Computer-assisted classification and discrimination of common phytoliths in the surface soils of different vegetation types in Northeast China

Abstract

Research into phytolith morphology provides essential information for understanding the relationship between phytoliths and vegetation and hence between vegetation and climate, increasing the precision and accuracy of the use of phytolith analysis in paleoenvironmental reconstruction. This study aims to provide an improved basis for the quantitative description and discrimination of the common phytoliths in the surface soils of forest and meadow steppe in northeast (NE) China. We sampled 15 sites from forest and meadow steppe and analyzed common phytoliths found in the surface soils and in their corresponding plant communities. An image processing procedure implemented in MATLAB was used to quantitatively distinguish phytoliths based on six textural characteristics and seven Hu invariant moments; a t-test and discriminant analysis were then used to analyze the results.The principal results are as follows: (i) The texture of elongate psilate phytoliths and the Hu invariant moments of tabular rectangle phytoliths in surface soils is significantly different between forest and meadow steppe. (ii) Texture characteristics can be used alone to differentiate tabular rectangle phytoliths in the plant communities from forest and meadow steppe, while both texture and Hu invariant moments should be used for elongate psilate. (iii) For both vegetation types, the texture characteristics are significantly different between phytoliths in the plant communities and those from the surface soils. (iv) Using image analysis, forest and meadow steppe can be differentiated by these common phytoliths in the plant communities and surface soils.