Abstract:

The analyses of spatial distribution patterns of plant populations are useful for understanding pattern types and intra/inter-specific relationships. One of the most frequently employed methods in detecting spatial distribution patterns of populations is the nearest neighbor analysis proposed by Clark and Evans in 1954. This method has been highly successful for analyzing spatial patterns at a single scale but is rarely used for analyzing distribution patterns at multiple scales. We present the extended nearest neighbor analysis (ENNA) in this paper to solve the scale-dependent problem associated with the traditional method of nearest neighbor analysis. The Clark-Evans index was modified by using a distance scale parameter d (m), described in the following equation: CE (d) =r dA /r dE = (1N d∑N di=1r di ) / (0.5A d/N d+0.051 4P d/N d+0.041P d/N d 3/2 ). Accordingly, the equation for testing the calculated CE index values against the significant deviation from 1 was changed into u (d) = (r dA -r dE ) /σ d, where the parameters, r dA, r dE, N d, r di, A d, P d, σ d, refer to the mean distance between an individual and its nearest neighbor (m), the expected mean distance of the individuals of a population randomly scattered (m), the number of individuals in the current sample plot, distance between individual i and its nearest neighbor (m), surface of the current sample plot (m 2), circumference of the current sample plot (m), and the standard deviation, respectively. The procedure of scaling transformation in this approach was similar to that of the sandbox experiment in fractal theory, and the rule for detecting the pattern type was the same as that in the traditional nearest neighbor analysis. The traditional nearest neighbor analysis is a special case for the extended nearest neighbor analysis in which the minimum value of the distance scale parameter (d) is used. An example using the data from a needle and broad-leaved mixed forest community at Heishiding Nature Reserve, Guangdong Province was presented to explain the procedure. Five typical plant populations of this community, Pinus massoniana, Symplocos laurina, Castanopsis nigrescens, Itea chinensis and Rhodomyrtus tomentosa, were chosen for the multi-scale analysis of spatial distribution patterns. The results showed that spatial patterns of all five populations were scale-dependent with varying degrees of intensity. The Pinus massoniana population was randomly distributed at most scales examined, which may have been caused by the random self-thinning process in the population. The population of Itea chinensis was clumped at all scales examined. A simulation with the aid of geographic information system (GIS) also revealed that the distribution patterns of Symplocos laurina, Castanopsis nigrescens, Itea chinensis and Rhodomyrtus tomentosa were mainly clumped or random with an increase of distance scale. These results demonstrated that the ENNA method presented in this paper could be used for multi-scale analysis of spatial distribution patterns of plant populations that could not be solved using the traditional nearest neighbor analysis.

Key words: Extended nearest neighbor analysis (ENNA), Spatial distribution patterns, Multi-scale analysis, GIS