Biodiversity offsets are an international emerging impact assessment tool, attempting to bridge the gap between biodiversity conservation and sustainable economic development. Offsets shall compensate for unavoidable ecological damage after appropriate prevention and mitigation measures have been taken so that there is no net loss and ultimately a net gain for biodiversity near the impact site. Worldwide, ecologists are being challenged in choosing the most appropriate ecosystem attributes for use in biodiversity offset models. Attributes ought to represent the key biodiversity features at a given site, be quantifiable, easy to measure, reliable, and sensitive to management actions. However, biodiversity is complex and not easy to describe or measure, especially in the context of offsetting. Determining which attributes are the most appropriate for this task is currently compromised by the lack of a theoretical framework. To ensure that offsetting does result in genuine biodiversity retention, attribute choice has to be based on a sound scientific basis.

To help establish such a foundation, this thesis first suggests a conceptual framework for attribute selection in forest ecosystems. Then ecosystem attributes commonly applied or suggested for the assessment of restoration success in forests are reviewed and a set that appears to be most suitable for application in biodiversity offsets is identified. Second, the performance of vegetation related attributes in terms of their predictability and information content are tested in a New Zealand restoration project using a chronosequence approach. Third, the surrogacy value of these vegetation measures for other species groups and ecosystem function is assessed. In particular, how well the recovery of aboveground attributes can predict the restoration of belowground attributes is assessed. This is critical, as typically the largest amount of site biodiversity occurs below-ground. Finally, a general set of attributes that will be applicable in most forest types is identified for biodiversity offset models. In addition, further recommendations for attribute selection within offsets models and how to manage uncertainty associated with them are given.

Results of this thesis suggest that: (i) Structural elements such as basal area and mean diameter are the most predictable attributes, providing important information about the successional development of forests. (ii) Compositional measures can be less predictable but provide the highest information content. Predictability of these measures can be optimised if early to mid-successional stages are used as a restoration target and if restoration includes active management such as planting. (iii) Vegetation measures do not correlate well with the recovery of belowground species groups, but further research is necessary to confirm this. (iv) This research emphasises that the re-creation of old growth forest attributes can take several centuries when starting from early successional stages. It might be also surrounded by a high uncertainty in respect to compositional development; in particular, when passive reestablishment of vegetation is applied as a restoration tool. Therefore, achieving a no net loss of biodiversity as required by biodiversity offsets might, in many cases, be doubtful when offsetting for the loss of old growth forest habitats.