms of sedimentary dynamics, the lacustrine-swamp facies are similar with the central lake facies, while the lake shore facies similar with alluvial flat facies. The lacustrine-swamp facies and central lake facies are mainly composed of fine silty sand, whereas the lake shore facies and alluvial flat facies composed of coarse silt. River bed facies consist of sand or coarse sand. The contents of coarse silt, medium sand and coarse sand are similar in the flooding deposits, characterized by multi-peak curves. The performance proves that the AP clustering algorithm can provide a new mean for inversion and zonation of sedimentary environment conditions.
-
-