A practical guide to species trend detection with unstructured data using local frequency scaling (Frescalo)

Romain Goury, Diana E. Bowler, Colin Harrower, Tamara Münkemüller, Jeanne Vallet, Jon Yearsley, Wilfried Thuiller, Oliver L. Pescott 

Accurately measuring biodiversity change remains a central challenge in ecology. Beyond the general idea of detection frameworks, which can help to estimate species trends under variable effort, other sampling-related biases in data collection remain a key challenge.


Long-term standardized ecological data are rare, and most available datasets exhibit considerable spatial and temporal variation in sampling effort (i.e., unstructured data). To derive reliable, unbiased estimates of biodiversity trends and to better understand the drivers of change, modelling approaches are likely to be essential.


Among the available methods, the local frequency scaling approach (Frescalo; Hill, 2012) has proven particularly effective at addressing these biases. By applying successive spatial and temporal corrections, Frescalo leverages emergent patterns in species assemblages to correct for variation in survey effort. Compared to other similar approaches, Frescalo is particularly well suited to long-term datasets and those with a high number of species. It is also a versatile method, allowing simultaneous estimation of temporal and spatial changes, or even providing diagnostics for survey design or bias assessment.


The method’s technical complexity, the level of ecological knowledge required, and the challenges of implementation raise a number of practical issues in the application of the method. In this paper, we present a clear and accessible explanation of the Frescalo methodology, offer a step-by-step roadmap to guide users, and highlight the wide range of applications it supports. To further facilitate its adoption, we also introduce an R package designed to simplify implementation.