147 Amelia J. Koch, Don A. Driscoll and J.B. Kirkpatrick Australian Forestry 2008 Vol. 71 No. 2 pp. 147–159 Estimating the accuracy of tree ageing methods in mature Eucalyptus obliqua forest, Tasmania Amelia J. Koch 1,2,3 , Don A. Driscoll 4 and J.B. Kirkpatrick 1 1 School of Geography and Environmental Science, University of Tasmania, Private Bag 78, Hobart, Tasmania, 7001, Australia 2 Email: ajkoch@utas.edu.au 3 Current address: Tasmanian Forest Practices Authority, 30 Patrick St, Hobart, Tasmania, 7000, Australia 4 Centre for Resource and Environmental Studies, Australian National University, W.K. Hancock Building (43) Biology Place, Canberra, ACT 0200, Australia Revised manuscript received 8 May 2008 Summary Estimates of tree age are important for making management decisions on the tree hollow resource because hollows suitable for fauna occur with greater frequency in older trees. The methods used to age trees vary with the practicalities of obtaining wood samples and the quality of the wood samples available. Ring counting is commonly used on smaller sample sizes when complete wood samples are available. When wood samples are incomplete, a combination of ring counting and extrapolation is often used. When no wood samples are obtained, growth models are generally used to estimate tree age. This paper uses all three methods, including three different growth models, to examine the accuracy of ageing trees. Simple regressions between tree age and diameter at breast height (over bark, dbh) provided the most accurate growth models at a site. Age estimates obtained from such models, however, had unacceptably large errors when few trees were used or when variable growth rates occurred. Under these circumstances, smaller errors margins were obtained from using a model based on tree dbh and site attributes than when averaging growth rates across sites. The estimated error in tree age estimates when using ring counting and extrapolation was about 10% of the tree age. The error of extrapolation increased with the amount of wood sample that was missing. Error margins were large for the oldest trees (average ± 42.4 y for trees > 350 y old) but less than ± 15 y for most (73%) of the trees estimated to be 100–300 y of age. These middle-aged trees are often the most useful to study when examining the rate of hollow production in eucalypts. Therefore, age estimates acquired in this way are generally accurate enough to be useful for making management decisions regarding the tree-hollow resource in production forests. Keywords: age; accuracy; habitats; growth rings; growth models; regression analysis; Eucalyptus obliqua Introduction Tree hollows provide important habitat for fauna (Gibbons and Lindenmayer 2002) and older trees are more likely to contain hollows (Mackowski 1987; Whitford 2002). The time required for a hollow suitable for faunal use to form (150 y: Mackowski 1987; 165 y: Wormington and Lamb 1999; 130 y: Whitford and Williams 2002) is substantially longer than the 60–120 y generally planned between harvesting operations in production forests (Whiteley 1999; Department of Natural Resources and Environment 2002a,b). Management of the hollow resource in production forest areas therefore requires an understanding of the age at which trees in different areas produce hollows suitable for use by animals. Trees can be aged according to the disturbance history of a site (Bradshaw and Rayner 1997b), by radiocarbon dating (Turner 1984), by tree ring counting (Banks 1997) or by using tree ring counts or tree diameter increment data to produce growth models (Lloyd and Lau 1986; Wormington and Lamb 1999; Gibbons et al. 2000; Moloney et al. 2002). The accuracy of tree age estimates can potentially influence the effectiveness of management prescriptions. For example, if there are large error margins and the average values are adopted by managers, hollows may be thought to occur in trees retained in harvested areas when they do not. It is therefore important to select an appropriate method for estimating tree age and to assess the accuracy of the technique used. Ring counting relies on the assumption that seasonal variation in growth conditions affects the density or size of the cells accumulated. Darker latewood bands can form when cold temperatures or moisture deficit cause seasonal periods of slower growth (Pilcher and Gray 1982; Leal et al. 2004), but can also result from defoliation by grazers and ire (Mazanec 1968; Mucha 1979). Consequently, their production is not always strictly annual, with some rings being locally absent, or ‘false’ rings being produced. The rate of false ring production can vary with the dominance status of the tree (Mucha 1979; Brookhouse 1997; Masiokas and Villaba 2004; Bar et al. 2006). Locally absent or ‘missing’ rings occur most frequently on trees that are old, suppressed, have poor crown development and in areas where environmental stresses are high (Brookhouse 1997; Lorimer et al. 1999; Jonsson et al. 2002; Bar et al. 2006; Waring and O’Hara 2006; Mayield et al. 2007; but see Rozas 2003). The annualarity of tree rings is usually assessed by cross-dating patterns of tree