An empirical resampling method for determining optimal high-pass filters used in correlation-based tree-ring crossdating

Visual crossdating of tree-ring series focusses on high-frequency variations. Automated correlation-based crossdating tools mimic this by transforming raw ring widths into indices that emphasise the high frequency signal, prior to calculating the goodness-of-fit between series. Here we present a resampling methodology to determine the relative merits of alternative simple high-pass filters and demonstrate it using two tree-ring data sets (British Isles oak, New Zealand kauri). Results indicate that: (a) high-pass filtering is a critical step; (b) the efficacy of alternative filters is variable, and; (c) efficacy appears to be species specific. These results have implications for crossdating in the two contexts investigated, and also for future software developments, especially the desirability of flexible implementations of high-pass filtering.