LAUSR

Spatiotemporal predictions of ecological phenomena are highly useful and significant in scientific and socio-economic applications. However, the inadequate availability of ecological time-series data often impedes the development of statistical predictions. On the other hand, considerable amounts of temporally discrete biological records (commonly known as ‘species occurrence records’) are being stored in public databases, and often include the location and date of the observation. In this paper, we describe an approach to develop spatiotemporal predictions based on the dates and locations found in species occurrence records. The approach is based on ‘time-series classification’, a field of machine learning, and consists of applying a machine-learning algorithm to classify between time series representing the environmental variation that precedes the occurrence records and time series representing the full range of environmental variation that is available in the location of the records. We exemplify the application of the approach for predicting the timing of emergence of fruiting bodies of two mushroom species (Boletus edulis and Macrolepiota procera) in Europe, from 2009 to 2015. Predictions made from this approach were superior to those provided by a ‘null’ model representing the average seasonality of the species. Given the increased availability and information contained in species occurrence records, particularly those supplemented with photographs, the range of environmental events that could be possible to predict using this approach is vast.