Análise comparativa de procedimentos amostrais em um remanescente de floresta nativa no município de Lavras (MG)

Abstract

The objectives were to: compare volumes obtained by simple random sampling, stratified random sampling, simple random sampling with unequal sample sizes, systematic sampling, systematic sampling with unequal samples sizes and systematic sampling with post stratification for pre established errors of 10% in ali samplings and 20% with exception of stratified samplings; to verify efficiency of random and systematic sampling at species level; to define species spacial distribution of anative forest; to compare the precision ofthe model tree and the mean stratified tree with asystematic sampling and with the population parameter; and to evaluate the efficiency of six sampling procedures to describe the phytosociology of atrack of semi-decidual seazonal montane forest. Data where collected from 126 parcels of400 m2 each, forming 2.000 m2 blocks by joining every five parcels, in which the sampling procedures took place. The procedures comparison used Student "t" test and the statistical procedure in relation to the parameter. The comparison ofestimated parameter from horizontal structure used Euclidean Distances and graphic visual analysis. According to "T" test, ali procedures revealed similar estimates of the true mean with a10% error except for those with unequal sizes; there was no exception with a 20% error. For small native tracks at the species levei, both random and systematic sampling yielded means statiscally similar to the population parameter. To attain less than 15% error it was necessary to census the population. There was a positive correlation between survey cost and precision but not with estimate ofinteress. Systematic sampling was 40% and 60% more efficient than the stratified mean tree method and the mean tree model respectively. It should be enfasized that volumes estimated by the there procedures were safistactory given that maximum and minimum survey errors were 6,60% and 10,38%. For phytosociology purposes, ali six procedures have adequate representation ofhorizontal' structure; however, the Euclidean Distances obtained with systematic sampling were superior to those based on randoness.