Mechanistic models project bird invasions with accuracy

Abstract

Invasive species pose a major threat to biodiversity and inflict massive economic costs. Effective management of bio-invasions depends on reliable predictions of areas at risk of invasion, as they allow early invader detection and rapid responses. Yet, considerable uncertainty remains as to how to predict best potential invasive distribution ranges. Using a set of mainly (sub)tropical birds introduced to Europe, we show that the true extent of the geographical area at risk of invasion can accurately be determined by using ecophysiological mechanistic models that quantify species' fundamental thermal niches. Potential invasive ranges are primarily constrained by functional traits related to body allometry and body temperature, metabolic rates, and feather insulation. Given their capacity to identify tolerable climates outside of contemporary realized species niches, mechanistic predictions are well suited for informing effective policy and management aimed at preventing the escalating impacts of invasive species.

Fig. 1. Relationships between realized and fundamental niches and the geographical areas at risk of invasion by introduced species.

Native-area species distribution models are based on species’ contemporary occurrences and may underpredict the area at risk of invasion (yellow areas). Purple dot-arrows illustrate an invasive species invading and spreading beyond the areas predicted as suitable based on the climates they occupy in their native range. Mechanistic ecophysiological models, in contrast, quantify species’ fundamental thermal niches by integrating their physiology, morphology and behavior with the microclimates they experience, predicting wider areas at risk of invasion (olive-green).

Fig. 2. Accuracy of invasion risk predictions.

Accuracy of correlative (light gray) versus mechanistic (dark gray) models in correctly predicting a invasive presences (‘sensitivity’) and b locations currently uninvaded (‘specificity’) by introduced birds across Europe (n = 20 introduced bird species). Each small black dot (jittered for visibility) denotes an introduced bird species. Boxplots indicate mean (large black dot) and interquartile range (i.e., 25th percentile–75th percentile), with whiskers corresponding to 1.5 times the interquartile range. Split-half violin plots visualize the probability density of the data at different accuracy values. Top gray lines indicate significant differences between correlative and mechanistic models as revealed by post-hoc Tukey tests. Correlative models: GLM (binomial Generalized Linear Models), BART (Bayesian Additive Regression Trees), and FNE (Fundamental Niche Ellipses). Mechanistic predictions were obtained through the NicheMapper platform using either species-level parameter estimates (NM_(sp)) or allowing for intraspecific variation (NM_(intra)).

Fig. 3. Forecasts of invasion risk.

Areas at risk of invasion by introduced a blue-crowned parakeets Thectocercus acuticaudatus (body mass of ~170 g) and b common waxbills Estrilda astrild (~9 g) across Europe. Gray indicates areas predicted to be climatically unsuitable while black areas are at risk of invasion according to correlative (GLM: generalized linear model, BART: Bayesian additive regression trees, FNE: fundamental niche ellipse) versus mechanistic (species-level NicheMapper) models. Red dots represent current invasive occurrences, used to independently evaluate model forecasts, expressed as the percentage of correctly predicted occurrences (sensitivity). Thresholds discriminating suitable versus unsuitable area are based on a 5% native-range omission rate for the correlative models and on a 4.6 times the basal metabolic rate limit for the mechanistic model. Response curves in the upper panel show how climate gradients (summarized into principal component axes) drive predicted suitability, while gray and black horizontal bars show the range of each climate axis in the native and invasive area, respectively. Estimated ellipses in the middle panel (FNE) represent 25, 75, and 95% confidence regions of the modeled fundamental niche, with the star as the niche center. Yellow points represent the climate conditions accessible to the species in the native range (‘M hypothesis’), blue dots the available native-range occurrences. Gray points indicate climates across Europe, red dots the invasive occurrences. In the lower panel, the black polygons are the species’ native ranges according to BirdLife, the blue dots are occurrences obtained through GBIF, and the yellow regions are the geographical backgrounds (‘M’) used to train the correlative models.

Fig. 4. Key biophysical traits.

Significant relationships between biophysical traits and mechanistic model predictive accuracy. Model performance is chiefly governed by a small number of key biophysical traits related to species’ capacity to generate and retain heat. Each dot represents a bird species, and data are slightly jittered along the x-axis for visibility. Solid black lines represent the mean estimate, gray shading indicates the 95% confidence interval of significant relationships.