An application of Embedded Markov chain for soil sequences: Case study in North Western part of Algeria.

Embedded Markov chain (EMC) has long history in geological domains, particularly to define the most representative sequences from statigraphic logs. In other words, what is viewed as a meaningless and disordered stratigraphic layer stack can be reorganized in a meaningful sequence by using EMC. This method was transposed in this paper to obtain soil sequences from data retrieved from soil map made by authors, covering a part of the region of Traras (N.W. of Algeria) and containing 13 major soil types. Each major soil type occupies at least one polygon in the map and allow to establish soil adjacencies, which have been tabulated in a matrix regardless to the direction. Three EMC methods have been tested, Walker, Harper and Türk using Strati-signal software and to erect soil relationship diagrams (SRD) representing the most significant links between soils. Significant test is the main difference between the above mentioned three EMC methods. It has been shown that Harper method is quite insensitive to small number of transitions. Besides, all three methods agreed for one soil sequence made by four soils: lithics leptosols- cambisols chormics- cambisols calcarics-fluvisols representing theoretical catena the most representative to the study area. This soil sequence is relevant to the study region and even to the whole Mediterranean region, and is commanded by the topography and the Mediterranean bioclimate. Walker SRD is the most realistic but the most difficult to interpret because of the high number of soil links, Harper SRD gives interesting results. Although the results didn't bring something new to the soil interpretation and soil pedogensis but EMC applied to a finer scale may highlights other hidden relationships between soils. [ABSTRACT FROM AUTHOR]