Crown age estimation of a monocotyledonous tree species Dracaena cinnabari using logistic regression

Unique woodlands of Dracaena cinnabari (DC) are at risk throughout most of their range (Socotra Island, Yemen) as a result of missing regeneration and overmaturity. Effective conservation measures depend on reliable predictions of future population dynamics, which depend on accurate data on current age structure. However, age determination of Dracaena sp. has long been a scientific challenge, because the common method of tree ring counts cannot be applied to this or to most other monocotyledonous trees. In the present study, the indirect method for crown age estimation proposed by Adolt and Pavlis (Trees 18:43–53, 2004) was further developed using a more appropriate statistical technique and an intuitive model formulation. This new technique is based on the relationship between the number of branching orders and the number of flowering events that result from a specific growth pattern. We used logistic regression to directly model annual flowering probability, the reciprocal value of which corresponds to the length of the interval between flowering events. Our methodology was applied to data sets collected at two ecologically distinct sites. In Firmihin, the time between flowering events decreases from 28?years between the first and second event to 10?years between the 25th and 26th event. The length of time between flower events in Skant, however, was estimated to be a constant value of 6.5?years. We propose the application of generalised mixed-effects models and methods of survey sampling to improve the accuracy of crown age estimation in DC. Our methodology may also be useful for age estimations of other tree species with similar growth patterns, such as Dracaena draco and Aloe dichotoma.