Abstract
Forest volume prediction is essential for many logistical purposes, such as transportation, harvesting and others. Therefore, the measurement of such a variable is indispensable for a forest enterprise. For such measurement, mathematical models are used to quantify the wood in the field. The aim is to evaluate the impact of incorporating locality information as indicator variables on the predictive accuracy of volumetric equation models, and to verify whether this incorporation influences the choice of the most appropriate model. A total of 11 volumetric models were used, divided into double entry and local models. For the qualitative variables the information of city, farm, stratum and field was used. It was observed the coefficient of determination, standard error of the estimate and various indexes related to the residuals, such as the mean absolute deviation, interquartile distance, range of variation, among others. In addition to the indices, the scatter plots, quantile-quantile and boxplot were analyzed. To evaluate the heteroscedasticity of the models, the Spearman correlation coefficient and the adjusted values of the models were calculated. The qualitative variables adjusted with a different intercept for each variable within the model and in a second moment the models were interacted with the qualitative variables.
Keywords:
Forest Inventory; Eucalyptus; Interaction
Thumbnail
