Mathematical Sciences: Statistical Inference in Paleoclimatology

点击翻译按钮获取中文摘要

9311071 West This project concerns statistical issues arising in data processing, analysis and inference based on geological records of environmental change. The current proposal results from discussion between the two investigators about the statistical issues and problems for models and methods of analysis, especially time series analysis, geared to the realistic assessment of patterns of time variation in sedimentary records relevant to climate change. This project will initiate a collaborative research and reciprocal training program to address these issues. The focus will be on the processes of uncertainty propagation and formal statistical inference about climatic changes based on raw records from deep lake sediments. Similar issues exist and deserve attention in the assessment of data from deep sea sediments, ice cores, and tree rings. Technically, the project concerns statistical calibration of raw data records, and time series analysis to assess patterns of variation in the geological records over time. The initial objective will be to develop an appreciation of the statistical issues as they permeate the various stages of the experimental processes in analyzing isotope and other proxy climatic variables from deep-lake sediment cores. Cross- disciplinary training for each investigator will e a central and continuing feature of the collaboration. This will be directed at bridging the gulf between, on one hand, current standards of practice in statistical analysis of paleoclimate data (e.g. Pisias and Moore 1981; Hagelberg, Pisias and Elgar 1991) and, on the other hand accessible and appropriate modern statistical methods and models. A reciprocal training and development schedule will provide the opportunity for each of the two investigators to become sufficiently proficient in the fundamentals of the other's field of expertise so as to initiate and enable a research program focussed on rigorous statistical analysis of paleoclima te data generated from deep lake sediment cores. Statistician Mike West will become familiar with the methods used to select sediments core sites for paleoclimatic analysis, the various methods of geochemical analysis of sediment cores (and notably the limitations of age assignment to sediments as a result of various natural processes such as erosion and redeposition, bioturbation, diagenesis and laboratory contamination), calibration of carbon dating assessments, and current methods of time series analysis used in paleoclimatology. Geologist Tom Johnson will become familiar with the capabilities and pitfalls of statistical methods, especially methods of time series analysis that are currently used in this area, and gain appreciation of the consequences for uncertainty assessment in geochronology. The investigators anticipate eventual development of Bayesian methods in statistical calibration at the various stages of the experimental process, and resulting time series analysis in assessing cyclical variability in calibrated records. Both investigators will also become familiar with the statistical issues and methods of calibration of 14C dates (together with associated uncertainty assessments) to calendar dates, and with the recent development of advanced statistical methods in the field. This project will eventually lead to focussed research efforts in developing Bayesian statistical models and methods appropriate for all stages of the experimental process in sediment based paleoclimatology, and that will ultimately apply in other geological and archaeological arenas. Most optimistically, it is hoped that this collaboration will seed the development of an internationally recognised centre for the statistical analysis of paleoclimate proxy records at Duke University. ***