Harnessing remote sensing data combinations for terrestrial wildlife modelling: A systematic review of image-based surveys

Remote sensing technologies are transforming the way ecologists collect data and expanding the possibilities for wildlife monitoring. In particular, the growing range of remote sensing image data is contributing to advances in modelling approaches. Combining multiple remote sensing data streams can provide opportunities to expand our understanding of the ecological processes represented in modelling. To further realise the potential of remote sensing data combination for wildlife monitoring, we must first understand the current applications. Here we present a systematic review that investigates how remote sensing image data are being applied to terrestrial wildlife population models in combination with other remote sensing technologies, and interrogates the processing, aggregation and scaling approaches employed in such models. We reviewed 125 articles published from 2014 to 2024 in the Scopus or Web of Science database. We found that camera trap data was most often combined with low and medium resolution satellite-derived spatial datasets. Higher resolution satellite products and fine-scale data from technologies such as uncrewed aerial vehicles (UAVs or drones) and light detection and ranging (LiDAR) sensors were less prevalent. We identified a continuing bias towards medium and large sized mammals, particularly ungulates and carnivores. Surprisingly there was almost no use of machine learning to automate the wildlife classification process when remote image data was used for terrestrial wildlife modelling. There is scope for ecologists to consider whether other combinations of data could increase the insights available from terrestrial wildlife population models, particularly in relation to fine-scale habitat structure. There is also potential for more rigorous consideration of scaling in terrestrial wildlife models. In particular, we recommend greater attention to the relationship between data scale and the processes that are represented.