Improving species distribution models for climate change studies: variable selection and scale

Statistical species distribution models (SDMs) are widely used to predict the potential changes in species distributions under climate change scenarios. We suggest that we need to revisit the conceptual framework and ecological assumptions on which the relationship between species distributions and environment is based. We present a simple conceptual framework to examine the selection of environmental predictors and data resolution scales. These vary widely in recent papers, with light inconsistently included in the models. Focusing on light as a necessary component of plant SDMs, we briefly review its dependence on aspect and slope and existing knowledge of its influence on plant distribution. Differences in light regimes between north‐ and south‐facing aspects in temperate latitudes can produce differences in temperature equivalent to moves 200 km polewards. Local topography may create refugia that are not recognized in many climate change SDMs using coarse‐scale data. We argue that current assumptions about the selection of predictors and data resolution need further testing. Application of these ideas can clarify many issues of scale, extent and choice of predictors, and potentially improve the use of SDMs for climate change modelling of biodiversity.     

Modelling biome shifts and tree cover change for 2050 in West Africa

Aim Africa is expected to face severe changes in climatic conditions. Our objectives are: (1) to model trends and the extent of future biome shifts that may occur by 2050, (2) to model a trend in tree cover change, while accounting for human impact, and (3) to evaluate uncertainty in future climate projections. Location West Africa. Methods We modelled the potential future spatial distribution of desert, grassland, savanna, deciduous and evergreen forest in West Africa using six bioclimatic models. Future tree cover change was analysed with generalized additive models (GAMs). We used climate data from 17 general circulation models (GCMs) and included human population density and fire intensity to model tree cover. Consensus projections were derived via weighted averages to: (1) reduce inter‐model variability, and (2) describe trends extracted from different GCM projections. Results The strongest predicted effect of climate change was on desert and grasslands, where the bioclimatic envelope of grassland is projected to expand into the desert by an area of 2 million km². While savannas are predicted to contract in the south (by 54 ± 22 × 10⁴ km²), deciduous and evergreen forest biomes are expected to expand (64 ± 13 × 10⁴ km² and 77 ± 26 × 10⁴ km²). However, uncertainty due to different GCMs was particularly high for the grassland and the evergreen biome shift. Increasing tree cover (1–10%) was projected for large parts of Benin, Burkina Faso, Côte d’Ivoire, Ghana and Togo, but a decrease was projected for coastal areas (1–20%). Furthermore, human impact negatively affected tree cover and partly changed the direction of the projected change from increase to decrease. Main conclusions Considering climate change alone, the model results of potential vegetation (biomes) show a ‘greening’ trend by 2050. However, the modelled effects of human impact suggest future forest degradation. Thus, it is essential to consider both climate change and human impact in order to generate realistic future tree cover projections.     

The performance of models relating species geographical distributions to climate is independent of trophic level

Species–climate ‘envelope’ models are widely used to evaluate potential climate change impacts upon species and biodiversity. Previous studies have used a variety of methods to fit models making it difficult to assess relative model performance for different taxonomic groups, life forms or trophic levels. Here we use the same climatic data and modelling approach for 306 European species representing three major taxa (higher plants, insects and birds), and including species of different life form and from four trophic levels. Goodness‐of‐fit measures showed that useful models were fitted for >96% of species, and that model performance was related neither to major taxonomic group nor to trophic level. These results confirm that such climate envelope models provide the best approach currently available for evaluating reliably the potential impacts of future climate change upon biodiversity.     

Potential changes in the distributions of latitudinally restricted Australian butterfly species in response to climate change

This study assessed potential changes in the distributions of Australian butterfly species in response to global warming. The bioclimatic program, BIOCLIM, was used to determine the current climatic ranges of 77 butterfly species restricted to Australia. We found that the majority of these species had fairly wide climatic ranges in comparison to other taxa, with only 8% of butterfly species having a mean annual temperature range spanning less than 3 °C. The potential changes in the distributions of 24 butterfly species under four climate change scenarios for 2050 were also modelled using BIOCLIM. Results suggested that even species with currently wide climatic ranges may still be vulnerable to climate change; under a very conservative climate change scenario (with a temperature increase of 0.8–1.4 °C by 2050) 88% of species distributions decreased, and 54% of species distributions decreased by at least 20%. Under an extreme scenario (temperature increase of 2.1–3.9 °C by 2050) 92% of species distributions decreased, and 83% of species distributions decreased by at least 50%. Furthermore, the proportion of the current range that was contained within the predicted range decreased from an average of 63% under a very conservative scenario to less than 22% under the most extreme scenario. By assessing the climatic ranges that species are currently exposed to, the extent of potential changes in distributions in response to climate change and details of their life histories, we identified species whose characteristics may make them particularly vulnerable to climate change in the future.     

The ability of climate envelope models to predict the effect of climate change on species distributions

Climate envelope models (CEMs) have been used to predict the distribution of species under current, past, and future climatic conditions by inferring a species' environmental requirements from localities where it is currently known to occur. CEMs can be evaluated for their ability to predict current species distributions but it is unclear whether models that are successful in predicting current distributions are equally successful in predicting distributions under different climates (i.e. different regions or time periods). We evaluated the ability of CEMs to predict species distributions under different climates by comparing their predictions with those obtained with a mechanistic model (MM). In an MM the distribution of a species is modeled based on knowledge of a species' physiology. The potential distributions of 100 plant species were modeled with an MM for current conditions, a past climate reconstruction (21 000 years before present) and a future climate projection (double preindustrial CO₂ conditions). Point localities extracted from the currently suitable area according to the MM were used to predict current, future, and past distributions with four CEMs covering a broad range of statistical approaches: Bioclim (percentile distributions), Domain (distance metric), GAM (general additive modeling), and Maxent (maximum entropy). Domain performed very poorly, strongly underestimating range sizes for past or future conditions. Maxent and GAM performed as well under current climates as under past and future climates. Bioclim slightly underestimated range sizes but the predicted ranges overlapped more with the ranges predicted with the MM than those predicted with GAM did. Ranges predicted with Maxent overlapped most with those produced with the MMs, but compared with the ranges predicted with GAM they were more variable and sometimes much too large. Our results suggest that some CEMs can indeed be used to predict species distributions under climate change, but individual modeling approaches should be validated for this purpose, and model choice could be made dependent on the purpose of a particular study.     

Predicting range expansion of the map butterfly in Northern Europe using bioclimatic models

The two main goals of this study are: (i) to examine the range shifts of a currently northwards expanding species, the map butterfly (Araschnia levana), in relation to annual variation in weather, and (ii) to test the capability of a bioclimatic envelope model, based on broad-scale European distribution data, to predict recent distributional changes (2000–2004) of the species in Finland. A significant relationship between annual maximum dispersal distance of the species and late summer temperature was detected. This suggests that the map butterfly has dispersed more actively in warmer rather than cooler summers, the most notable dispersal events being promoted by periods of exceptionally warm weather and southerly winds. The accuracy of the broad-scale bioclimatic model built for the species with European data using Generalized Additive Models (GAM) was good based on split-sample evaluation for a single period. However, the model’s performance was poor when applied to predict range shifts in Finland. Among the many potential explanations for the poor success of the transferred bioclimatic model, is the fact that bioclimatic envelope models do not generally account for species dispersal. This and other uncertainties support the view that bioclimatic models should be applied with caution when they are used to project future range shifts of species.     

广义模型及分类回归树在物种分布模拟中的应用与比较

比较3个应用较广的模拟物种地理分布模型:广义线性模型(GLM)、广义加法模型(GAM)与分类回归树(CART)对中国树种地理分布模拟的优劣,以提出更为合适的模拟物种地理分布模型,并用于预测气候变化对物种地理分布的影响。3个模型对中国15种树种地理分布的模拟研究表明:除对油松、辽东栎分布的模拟精度稍差外,对其余树种分布的模拟精度均较高,其中以GAM模型最好。结合地理信息系统(GIS),比较分析了这3个模型对青冈、木荷、红松和油松4种树种的地理分布模拟效果,结果亦表明:这3个模型均能很好模拟青冈和木荷的地理分布,而GLM模型对红松分布的模拟结果不太理想,3个模型对油松分布的模拟结果均不甚理想,其中以GLM模型最差。基于3个模型对未来气候变化下青冈与蒙古栎地理分布的预测表明:GLM模型与GAM模型对青冈分布的预测结果较为接近,青冈在未来气候变化情景下向西和向北扩展,而CART模型预测青冈在未来气候变化情景下除有向西、向北扩展趋势外,广东和广西南部的青冈分布区将消失;3个模型均预测蒙古栎在未来气候变化情景下向西扩展,扩展面积的大小为:模型的模拟面积>模型>模型。     

Impact of landscape predictors on climate change modelling of species distributions: a case study with Eucalyptus fastigata in southern New South Wales,Australia

Aim We consider three questions. (1) How different are the predicted distribution maps when climate‐only and climate‐plus‐terrain models are developed from high‐resolution data? (2) What are the implications of differences between the models when predicting future distributions under climate change scenarios, particularly for climate‐only models at coarse resolution? (3) Does the use of high‐resolution data and climate‐plus‐terrain models predict an increase in the number of local refugia? Location South‐eastern New South Wales, Australia. Methods We developed two species distribution models for Eucalyptus fastigata under current climate conditions using generalized additive modelling. One used only climate variables as predictors (mean annual temperature, mean annual rainfall, mean summer rainfall); the other used both climate and landscape (June daily radiation, topographic position, lithology, nutrients) variables as predictors. Predictions of the distribution under current climate and climate change were then made for both models at a pixel resolution of 100 m. Results The model using climate and landscape variables as predictors explained a significantly greater proportion of the deviance than the climate‐only model. Inclusion of landscape variables resulted in the prediction of much larger areas of existing optimal habitat. An overlay of predicted future climate on the current climate space indicated that extrapolation of the statistical models was not occurring and models were therefore more robust. Under climate change, landscape‐defined refugia persisted in areas where the climate‐only model predicted major declines. In areas where expansion was predicted, the increase in optimal habitat was always greater with landscape predictors. Recognition of extensive optimal habitat conditions and potential refugia was dependent on the use of high‐resolution landscape data. Main conclusions Using only climate variables as predictors for assessing species responses to climate change ignores the accepted conceptual model of plant species distribution. Explicit statements justifying the selection of predictors based on ecological principles are needed. Models using only climate variables overestimate range reduction under climate change and fail to predict potential refugia. Fine‐scale‐resolution data are required to capture important climate/landscape interactions. Extrapolation of statistical models to regions in climate space outside the region where they were fitted is risky.     

Climate change and the future distributions of aquatic macrophytes across boreal catchments

Aim Aquatic–terrestrial ecotones are vulnerable to climate change, and degradation of the emergent aquatic macrophyte zone would have severe ecological consequences for freshwater, wetland and terrestrial ecosystems. Our aim was to uncover future changes in boreal emergent aquatic macrophyte zones by modelling the occurrence and percentage cover of emergent aquatic vegetation under different climate scenarios in Finland by the 2050s. Location Finland, northern Europe. Methods Data derived from different GIS sources were used to estimate future emergent aquatic macrophyte distributions in all catchments in Finland (848 in total). We used generalized additive models (GAM) with a full stepwise selection algorithm and Akaike information criterion to explore the main environmental determinates (climate and geomorphology) of emergent aquatic macrophyte distributions, which were derived from the national subclass of CORINE land‐cover classification. The accuracy of the distribution models (GAMs) was cross‐validated, using percentage of explained deviance and the area under the curve derived from the receiver‐operating characteristic plots. Results Our results indicated that emergent aquatic macrophytes will expand their distributions northwards from the current catchments and percentage cover will increase in all of the catchments in all climate scenarios. Growing degree‐days was the primary determinant affecting distributions of emergent aquatic macrophytes. Inclusion of geomorphological variables clearly improved model performance in both model exercises compared with pure climate variables. Main conclusions Emergent aquatic macrophyte distributions will expand due to climate change. Many emergent aquatic plant species have already expanded their distributions during the past decades, and this process will continue in the years 2051–80. Emergent aquatic macrophytes pose an increasing overgrowth risk for sensitive macrophyte species in boreal freshwater ecosystems, which should be acknowledged in management and conservation actions. We conclude that predictions based on GIS data can provide useful ‘first‐filter’ estimates of changes in aquatic–terrestrial ecotones.     

The data void in modeling current and future distributions of tropical species

Conserving biodiversity in the face of climate change requires a predictive ecology of species distributions. Nowhere is this need more acute than in the tropics, which harbor the majority of Earth's species and face rapid and large climate and land‐use changes. However, the study of species distributions and their responses to climate change in high diversity tropical regions is potentially crippled by a lack of basic data. We analyzed a database representing more than 800 000 unique geo‐referenced natural history collections to determine what fraction of tropical plant species has sufficient numbers of available collections for use in the habitat or niche models commonly used to predict species responses to climate change. We found that more than nine out of 10 species from the three principle tropical realms are so poorly collected (n < 20 records) that they are essentially invisible to modern modeling and conservation tools. In order to predict the impact of climate change on tropical species, efforts must be made to increase the amount of data available from tropical countries through a combination of collecting new specimens and digitizing existing records.