Focusing modelling on the tracheid development period – An alternative method for treatment of xylogenesis intra-annual data

Intra-annual repeated micro-sampling of the developing tree ring is getting more and more applied in xylogenesis studies. Variability in growth magnitude, notably due to different sampling positions on the stem, encouraged application of standardization and modelling techniques. Among these, methods using Gompertz equation had become widely spread, but tests made with black spruce revealed a frequent occurrence of crossovers between the cumulative number of cells in enlargement and the cumulative number of cells in wall thickening. This was due to a localized problem in the fitting for values close to the asymptote and was a major problem for estimating the timing of each individual cell development phases, which is an interesting application of these data. In this paper, a new method, based on a different approach, has been developed in order to avoid that problem and applied to intra-annual growth curves from four sites in Quebec (Canada). Since tracheid development analysis allows discriminating between active and inactive period of a phase, modelling can be restricted on the active period alone. The new method did not cause crossovers between the fitted curves. Therefore, it has been considered appropriate for estimating the timing for each individual cell in the whole range of data. Since resulting functions are polynomials from degree 1 to 3, possible studies concerning general tendency should be easy to lead. Also, the method has been tested with different sampling frequencies. To do this, number of observations from weekly samplings has been halved to simulate a semi-monthly sampling frequency and a comparison of the results from the new method applied on each version of the datasets has been tested. Generally, the simulated semi-monthly sampled dataset did not give significantly different results from the original weekly sampled dataset, in terms of general tendency and predicted intercept time in the extremities of the data range. This is very encouraging for situations when only semi-monthly sampling is available.