| Vegetation is greatly related with human being's living. Human have been researching vegetation all the time. In field, vegetation is also clustered, so vegetation research can be considered as the plant community research. In vegetation investigation, releve method is commonly used, but how to determine the quadrat size is a problem. This size is also largely correlated to the finding of some ecological boundary. Nowadays, it is a hot topic that researching the boundary of plant community in landscape ecology and finding the regeneration capability of plant community. Therefore, how to determine the quadrat size in releve method and finding the plant community boundary furthermore is a important research context in community ecology.In traditional releve method, minimum area method is broadly applied but has many shortcomings, such as non-random sampling and subjectivity to find the asymptote point of species area curve. Furthermore, common used boundary-finding method suffers some crucial disadvantages, such as easily affected by quadrat size, never giving an objective standard to find the clustering stop branches with significant test.Our paper puts forward a new method to find the suitable quadrat size and community boundary in vegetation, based on the fast-developed computer skill and mathematics and statistics knowledge. Using the finite negative binomial distribution of species in plot site, it is deduced the probability function of the species emerging in certain-size quadrat. What's more, the determined quadrat size is used as the initial size to combine the quadrat in plot, and then, find the plant community boundary in research area. During this process, spatial correlation is greatly noticed, combining neighboring quadrats can deal with this correlation, while community mosaic and succession are also paid attention to, combining the paired areas based on the formal step can adjust our result to avoid mosaic situation. By the way, our method uses a randomized test---permutation test to give significant test of every combining step, so our result is objective.In order to test the suitability of our method, this paper take Barro Colorado Island (BCI) as an example to utilize and attest our method in field research, our approach gets some conclusions as follows:First, our approach can find the plant community boundary in reality. The BCI plot can be divided into two parts.Second, our method is easy, realize the visualization and can decrease the effect of quadrat size to our result as much as possible. Through calculating the occurrence rate of the species with high dominance in plot, our method can give out the suitable quadrat size in another perspective. Compared with Bayesian wombing method which uses hierarchical model to give the post suitable quadrat, our approach is apparently simple and effective, researchers can objectively change the emerging standard to control the suitable quadrat.Third, our method is less influenced by combination sequence. Every combination, the most significant quadrat pair is selected out and combined which can avoid the chaos of combination and make our result steadier. If every circulation, all the quadrat pairs which pass the permutation test are combined, the number of divided parts will be shortened.Finally, our method can objectively give out the significant boundary of plant community. Compared with the traditional combining method, our approach never give out the finalnumber of divided parts subjectively, instead, it use permutation test to give the significant test to every combination. Therefore, in our combination, human empirical error can be lowed down as much as possible and our result is also more accurate.To sum up, our method can find out the community boundary according to the defined standard, so it is helpful to further research the community structure and function. However, in practice, our method always affected by quadrat size and selection of similarity indices, but these shortcomings can be avoided by our method which determines the optimum quadrat size and the selection of similarity indices. |
