388 DENDROCHRONOLOGY on dating discrete events such as volcanic eruptions, earthquakes, landslides, avalanches, forest fires, floods, severe frosts etc through history, giving important background information about the frequency and intensity of these phenomena, and in many instances giving the first calendar dates for them These events may be represented in the tree-ring series by sudden growth rate changes, scars, or abnormal cells Longer-term variations and rates of change for such phenomena as river-flow changes, lake-level changes, saltwater ingression, changes in ocean currents through time (mapped by dating driftwood), glacial advance and retreat (reflected in subfossil trunks from valley sides), erosion rates, human forest clearance and the like have all been subjects for dendrochronological study Periods of building of prehistoric lakeside settlements have been found to be coincident over areas of central Europe, perhaps reflecting migration patterns in response to climatic changes Dendroclimatology It was not long before people realized that well replicated chronologies were themselves a proxy dataset of the major climatic influences on growth, and people looked for the best methods of extracting this information at annual, decadal, century and millennial time-scales In some areas the relationships between climatic factors and growth are relatively simple, and the dominant limiting factor to growth exhibits a clear correlation with the width of the annual ring In semi-arid areas for example, the ring width is generally a reflection of rainfall levels Some basic environmental information can be gained simply by looking at long-term trends in the ring width itself For example, different centuries show quite different growth rates for similarly biologically aged oaks preserved in river gravels in Germany, used to construct a multimillennial chronology Even without any complex calibration of the climate–growth relationship, it is possible to draw broad inferences about changing conditions at particular times, with narrow mean ring widths representing poorer growth conditions and wider mean ring widths more favourable growth conditions The sensitivity of trees to external changes in their environment changes with the site conditions In more extreme environments it is possible that rings may be missing, or partially missing around the circumference, or if unfavourable conditions are experienced during the normal growth period, a slowing down and then resurgence of growth may produce apparent ‘false’ rings These problems are generally detected when ‘cross-matching’ the samples during chronology construction Unlike many scientific studies, sampling for dendroclimatological reconstruction relies on careful selection of the trees, not random selection The reasons for this become clear when one considers the nature of the ring width itself The ring width in any given growth season is the result of a number of factors: the biological age of the tree, environmental factors unique to that tree, stand-wide influences atypical of growth elsewhere, a large number of other factors, and of course the regional climatic signal that is of interest in these studies One can minimize the influence of non-climatic factors and maximize the climatic signal by careful site selection Trees growing at the margins of the population, whether that be altitudinally or latitudinally are generally most sensitive to climatic factors By choosing the dominant trees with no obvious signs of damage or disease, one enhances the relationship between ring width and regional climatic influence By increasing the number of samples one reduces the individual tree responses and enhances the ‘climatic signal’ The ‘signal to noise ratio’ has been studied in detail and will vary from location to location and species to species, but as a general guide a minimum of 15 to 20 trees seems to produce a representative sample The biological age of the tree generally has an effect on ring width, with a natural tendency to put on narrower rings as the tree increases in girth, even though overall productivity levels may be very similar This so called ‘age trend’ can be readily removed by fitting curves to the overall series and then taking account of the difference of the observed ring width from the theoretical mean value at a given time This process, known as ‘standardization’, results in the production of a series of indices of growth, and makes direct comparison between trees of different biological age more readily achievable Just how best to derive this series of indices for subsequent analysis has been the subject of a vast literature within dendroclimatology, and each new study really needs to justify the particular methods employed to remove age-related trends The problem of course is that the more closely one fits curves to the original data, the more climatic information one may be subtracting, and whilst the great advantage of dendroclimatology is that it has the potential to produce annual resolution in the results, longer-term trends in growth may also reflect climatic information, and may be lost in the process of standardization Even so, the resulting series is often highly autoregressive, that is to say that growth in any one year often reflects the overall vigour of the tree for the few prior years, and this autoregression needs to be removed to enhance