Search

A scanningt-test algorithm for detecting multiple time-scale abrupt changes in the level of a time-series was used to analyze an 8000 year time series of annual precipitation which was reconstructed from tree rings for the Nevada Climate Division 3 in the western USA. The tree ring samples were gathered from eight states in the southwestern USA. Twenty-two change-points were identified by the algorithm and these were used to partition the tree-ring series into twenty-three relatively Wet/Normal/Dry episodes. These twenty-three episodes were collaborated by a coherency analysis of abrupt changes between the precipitation reconstruction series and the TIC/δ18O records from cored sediments of Pyramid Lake in Nevada, and by comparison with published results from related studies. These episodes were also compared with studies of the global climate change and with records of climate change in China during the same periods. The results suggest that the precipitation reconstruction series is quite valuable for climate-change research on multi-centurial time-scales in the western USA, and that the scanningt-test and coherency detection algorithms may have a wide use for detecting multiple time-scale abrupt changes in a long time series.

Two computational cases that have analytic solutions are employed for studying the adaptive grid technique based on the variational principle. The results show that for the computational case of traveling shock waves the weight function, with the 2nd-order derivation terms taken into consideration, can more effectively reduce the error than one with gradient terms. For the case of cyclonic frontogenesis, weight functions only related to the gradient are unable to enhance the computational accuracy while ones with the wind field and frontogenesis function taken into consideration can more reasonably arrange the grid. Compared with analytic solutions, the adaptive grid technique suggested in this paper can improve computational accuracy and it displays the prominent advantage of saving memory.