Measuring wood density of different tree species using the micro-drilling resistance method

Abstract

In order to reduce modeling workload and improve the accuracy of wood density measurement, a modeling method for establishing an overall model for multiple tree species was studied. Firstly, the wood cores and drill resistance of 9 tree species were collected. The absolute dry density, wood moisture content, and average drill resistance for each wood core were calculated. Secondly, the random forest algorithm was used to rank the relative importance of each factor affecting wood density, and factors with relative importance higher than 0.05 were selected as independent variables for building the overall mathematical model for total tree species and sub model for individual tree species. The results showed that: (1) the relative importance of tree species and drill resistance on wood density was higher than 0.05; (2) the relative importance of moisture content and drill usage frequency (less than 150 times) on wood density was lower than 0.05; (3) the average estimation accuracy of overall model was 91.42%, while that of the sub model was only 90.44%; (4) among the 9 tree species, the standard deviation of the overall model for 5 tree species was lower than that of the sub model. The results showed that the accuracy and stability of the overall model were superior to those of the sub models, and it is feasible to establish an overall model to estimate wood density.