Abstract
1. Animals are important vectors for transporting seeds, nutrients and microbes across landscapes. However, models that quantify the magnitude of these ecosystem services across a broad range of taxa often rely on generalised mass-based scaling parameters for gut passage time. This relationship is weak and fundamentally breaks down when considering individual species, indicating that current models may incorrectly attribute or estimate the magnitude of dispersal.
2. We collated a large dataset of gut passage time for endothermic animals measured using undigested markers (n = 391 species). For each species, we compiled trait data, including body mass, morphology, gut physiology, diet and phylogeny. We then compared the ability of five statistical models (constant, generalised least squares, phylogenetic generalised least squares, general linear model and random forest) to estimate the time of first marker appearance (transit time; TT) and mean marker retention time (MRT) for particle and solute markers in mammals and birds separately.
3. For mammals, we found that the inclusion of additional traits appreciably reduced the median root-mean
squared error across all markers in a leave-one-out cross validation. For birds, however, additional traits did not significantly improve our ability to predict gut passage time across markers. This may have occurred due to the smaller number of bird species included in our analysis or the absence of important explanatory factors such as differences in gastrointestinal morphology.
4. Using the MRTparticle random forest model from this study, we updated two trait-based dispersal models for seed and nutrient movement by mammals. The magnitude of dispersal in our updated predictions ranged from 66% to 176% of the original model formulation for different scenarios, highlighting the importance of gut passage time for dispersal models. Furthermore, the contribution by individual or groups of species was found sizeably altered in our updated models.
5. Future modelling studies of dispersal by mammals, for which empirical estimates of gut passage time are absent, will benefit from predicting gut passage time using statistical models that incorporate traits including animal morphology, diet and gut physiology.
Abraham, Andrew J