Preserving long-term fluctuations in standardisation of tree-ring series by the adaptative regional growth curve (ARGC)

This paper presents a new method for the standardisation of tree-ring series, which attempts to remove the age effect from the low-frequency variations in the series. standardisation techniques based on the biological growth trend (RCS) only remove the trend linked to the age of the tree. However, in some trees, the trend is substantially different from the regional curve, and when the site fertility is not taken into account, the standardisation process may induce significant biases in the RCS standardised curve. An artificial neural network is used here to estimate an adaptive regional growth curve (ARGC) model. For each population or group of populations, the predictors are, in addition to the age (used by RCS) of each ring, the initial and the maximum growth rates (measured by the ring increments) of each tree. We have compared this method to the RCS method, using 20 Pinus cembra sites covering the Southern French Alps. The results show that the ARGC standardisation performs better for growth trend analysis and, by inference, for climate reconstruction.