Are stacked species distribution models accurate at predicting multiple levels of diversity along a rainfall gradient?

We use observed patterns of species richness and composition of ant communities along a 1000 mm rainfall gradient in northern Australian savanna to assess the accuracy of species richness and turnover predictions derived from stacked species distribution models (S‐SDMs) and constrained by macroecological models (MEMs). We systematically sampled ants at 15 sites at 50 km intervals along the rainfall gradient in 2012 and 2013. Using the observed data, we created MEMs of species richness, composition and turnover. We built distribution models for 135 of the observed species using data from museum collections and online databases. We compared two approaches of stacking SDMs and three modelling algorithms to identify the most accurate way of predicting richness and composition. We then applied the same beta diversity metrics to compare the observed versus predicted patterns. Stacked SDMs consistently over‐predicted local species richness, and there was a mismatch between the observed pattern of richness estimated from the MEM, and the pattern predicted by S‐SDMs. The most accurate richness and turnover predictions occurred when the stacked models were rank‐ordered by their habitat suitability and constrained by the observed MEM richness predictions. In contrast with species richness, the predictions obtained by the MEM of community similarity, composition and turnover matched those predicted by the S‐SDMs. S‐SDMs regulated by MEMs may therefore be a useful tool in predicting compositional patterns despite being unreliable estimators of species richness. Our results highlight that the choice of species distribution model, the stacking method used, and underlying macroecological patterns all influence the accuracy of community assembly predictions derived from S‐SDMS.     

Do species distribution models explain spatial structure within tree species ranges?

Aim To evaluate the ability of species distribution models (SDMs) to predict the spatial structure of tree species within their geographical ranges (how trees are distributed within their ranges). Location Continental Spain. Methods We used an extensive dataset consisting of c. 90,000 plots (1 plot km^?2) where presence/absence data for 23 common Mediterranean and Atlantic tree species had been surveyed. We first generated SDMs relating the presence or absence of each species to a set of 16 environmental predictors, following a stepwise modelling process based on maximum likelihood methods. Superimposing spatial correlograms generated from the predictions of the SDMs over those generated from the raw data allowed a model–observation comparison of the nature, scale and intensity (level of aggregation) of spatial structure with the species ranges. Results SDMs predicted accurately the nature and scale of the spatial structure of trees. However, for most species, the observed intensity of spatial structure (level of aggregation of species in space) was substantially greater than that predicted by the SDMs. On average, the intensity of spatial aggregation was twice that predicted by SDMs. In addition, we also found a negative correlation between intensity of aggregation and species range size. Main conclusions Standard SDM predictions of spatial structure patterns differ among species. SDMs are apparently able to reproduce both the scale and intensity of species spatial structure within their ranges. However, one or more missing processes not included in SDMs results in species being substantially more aggregated in space than can be captured by the SDMs. This result adds to recent calls for a new generation of more biologically realistic SDMs. In particular, future SDMs should incorporate ecological processes that are likely to increase the intensity of spatial aggregation, such as source–sink dynamics, fine‐scale environmental heterogeneity and disequilibrium.     

Predicting spread of invasive macrophytes in New Zealand lakes using indirect measures of human accessibility

1. Models predicting invasive macrophyte spread between lakes provide an important tool for focusing proactive management efforts to lakes deemed susceptible to invasion. However, challenges to forecasting macrophyte spread include wide physiological tolerances of invasive macrophytes and a lack of information on the relative importance of the various human vectors (e.g. boating traffic). In New Zealand, three invasive species that reproduce vegetatively, Ceratophyllum demersum, Lagarosiphon major, Egeria densa, and a single species that reproduces sexually, Utricularia gibba, are currently spreading across the lake landscape at a great cost to the local ecology and economy. 2. In this study, we first examined whether variables that indirectly describe weed spread via human access and use, as well as a lake’s position in the landscape, could describe the distribution of these four species using a boosted regression trees (BRT) modelling approach. Then, as these invasive species have not reached their full invasion potential, we examined how giving more influence to infected lakes at the edge of the invasion front, and including all lakes across New Zealand as background samples, simulating ‘absences beyond the invasion front’, influenced our ability to forecast the potential for new lakes to be invaded. 3. The BRT models identified that variables characterising human access and use, as well as lake position, were associated with the occurrence of the three vegetatively reproducing macrophytes. Weed occurrence was more likely when there was a highway in the vicinity, human population density was high and if the lake was large (c. 55 km²). But in the single case of U. gibba, temperature was the variable that best explained occurrence. This is consistent with the suggestion that U. gibba is predominantly dispersed by waterbirds, rather than human activity. 4. But for all four species, the BRT models based on the recorded observations alone predicted observed invasions with low prediction probabilities and did not forecast further spread. By contrast, when observations at the edge of the invasion front were upweighted, and additional background lakes implemented into the model, recorded observations were predicted and additional lakes were forecast to be at risk, suggesting that these models better captured the current and potential distribution of these macrophyte species. 5. The use of variables that characterise weed spread could provide similar insights into other systems where survey information on the nature, strength and direction of invasion vectors is lacking. Furthermore, when weighting the data, many lakes across New Zealand were forecasted to be at risk of invasion. The advantage of weighing the presence data was that insights into the potential for a species to spread were obtained. The probabilistic estimates of risk, as derived from the models, together with other information for prioritising lakes, can be used to focus surveillance and protection efforts.