Estimating Alpha, Beta, and Gamma Diversity Through Deep Learning

Abstract

The reliable mapping of species richness is a crucial step for the identification of areas of high conservation priority, alongside other value and threat considerations. This is commonly done by overlapping range maps of individual species, which requires dense availability of occurrence data or relies on assumptions about the presence of species in unsampled areas deemed suitable by environmental niche models. Here, we present a deep learning approach that directly estimates species richness, skipping the step of estimating individual species ranges. We train a neural network model based on species lists from inventory plots, which provide ground truth data for supervised machine learning. The model learns to predict species richness based on spatially associated variables, including climatic and geographic predictors, as well as counts of available species records from online databases. We assess the empirical utility of our approach by producing independently verifiable maps of alpha, beta, and gamma plant diversity at high spatial resolutions for Australia, a continent with highly heterogeneous diversity patterns. Our deep learning framework provides a powerful and flexible new approach for estimating biodiversity patterns, constituting a step forward toward automated biodiversity assessments.

Global patterns of vascular plant alpha diversity

“…Finally, high species richness at fine grains might also depend on plant size, as many small plants can coexist in a given grain size. Such conditions mainly occur in grasslands, e.g., in Eastern Australia, where this mechanism has been invoked to explain differences in beta diversity among vegetation types 56 .…”

Deep Learning from Phylogenies for Diversification Analyses

“…This definition is the one used in the present article. Applying a regression task with deep learning (learning model parameter values from simulations) is increasingly used in several fields including population genetics (Sheehan and Song 2016; Sanchez et al 2020; Avecilla et al 2022), phylogenetic reconstruction (Nesterenko et al 2022), macroecology (Andermann et al 2022) and physiology (Kroll et al 2021). In macroevolution, an early progress of using machine learning (Bokma 2006) consisted in training an artificial neural network on the axes of a principal component analysis of phylogenetic branching times to infer speciation and extinction rates.…”

Patterns and drivers of plant diversity across Australia

“…This could provide new insight into the biogeographic processes responsible for plant diversity patterns in Australia, as well as improved spatial predictions of diversity patterns, potentially harnessing geographically weighted modelling techniques (Fotheringham et al 2002, Alves et al 2018). Machine learning modelling approaches could also be worth exploring for improving the accuracy of diversity models for Australian plants, though these approaches may reduce interpretability and potentially increase extrapolation error in environments where plot coverage is poor (Elith et al 2008, Andermann et al 2022). It is also important to note that our exploration of potential environmental drivers of plant community diversity patterns harnessed spatial environmental layers that have their own bias, errors and uncertainty associated with them (Xu and Hutchinson 2010, Storlie et al 2013).…”

“…To date, Australia has had no national, standardised plant community survey dataset that combines available data from the state and territory agencies (Gellie et al 2018). Plot‐based analyses of plant diversity patterns in Australia have, therefore, been restricted to particular regions (Austin et al 1996, Hunter 2005, Guerin et al 2013, Mokany et al 2014, McCarthy et al 2018) or have utilised a relatively small number of survey plots (Rice and Westoby 1983, Andermann et al 2022). The TERN AusPlots initiative has surveyed ca 870 plots across Australia to date using standard methods; however, these data still have substantial spatial, environmental and taxonomic gaps (Guerin et al 2021).…”