Critical habitat is defined scientifically as the subset of habitat necessary for the long-term persistence of a given species. Based on this definition, loss of any part of the critical habitat would result in extinction of the species. In the United States, Australia, and Canada, critical habitat of threatened species is protected, to various degrees, under endangered species legislation. Effective protection of critical habitat depends on it being identified accurately. Where there is potential for conflict with landowners and industry stakeholders, accurate critical habitat identifications are more defensible in court, and minimise the opportunity costs of protecting areas that may not be as beneficial to species persistence. However, obtaining the data required to accurately identify critical habitat can take up considerable time and resources that may otherwise be spent on conservation actions. At the same time, delaying protection of critical habitat to improve knowledge can result in further habitat loss. In this thesis, I review key concepts and challenges surrounding the identification of critical habitat, and develop decision tools to assist in deciding when and how to identify it.In Chapter 2, I present a systematic review of critical habitat documentation from the United States, Canada, and Australia to identify the types of data and criteria that have been used to identify critical habitat in the last decade. Contrary to scientific recommendations that long-term species persistence should be used as the criterion for identifying critical habitat, information about the location of species occurrences and particular habitat features were used instead to identify critical habitats for most of the species reviewed. Insufficient data and the desire to avoid potential opposition from landowners are likely to be the main reasons for the use of such approaches to critical habitat identification. Chapter 3 continues with an examination of the merits of the different criteria that could be used to inform critical habitat identification, and the types of errors associated with each. I also considered the potential consequences of the errors, and recommended that more explicit recognition of the potential for errors is important in minimising their negative consequences for species persistence.While the accuracy of critical habitat identification may be improved by collecting more data, delaying protection to do so may result in additional habitat loss if habitats are left unprotected in the meantime. In Chapter 4, I used an optimisation approach to examine this trade-off between the benefits of delaying protection to improve accuracy and the costs of additional habitat loss in the interim. I modelled the change in the proportion of habitat correctly identified over time as a function of both accuracy and habitat loss, and determined the optimal amount of time to spend iii learning that maximises this value. I found that at low rates of habitat loss, slow learning rates resulted in a longer optimal lea...