Process,correlation and parameter fitting in species distribution models: a response to Kriticos et al

In a recent article (Dormann et al., 2012, Journal of Biogeography, 39, 2119–2131), we compared different approaches to species distribution modelling and depicted modelling approaches along an axis from purely ‘correlative’ to ‘forward process‐based’ models. In their correspondence, Kriticos et al. (2013, Journal of Biogeography, doi: 10.1111/j.1365‐2699.2012.02791.x ) challenge this view, claiming that our continuum representation neglects differences among models and does not consider the ability of fitted process‐based models to combine the advantages of both process‐based and correlative modelling approaches. Here we clarify that the continuum view resulted from recognition of the manifold differences between models. We also reinforce the point that the current trend towards combining different modelling approaches may lead not only to the desired combination of the advantages but also to the accumulation of the disadvantages of those approaches. This point has not been made sufficiently clear previously.     

Forecasting species range dynamics with process‐explicit models: matching methods to applications

Knowing where species occur is fundamental to many ecological and environmental applications. Species distribution models (SDMs) are typically based on correlations between species occurrence data and environmental predictors, with ecological processes captured only implicitly. However, there is a growing interest in approaches that explicitly model processes such as physiology, dispersal, demography and biotic interactions. These models are believed to offer more robust predictions, particularly when extrapolating to novel conditions. Many process–explicit approaches are now available, but it is not clear how we can best draw on this expanded modelling toolbox to address ecological problems and inform management decisions. Here, we review a range of process–explicit models to determine their strengths and limitations, as well as their current use. Focusing on four common applications of SDMs – regulatory planning, extinction risk, climate refugia and invasive species – we then explore which models best meet management needs. We identify barriers to more widespread and effective use of process‐explicit models and outline how these might be overcome. As well as technical and data challenges, there is a pressing need for more thorough evaluation of model predictions to guide investment in method development and ensure the promise of these new approaches is fully realised.     

Modeling bird species distribution change in fire prone Mediterranean landscapes: incorporating species dispersal and landscape dynamics

Species distribution models (SDMs) have traditionally been founded on the assumption that species distributions are in equilibrium with environmental conditions and that these species–environment relationships can be used to estimate species responses to environmental changes. Insight into the validity of this assumption can be obtained from comparing the performance of correlative species distribution models with more complex hybrid approaches, i.e. correlative and process‐based models that explicitly include ecological processes, thereby accounting for mismatches between habitat suitability and species occupancy patterns. Here we compared the ability of correlative SDMs and hybrid models, which can accommodate non‐equilibrium situations arising from dispersal constraints, to reproduce the distribution dynamics of the ortolan bunting Emberiza hortulana in highly dynamic, early successional, fire driven Mediterranean landscapes. Whereas, habitat availability was derived from a correlative statistical SDM, occupancy was modeled using a hybrid approach combining a grid‐based, spatially‐explicit population model that explicitly included bird dispersal with the correlative model. We compared species occupancy patterns under the equilibrium assumption and different scenarios of species dispersal capabilities. To evaluate the predictive capability of the different models, we used independent species data collected in areas affected to different degree by fires. In accordance with the view that disturbance leads to a disparity between the suitable habitat and the occupancy patterns of the ortolan bunting, our results indicated that hybrid modeling approaches were superior to correlative models in predicting species spatial dynamics. Furthermore, hybrid models that incorporated short dispersal distances were more likely to reproduce the observed changes in ortolan bunting distribution patterns, suggesting that dispersal plays a key role in limiting the colonization of recently burnt areas. We conclude that SDMs used in a dynamic context can be significantly improved by using combined hybrid modeling approaches that explicitly account for interactions between key ecological constraints such as dispersal and habitat suitability that drive species response to environmental changes.     

Community versus single‐species distribution models for British plants

Aim Species distribution models are increasingly used to predict the impacts of global change on whole ecological communities by modelling the individualistic niche responses of large numbers of species. However, it is not clear whether this single‐species ensemble approach is preferable to community‐wide strategies that represent interspecific associations or shared responses to environmental gradients. Here, we test the performance of two multi‐species modelling approaches against equivalent single‐species models. Location Great Britain. Methods Single‐ and multi‐species distribution models were fitted for 701 native British plant species at a 10‐km grid scale. Two machine learning methods were used – classification and regression trees (CARTs) and artificial neural networks (ANNs). The single‐species versions are widely used in ecology but their multivariate extensions are less well known and have not previously been evaluated against one another. We compared their abilities to predict species distributions, community compositions and species richness in an independent geographical region reserved from model‐fitting. Results The single‐ and multi‐species models performed similarly, although the community models gave slightly poorer predictive accuracy by all measures. However, from the point of view of the whole community they were much simpler than the array of single‐species models, involving orders of magnitude fewer parameters. Multi‐species approaches also left greater residual spatial autocorrelation than the individualistic models and, contrary to expectation, were relatively less accurate for rarer species. However, the fitted multi‐species response curves had lower tendency for pronounced discontinuities that are unlikely to be a feature of realized niche responses. Main conclusions Although community distribution models were slightly less accurate than single‐species models, they offered a highly simplified way of modelling spatial patterns in British plant diversity. Moreover, an advantage of the multi‐species approach was that the modelling of shared environmental responses resolved more realistic response curves. However, there was a slight tendency for community models to predict rare species less accurately, which is potentially disadvantageous for conservation applications. We conclude that multi‐species distribution models may have potential for understanding and predicting the structure of ecological communities, but were slightly inferior to single‐species ensembles for our data.     

Phenology predicts the native and invasive range limits of common ragweed

Accurate models for species' distributions are needed to forecast the progress and impacts of alien invasive species and assess potential range‐shifting driven by global change. Although this has traditionally been achieved through data‐driven correlative modelling, robustly extrapolating these models into novel climatic conditions is challenging. Recently, a small number of process‐based or mechanistic distribution models have been developed to complement the correlative approaches. However, tests of these models are lacking, and there are very few process‐based models for invasive species. We develop a method for estimating the range of a globally invasive species, common ragweed (Ambrosia artemisiifolia L.), from a temperature‐ and photoperiod‐driven phenology model. The model predicts the region in which ragweed can reach reproductive maturity before frost kills the adult plants in autumn. This aligns well with the poleward and high‐elevation range limits in its native North America and in invaded Europe, clearly showing that phenological constraints determine the cold range margins of the species. Importantly, this is a ‘forward’ prediction made entirely independently of the distribution data. Therefore, it allows a confident and biologically informed forecasting of further invasion and range shifting driven by climate change. For ragweed, such forecasts are extremely important as the species is a serious crop weed and its airborne pollen is a major cause of allergy and asthma in humans. Our results show that phenology can be a key determinant of species' range margins, so integrating phenology into species distribution models offers great potential for the mechanistic modelling of range dynamics.     

Predicting extinction risks under climate change: coupling stochastic population models with dynamic bioclimatic habitat models

Species responses to climate change may be influenced by changes in available habitat, as well as population processes, species interactions and interactions between demographic and landscape dynamics. Current methods for assessing these responses fail to provide an integrated view of these influences because they deal with habitat change or population dynamics, but rarely both. In this study, we linked a time series of habitat suitability models with spatially explicit stochastic population models to explore factors that influence the viability of plant species populations under stable and changing climate scenarios in South African fynbos, a global biodiversity hot spot. Results indicate that complex interactions between life history, disturbance regime and distribution pattern mediate species extinction risks under climate change. Our novel mechanistic approach allows more complete and direct appraisal of future biotic responses than do static bioclimatic habitat modelling approaches, and will ultimately support development of more effective conservation strategies to mitigate biodiversity losses due to climate change.     

Predicting impacts of climate change on biodiversity: a role for semi‐mechanistic community‐level modelling

Aim Robust and reliable predictions of the effects of climate change on biodiversity are required in formulating conservation and management strategies that best retain biodiversity into the future. Significant challenges in modelling climate change impacts arise from limitations in our current knowledge of biodiversity. Community‐level modelling can complement species‐level approaches in overcoming these limitations and predicting climate change impacts on biodiversity as a whole. However, the community‐level approaches applied to date have been largely correlative, ignoring the key processes that influence change in biodiversity over space and time. Here, we suggest that the development of new ‘semi‐mechanistic’ community‐level models would substantially increase our capacity to predict climate change impacts on biodiversity. Location Global. Methods Drawing on an expansive review of biodiversity modelling approaches and recent advances in semi‐mechanistic modelling at the species level, we outline the main elements of a new semi‐mechanistic community‐level modelling approach. Results Our quantitative review revealed a sharp divide between mechanistic and non‐mechanistic biodiversity modelling approaches, with very few semi‐mechanistic models developed to date. Main conclusions We suggest that the conceptual framework presented here for combining mechanistic and non‐mechanistic community‐level approaches offers a promising means of incorporating key processes into predictions of climate change impacts on biodiversity whilst working within the limits of our current knowledge.     

Benchmarking novel approaches for modelling species range dynamics

Increasing biodiversity loss due to climate change is one of the most vital challenges of the 21st century. To anticipate and mitigate biodiversity loss, models are needed that reliably project species’ range dynamics and extinction risks. Recently, several new approaches to model range dynamics have been developed to supplement correlative species distribution models (SDMs), but applications clearly lag behind model development. Indeed, no comparative analysis has been performed to evaluate their performance. Here, we build on process‐based, simulated data for benchmarking five range (dynamic) models of varying complexity including classical SDMs, SDMs coupled with simple dispersal or more complex population dynamic models (SDM hybrids), and a hierarchical Bayesian process‐based dynamic range model (DRM). We specifically test the effects of demographic and community processes on model predictive performance. Under current climate, DRMs performed best, although only marginally. Under climate change, predictive performance varied considerably, with no clear winners. Yet, all range dynamic models improved predictions under climate change substantially compared to purely correlative SDMs, and the population dynamic models also predicted reasonable extinction risks for most scenarios. When benchmarking data were simulated with more complex demographic and community processes, simple SDM hybrids including only dispersal often proved most reliable. Finally, we found that structural decisions during model building can have great impact on model accuracy, but prior system knowledge on important processes can reduce these uncertainties considerably. Our results reassure the clear merit in using dynamic approaches for modelling species’ response to climate change but also emphasize several needs for further model and data improvement. We propose and discuss perspectives for improving range projections through combination of multiple models and for making these approaches operational for large numbers of species.     

sdm: a reproducible and extensible R platform for species distribution modelling

sdm is an object‐oriented, reproducible and extensible, platform for species distribution modelling. It uses individual species and community‐based approaches, enabling ensembles of models to be fitted and evaluated, to project species potential distributions in space and time. It provides a standardized and unified structure for handling species distributions data and modelling techniques, and supports markedly different modelling approaches, including correlative, process‐based (mechanistic), agent‐based, and cellular automata. The object‐oriented design of software is such that scientists can modify existing methods, extend the framework by developing new methods or modelling procedures, and share them to be reproduced by other scientists. sdm can handle spatial and temporal data for single or multiple species and uses high performance computing solutions to speed up modelling and simulations. The framework is implemented in R, providing a flexible and easy‐to‐use GUI interface.     

Prediction of plant species distribution in lowland river valleys in Belgium: modelling species response to site conditions

In ecological modelling, limitations in data and their applicability for predictive modelling are more rule than exception. Often modelling has to be performed on sub-optimal data, as explicit and controlled collection of (more) appropriate data would not be feasible. An example of predictive ecological modelling is given with application of generalized additive and generalized linear models fitted to presence–absence records of plant species and site condition data from four nutrient-poor Flemish lowland valleys. Standard regression procedures are used for modelling, although explanatory and response data do not meet all the assumptions implicit in these procedures. Data were non-randomly collected and are spatially autocorrelated; model residuals retain part of that correlation. The scale of most site-condition records does not match the scale of the response variable (species distribution). Hence, interpolated and up-scaled explanatory variables are used. Data are aggregated from distinct phytogeographical regions to allow for generalized models, applicable to a wider population of river valleys in the same region. Nevertheless, ecologically sound models are obtained, which predict well the distribution of most plant species for the Flemish river valleys considered.     

Integrating modelling of biodiversity composition and ecosystem function

There is increasing reliance on ecological models to improve our understanding of how ecological systems work, to project likely outcomes under alternative global change scenarios and to help develop robust management strategies. Two common types of spatiotemporally explicit ecological models are those focussed on biodiversity composition and those focussed on ecosystem function. These modelling disciplines are largely practiced separately, with separate literature, despite growing evidence that natural systems are shaped by the interaction of composition and function. Here we call for the development of new modelling approaches that integrate composition and function, accounting for the important interactions between these two dimensions, particularly under rapid global change. We examine existing modelling approaches that have begun to combine elements of composition and function, identifying their potential contribution to fully integrated modelling approaches. The development and application of integrated models of composition and function face a number of important challenges, including biological data limitations, system knowledge and computational constraints. We suggest a range of promising avenues that could help researchers overcome these challenges, including the use of virtual species, macroecological relationships and hybrid correlative‐mechanistic modelling. Explicitly accounting for the interactions between composition and function within integrated modelling approaches has the potential to improve our understanding of ecological systems, provide more accurate predictions of their future states and transform their management. Synthesis There is increasing attention from researchers and policy makers around the world on both assessing and projecting the state of the planet's biodiversity, its ecosystems and the essential services they provide to society. However, existing modelling approaches largely ignore the interactions between biodiversity composition and ecosystem function. We highlight the key challenges and potential solutions to developing integrated models of composition and function. Such models will require a new effort and focus from ecologists, yet the benefits are likely to be substantial, including better informing the management of natural systems at regional, national and international scales.     

Probing the role of interfacial waters in protein–DNA recognition using a hybrid implicit/explicit solvation model

When proteins bind to their DNA target sites, ordered water molecules are often present at the protein–DNA interface bridging protein and DNA through hydrogen bonds. What is the role of these ordered interfacial waters? Are they important determinants of the specificity of DNA sequence recognition, or do they act in binding in a primarily nonspecific manner, by improving packing of the interface, shielding unfavorable electrostatic interactions, and solvating unsatisfied polar groups that are inaccessible to bulk solvent? When modeling details of structure and binding preferences, can fully implicit solvent models be fruitfully applied to protein–DNA interfaces, or must the individualistic properties of these interfacial waters be accounted for? To address these questions, we have developed a hybrid implicit/explicit solvation model that specifically accounts for the locations and orientations of small numbers of DNA‐bound water molecules, while treating the majority of the solvent implicitly. Comparing the performance of this model with that of its fully implicit counterpart, we find that explicit treatment of interfacial waters results in a modest but significant improvement in protein side‐chain placement and DNA sequence recovery. Base‐by‐base comparison of the performance of the two models highlights DNA sequence positions whose recognition may be dependent on interfacial water. Our study offers large‐scale statistical evidence for the role of ordered water for protein–DNA recognition, together with detailed examination of several well‐characterized systems. In addition, our approach provides a template for modeling explicit water molecules at interfaces that should be extensible to other systems. Proteins 2013; 81:1318–1329. © 2013 Wiley Periodicals, Inc.     

Community‐level vs species‐specific approaches to model selection

A topic of particular current interest is community‐level approaches to species distribution modelling (SDM), i.e. approaches that simultaneously analyse distributional data for multiple species. Previous studies have looked at the advantages of community‐level approaches for parameter estimation, but not for model selection – the process of choosing which model (and in particular, which subset of environmental variables) to fit to data. We compared the predictive performance of models using the same modelling method (generalised linear models) but choosing the subset of variables to include in the model either simultaneously across all species (community‐level model selection) or separately for each species (species‐specific model selection). Our results across two large presence/absence tree community datasets were inconclusive as to whether there was an overall difference in predictive performance between models fitted via species‐specific vs community‐level model selection. However, we found some evidence that a community approach was best suited to modelling rare species, and its performance decayed with increasing prevalence. That is, when data were sparse there was more opportunity for gains from “borrowing strength” across species via a community‐level approach. Interestingly, we also found that the community‐level approach tended to work better when the model selection problem was more difficult, and more reliably detected “noise” variables that should be excluded from the model.     

How to understand species’ niches and range dynamics: a demographic research agenda for biogeography

Range dynamics causes mismatches between a species’ geographical distribution and the set of suitable environments in which population growth is positive (the Hutchinsonian niche). This is because source–sink population dynamics cause species to occupy unsuitable environments, and because environmental change creates non‐equilibrium situations in which species may be absent from suitable environments (due to migration limitation) or present in unsuitable environments that were previously suitable (due to time‐delayed extinction). Because correlative species distribution models do not account for these processes, they are likely to produce biased niche estimates and biased forecasts of future range dynamics. Recently developed dynamic range models (DRMs) overcome this problem: they statistically estimate both range dynamics and the underlying environmental response of demographic rates from species distribution data. This process‐based statistical approach qualitatively advances biogeographical analyses. Yet, the application of DRMs to a broad range of species and study systems requires substantial research efforts in statistical modelling, empirical data collection and ecological theory. Here we review current and potential contributions of these fields to a demographic understanding of niches and range dynamics. Our review serves to formulate a demographic research agenda that entails: (1) advances in incorporating process‐based models of demographic responses and range dynamics into a statistical framework, (2) systematic collection of data on temporal changes in distribution and abundance and on the response of demographic rates to environmental variation, and (3) improved theoretical understanding of the scaling of demographic rates and the dynamics of spatially coupled populations. This demographic research agenda is challenging but necessary for improved comprehension and quantification of niches and range dynamics. It also forms the basis for understanding how niches and range dynamics are shaped by evolutionary dynamics and biotic interactions. Ultimately, the demographic research agenda should lead to deeper integration of biogeography with empirical and theoretical ecology.     

The effect of sample size on the accuracy of species distribution models: considering both presences and pseudo‐absences or background sites

Most high‐performing species distribution modelling techniques require both presences, and either absences or pseudo‐absences or background points. In this paper, we explore the effect of sample size, towards developing improved strategies for modelling. We generated 1800 virtual species with three levels of prevalence using ten modelling techniques, while varying the number of training presences (NTP) and the number of random points (NRP representing pseudo‐absences or background sites). For five of the ten modelling techniques we built two versions of models: one with an equal total weight (ETW) setting where the total weight for pseudo‐absence is equivalent to the total weight for presence, and another with an unequal total weight (UTW) setting where the total weight for pseudo‐absence is not required to be equal to the total weight for presence. We compared two strategies for NRP: a small multiplier strategy (i.e. setting NRP at a few times as large as NTP), and a large number strategy (i.e. using numerous random points). We produced ensemble models (by averaging the predictions from 30 models built with the same set of training presences and different sets of random points in equivalent numbers) for three NTP magnitudes and two NRP strategies. We found that model accuracy altered as NRP increased with four distinct patterns of performance: increasing, decreasing, arch‐shaped and horizontal. In most cases ETW improved model performance. Ensemble models had higher accuracy than the corresponding single models, and this improvement was pronounced when NTP was low. We conclude that a large NRP is not always an appropriate strategy. The best choice for NRP will depend on the modelling techniques used, species prevalence and NTP. We recommend building ensemble models instead of single models, using the small multiplier strategy for NRP with ETW, especially when only a small number of species presence records are available.     

Validation of species–climate impact models under climate change

Increasing concern over the implications of climate change for biodiversity has led to the use of species–climate envelope models to project species extinction risk under climate‐change scenarios. However, recent studies have demonstrated significant variability in model predictions and there remains a pressing need to validate models and to reduce uncertainties. Model validation is problematic as predictions are made for events that have not yet occurred. Resubstituition and data partitioning of present‐day data sets are, therefore, commonly used to test the predictive performance of models. However, these approaches suffer from the problems of spatial and temporal autocorrelation in the calibration and validation sets. Using observed distribution shifts among 116 British breeding‐bird species over the past ～20 years, we are able to provide a first independent validation of four envelope modelling techniques under climate change. Results showed good to fair predictive performance on independent validation, although rules used to assess model performance are difficult to interpret in a decision‐planning context. We also showed that measures of performance on nonindependent data provided optimistic estimates of models' predictive ability on independent data. Artificial neural networks and generalized additive models provided generally more accurate predictions of species range shifts than generalized linear models or classification tree analysis. Data for independent model validation and replication of this study are rare and we argue that perfect validation may not in fact be conceptually possible. We also note that usefulness of models is contingent on both the questions being asked and the techniques used. Implementations of species–climate envelope models for testing hypotheses and predicting future events may prove wrong, while being potentially useful if put into appropriate context.     

Pushing the limits in marine species distribution modelling: lessons from the land present challenges and opportunities

Aim Species distribution models (SDMs) have been used to address a wide range of theoretical and applied questions in the terrestrial realm, but marine‐based applications remain relatively scarce. In this review, we consider how conceptual and practical issues associated with terrestrial SDMs apply to a range of marine organisms and highlight the challenges relevant to improving marine SDMs. Location We include studies from both marine and terrestrial systems that encompass many geographic locations around the globe. Methods We first performed a literature search and analysis of marine and terrestrial SDMs in ISI Web of Science to assess trends and applications. Using knowledge from terrestrial applications, we critically evaluate the application of SDMs in marine systems in the context of ecological factors (dispersal, species interactions, aggregation and ontogenetic shifts) and practical considerations (data quality, alternative modelling approaches and model validation) that facilitate or create difficulties for model application. Results The relative importance of ecological factors to be considered when applying SDMs varies among terrestrial and marine organisms. Correctly incorporating dispersal is frequently considered an important issue for terrestrial models, but because there is greater potential for dispersal in the ocean, it is often less of a concern in marine SDMs. By contrast, ontogenetic shifts and feeding have received little attention in terrestrial SDM applications, but these factors are important to many marine SDMs. Opportunities also exist for applying more advanced SDM approaches in the marine realm, including mechanistic ecophysiological models, where water balance and heat transfer equations are simpler for some marine organisms relative to their terrestrial counterparts. Main conclusions SDMs have generally been under‐utilized in the marine realm relative to terrestrial applications. Correlative SDM methods should be tested on a range of marine organisms, and we suggest further development of methods that address ontogenetic shifts and feeding interactions. We anticipate developments in, and cross‐fertilization between, coupled correlative and process‐based SDMs, mechanistic eco‐physiological SDMs, and spatial population dynamic models for climate change and species invasion applications in particular. Comparisons of the outputs of different model types will provide insight that is useful for improved spatial management of marine species.     

A niche for biology in species distribution models

Why species are found where they are is a central question in biogeography. The most widely used tool for understanding the controls on distribution is species distribution modelling. Species distribution modelling is now a well‐established method in both the theoretical and applied ecological literature. In this special issue we examine the current state of the art in species distribution modelling and explore avenues for including more biological processes in such models. In particular we focus on physiological, demographic, dispersal, competitive and ecological‐modulation processes. This overview highlights opportunities for new species distribution model concepts and developments, as well as a statistical agenda for implementing such models.     

Predicting fundamental and realized distributions based on thermal niche: A case study of a freshwater turtle

Species distribution models (SDM) have been broadly used in ecology to address theoretical and practical problems. Currently, there are two main approaches to generate SDMs: (i) correlative, which is based on species occurrences and environmental predictor layers and (ii) process-based models, which are constructed based on species' functional traits and physiological tolerances. The distributions estimated by each approach are based on different components of species niche. Predictions of correlative models approach species realized niches, while predictions of process-based are more akin to species fundamental niche. Here, we integrated the predictions of fundamental and realized distributions of the freshwater turtle Trachemys dorbigni. Fundamental distribution was estimated using data of T. dorbigni's egg incubation temperature, and realized distribution was estimated using species occurrence records. Both types of distributions were estimated using the same regression approaches (logistic regression and support vector machines), both considering macroclimatic and microclimatic temperatures. The realized distribution of T. dorbigni was generally nested in its fundamental distribution reinforcing theoretical assumptions that the species' realized niche is a subset of its fundamental niche. Both modelling algorithms produced similar results but microtemperature generated better results than macrotemperature for the incubation model. Finally, our results reinforce the conclusion that species realized distributions are constrained by other factors other than just thermal tolerances.