Outstanding Challenges in the Transferability of Ecological Models

Coding for Life: Designing a Platform for Projecting and Protecting Global Biodiversity

Time is running out to limit further devastating losses of biodiversity and nature's contributions to humans. Addressing this crisis requires accurate predictions about which species and ecosystems are most at risk to ensure efficient use of limited conservation and management resources. We review existing biodiversity projection models and discover problematic gaps. Current models usually cannot easily be reconfigured for other species or systems, omit key biological processes, and cannot accommodate feedbacks with Earth system dynamics. To fill these gaps, we envision an adaptable, accessible, and universal biodiversity modeling platform that can project essential biodiversity variables, explore the implications of divergent socioeconomic scenarios, and compare conservation and management strategies. We design a roadmap for implementing this vision and demonstrate that building this biodiversity forecasting platform is possible and practical.

Importance of Spatial Autocorrelation in Machine Learning Modeling of Polymetallic Nodules, Model Uncertainty and Transferability at Local Scale

Machine learning spatial modeling is used for mapping the distribution of deep-sea polymetallic nodules (PMN). However, the presence and influence of spatial autocorrelation (SAC) have not been extensively studied. SAC can provide information regarding the variable selection before modeling, and it results in erroneous validation performance when ignored. ML models are also problematic when applied in areas far away from the initial training locations, especially if the (new) area to be predicted covers another feature space. Here, we study the spatial distribution of PMN in a geomorphologically heterogeneous area of the Peru Basin, where SAC of PMN exists. The local Moran’s I analysis showed that there are areas with a significantly higher or lower number of PMN, associated with different backscatter values, aspect orientation, and seafloor geomorphological characteristics. A quantile regression forests (QRF) model is used using three cross-validation (CV) techniques (random-, spatial-, and cluster-blocking). We used the recently proposed “Area of Applicability” method to quantify the geographical areas where feature space extrapolation occurs. The results show that QRF predicts well in morphologically similar areas, with spatial block cross-validation being the least unbiased method. Conversely, random-CV overestimates the prediction performance. Under new conditions, the model transferability is reduced even on local scales, highlighting the need for spatial model-based dissimilarity analysis and transferability assessment in new areas.

Climatic niche shifts in introduced species

Predictions of future biological invasions often rely on the assumption that introduced species establish only under climatic conditions similar to those in their native range. To date, 135 studies have tested this assumption of 'niche conservatism', yielding contradictory results. Here we revisit this literature, consider the evidence for niche shifts, critically assess the methods used, and discuss the authors' interpretations of niche shifts. We find that the true frequency of niche shifts remains unknown because of diverging interpretations of similar metrics, conceptual issues biasing conclusions towards niche conservatism, and the use of climatic data that may not be biologically meaningful. We argue that these issues could be largely addressed by focussing on trends or relative degrees of niche change instead of dichotomous classifications (shift versus no shift), consistently and transparently including non-analogous climates, and conducting experimental studies on mismatches between macroclimates and microclimates experienced by the study organism. Furthermore, an observed niche shift may result either from species filling a greater part of their fundamental niche during the invasion (a 'realised niche shift') or from rapid evolution of traits adapting species to novel climates in the introduced range (a 'fundamental niche shift'). Currently, there is no conclusive evidence distinguishing between these potential mechanisms of niche shifts. We outline how these questions may be addressed by combining computational analyses and experimental evidence.

Assessing Pathways of Climate Change Effects in SpaDES: An Application to Boreal Landbirds of Northwest Territories Canada

Distributions of landbirds in Canadian northern forests are expected to be affected by climate change, but it remains unclear which pathways are responsible for projected climate effects. Determining whether climate change acts indirectly through changing fire regimes and/or vegetation dynamics, or directly through changes in climatic suitability may allow land managers to address negative trajectories via forest management. We used SpaDES, a novel toolkit built in R that facilitates the implementation of simulation models from different areas of knowledge to develop a simulation experiment for a study area comprising 50 million ha in the Northwest Territories, Canada. Our factorial experiment was designed to contrast climate effects pathways on 64 landbird species using climate-sensitive and non-climate sensitive models for tree growth and mortality, wildfire, and landbirds. Climate-change effects were predicted to increase suitable habitat for 73% of species, resulting in average net gain of 7.49 million ha across species. We observed higher species turnover in the northeastern, south-central (species loss), and western regions (species gain). Importantly, we found that most of the predicted differences in net area of occupancy across models were attributed to direct climate effects rather than simulated vegetation change, despite a similar relative importance of vegetation and climate variables in landbird models. Even with close to a doubling of annual area burned by 2100, and a 600 kg/ha increase in aboveground tree biomass predicted in this region, differences in landbird net occupancy across models attributed to climate-driven forest growth were very small, likely resulting from differences in the pace of vegetation and climate changes, or vegetation lags. The effect of vegetation lags (i.e., differences from climatic equilibrium) varied across species, resulting in a wide range of changes in landbird distribution, and consequently predicted occupancy, due to climate effects. These findings suggest that hybrid approaches using statistical models and landscape simulation tools could improve wildlife forecasts when future uncoupling of vegetation and climate is anticipated. This study lays some of the methodological groundwork for ecological adaptive management using the new platform SpaDES, which allows for iterative forecasting, mixing of modeling paradigms, and tightening connections between data, parameterization, and simulation.

Prediction of habitat suitability for Patrinia sibirica Juss. in the Southern Urals

The paper presents the results of predictions of the habitat persistence for rare relict of the Pleistocene floristic complex Patrinia sibirica (L.) Juss. in the Southern Urals under various forecasted climate change scenarios. Climate variables from CHELSA BIOCLIM, elevation data (GMTED2010) and coarse fragment content in the top level of soil were used as predictors for modeling in the MaxEnt software. The impact of climate change on P. sibirica habitats under the RCP4.5 and RCP8.5 scenarios calculated from an ensemble of four general circulation models has been analyzed. The modeling has shown that the changes in the habitat suitability depend on the altitude. Deterioration of the habitats could be attributed to a temperature increase in mountain forest locations, and to a precipitation of driest quarter increase in mountain forest-steppe locations. In both cases, this leads to the expansion of forest and shrub vegetation. Monitoring of the habitat persistence of P. sibirica and other relict species of the Pleistocene floristic complex can play a major role in predictions, as their massive decline would constitute that climatic changes exceed the ranges of their fluctuations in the Holocene.

Data integration methods to account for spatial niche truncation effects in regional projections of species distribution

Many species distribution models (SDMs) are built with precise but geographically restricted presence-absence data sets (e.g., a country) where only a subset of the environmental conditions experienced by a species across its range is considered (i.e., spatial niche truncation). This type of truncation is worrisome because it can lead to incorrect predictions e.g., when projecting to future climatic conditions belonging to the species niche but unavailable in the calibration area. Data from citizen-science programs, species range maps or atlases covering the full species range can be used to capture those parts of the species' niche that are missing regionally. However, these data usually are too coarse or too biased to support regional management. Here, we aim to (1) demonstrate how varying degrees of spatial niche truncation affect SDMs projections when calibrated with climatically truncated regional data sets and (2) test the performance of different methods to harness information from larger-scale data sets presenting different spatial resolutions to solve the spatial niche truncation problem. We used simulations to compare the performance of the different methods, and applied them to a real data set to predict the future distribution of a plant species (Potentilla aurea) in Switzerland. SDMs calibrated with geographically restricted data sets expectedly provided biased predictions when projected outside the calibration area or time period. Approaches integrating information from larger-scale data sets using hierarchical data integration methods usually reduced this bias. However, their performance varied depending on the level of spatial niche truncation and how data were combined. Interestingly, while some methods (e.g., data pooling, downscaling) performed well on both simulated and real data, others (e.g., those based on a Poisson point process) performed better on real data, indicating a dependency of model performance on the simulation process (e.g., shape of simulated response curves). Based on our results, we recommend to use different data integration methods and, whenever possible, to make a choice depending on model performance. In any case, an ensemble modeling approach can be used to account for uncertainty in how niche truncation is accounted for and identify areas where similarities/dissimilarities exist across methods.

Drivers of Solidago species invasion in Central Europe-Case study in the landscape of the Carpathian Mountains and their foreground

AIM

The invasion process is a complex, context-dependent phenomenon; nevertheless, it can be described using the PAB framework. This framework encompasses the joint effect of propagule pressure (P), abiotic characteristics of the environment (A), and biotic characteristics of both the invader and recipient vegetation (B). We analyzed the effectiveness of proxies of PAB factors to explain the spatial pattern of Solidago canadensis and S. gigantea invasion using invasive species distribution models.

LOCATION

Carpathian Mountains and their foreground, Central Europe.

METHODS

The data on species presence or absence were from an atlas of neophyte distribution based on a 2 × 2 km grid, covering approximately 31,200 km2 (7,752 grid cells). Proxies of PAB factors, along with data on historical distribution of invaders, were used as explanatory variables in Boosted Regression Trees models to explain the distribution of invasive Solidago. The areas with potentially lower sampling effort were excluded from analysis based on a target species approach.

RESULTS

Proxies of the PAB factors helped to explain the distribution of both S. canadensis and S. gigantea. Distributions of both species were limited climatically because a mountain climate is not conducive to their growth; however, the S. canadensis distribution pattern was correlated with proxies of human pressure, whereas S. gigantea distribution was connected with environmental characteristics. The varied responses of species with regard to distance from their historical distribution sites indicated differences in their invasion drivers.

MAIN CONCLUSIONS

Proxies of PAB are helpful in the choice of explanatory variables as well as the ecological interpretation of species distribution models. The results underline that human activity can cause variation in the invasion of ecologically similar species.

Real-time predictions of seabird distribution improve oil spill risk assessments

Current knowledge of the distribution of sensitive seabirds is inadequate to safeguard seabird populations from impacts of oil spills in the Arctic. This gap is mainly driven by the fact that statistical models applied to survey data are coarse-scale and static with limited documentation of the distributional dynamics and patchiness of seabirds relevant to risk assessments related to oil spills. This paper describes a dynamic modelling framework solution for prediction of fine-scale densities and movements of seabirds in close-to-real time using fully integrated 3-D hydrodynamic models, dynamic habitat suitability models and agent-based models. The modelling framework has been developed and validated for the swimming migration of Brünnich's Guillemot Uria lomvia in the Barents Sea. The results document that the distributional dynamics of Brünnich's Guillemot and other seabird species to a large degree can be simulated with in-situ state variables and patterns reflecting the physical meteorology and oceanography and habitat suitability.

Emerging themes in Population Consequences of Disturbance models

Assessing the non-lethal effects of disturbance from human activities is necessary for wildlife conservation and management. However, linking short-term responses to long-term impacts on individuals and populations is a significant hurdle for evaluating the risks of a proposed activity. The Population Consequences of Disturbance (PCoD) framework conceptually describes how disturbance can lead to changes in population dynamics, and its real-world application has led to a suite of quantitative models that can inform risk assessments. Here, we review PCoD models that forecast the possible consequences of a range of disturbance scenarios for marine mammals. In so doing, we identify common themes and highlight general principles to consider when assessing risk. We find that, when considered holistically, these models provide valuable insights into which contextual factors influence a population's degree of exposure and sensitivity to disturbance. We also discuss model assumptions and limitations, identify data gaps and suggest future research directions to enable PCoD models to better inform risk assessments and conservation and management decisions. The general principles explored can help wildlife managers and practitioners identify and prioritize the populations most vulnerable to disturbance and guide industry in planning activities that avoid or mitigate population-level effects.

Assessing the geographic specificity of pH prediction by classification and regression trees

Soil pH effects a wide range of critical biogeochemical processes that dictate plant growth and diversity. Previous literature has established the capacity of classification and regression trees (CARTs) to predict soil pH, but limitations of CARTs in this context have not been fully explored. The current study collected soil pH, climatic, and topographic data from 100 locations across New York’s Temperate Deciduous Forests (in the United States of America) to investigate the extrapolative capacity of a previously developed CART model as compared to novel CART and random forest (RF) models. Results showed that the previously developed CART underperformed in terms of predictive accuracy (RRMSE = 14.52%) when compared to a novel tree (RRMSE = 9.33%), and that a novel random forest outperformed both models (RRMSE = 8.88%), though its predictions did not differ significantly from the novel tree (p = 0.26). The most important predictors for model construction were climatic factors. These findings confirm existing reports that CART models are constrained by the spatial autocorrelation of geographic data and encourage the restricted application of relevant machine learning models to regions from which training data was collected. They also contradict previous literature implying that random forests should meaningfully boost the predictive accuracy of CARTs in the context of soil pH.

Biological invasion risk assessment of Tuta absoluta: mechanistic versus correlative methods

The capacity to assess invasion risk from potential crop pests before invasion of new regions globally would be invaluable, but this requires the ability to predict accurately their potential geographic range and relative abundance in novel areas. This may be unachievable using de facto standard correlative methods as shown for the South American tomato pinworm Tuta absoluta, a serious insect pest of tomato native to South America. Its global invasive potential was not identified until after rapid invasion of Europe, followed by Africa and parts of Asia where it has become a major food security problem on solanaceous crops. Early prospective assessment of its potential range is possible using physiologically based demographic modeling that would have identified knowledge gaps in T. absoluta biology at low temperatures. Physiologically based demographic models (PBDMs) realistically capture the weather-driven biology in a mechanistic way allowing evaluation of invasive risk in novel areas and climes including climate change. PBDMs explain the biological bases for the geographic distribution, are generally applicable to species of any taxa, are not limited to terrestrial ecosystems, and hence can be extended to support ecological risk modeling in aquatic ecosystems. PBDMs address a lack of unified general methods for assessing and managing invasive species that has limited invasion biology from becoming a more predictive science.

Distinct interspecific and intraspecific vulnerability of coastal species to global change

Characterising and predicting species responses to anthropogenic global change is one of the key challenges in contemporary ecology and conservation. The sensitivity of marine species to climate change is increasingly being described with forecasted species distributions, yet these rarely account for population level processes such as genomic variation and local adaptation. This study compares inter- and intraspecific patterns of biological composition to determine how vulnerability to climate change, and its environmental drivers, vary across species and populations. We compare species trajectories for three ecologically important southern African marine invertebrates at two time points in the future, both at the species level, with correlative species distribution models, and at the population level, with gradient forest models. Reported range shifts are species-specific and include both predicted range gains and losses. Forecasted species responses to climate change are strongly influenced by changes in a suite of environmental variables, from sea surface salinity and sea surface temperature, to minimum air temperature. Our results further suggest a mismatch between future habitat suitability (where species can remain in their ecological niche) and genomic vulnerability (where populations retain their genomic composition), highlighting the inter- and intraspecific variability in species' sensitivity to global change. Overall, this study demonstrates the importance of considering species and population level climatic vulnerability when proactively managing coastal marine ecosystems in the Anthropocene.

Climatic suitability of the eastern paralysis tick, Ixodes holocyclus, and its likely geographic distribution in the year 2050

The eastern paralysis tick, Ixodes holocyclus is one of two ticks that cause potentially fatal tick paralysis in Australia, and yet information on the full extent of its present or potential future spatial distribution is not known. Occurrence data for this tick species collected over the past two decades, and gridded environmental variables at 1 km² resolution representing climate conditions, were used to derive correlative ecological niche models to predict the current and future potential distribution. Several hundreds of candidate models were constructed with varying combinations of model parameters, and the best-fitting model was chosen based on statistical significance, omission rate, and Akaike Information Criterion (AICc). The best-fitting model matches the currently known distribution but also extends through most of the coastal areas in the south, and up to the Kimbolton peninsula in Western Australia in the north. Highly suitable areas are present around south of Perth, extending towards Albany, Western Australia. Most areas in Tasmania, where the species is not currently present, are also highly suitable. Future spatial distribution of this tick in the year 2050 indicates moderate increase in climatic suitability from the present-day prediction but noticeably also moderate to low loss of climatically suitable areas elsewhere.

Individual environmental niches in mobile organisms

Individual variation is increasingly recognized as a central component of ecological processes, but its role in structuring environmental niche associations remains largely unknown. Species' responses to environmental conditions are ultimately determined by the niches of single individuals, yet environmental associations are typically captured only at the level of species. Here, we develop scenarios for how individual variation may combine to define the compound environmental niche of populations, use extensive movement data to document individual environmental niche variation, test associated hypotheses of niche configuration, and examine the consistency of individual niches over time. For 45 individual white storks (Ciconia ciconia; 116 individual-year combinations), we uncover high variability in individual environmental associations, consistency of individual niches over time, and moderate to strong niche specialization. Within populations, environmental niches follow a nested pattern, with individuals arranged along a specialist-to-generalist gradient. These results reject common assumptions of individual niche equivalency among conspecifics, as well as the separation of individual niches into disparate parts of environmental space. These findings underscore the need for a more thorough consideration of individualistic environmental responses in global change research.

Do Review Papers on Bird–Vegetation Relationships Provide Actionable Information to Forest Managers in the Eastern United States?

Forest management planning requires the specification of measurable objectives as desired future conditions at spatial extents ranging from stands to landscapes and temporal extents ranging from a single growing season to several centuries. Effective implementation of forest management requires understanding current conditions and constraints well enough to apply the appropriate silvicultural strategies to produce desired future conditions, often for multiple objectives, at varying spatial and temporal extents. We administered an online survey to forest managers in the eastern US to better understand how wildlife scientists could best provide information to help meet wildlife-related habitat objectives. We then examined more than 1000 review papers on bird–vegetation relationships in the eastern US compiled during a systematic review of the primary literature to see how well this evidence-base meets the information needs of forest managers. We identified two main areas where wildlife scientists could increase the relevance and applicability of their research. First, forest managers want descriptions of wildlife species–vegetation relationships using the operational metrics of forest management (forest type, tree species composition, basal area, tree density, stocking rates, etc.) summarized at the operational spatial units of forest management (stands, compartments, and forests). Second, forest managers want information about how to provide wildlife habitats for many different species with varied habitat needs across temporal extents related to the ecological processes of succession after harvest or natural disturbance (1–2 decades) or even longer periods of stand development. We provide examples of review papers that meet these information needs of forest managers and topic-specific bibliographies of additional review papers that may contain actionable information for foresters who wish to meet wildlife management objectives. We suggest that wildlife scientists become more familiar with the extensive grey literature on forest bird–vegetation relationships and forest management that is available in natural resource management agency reports. We also suggest that wildlife scientists could reconsider everything from the questions they ask, the metrics they report on, and the way they allocate samples in time and space, to provide more relevant and actionable information to forest managers.

A Test of Species Distribution Model Transferability Across Environmental and Geographic Space for 108 Western North American Tree Species

Predictions from species distribution models (SDMs) are commonly used in support of environmental decision-making to explore potential impacts of climate change on biodiversity. However, because future climates are likely to differ from current climates, there has been ongoing interest in understanding the ability of SDMs to predict species responses under novel conditions (i.e., model transferability). Here, we explore the spatial and environmental limits to extrapolation in SDMs using forest inventory data from 11 model algorithms for 108 tree species across the western United States. Algorithms performed well in predicting occurrence for plots that occurred in the same geographic region in which they were fitted. However, a substantial portion of models performed worse than random when predicting for geographic regions in which algorithms were not fitted. Our results suggest that for transfers in geographic space, no specific algorithm was better than another as there were no significant differences in predictive performance across algorithms. There were significant differences in predictive performance for algorithms transferred in environmental space with GAM performing best. However, the predictive performance of GAM declined steeply with increasing extrapolation in environmental space relative to other algorithms. The results of this study suggest that SDMs may be limited in their ability to predict species ranges beyond the environmental data used for model fitting. When predicting climate-driven range shifts, extrapolation may also not reflect important biotic and abiotic drivers of species ranges, and thus further misrepresent the realized shift in range. Future studies investigating transferability of process based SDMs or relationships between geodiversity and biodiversity may hold promise.

A machine learning approach to integrating genetic and ecological data in tsetse flies (Glossina pallidipes) for spatially explicit vector control planning

Vector control is an effective strategy for reducing vector-borne disease transmission, but requires knowledge of vector habitat use and dispersal patterns. Our goal was to improve this knowledge for the tsetse species Glossina pallidipes, a vector of human and animal African trypanosomiasis, which are diseases that pose serious health and socioeconomic burdens across sub-Saharan Africa. We used random forest regression to (i) build and integrate models of G. pallidipes habitat suitability and genetic connectivity across Kenya and northern Tanzania and (ii) provide novel vector control recommendations. Inputs for the models included field survey records from 349 trap locations, genetic data from 11 microsatellite loci from 659 flies and 29 sampling sites, and remotely sensed environmental data. The suitability and connectivity models explained approximately 80% and 67% of the variance in the occurrence and genetic data and exhibited high accuracy based on cross-validation. The bivariate map showed that suitability and connectivity vary independently across the landscape and was used to inform our vector control recommendations. Post hoc analyses show spatial variation in the correlations between the most important environmental predictors from our models and each response variable (e.g., suitability and connectivity) as well as heterogeneity in expected future climatic change of these predictors. The bivariate map suggests that vector control is most likely to be successful in the Lake Victoria Basin and supports the previous recommendation that G. pallidipes from most of eastern Kenya should be managed as a single unit. We further recommend that future monitoring efforts should focus on tracking potential changes in vector presence and dispersal around the Serengeti and the Lake Victoria Basin based on projected local climatic shifts. The strong performance of the spatial models suggests potential for our integrative methodology to be used to understand future impacts of climate change in this and other vector systems.

AlleleShift: an R package to predict and visualize population-level changes in allele frequencies in response to climate change

Background

At any particular location, frequencies of alleles that are associated with adaptive traits are expected to change in future climates through local adaption and migration, including assisted migration (human-implemented when climate change is more rapid than natural migration rates). Making the assumption that the baseline frequencies of alleles across environmental gradients can act as a predictor of patterns in changed climates (typically future but possibly paleo-climates), a methodology is provided by AlleleShift of predicting changes in allele frequencies at the population level.

Methods

The prediction procedure involves a first calibration and prediction step through redundancy analysis (RDA), and a second calibration and prediction step through a generalized additive model (GAM) with a binomial family. As such, the procedure is fundamentally different to an alternative approach recently proposed to predict changes in allele frequencies from canonical correspondence analysis (CCA). The RDA step is based on the Euclidean distance that is also the typical distance used in Analysis of Molecular Variance (AMOVA). Because the RDA step or CCA approach sometimes predict negative allele frequencies, the GAM step ensures that allele frequencies are in the range of 0 to 1.

Results

AlleleShift provides data sets with predicted frequencies and several visualization methods to depict the predicted shifts in allele frequencies from baseline to changed climates. These visualizations include 'dot plot' graphics (function shift.dot.ggplot), pie diagrams (shift.pie.ggplot), moon diagrams (shift.moon.ggplot), 'waffle' diagrams (shift.waffle.ggplot) and smoothed surface diagrams of allele frequencies of baseline or future patterns in geographical space (shift.surf.ggplot). As these visualizations were generated through the ggplot2 package, methods of generating animations for a climate change time series are straightforward, as shown in the documentation of AlleleShift and in the supplemental videos.

Availability

AlleleShift is available as an open-source R package from https://cran.r-project.org/package=AlleleShift and https://github.com/RoelandKindt/AlleleShift. Genetic input data is expected to be in the adegenet::genpop format, which can be generated from the adegenet::genind format. Climate data is available from various resources such as WorldClim and Envirem.

How can we fully realize the potential of mathematical and biological models to reintegrate biology?

Both mathematical models and biological model systems stand as tractable representations of complex biological systems or behaviors. They facilitate research and provide insights, and they can describe general rules. Models that represent biological processes or formalize general hypotheses are essential to any broad understanding. Mathematical or biological models necessarily omit details of the natural systems and thus may ultimately be "incorrect" representations. A key challenge is that tractability requires relatively simple models but simplification can result in models that are incorrect in their qualitative, broad implications if the abstracted details matter. Our paper discusses this tension, and how we can improve our inferences from models. We advocate for further efforts dedicated to model development, improvement, and acceptance by the scientific community, all of which may necessitate a more explicit discussion of the purpose and power of models. We argue that models should play a central role in reintegrating biology as a way to test our integrated understanding of how molecules, cells, organs, organisms, populations, and ecosystems function.

Combining Regional Habitat Selection Models for Large-Scale Prediction: Circumpolar Habitat Selection of Southern Ocean Humpback Whales

Machine learning algorithms are often used to model and predict animal habitat selection—the relationships between animal occurrences and habitat characteristics. For broadly distributed species, habitat selection often varies among populations and regions; thus, it would seem preferable to fit region- or population-specific models of habitat selection for more accurate inference and prediction, rather than fitting large-scale models using pooled data. However, where the aim is to make range-wide predictions, including areas for which there are no existing data or models of habitat selection, how can regional models best be combined? We propose that ensemble approaches commonly used to combine different algorithms for a single region can be reframed, treating regional habitat selection models as the candidate models. By doing so, we can incorporate regional variation when fitting predictive models of animal habitat selection across large ranges. We test this approach using satellite telemetry data from 168 humpback whales across five geographic regions in the Southern Ocean. Using random forests, we fitted a large-scale model relating humpback whale locations, versus background locations, to 10 environmental covariates, and made a circumpolar prediction of humpback whale habitat selection. We also fitted five regional models, the predictions of which we used as input features for four ensemble approaches: an unweighted ensemble, an ensemble weighted by environmental similarity in each cell, stacked generalization, and a hybrid approach wherein the environmental covariates and regional predictions were used as input features in a new model. We tested the predictive performance of these approaches on an independent validation dataset of humpback whale sightings and whaling catches. These multiregional ensemble approaches resulted in models with higher predictive performance than the circumpolar naive model. These approaches can be used to incorporate regional variation in animal habitat selection when fitting range-wide predictive models using machine learning algorithms. This can yield more accurate predictions across regions or populations of animals that may show variation in habitat selection.

Scale gaps in landscape phenology: challenges and opportunities

Phenology, or the timing of life history events, can be heterogeneous across biological communities and landscapes and can vary across a wide variety of spatiotemporal scales. Here, we synthesize information from landscape phenology studies across different scales of measurement around a set of core concepts. We highlight why phenology is scale dependent and identify gaps in the spatiotemporal scales of phenological observations and inferences. We discuss the consequences of these gaps and describe opportunities to address the inherent sensitivities of phenological metrics to measurement scale. Although most studies we review and discuss are focused on plants, our work provides a broadly relevant overview of the role of observation scale in landscape phenology and a general approach for measuring and reporting scale dependence.

Identifying high-density areas of oysters using species distribution modeling: Lessons for conservation of the native Ostrea edulis and management of the invasive Magallana (Crassostrea) gigas in Sweden

AIM

Understanding spatial patterns of the distribution of adult native oyster, Ostrea edulis, and the invasive Magallana (Crassostrea) gigas is important for management of these populations. The aim of this study was to use ensemble SDM's to (a) identify and predict conservation hotspots, (b) assess the current level of protection for O. edulis, and (c) quantify the amount of overlap between the two species where interactions with M. gigas are most likely.

LOCATION

Skagerrak, Sweden.

METHODS

We used data collected by video at depths from 0.5 to 10 m in 436 sites. Models of occurrence and densities >1 m-2 were fitted and assessed using ensemble methods ("biomod2" package). Models of high-density hotspots were used to predict, map, and quantify areal extent of the species in order to assess the degree of overlap with protected areas and the potential for interactions between the two species.

RESULTS

Both species were widely distributed in the region. Observations of high-density habitats, mainly occurring at depths of ≈3 and 0.5 m for O. edulis and M. gigas, respectively, were found in 4% and 2% of the sites. Models provided useful predictions for both species (AUC = 0.85-0.99; sensitivity = 0.74-1.0; specificity = 0.72-0.97). High-density areas occupy roughly 15 km2 each with substantial overlap between species. 50% of these are protected only by fisheries regulations, 44% are found in Natura 2000 reserves and 6% of the predicted O. edulis enjoys protection in a national park.

MAIN CONCLUSIONS

Data collection by video in combination with SDM's provides a realistic approach for large-scale quantification of spatial patterns of marine population and habitats. O. edulis and M. gigas are common in the area, but a large proportion of the most valuable O. edulis habitats are not found in protected areas. The overlap between species suggests that efforts to manage the invasive M. gigas need to be integrated with management actions to conserve the native O. edulis.

Determinants of habitat suitability models transferability across geographically disjunct populations: Insights from Vipera ursinii urs inii

Transferability of habitat suitability models (HSMs), essential to accurately predict outside calibration conditions, has been seldom investigated at intraspecific level. We targeted Vipera ursinii ursinii, a meadow viper from southeastern France and central Italy, to assess determinants of transferability among geographically disjunct populations. We fitted HSMs upon occurrences of the Italian and French populations separately, as well as on the combined occurrence dataset. Internal transferability of HSMs, on spatially independent test data drawn from the calibration region, and their external transferability on the geographically disjunct populations were evaluated according to (a) use of full or spatially rarefied presence datasets; (b) ecology-driven or statistics-driven filtering of predictors; (c) modeling algorithm, testing generalized additive models and gradient boosting models; and (d) multivariate environmental novelty within test data. Niche overlap between French and Italian populations was also tested. Niche overlap was low, but niche divergence between the two populations' clusters was not corroborated. Nonetheless, wider niche breadth and heterogeneity of background environmental conditions characterizing the French populations led to low intercluster transferability. Although models fitted on the combined datasets did not attain consistently higher internal transferability than those separately fitted for the French and Italian populations, ensemble projection from the HSMs fitted on the joint occurrences produced more consistent suitability predictions across the full range of V. u. ursinii. Spatial thinning of occurrences ameliorated internal transferability but did not affect external transferability. The two approaches to predictors filtering did not differ in transferability of the respective HSMs but led to discrepant estimated environment-occurrence relationships and spatial predictions, while the two algorithms attained different relative rankings depending on the considered prediction task. Multivariate novelty of projection sites was negatively correlated to both internal transferability and external transferability. Our findings clarify issues researchers should keep in mind when using HSMs to get predictions across geographically disjunct populations.

Is New Always Better? Frontiers in Global Climate Datasets for Modeling Treeline Species in the Himalayas

Comparing and evaluating global climate datasets and their effect on model performance in regions with limited data availability has received little attention in ecological modeling studies so far. In this study, we aim at comparing the interpolated climate dataset Worldclim 1.4, which is the most widely used in ecological modeling studies, and the quasi-mechanistical downscaled climate dataset Chelsa, as well as their latest versions Worldclim 2.1 and Chelsa 1.2, with regard to their suitability for modeling studies. To evaluate the effect of these global climate datasets at the meso-scale, the ecological niche of Betula utilis in Nepal is modeled under current and future climate conditions. We underline differences regarding methodology and bias correction between Chelsa and Worldclim versions and highlight potential drawbacks for ecological models in remote high mountain regions. Regarding model performance and prediction plausibility under current climatic conditions, Chelsa-based models significantly outperformed Worldclim-based models, however, the latest version of Chelsa contains partially inherent distorted precipitation amounts. This study emphasizes that unmindful usage of climate data may have severe consequences for modeling treeline species in high-altitude regions as well as for future projections, if based on flawed current model predictions. The results illustrate the inevitable need for interdisciplinary investigations and collaboration between climate scientists and ecologists to enhance climate-based ecological model quality at meso- to local-scales by accounting for local-scale physical features at high temporal and spatial resolution.

Satellite-based habitat monitoring reveals long-term dynamics of deer habitat in response to forest disturbances

Disturbances play a key role in driving forest ecosystem dynamics, but how disturbances shape wildlife habitat across space and time often remains unclear. A major reason for this is a lack of information about changes in habitat suitability across large areas and longer time periods. Here, we use a novel approach based on Landsat satellite image time series to map seasonal habitat suitability annually from 1986 to 2017. Our approach involves characterizing forest disturbance dynamics using Landsat-based metrics, harmonizing these metrics through a temporal segmentation algorithm, and then using them together with GPS telemetry data in habitat models. We apply this framework to assess how natural forest disturbances and post-disturbance salvage logging affect habitat suitability for two ungulates, roe deer (Capreolus capreolus) and red deer (Cervus elaphus), over 32 yr in a Central European forest landscape. We found that red and roe deer differed in their response to forest disturbances. Habitat suitability for red deer consistently improved after disturbances, whereas the suitability of disturbed sites was more variable for roe deer depending on season (lower during winter than summer) and disturbance agent (lower in windthrow vs. bark-beetle-affected stands). Salvage logging altered the suitability of bark beetle-affected stands for deer, having negative effects on red deer and mixed effects on roe deer, but generally did not have clear effects on habitat suitability in windthrows. Our results highlight long-lasting legacy effects of forest disturbances on deer habitat. For example, bark beetle disturbances improved red deer habitat suitability for at least 25 yr. The duration of disturbance impacts generally increased with elevation. Methodologically, our approach proved effective for improving the robustness of habitat reconstructions from Landsat time series: integrating multiyear telemetry data into single, multi-temporal habitat models improved model transferability in time. Likewise, temporally segmenting the Landsat-based metrics increased the temporal consistency of our habitat suitability maps. As the frequency of natural forest disturbances is increasing across the globe, their impacts on wildlife habitat should be considered in wildlife and forest management. Our approach offers a widely applicable method for monitoring habitat suitability changes caused by landscape dynamics such as forest disturbance.

A 'How to' guide for interpreting parameters in habitat-selection analyses

Habitat‐selection analyses allow researchers to link animals to their environment via habitat‐selection or step‐selection functions, and are commonly used to address questions related to wildlife management and conservation efforts. Habitat‐selection analyses that incorporate movement characteristics, referred to as integrated step‐selection analyses, are particularly appealing because they allow modelling of both movement and habitat‐selection processes.

Despite their popularity, many users struggle with interpreting parameters in habitat‐selection and step‐selection functions. Integrated step‐selection analyses also require several additional steps to translate model parameters into a full‐fledged movement model, and the mathematics supporting this approach can be challenging for many to understand.

Using simple examples, we demonstrate how weighted distribution theory and the inhomogeneous Poisson point process can facilitate parameter interpretation in habitat‐selection analyses. Furthermore, we provide a ‘how to’ guide illustrating the steps required to implement integrated step‐selection analyses using the amt package

By providing clear examples with open‐source code, we hope to make habitat‐selection analyses more understandable and accessible to end users.

Continent-Wide Tree Species Distribution Models May Mislead Regional Management Decisions: A Case Study in the Transboundary Biosphere Reserve Mura-Drava-Danube

The understanding of spatial distribution patterns of native riparian tree species in Europe lacks accurate species distribution models (SDMs), since riparian forest habitats have a limited spatial extent and are strongly related to the associated watercourses, which needs to be represented in the environmental predictors. However, SDMs are urgently needed for adapting forest management to climate change, as well as for conservation and restoration of riparian forest ecosystems. For such an operative use, standard large-scale bioclimatic models alone are too coarse and frequently exclude relevant predictors. In this study, we compare a bioclimatic continent-wide model and a regional model based on climate, soil, and river data for central to south-eastern Europe, targeting seven riparian foundation species—Alnus glutinosa, Fraxinus angustifolia, F. excelsior, Populus nigra, Quercus robur, Ulmus laevis, and U. minor. The results emphasize the high importance of precise occurrence data and environmental predictors. Soil predictors were more important than bioclimatic variables, and river variables were partly of the same importance. In both models, five of the seven species were found to decrease in terms of future occurrence probability within the study area, whereas the results for two species were ambiguous. Nevertheless, both models predicted a dangerous loss of occurrence probability for economically and ecologically important tree species, likely leading to significant effects on forest composition and structure, as well as on provided ecosystem services.

On the Interpretations of Joint Modeling in Community Ecology

Explaining and modeling species communities is more than ever a central goal of ecology. Recently, joint species distribution models (JSDMs), which extend species distribution models (SDMs) by considering correlations among species, have been proposed to improve species community analyses and rare species predictions while potentially inferring species interactions. Here, we illustrate the mathematical links between SDMs and JSDMs and their ecological implications and demonstrate that JSDMs, just like SDMs, cannot separate environmental effects from biotic interactions. We provide a guide to the conditions under which JSDMs are (or are not) preferable to SDMs for species community modeling. More generally, we call for a better uptake and clarification of novel statistical developments in the field of biodiversity modeling.

Where did they not go? Considerations for generating pseudo-absences for telemetry-based habitat models

BACKGROUND

Habitat suitability models give insight into the ecological drivers of species distributions and are increasingly common in management and conservation planning. Telemetry data can be used in habitat models to describe where animals were present, however this requires the use of presence-only modeling approaches or the generation of 'pseudo-absences' to simulate locations where animals did not go. To highlight considerations for generating pseudo-absences for telemetry-based habitat models, we explored how different methods of pseudo-absence generation affect model performance across species' movement strategies, model types, and environments.

METHODS

We built habitat models for marine and terrestrial case studies, Northeast Pacific blue whales (Balaenoptera musculus) and African elephants (Loxodonta africana). We tested four pseudo-absence generation methods commonly used in telemetry-based habitat models: (1) background sampling; (2) sampling within a buffer zone around presence locations; (3) correlated random walks beginning at the tag release location; (4) reverse correlated random walks beginning at the last tag location. Habitat models were built using generalised linear mixed models, generalised additive mixed models, and boosted regression trees.

RESULTS

We found that the separation in environmental niche space between presences and pseudo-absences was the single most important driver of model explanatory power and predictive skill. This result was consistent across marine and terrestrial habitats, two species with vastly different movement syndromes, and three different model types. The best-performing pseudo-absence method depended on which created the greatest environmental separation: background sampling for blue whales and reverse correlated random walks for elephants. However, despite the fact that models with greater environmental separation performed better according to traditional predictive skill metrics, they did not always produce biologically realistic spatial predictions relative to known distributions.

CONCLUSIONS

Habitat model performance may be positively biased in cases where pseudo-absences are sampled from environments that are dissimilar to presences. This emphasizes the need to carefully consider spatial extent of the sampling domain and environmental heterogeneity of pseudo-absence samples when developing habitat models, and highlights the importance of scrutinizing spatial predictions to ensure that habitat models are biologically realistic and fit for modeling objectives.

Improved prediction of Canada lynx distribution through regional model transferability and data efficiency

The application of species distribution models (SDMs) to areas outside of where a model was created allows informed decisions across large spatial scales, yet transferability remains a challenge in ecological modeling. We examined how regional variation in animal-environment relationships influenced model transferability for Canada lynx (Lynx canadensis), with an additional conservation aim of modeling lynx habitat across the northwestern United States. Simultaneously, we explored the effect of sample size from GPS data on SDM model performance and transferability. We used data from three geographically distinct Canada lynx populations in Washington (n = 17 individuals), Montana (n = 66), and Wyoming (n = 10) from 1996 to 2015. We assessed regional variation in lynx-environment relationships between these three populations using principal components analysis (PCA). We used ensemble modeling to develop SDMs for each population and all populations combined and assessed model prediction and transferability for each model scenario using withheld data and an extensive independent dataset (n = 650). Finally, we examined GPS data efficiency by testing models created with sample sizes of 5%-100% of the original datasets. PCA results indicated some differences in environmental characteristics between populations; models created from individual populations showed differential transferability based on the populations' similarity in PCA space. Despite population differences, a single model created from all populations performed as well, or better, than each individual population. Model performance was mostly insensitive to GPS sample size, with a plateau in predictive ability reached at ~30% of the total GPS dataset when initial sample size was large. Based on these results, we generated well-validated spatial predictions of Canada lynx distribution across a large portion of the species' southern range, with precipitation and temperature the primary environmental predictors in the model. We also demonstrated substantial redundancy in our large GPS dataset, with predictive performance insensitive to sample sizes above 30% of the original.

Ecological forecasts reveal limitations of common model selection methods: predicting changes in beaver colony densities

Over the past two decades, there have been numerous calls to make ecology a more predictive science through direct empirical assessments of ecological models and predictions. While the widespread use of model selection using information criteria has pushed ecology toward placing a higher emphasis on prediction, few attempts have been made to validate the ability of information criteria to correctly identify the most parsimonious model with the greatest predictive accuracy. Here, we used an ecological forecasting framework to test the ability of information criteria to accurately predict the relative contribution of density dependence and density-independent factors (forage availability, harvest, weather, wolf [Canis lupus] density) on inter-annual fluctuations in beaver (Castor canadensis) colony densities. We modeled changes in colony densities using a discrete-time Gompertz model, and assessed the performance of four models using information criteria values: density-independent models with (1) and without (2) environmental covariates; and density-dependent models with (3) and without (4) environmental covariates. We then evaluated the forecasting accuracy of each model by withholding the final one-third of observations from each population and compared observed vs. predicted densities. Information criteria and our forecasting accuracy metrics both provided strong evidence of compensatory density dependence in the annual dynamics of beaver colony densities. However, despite strong within-sample performance by the most complex model (density-dependent with covariates) as determined using information criteria, hindcasts of colony densities revealed that the much simpler density-dependent model without covariates performed nearly as well predicting out-of-sample colony densities. The hindcast results indicated that the complex model over-fit our data, suggesting that parameters identified by information criteria as important predictor variables are only marginally valuable for predicting landscape-scale beaver colony dynamics. Our study demonstrates the importance of evaluating ecological models and predictions with long-term data and revealed how a known limitation of information criteria (over-fitting of complex models) can affect our interpretation of ecological dynamics. While incorporating knowledge of the factors that influence animal population dynamics can improve population forecasts, we suggest that comparing forecast performance metrics can likewise improve our knowledge of the factors driving population dynamics.

Legacy effects of fish but not elevation influence lake ecosystem response to environmental change

How communities reorganize during climate change depends on the distribution of diversity within ecosystems and across landscapes. Understanding how environmental and evolutionary history constrain community resilience is critical to predicting shifts in future ecosystem function. The goal of our study was to understand how communities with different histories respond to environmental change with regard to shifts in elevation (temperature, nutrients) and introduced predators. We hypothesized that community responses to the environment would differ in ways consistent with local adaptation and initial trait structure. We transplanted plankton communities from lakes at different elevations with and without fish in the Sierra Nevada Mountains in California to mesocosms at different elevations with and without fish. We examined the relative importance of the historical and experimental environment on functional (size structure, effects on lower trophic levels), community (zooplankton composition, abundance and biomass) and population (individual species abundance and biomass) responses. Communities originating from different elevations produced similar biomass at each elevation despite differences in species composition; that is, the experimental elevation, but not the elevation of origin, had a strong effect on biomass. Conversely, we detected a legacy effect of predators on plankton in the fishless environment. Daphnia pulicaria that historically coexisted with fish reached greater biomass under fishless conditions than those from fishless lakes, resulting in greater zooplankton community biomass and larger average size. Therefore, trait variation among lake populations determined the top-down effects of fish predators. In contrast, phenotypic plasticity and local diversity were sufficient to maintain food web structure in response to changing environmental conditions associated with elevation.

Re-Visiting the Incidence of Environmental Factors on a Pre-Imaginal Population of the Red Gum Lerp Psyllid, Glycaspis brimblecombei Moore

The red gum lerp psyllid, Glycaspis brimblecombei Moore (Hemiptera: Aphalaridae), is an invasive pest of Eucalyptus trees worldwide, responsible for serious damage, including the death of plants. Knowledge about the incidence of climatic factors on the insect development are essential to define useful strategies for controlling this pest. To this aim, G. brimblecombei has been sampled by two different methods from April 2012 to February 2013 in eastern Sicily on Eucalyptus camaldulensis in nine different sites, where the main climatic data (air temperature, relative humidity, and precipitation) have been also registered. The Glycaspis brimblecombei population showed a similar trend in all nine sites, positively correlated only with air temperature, but a negative correlation has emerged with precipitation and relative humidity. The results show the need for a deeper understanding of the role played by other abiotic (such as different concentrations of CO2) and biotic (e.g., the antagonistic action of natural enemies, competition with other pests, etc.) factors. The greater sensitivity, even at low densities of psyllid, of sampling methods based on the random collection of a fixed number of leaves compared to methods based on the collection of infested leaves in a fixed time interval has been also outlined.

Shade trees preserve avian insectivore biodiversity on coffee farms in a warming climate

AIM

Coffee is an important export for many developing countries, with a global annual trade value of $100 billion, but it is threatened by a warming climate. Shade trees may mitigate the effects of climate change through temperature regulation that can aid in coffee growth, slow pest reproduction, and sustain avian insectivore diversity. The impact of shade on bird diversity and microclimate on coffee farms has been studied extensively in the Neotropics, but there is a dearth of research in the Paleotropics.

LOCATION

East Africa.

METHODS

We created current and future regional Maxent models for avian insectivores in East Africa using Worldclim temperature data and observations from the Global Biodiversity Information Database. We then adjusted current and future bioclimatic layers based on mean differences in temperature between shade and sun coffee farms and projected the models using these adjusted layers to predict the impact of shade tree removal on climatic suitability for avian insectivores.

RESULTS

Existing Worldclim temperature layers more closely matched temperatures under shade trees than temperatures in the open. Removal of shade trees, through warmer temperatures alone, would result in reduction of avian insectivore species by over 25%, a loss equivalent to 50 years of climate change under the most optimistic emissions scenario. Under the most extreme climate scenario and removal of shade trees, insectivore richness is projected to decline from a mean of 38 to fewer than 8 avian insectivore species.

MAIN CONCLUSIONS

We found that shade trees on coffee farms already provide important cooler microclimates for avian insectivores. Future temperatures will become a regionally limiting factor for bird distribution in East Africa, which could negatively impact control of coffee pests, but the effect of climate change can be potentially mediated through planting and maintaining shade trees on coffee farms.

Habitat selection patterns are density dependent under the ideal free distribution

Despite being widely used, habitat selection models are rarely reliable and informative when applied across different ecosystems or over time. One possible explanation is that habitat selection is context-dependent due to variation in consumer density and/or resource availability. The goal of this paper is to provide a general theoretical perspective on the contributory mechanisms of consumer and resource density-dependent habitat selection, as well as on our capacity to account for their effects. Towards this goal we revisit the ideal free distribution (IFD), where consumers are assumed to be omniscient, equally competitive and freely moving, and are hence expected to instantaneously distribute themselves across a heterogeneous landscape such that fitness is equalised across the population. Although these assumptions are clearly unrealistic to some degree, the simplicity of the structure in IFD provides a useful theoretical vantage point to help clarify our understanding of more complex spatial processes. Of equal importance, IFD assumptions are compatible with the assumptions underlying common habitat selection models. Here we show how a fitness-maximising space use model, based on IFD, gives rise to resource and consumer density-dependent shifts in consumer distribution, providing a mechanistic explanation for the context-dependent outcomes often reported in habitat selection analysis. Our model suggests that adaptive shifts in consumer distribution patterns would be expected to lead to nonlinear and often non-monotonic patterns of habitat selection. These results indicate that even under the simplest of assumptions about adaptive organismal behaviour, habitat selection strength should critically depend on system-wide characteristics. Clarifying the impact of adaptive behavioural responses may be pivotal in making meaningful ecological inferences about observed patterns of habitat selection and allow reliable transferability of habitat selection predictions across time and space.

Racing against change: understanding dispersal and persistence to improve species' conservation prospects

Climate change is contributing to the widespread redistribution, and increasingly the loss, of species. Geographical range shifts among many species were detected rapidly after predictions of the potential importance of climate change were specified 35 years ago: species are shifting their ranges towards the poles and often to higher elevations in mountainous areas. Early tests of these predictions were largely qualitative, though extraordinarily rapid and broadly based, and statistical tests distinguishing between climate change and other global change drivers provided quantitative evidence that climate change had already begun to cause species’ geographical ranges to shift. I review two mechanisms enabling this process, namely development of approaches for accounting for dispersal that contributes to range expansion, and identification of factors that alter persistence and lead to range loss. Dispersal in the context of range expansion depends on an array of processes, like population growth rates in novel environments, rates of individual species movements to new locations, and how quickly areas of climatically tolerable habitat shift. These factors can be tied together in well-understood mathematical frameworks or modelled statistically, leading to better prediction of extinction risk as climate changes. Yet, species' increasing exposures to novel climate conditions can exceed their tolerances and raise the likelihood of local extinction and consequent range losses. Such losses are the consequence of processes acting on individuals, driven by factors, such as the growing frequency and severity of extreme weather, that contribute local extinction risks for populations and species. Many mechanisms can govern how species respond to climate change, and rapid progress in global change research creates many opportunities to inform policy and improve conservation outcomes in the early stages of the sixth mass extinction.