Variability of Persistent Temporal Correlation in Climate Data
This dissertation examines manifestations of persistent memory in climate data. Persistence is characterized by a slow power-law decay in the autocorrelations of a time series. Its existence implies that the influence of past values in a time series extend into the distant future. It has numerous theoretical implications, notably that it changes the asymptotic decay in the variance of sample means, which can substantially impact the uncertainty in climate mean states. Its intensity can vary over space, time, and other dimensions, e.g. tree species. Variation in its intensity can be used for practical applications such as discriminating between steady and intermittent rainfall and assessing the calibration period needed for paleoclimate proxy data. This work explores three major areas in which persistence can be leveraged to better understand the complexities of climate data. The first is in tree ring width data, which are among the best proxies for reconstructing paleoclimate records. The persistent correlations found in tree ring data suggest that the behavior of tree ring growth observed in a short calibration period may be similar to the general behavior of tree ring growth in a much longer period; therefore, the limited calibration period may be more useful than previously thought. The second area is in the quantification of uncertainty in the mean states of climate data. A framework for quantifying uncertainty in climate means is presented which can account for both classical short-range correlations and long-term persistent correlations. The final area is in the detection of subtle changes in tropical rainfall patterns. Persistence is used to illuminate recent changes in the temporal clustering patterns of rainfall in the tropical belt; the detected changes could have critical implications for the water resource management of the affected regions.
Tung, Purdue University.
Off-Campus Purdue Users:
To access this dissertation, please log in to our