科研之友
微信
新浪微博
Facebook
分享链接
摘要
The objective of this study was to increase the accuracy of estimates of tree height and to reduce the need for measurement in field of height, leading to reduction of costs on forest inventory through the construction and validation of a model for estimating the height of trees in stands of eucalyptus using artificial neural networks. The data used in the experiment consisted of three clones, comprising nearly 3,000 trees on 145 permanent plots with an average area of 215 m(2), measured on six occasions (ages). The variables used to estimate the total tree height were divided into quantitative and qualitative. The quantitative variables were: age (months), shell diameter at 1.30 m height from the ground surface (dbh) and average dominant height in the plot. The qualitative variables was the soil type in their respective classes. For validation and application of the proposed methodology two situations were considered: (a) when there is the introduction of new genetic material and there is no information about the hypsometric relation thereof, and (b) when the trend of growth in height of the stands implanted obtained by the existence of measurements on inventory plots is already known. Values of correlation coefficient higher than 0.99 were achieved with the tested methodologies. The methods were effective to achieve the proposed objectives, ensuring high precision of the estimates obtained through artificial neural networks.
- 出版日期2013-8
