A new dissimilarity measure and a new optimality criterion in phytosociological classification

A new dissimilarity measure, Uppsala dissimilarity, is proposed. It is a Manhattan-type measure in between the Canberra and Gower measures, based on the differences between scores in relevés compared, but it also takes both the sums of scores and the difference between maximum and minimum score into account. The measure is considered realistic for phytosociological material.A new optimality criterion has been developed after unsatisfactory results had been obtained with the DOL criterion (Popma et al. 1983) which was developed previously by our group. Problems with DOL were especially met when the criterion was applied to the distribution of only one species over the cluster array obtained. The new criterion takes both internal cluster homogeneity and between-cluster dissimilarity into account. Between-cluster dissimilarity is calculated for all other clusters and not only for the nearest neighbour, as in DOL. The new criterion has both an unweighted form: SOM, and a form with weighting for cluster size: SWOM.This new criterion was successfully applied to the evaluation of the sharpness of distribution of individual species over cluster arrays, under the name of SIM: species indication measure and SWIM, species weighted indication measure.The measures were applied to some test data. Differences between the unweighted and weighted forms were found which could not be easily interpreted.Some remarks are made on the coherence of d-SAHN and h-SAHN approaches in agglomerative clustering within the new strategy proposed.Abbreviations DOL = Detection of Optimal Level - S(W)IM = Species (Weighted) Indication Measure - S(W)OM = Standardized (Weighted) Optimality Measure - UD = Uppsala Dissimilarity measure - WPGMA = Weighted Pair-Group Method Average linking clustering - SAHN = Sequential Agglomerative Hierarchical Non-overlapping clustering     

Monitoring plant community change: an application of quadrat classification + discriminant analysis

A procedure for monitoring plant community change was described using data from 189 quadrats (each 0.09 m² in area) from or near 11 Carex exserta meadow sites in the high Sierra Nevada, California, USA. Initially the quadrats were agglomerated into five clusters by the flexible clustering strategy (beta=–0.25) with the standard absolute distance resemblance function. Data for each quadrat were cover percentages for C. exserta, other plants, litter, soil, gravel, and rock. The five clusters appeared to define a cover gradient, from quadrats with mostly gravel and rock to those with mostly C. exserta, and were accordingly designated pioneer, low seral, mid-seral, high seral, and climax.Classification functions (from discriminant analysis) are used with values of the variables to classify individual quadrats on sites used to monitor change. A site is characterized at repeated observations by the proportions of quadrats in each class. Within-class (low seral vs. low seral) rather than between-class (pioneer vs. low seral) tests are made for presence of change. Confidence intervals for differences in proportions of quadrats or individual quadrat probabilities of class membership are computed. If the confidence intervals do not cover zero, values for time one versus time two differ significantly.     

闽北地区四代三化螟发生动态Fuzzy预测模型研究

根据Fuzzy数学原理和三化螟生物学特性,组建了四代三化螟发生动态综合预测模型经对福建省将乐县16年四代三化螟发生期、发生量回报及1995年的预报,拟合率及正确率均达100％。     

Blurry artifacts: a retraction

Abstract. This note is to apologize for an error in the computer program used to evaluate the random data used in Fuzzy Set Ordination according to Zhang & Oxley. After correction of this error no artifacts could be detected any longer. However, the basic conclusion of the earlier critical note still stands: if one selects environmental variables after analyzing the results of a multivariate gradient analysis, and then uses these variables as input into a multiple univariate gradient analysis, the results are expected to be comparable.     

Comparison of RAPD and RFLP genetic markers in determining genetic similarity among Brassica oleracea L. genotypes

Genetic similarity among 45 Brassica Oleracea genotypes was compared using two molecular markers, random amplified polymorphic DNA (RAPD) and restriction fragment length polymorphisms (RFLPs). The genotypes included 37 broccolis (var. italica), five cauliflowers (var. botrytis) and three cabbages (var. capitata) which represented a wide range of commercially-available germplasm, and included open-pollinated cultivars, commercial hybrids, and inbred parents of hybrid cultivars. Fifty-six polymorphic RFLP bands and 181 polymorphic RAPD bands were generated using 15 random cDNA probes and 62 10-mer primers, respectively. The objectives were to compare RFLP and RAPD markers with regard to their (1) sampling variance, (2) rank correlations of genetic distance among sub-samples, and (3) inheritance. A bootstrap procedure was used to generate 200 random samples of size n (n=2,3,5,... 55) independently from the RAPD and RFLP data sets. The coefficient of variance (CV) was estimated for each sample. Pooled regressions of the coefficient of variance on bootstrap sample size indicated that the rate of decrease in CV with increasing sample size was the same for RFLPs and RAPDs. The rank correlation between the Nei-Li genetic similarity values for all pairs of genotypes (990) based on RFLP and RAPD data was 0.745. Differences were observed between the RFLP and RAPD dendrograms of the 45 genotypes. Overlap in the distributions of rank correlations between independent sub-samples from the RAPD data set, compared to correlations between RFLP and RAPD sub-samples, suggest that observed differences in estimation of genetic similarity between RAPDs and RFLPs is largely due to sampling error rather than due to DNA-based differences in how RAPDs and RFLPs reveal polymorphisms. A crossing algorithm was used to generate hypothetical banding patterns of hybrids based on the genotypes of the parents. The results of this study indicate that RAPDs provide a level of resolution equivalent to RFLPs for detemination of the genetic relationships among genotypes.     

A neural network model for the prediction of membrane-spanning amino acid sequences

The architecture and weights of an artificial neural network model that predicts putative transmembrane sequences have been developed and optimized by the algorithm of structure evolution. The resulting filter is able to classify membrane/nonmembrane transition regions in sequences of integral human membrane proteins with high accuracy. Similar results have been obtained for both training and test set data, indicating that the network has focused on general features of transmembrane sequences rather than specializing on the training data. Seven physicochemical amino acid properties have been used for sequence encoding. The predictions are compared to hydrophobicity plots.     

棉蚜种群动态预报模型

对1980-1989年河南棉区棉蚜种群动态分析表明,棉蚜为害可分为苗蚜和伏蚜两个时期,根据环境因子(气温、相对湿度、降雨量)和不同危害期,建立了2个模糊预报模型,并取得了较满意的预测效果。