A quantitative review of abundance-based species distribution models

Abstract

The contributions of species to ecosystem functions or services depend not only on their presence in a given community, but also on their local abundance. Progress in predictive spatial modelling has largely focused on species occurrence, rather than abundance. As such, limited guidance exists on the most reliable methods to explain and predict spatial variation in abundance. We analysed the performance of 68 abundance-based species distribution models fitted to 800,000 standardised abundance records for more than 800 terrestrial bird and reef fish species. We found high heterogeneity in performance of abundance-based models. While many models performed poorly, a subset of models consistently reconstructed range-wide abundance patterns. The best predictions were obtained using random forests for frequently encountered and abundant species, and for predictions within the same environmental domain as model calibration. Extending predictions of species abundance outside of the environmental conditions used in model training generated poor predictions. Thus, interpolation of abundances between observations can help improve understanding of spatial abundance patterns, but extrapolated predictions of abundance, e.g. under climate change, have a much greater uncertainty. Our synthesis provides a roadmap for modelling abundance patterns, a key property of species' distributions that underpins theoretical and applied questions in ecology and conservation.

Predict first–assemble later versus assemble first–predict later: Revisiting the dilemma for functional biogeography

“…This was true for our case studies on plants across large environmental gradients (Supporting Information S2), but also for our simulation experiments, highlighting that these results are likely to be quite general. Random Forest algorithms are widely recognized as a common and effective method for ecological predictions (Beaumont et al, 2016; Elith et al, 2006; Kosicki, 2020; Mi et al, 2017; Waldock et al, 2021). While acknowledging their established performance, we also explored the application of GAM models, as they are frequently employed in ecological research (see Supporting Information S3).…”

“…Using massive set of plant community (4463 grassland plots) and trait data (species-leaf-area, leaf-nitrogen-content, and plant-height for >800 species) covering the whole French Alps, we compared the two approaches for predicting community-(weighted) mean traits (CWM/ CM) and (un)weighted functional dispersion (FDis/uFDis), accounting or not for abundance, and tested their interpolation versus extrapolation capabilities. We employed Random Forests for the species and trait indices predictors, as it is a widely used statistical model for this type of data (Ahmed et al, 2021;Elith et al, 2006;Hill et al, 2017;Kosicki, 2020;Marchi et al, 2016;Waldock et al, 2021) and we indeed found it was performing better than other tested predictors (see Supporting Information S3). To validate our findings on the empirical data, we also carried out the same comparative analyses on four simulated data replicates, where species-environment relationships were not affected by unknown confounding factors that could happen with in-situ data.…”

Stability of crop pollinator occurrence is influenced by bee community composition

“…Furthermore, whilst there is known to be a positive relationship between species occupancy and abundance (Holt et al, 2002), this may not be the case for all species and all contexts, and may also depend on the scale at which occupancy is considered. We also looked specifically at bee species occupancy, but ecosystem service function depends not only on species occurrence, but also their local abundance (Waldock et al, 2021). We make the assumption here that pollinator occurrence and abundance are closely related, based upon prior evidence that inter-annual changes in citizen science collected distribution records are a reasonable proxy for interannual changes in abundance (Mason et al, 2018).…”