STUDY OF SELECTION OF WEEDS COMMUNITY BY HERBICIDES USING MULTIVARIATE ANALYSIS TECHNIQUES

Abstract

For decision making on which strategy is improve for weed management, statistical tools such as multivariate analysis. Especially with a large amount of data, can be used. The objective of this study was to evaluate the feasibility of the use of multivariate analysis for studies on the selection of weed flora as a result of the application of different herbicides. The experiment was in field, in randomized blocks, with four repetition, the area was divided into 32 plots with 3 x 5 meters in length with 8 plots per block. The experimental treatments were: 1-glyphosate, 2- 2,4D; 3-glyphosate + 2,4D; 4- carfentrazone-ethyl; 5-glyphosate + carfentrazone-ethyl; 6-haloxyfop-methyl; 7-clethodim; 8-ammonium glufosinate. The weed community was surveyed in each plot, before application and at 30 days after application of the treatments. The data were submitted to cluster analysis and main components. The most discriminating weed species were Alternanthera tenella and Commelina benghalensis, followed by the grasses Panicum maximum, Cenchrus echinatus and Eleusine indica. The results of cluster analysis were very similar to those obtained by principal component analysis. Multivariate statistical techniques were able to group chemical treatments according to specific composition regardless of the phytosociological index considered.