L 1 l 1 l 1 l 1 l 1 l 1 l 1 l 1 l 1 l 1 l 1 l 1 0 2000 4000 6000 8000 10000 12000 Age a BP

Figure 1.6. Tree-ring data: record of atmospheric radiocarbon content over the past 12,410 years. The tree-ring radiocarbon equilibrates with atmospheric radiocarbon via the photosynthetic cycle. The 14C radioactivity was measured by counting the 3 particles on CO2 produced by combusting the wood material. Original sampling resolution was yearly (individual tree-rings) and lower; data shown are 5-year averages (n = 2483). The values are presented in delta notation (Fig. 1.3) with the oxalic acid standard of the National Bureau of Standards, for conventional reasons "A" is used instead of "5." The timescale (given as years before present (b.p.) where "present" is, as in "radiocarbon terminology," the year 1950) is based on a counted tree-ring chronology, established by matching radiocarbon patterns from individual trees. Since the age spans of the trees overlap, it is possible to go back in time as far as shown (and beyond). Since the radiocarbon data act as a proxy for solar activity (high A14C means low solar irradiance), it is possible to analyse Sun-climate connections by studying correlations between A14C and climate proxy records. (Data from Reimer et al. 2004.)

T1 = T2, is here of interest; higher-order moments are neglected. Lag-1 scatterplots (x(i — 1) versus x(i)) of the climate time series, using de-trended {i(i),x(i)}™=1 as realizations of the noise process, explore the autocovariance structure (Fig. 1.12). It is evident that all examples exhibit a more or less pronounced orientation of the points along the 1:1 line. This indicates positive serial dependence, or "memory," also called persistence in the atmospheric sciences. The reason for that memory effect is twofold. First, it is characteristic for many types of climatic fluctuations (Wilks 1995). Second, it can be induced by the sampling of the data. A record sampled at high resolution has often stronger persistence than when sampled at low resolution (see next section).

The lag-1 scatterplots (Fig. 1.12) reflect also the right-skewed shape of many of the distributions (more spreading towards right-up) and let some outliers appear.

o oo

0 0

Post a comment