The structure of the end-Cretaceous dinosaur fossil record in North America.

Whether non-avian dinosaurs were in decline prior to their extinction 66 million years ago remains a contentious topic. This uncertainty arises from spatiotemporal sampling inconsistency and data absence, which cause challenges in distinguishing between genuine biological trends and sampling artifacts. Consequently, there is an inherent interest in better quantifying the quality of the data and concomitant biases of the dinosaur fossil record. To elucidate the structure of this record and the nature of the biases impacting it, we integrate paleoclimatic, geographic, and fossil data within a Bayesian occupancy modeling framework to simultaneously estimate the probability of dinosaurs occupying and being detected in sites across North America throughout the latest Cretaceous for the first time. We find that apparent declines in occupancy generated from the raw fossil record do not match modeled occupancy probability, which generally remained stable throughout the latest Cretaceous. Instead, they coincide with decreased probability of detecting dinosaur occurrences, despite high overall sampling during this interval. By incorporating model covariates, we additionally reveal that detection probability is directly and significantly influenced by the available area of geological outcrop and modern land cover. Our findings offer evidence that traditional comparisons of diversity estimates between time intervals are likely inaccurate due to underlying structural issues in the geological record operating at both local and regional scales. This study underscores the utility of occupancy modeling as a novel approach in paleobiology for quantifying the impact of heterogeneous sampling on the available fossil record.