sents the altitudinal or latitudinal limit to tree growth (Kullman 1998, Körner 1999). From a review of published data, Grace (1988) concluded that “a 1°C increase in a north temperate climate may be expected to increase plant productivity by about 10 percent, providing that other factors like water or nutrients do not become limiting.” Controlled experiments dealing with the effect of temperature on plant growth are mostly performed on herbaceous species or seedlings (Junttila 1986, Loveys et al. 2002), and it is difficult to extrapolate those findings to the spatial and temporal scales considered by dendroclimatologists. For example, consider the evidence for treeline shifts in many areas of the world (MacDonald et al. 1998, Esper and Schweingruber 2004, Millar et al. in press). Such observations do not easily lend themselves to experimental testing of causal mechanisms. It has been argued that treeline position is not highly sensitive to interdecadal temperature change (Paulsen et al. 2000), but rather reflects environmental variability over several hundred years (Lloyd and Graumlich 1997, Körner 1999). Local disturbances, site conditions, and regional climatic regimes also influence the degree of sensitivity and rate of response of treelines to temperature changes (Kjällgren and Kullman 2002, Daniels and Veblen 2003).
The biological connection between temperature and tree ring variations on hourly to annual timescales has been investigated in the field using specialized instruments called dendrometers (Biondi et al. 2005), together with wood anatomy observations (Deslauriers et al. 2003a). For European and North American conifers living in cold environments, ring formation mostly occurs from May to the beginning of August, peaking around the time of maximum day length (Rossi et al. 2006 and references therein). By monitoring stem size of Pinus cembra and temperature during the growing season for two full years in the Alps, it was found that radial expansion ceased whenever air temperature fell below 5°C (Körner 1999). Night temperature was more important than day temperature for controlling radial growth of balsam fir at about 50°N latitude (Deslauriers et al. 2003b). At longer timescales (monthly to decadal), a number of dendroclimatic studies have identified a positive, linear relationship between mean July temperature and ring-width chronologies of Pinus sylvestris in northern Fennoscandia (Mikola 1962, Kalela-Brundin 1999, Helama et al. 2002).
In terms of causal mechanisms, tree ring records are likely to be the result of multivariate, and often nonlinear, biophysical processes. Models based on ecological or physiological concepts have been proposed to account for such processes (Fritts et al. 1991, Hunt Jr. et al. 1991, Scuderi et al. 1993, Berninger et al. 2004, Misson 2004). An intriguing hypothesis for the ability of treeline pine species to record slowly changing surface temperatures involves the fact that needles formed in one growing season remain alive and functioning for 10–30 years (LaMarche 1974). The mechanistic bases for the statistical models used to extract climate signals from tree ring data have been summarized in simulation models focusing on the activity of the tissue that forms wood, the vascular cambium (Vaganov et al. 2006). Also note that linear relationships between tree ring records and climate are at least equal to, and often exceed, those found for other proxies (Jones et al. 1998). Statistical techniques more responsive to nonlinear interactions have so far provided relatively small improvements for explaining climatic variance (Hughes 2002 and references therein).
All proxy records of climate are obtained from samples that are not randomly selected (Cronin 1999). Part of the researcher’s ability consists of identifying sites where proxy records are as long, continuous, and representative of the target climatic variable as possible. Guidelines have been specified in the tree ring literature