Modified climate with long term memory in tree ring proxies

Long term memory (LTM) scaling behavior in worldwide tree-ring proxies and subsequent climate reconstructions is analyzed for and compared with the memory structure inherent to instrumental temperature and precipitation data. Detrended fluctuation analysis is employed to detect LTM, and its scaling exponent a is used to evaluate LTM. The results show that temperature and precipitation reconstructions based on ring width measurements (mean alpha = 0.8) contain more memory than records based on maximum latewood density (mean alpha = 0.7). Both exceed the memory inherent to regional instrumental data (alpha = 0.6 for temperature, alpha = 0.5 for precipitation) in the time scales ranging from 1 year up to 50 years. We compare memory-free (alpha = 0.5) pseudo-instrumental precipitation data with pseudo-reconstructed precipitation data with LTM (alpha > 0.5), and demonstrate the biasing influences of LTM on climate reconstructions. We call for attention to statistical analysis with regard to the variability of proxy-based chronologies or reconstructions, particularly with respect to the contained (i) trends, (ii) past warm/cold period and wet/dry periods; and (iii) extreme events.