In many areas of the world where there is a strong seasonality in either temperature or rainfall, many species of trees form clear annual growth rings around their circumference. As more and more rings grow over the lifetime of these trees they form a record of year-by-year changes in climate, often stretching across hundreds and sometimes thousands of years. These records are the primary resource of the science of dendroclimatology - the study of the relationships between climate and tree-growth parameters and their use in the reconstruction of past climates.

At a few locations in the world, instrumental records of climate variables such as temperature or precipitation, reach back several hundred years. Over much of the world's well-populated land masses, however, there are extensive records only for a century or so. In many large and more remote regions such as northern Russia or northern Canada only a very sparse network of climate stations exists and most of their records span only a few decades. Our knowledge of `natural' climate variability is limited by the shortness of these records.

Tree rings are one important source of what are called proxy-climate indicators or palaeoclimate data. Tree-ring data, such as ring widths or measures of ring density, hold a special place among such palaeoclimate records because they provide information which is annually resolved, continuous and, importantly, they provide climate information which is dated to the precise year. Extracting this information, however, is not straightforward.

Tree growth is actually controlled by a complicated mix of climate-related factors. These include soil and air temperatures, soil moisture conditions, sunshine, wind etc. The size and density of cells within different parts of the annual growth ring, and the width of the ring itself, vary according to how the combination of all these climate factors change throughout the growing season. To complicate matters further, tree growth in one year is influenced to varying degrees by the nature of growth in one or more previous years and even by the climate conditions that prevailed outside the growing season. Ring growth over a number of years is also affected by non-climate-related factors that include tree age, competition from other plants, soil fertility, attacks by herbivorous insects and even changes in the composition of the atmosphere. It is the task of the dendroclimatologist to try to identify and separate these various influences in an effort to isolate information about some particular aspect of changing climate.

By sampling selected trees in carefully chosen locations it is possible to simplify the identification of specific climate information. For example, trees growing at high latitudes or high altitudes are most sensitive to changing temperatures while the growth of trees in semi arid environments responds strongly to changing soil water conditions and so provides information on precipitation.

Information about the past growth of trees is simply acquired by taking small-diameter radial cores which do not injure the tree. These are mounted, prepared and the ring widths measured and stored. By carefully comparing, correctly aligning and then averaging the yearly growth measurements from many trees in an area, a chronology can be constructed which reaches, unbroken, from the present back to the earliest growth years of the oldest trees. In special situations, matching a chronology with overlapping series of measurements from dead trees, perhaps preserved in old buildings or naturally in river gravels, peat bogs or lakes, enables chronologies to be extended backwards, sometimes for thousands of years, even though the individual trees that make up the chronology may have lived for only a few hundred years.

By statistically comparing tree-ring chronologies with modern climate records, equations can be developed which can be used in conjunction with the tree-ring data to reconstruct past climate values, with a degree of confidence that depends on the strength of the tree- growth/climate link and the statistical quality of the chronology back in time.