Periodic Reporting for period 1 - FREYA (Forecasting RangE dYnamics of Alien species under climate change.)
Reporting period: 2016-06-01 to 2018-05-31
Humans are redistributing the world’s biodiversity both indirectly by changing habitats and the climate, and directly by transporting species across natural biogeographical boundaries. Predicting how species will respond to such changes is paramount for formulating effective conservation measures aimed at protecting native biodiversity, economy and human health. Indeed, introduced species can become invasive, and be a main threat to biodiversity and also impose large costs to the economy. Unfortunately, our capacity to accurately predict under which environmental conditions that introduced species will become invasive remains limited. This may be at least partly due to the fact that forecasts of invasion risk typically result from relatively simple correlative models, which integrate information on where a species currently occurs with the dominant climate at these occurrence locations. While this approach generally works well when making forecasts within the area where a species is native, correlations between climate and species occurrence tend to break down when attempting predictions to other areas and time-frames – i.e. when trying to forecast invasion risk under changing climates. Robust predictions of invasion risk may be formulated based on more fundamental, physiologically-informed distribution models. Such mechanistic models do not rely on correlations but use species’ functional traits to characterize the range of environmental conditions tolerable by species. Yet, both real and perceived difficulties in obtaining sufficiently detailed information on species’ ecology and life-history have held back research success in this area. Therefore, FREYA had two overarching Objectives, namely (1) to evaluate predictions of range dynamics derived from correlative versus mechanistic niche modelling techniques, using a multi-species approach, and (2) to assess how ecological and evolutionary processes influence forecasts of invasion risk, using a well-known avian invader as model species. The first objective focusses on a large number of non-native invasive birds while for the second objective, the project concentrates on the ring-necked parakeet. Main conclusions are that overall, mechanistic-physiological model predictive accuracy was moderate to low, as our invasion risk forecasts were prone to both omission and commission errors. Sensitivity analyses revealed a set of key functional traits strongly influencing model accuracy. For comparatively larger avian invaders, estimates of basal metabolic rates, body temperate and body mass are crucial while for smaller birds, feather characteristics such as feather length and plumage depth are important as well. The research on ring-necked parakeets showed that introduced populations were of predominantly Asian ancestry, with differentiation of African native populations occurring through historical evolutionary processes. Within Europe, we identified linear correlations between allele frequencies and environmental variables across a climate spectrum, suggesting rapid selection in response to climatic change within introduced ring-necked parakeet populations, further complicating the modelling of invasion risk. Indeed, mechanistic-physiological models based on European parakeet data (as opposed to native range, Asian data) result in generally more accurate predictions of the current European distribution of this prominent invader. Together, results obtained indicate that complex and parameter hungry mechanistic modelling approaches such as the one applied here may be better suited to uncover processes driving species invasions, rather than for obtaining highly accurate spatial predictions of where invaders are likely to establish and spread.
The work performed can be subdivided into (1) collection of occurrence data on invasive bird species. Data was gathered from wide range of sources to optimally characterize species distributions both in their native and invasive areas. The bulk of the data is integrated from online open-access repositories such as GBIF and eBird, yet much effort was invested in reviewing both grey and published literature in order to collect additional occurrence data and verify the reliability of the data (i.e. do reported data indicate an established population or temporary, local escapes); (2) collecting species physiological and life-history data. Much information needed for physiological modelling is not readily available, and has therefore been measured on a set of museum specimens available from the Royal Belgian Institute of Natural Science. This includes data on bird feather lengths and depths, feather solar reflectivity (to quantify the amount of heat captured from the sun), and the lengths and widths of bird head, neck, torso and legs. Data that could not be collected empirically were obtained through a wide-reaching literature review. This includes key physiological data such as metabolic rates, metabolic activity multipliers, body fat percentages and life-history data such timing of breeding and moulting; (3) creating open-access modelling scripts. Physiological niche models can be run using the patented NicheMapper software, and while this software package is indeed elegant and flexible, it does operate as a difficult to handle ‘black box’ (i.e. a set of Fortran codes embedded in Windows.exe files). Much time and effort was invested in creating a set of R scripts (R is the most commonly used statistical modelling framework in ecology), allowing to flexible and transparently run mechanistic models from within R; (4) actual modelling of species. Very computer-intensive analyses have been conducted, mainly to carry out sensitivity analyses aimed at finding out (a) which variables that are most important for mechanistic-physiological models and (b) to establish optimal modelling strategies (e.g. selection of background area to correctly evaluate model performance); (5) for ring-necked parakeets, based on information from genomic analyses indicated possible rapid adaptation to colder European climates, bespoke mechanistic models using European parakeet data were implemented.
Mechanistic niche models are often proposed as a promising avenue for obtaining reliable forecasts of species distributions under changing climates and for invasive species colonizing novel areas. Several applications of this methodology have indeed suggested that predictions based on physiology may be able to help where traditional, easier-to-implement correlative models fail. We have confirmed that physiological models can indeed adequately characterize the set of environmental conditions under which species can persist, yet we show that it is very difficult to correctly quantify or estimate relevant parameter values for the large number of variables needed by these methods. Using a thorough sensitivity analyses, we found that the parameter space is vast and wide. Even a conservative range estimates for key parameter estimates can result in widely varying model predictions on invasion risk. This finding indicates that (1) mechanistic models are able to supplement, not replace, correlative models of invasion risk and offer a way forward to explicitly test which fundamental life-history traits that underlie species invasion success, and (2) gathering more basic data on species’ behavioural, morphological and physiological traits is a prerequisite for further advancing predictive ecology.