(Schweingruber 1988, Fritts and Swetnam 1989) to ensure that sample (site, tree, and core) selection is based on a priori rather than a posteriori criteria. For instance, sites are selected in remote areas where tree density is low in order to minimize the impact of stand dynamics and intertree competition (Biondi et al. 1994). The influence of varying local conditions on dendroclimatic records has been studied for elevation, slope, and exposure (Kienast and Schweingruber 1986, Villalba et al. 1994, Buckley et al. 1997, Tardif et al. 2003, Piovesan et al. 2005), topographic convergence and potential relative radiation (Bunn et al. 2005), and flooding patterns (Tardif and Bergeron 1997). Sampled trees should not show signs of disturbance factors such as insect infestation, grazing, fire damage, human utilization, fungal infestation, or mistletoe attack (Schweingruber 1988, Fritts and Swetnam 1989). Overall, as in any other field-based investigation of environmental change, defining the research question is the premise to a proper selection of materials and methods (Bräker 2002).
To ensure reliable results, tree ring science places great emphasis on replication (Wigley et al. 1984, Fritts and Swetnam 1989). At least 10–20 trees per species are sampled at a site, mostly by taking increment cores, and each tree is cored following specific guidelines (Grissino-Mayer 2003). All collected samples are transported back to the laboratory, where they are compared to one another. The method of crossdating (or pattern matching) is used to assign calendar years to the individual rings (Baillie and Pilcher 1973, Wigley et al. 1987, Yamaguchi 1991). Initially based on a visual comparison (Stokes and Smiley 1996), crossdating is quality controlled by means of numerical techniques once the ring widths are measured (Holmes 1983, Grissino-Mayer 1997). The precision and accuracy of crossdating have allowed the refinement of radiocarbon dating techniques (LaMarche and Harlan 1973, Friedrich et al. 2004). Tree ring chronology development follows rules that are common to all applications of tree ring science, and is completely independent of any climatic data. Samples or portions of samples that cannot be crossdated with the rest of the specimens are not included in the final chronology. Recommendations to archive all collected materials, so that they remain available for future study, have been published (Eckstein et al. 1984). The Laboratory of Tree-Ring Research at the University of Arizona still has wood samples, field notes, and measurements that were taken a century ago by A.E. Douglass, the Tucson astronomer who proposed many of the dendrochronological methods still in use today (Webb 1983).
After crossdating, tree ring parameters other than width (such as density, stable isotopic composition, cell size and wall thickness, resin duct density, and trace metal concentrations) can be measured. Dendroclimatic studies of past surface temperature are mostly based on ring width or maximum latewood density; the latter usually has a higher correlation with temperature, especially during the summer (Conkey 1986, Briffa et al. 2002). Maximum latewood density is also correlated with ring anatomy as measured by cell number, cell diameter, and cell wall thickness (Wang et al. 2002). Measurements made on crossdated wood samples from the same species and site are typically combined into a master chronology (Fritts 1976, Cook and Kairiukstis 1990). This process is aimed at increasing the climatic signal by reducing the importance of individual sample noise. In general, the number of specimens required to obtain a robust chronology increases as the common variance among specimens decreases (Fritts