The MESOSCAN dataset: a collection of mesofauna images from UK forest and peatland soils

Mesofauna such as Acari and Collembola represent a key bioindicator of soil health and condition, but limited availability of taxonomic expertise and the time-intensive nature of their identification constrain our ability to characterize their community structure and function in forest and peatland ecosystems. As these limitations stem largely from challenges in species identification, novel machine learning (ML) approaches offer promising tools to address them, but in turn require substantive collections of annotated images. Recent large-scale datasets have limited mesofauna images, and to address this paucity in training data we have developed the complementary MESOSCAN dataset with samples taken from UK forest and peatland soils, using an optimized macro-photography system to capture high-resolution and multi-orientation images of specimens to provide broad representation of the main UK mesofauna groups. We validate this resulting dataset by showing how it can be used to construct a simple baseline ML model to confirm good predictive accuracy.