Quantifying biomass, median body size, and biodiversity
With these videos, we identified and characterised protist species in the communities using the R-package BEMOVI (Altermatt et al., 2015; Pennekamp et al., 2015; Pennekamp & Schtickzelle, 2013). We first extracted moving particles’ traits (e.g., speed, shape, size) in the videos and used such traits to exclude particles that were not protists to obtain an average abundance of protist individuals per volume. We used protist biomass as the focal ecosystem function. Specifically, we measured the total area of protists (as area per volume medium) and used this “bioarea” as a proxy of biomass (hereafter referred to as “biomass”)–a fair assumption given that protists have a roundish shape (see also other work using the same proxy; e.g., Jacquet et al., 2020; Pennekamp et al., 2018). The area of protists gave us a metric of body size, from which we calculated the median body size of individuals in each patch at each time point (the median gives an average metric robust to different types of distributions). We then used a support vector machine model to identify protist species (Cortes et al., 1995; r-package “e1071”: Dimitriadou et al., 2006), using traits in species monocultures as predictor variables. Finally, we calculated patch biodiversity using the Shannon Index to account for relative abundances.