This study was done in the eastern administrative region of Belo Horizonte-MG to compare the inventory types and sampling methods, and testing the size and design of sample unit. Three methods denominated as method I, method II, and the method III, were used to evaluate the quality of street arborization. Efficiency of the methods was evaluated by determining the time and resources required for execution. The method I was 2.4 and 1.6 times more rapid than method III and the method II, respectively. The execution cost of the method III was 5.4 and 2.6 times higher than the method I and II, respectively, while that of method II was twice of the method I. The efficiency evaluation of the sample size and design normally used for inventory arborization of roads of Belo Horizonte was compared by evaluating the individual trees on the streets of the zones called Colégio Batista, Floresta, Horto, Sagrada Família and Santa Tereza, through a quali-quantitative inventory, which served as the base for the comparing plot size and design. Plots of 200x200 m, 100x400 m, 300x300 m, 150x600 m, 400x400 m and 200x800 m, distributed over the study area were sampled. The number of trees per street-km was the main variable to calculate the variance and the
minimum sample units. In all the five zones about 7200 trees were identified. The smallest plots (4 ha) showed maximum standard errors of the means, which are important parameters to determine the degree of statistical precision and are indicatives of higher or lower efficiency. The plots of bigger size (16 ha) showed good results in terms of precision and also showed the necessity of greater sampling intensity. The square plots of 300x300 m showed least standard error of means, least sampling error, least coefficient of variation, and least deviation from the mean at a given confidence interval, thus appeared to be the best sampling units. The results from this and the other studies suggest that the squares or rectangular plots of about 10 ha. are best suited to
evaluate street arborization. To compare random and systematic sampling procedure, the region map (1:10.000) was quadrated into 300x300m, sample units and the sampling was done either randomly and systematically. The ttest for the independent data confirmed a significant difference between methods. Relative efficiency concept was used to determine the most efficient method, and calculated on the bases of standard error and time spent on plot measurement excluding the time spent on moving between the plots. The systemic sampling showed relative efficiency of 34.17% in relation to the random sampling, showing that the latter was of higher relative efficiency. Although the random sampling showed greater relative efficiency, the systemic
process also can be used to obtain good results.
