Filtering the Effects of Competition from Ring-Width Series

Abstract

Spline functions were examined and compared with conventional polynomials for use in filtering nonclimatic variance from tree-ring width series. Both types of curve were fitted to ring-width series exhibiting particularly marked effects of competition and release from competition with neighboring trees during the last 100 years. Available climatic data from that interval were used to statistically evaluate the capabilities of each type of function for removing nonclimatic effects and preserving the climatic signal. The results suggest that both types of function can be used successfully for those purposes, though in the presence of extremely rapid changes in growth rate due to a sudden release from competition it may be necessary to divide the ring-width series into two segments and fit a separate curve to each segment. Tightly-fit polynomials seem to be about as effective as splines, but are less desirable from a computational point of view as, for example, when the magnitude of the (negative) exponents of the coefficients exceeds computer capacity. Further, a spline function can be prescribed from its frequency-response characteristics, so it is possible to specify in advance the extent to which it will filter out any potential climatic cycles.