Urbanization has clear and significant effects on biodiversity, leading to habitat loss and degradation, homogenisation, and introduction of exotic species. As the global population increases and urbanization continues to threaten natural ecosystems, it becomes more vital to mitigate biodiversity loss by making well informed management decisions. A tool that is being used increasingly to understand how the land cover affects species, thereby assisting these management decisions, is the joint species distribution model. Joint species distribution models allow for species- and community level inferences; however, the ecological data they take is prone to biases due to variation in detectability of the species or environment. This can lead to inaccurate inferences on the drivers of occupancy and significant underestimations of species richness, both of which can hinder efforts to maintain biodiversity as urbanization and other global change drivers take effect. Nonetheless, few studies consider imperfect detection when analysing urban ecological data.

Here, I conduct a historical literature review of the development of joint species distribution models to give context to the current state of this methodology, showing where each component of the model has its origin and when each line of research has converged. Following this, I develop a framework that accounts for imperfect detection in joint species distribution models, as well as providing a stepby- step user guide for future reference and to aid in usability of the framework described, using Just Another Gibbs Sampler (JAGS) and R. Finally, I use the developed framework in the context of bird and butterfly communities in Singapore with the aim of understanding how urbanization and detectability affect both bird and butterfly occupancy. By doing this I fill the gap in accounting for imperfect detection in joint species distribution models in urban ecosystems. I make comparisons between my analysis and models that overlook detectability by using information criteria, and I compare occupancy and detection responses of birds and butterflies to land cover variables including various vegetation types and traffic density. In doing so, I found that accounting for detectability increased prediction accuracy of the model, and that when detectability was overlooked while making species richness predictions, the values were considerably lower than when detectability was accounted for. This work shows the importance of accounting for detectability when making predictions of species richness, and it describes the species– and community–environment relationships that can guide effective conservation decisions in urban ecosystems. Using this framework can help with the identification of effective indicator species, as well as guide urban development to minimise loss of essential habitat and consequently urban biodiversity.