Reducing the loss of biodiversity is key to ensure the future well being of the planet. Indicators to measure the state of biodiversity should come from primary data that are collected using consistent field methods across several sites, longitudinal, and derived using sound statistical methods that correct for observation/detection bias. In this paper we analyze camera trap data collected between 2008 and 2012 at a site in Costa Rica (Volcan Barva transect) as part of an ongoing tropical forest global monitoring network (Tropical Ecology Assessment and Monitoring Network). We estimated occupancy dynamics for 13 species of mammals, using a hierarchical modeling approach. We calculated detection-corrected species richness and the Wildlife Picture Index, a promising new indicator derived from camera trap data that measures changes in biodiversity from the occupancy estimates of individual species. Our results show that 3 out of 13 species showed significant declines in occupancy over 5 years (lowland paca, Central American agouti, nine-banded armadillo). We hypothesize that hunting, competition and/or increased predation for paca and agouti might explain these patterns. Species richness and the Wildlife Picture Index are relatively stable at the site, but small herbivores that are hunted showed a decline in diversity of about 25%. We demonstrate the usefulness of longitudinal camera trap deployments coupled with modern statistical methods and advocate for the use of this approach in monitoring and developing global and national indicators for biodiversity change.