A methodological framework to optimize models predicting critical dates of xylem phenology based on dendrometer data

Dendrometer measurements are a frequently used alternative to the laborious and time consuming microcoring to investigate intra-annual growth dynamics of trees. However, since dendrometer data not only comprise cambial growth, but also hydrological fluctuations, both signals need to be disentangled to derive critical dates of xylem phenology from dendrometer data. For this purpose, various approaches can be found in the literature, however a systematic comparison of the different options is still missing. In this study we present a methodo-logical framework to evaluate the accuracy of different mathematical fittings to derive tree-ring phenology from dendrometer data and apply this approach to a data set comprised of three conifer species where high-resolution band dendrometer measurements and microcore sampling have been done in parallel. Based on our study we provide evidence that the most common approaches to derive onset and cessation of xylem cell enlargement from dendrometer data, i.e. applying absolute and relative thresholds to deterministic models fitted to dendrometer data, caused systematic deviations from the reference of xylogenesis observations, but could be improved by small adaptations. The most precise and unbiased predictions of xylem growth phenology were obtained by fitting the three-parameter Gompertz model to the dendrometer data and applying a relative threshold of 3.5% of annual increment to the model predictions for growth onset and an absolute threshold of 4.5 mu m day-1 based on the first model derivative for growth cessation. Our framework enables an improved usage of dendrometer data for the prediction of the onset and cessation of xylem cell enlargement, which are important ecological indicators to quantify the effects environmental changes on forest growth and the terrestrial carbon cycle.

Automated summary

This study presents a methodological framework for evaluating the accuracy of different mathematical fittings to derive tree-ring phenology from dendrometer data. The results show that fitting the Gompertz model to the dendrometer data and applying relative and absolute thresholds to the model predictions can provide the most precise and unbiased predictions of tree growth phenology.