Résumé
Studies of intra-annual dynamics of cambial activity and wood formation share very similar data because they are based on classic concepts of xylem development and well-documented techniques of sample preparation. However, the way the data are produced is specific to each study or group of authors because there is not a real agreement or consensus about critical variable definitions and data processing. The consequence is that despite very similar data, these studies are difficult to compare. In order to improve this point, we proposed objective definitions of the main phenological events (critical dates) occurring during wood formation, i.e. the beginning and ending of the cell enlarging and maturing phases and the beginning of the cell mature phase. The beginning of a given phase was defined as the date at which 50% of the observed radial files show at least one first cell in this phase, while the end of a given phase was defined as the date at which 50% of the observed radial files show at most one last cell in this phase. Dedicated R package CAVIAR was developed in order to apply these definitions to the computation of critical dates, as well as to their visualisation and analysis. Several statistical tests (Student's t-test, Wilcoxon's rank sum test and bootstrap tests on means and medians) were evaluated for comparing critical dates between two groups. Based on theoretical considerations and on study case analyses, it was shown that the bootstrap test on median is the best-suited statistical test for comparing critical dates when considering the peculiarities of the data at hand. Finally, the application of the presented definitions and methods was illustrated by the study of a real dataset obtained from five silver firs and five Scots pines grown under the same conditions and sampled weekly during the 2006 vegetation period. The results of this application showed that wood formation in pines started earlier, finished later and lasted longer than in firs, as documented by previous studies. We believe that the development of such objective definitions accompanied by ready-to-use tools for data processing, analysis and visualisation will reinforce the discipline by contributing efficiently to standardise concepts and methodologies