Consensus Forecasting of Species Distributions: The Effects of Niche Model Performance and Niche Properties

Ensemble forecasting is advocated as a way of reducing uncertainty in species distribution modeling (SDM). This is because it is expected to balance accuracy and robustness of SDM models. However, there are little available data regarding the spatial similarity of the combined distribution maps generated by different consensus approaches. Here, using eight niche-based models, nine split-sample calibration bouts (or nine random model-training subsets), and nine climate change scenarios, the distributions of 32 forest tree species in China were simulated under current and future climate conditions. The forecasting ensembles were combined to determine final consensual prediction maps for target species using three simple consensus approaches (average, frequency, and median [PCA]). Species’ geographic ranges changed (area change and shifting distance) in response to climate change, but the three consensual projections did not differ significantly with respect to how much or in which direction, but they did differ with respect to the spatial similarity of the three consensual predictions. Incongruent areas were observed primarily at the edges of species’ ranges. Multiple stepwise regression models showed the three factors (niche marginality and specialization, and niche model accuracy) to be related to the observed variations in consensual prediction maps among consensus approaches. Spatial correspondence among prediction maps was the highest when niche model accuracy was high and marginality and specialization were low. The difference in spatial predictions suggested that more attention should be paid to the range of spatial uncertainty before any decisions regarding specialist species can be made based on map outputs. The niche properties and single-model predictive performance provide promising insights that may further understanding of uncertainties in SDM.     

Disruptive selection as a driver of evolutionary branching and caste evolution in social insects

Theory suggests that evolutionary branching via disruptive selection may be a relatively common and powerful force driving phenotypic divergence. Here, we extend this theory to social insects, which have novel social axes of phenotypic diversification. Our model, built around turtle ant (Cephalotes) biology, is used to explore whether disruptive selection can drive the evolutionary branching of divergent colony phenotypes that include a novel soldier caste. Soldier evolution is a recurrent theme in social insect diversification that is exemplified in the turtle ants. We show that phenotypic mutants can gain competitive advantages that induce disruptive selection and subsequent branching. A soldier caste does not generally appear before branching, but can evolve from subsequent competition. The soldier caste then evolves in association with specialized resource preferences that maximize defensive performance. Overall, our model indicates that resource specialization may occur in the absence of morphological specialization, but that when morphological specialization evolves, it is always in association with resource specialization. This evolutionary coupling of ecological and morphological specialization is consistent with recent empirical evidence, but contrary to predictions of classical caste theory. Our model provides a new theoretical understanding of the ecology of caste evolution that explicitly considers the process of adaptive phenotypic divergence and diversification.     

Spatially autocorrelated sampling falsely inflates measures of accuracy for presence‐only niche models

Aim Environmental niche models that utilize presence‐only data have been increasingly employed to model species distributions and test ecological and evolutionary predictions. The ideal method for evaluating the accuracy of a niche model is to train a model with one dataset and then test model predictions against an independent dataset. However, a truly independent dataset is often not available, and instead random subsets of the total data are used for ‘training’ and ‘testing’ purposes. The goal of this study was to determine how spatially autocorrelated sampling affects measures of niche model accuracy when using subsets of a larger dataset for accuracy evaluation. Location The distribution of Centaurea maculosa (spotted knapweed; Asteraceae) was modelled in six states in the western United States: California, Oregon, Washington, Idaho, Wyoming and Montana. Methods Two types of niche modelling algorithms – the genetic algorithm for rule‐set prediction (GARP) and maximum entropy modelling (as implemented with Maxent) – were used to model the potential distribution of C. maculosa across the region. The effect of spatially autocorrelated sampling was examined by applying a spatial filter to the presence‐only data (to reduce autocorrelation) and then comparing predictions made using the spatial filter with those using a random subset of the data, equal in sample size to the filtered data. Results The accuracy of predictions from both algorithms was sensitive to the spatial autocorrelation of sampling effort in the occurrence data. Spatial filtering led to lower values of the area under the receiver operating characteristic curve plot but higher similarity statistic (I) values when compared with predictions from models built with random subsets of the total data, meaning that spatial autocorrelation of sampling effort between training and test data led to inflated measures of accuracy. Main conclusions The findings indicate that care should be taken when interpreting the results from presence‐only niche models when training and test data have been randomly partitioned but occurrence data were non‐randomly sampled (in a spatially autocorrelated manner). The higher accuracies obtained without the spatial filter are a result of spatial autocorrelation of sampling effort between training and test data inflating measures of prediction accuracy. If independently surveyed data for testing predictions are unavailable, then it may be necessary to explicitly account for the spatial autocorrelation of sampling effort between randomly partitioned training and test subsets when evaluating niche model predictions.     

Examining the potential effects of species aggregation on the network structure of food webs

One of the key measures that have been used to describe the topological properties of complex networks is the “degree distribution”, which is a measure that describes the frequency distribution of number of links per node. Food webs are complex ecological networks that describe the trophic relationships among species in a community, and the topological properties of empirical food webs, including degree distributions, have been examined previously. Previously, the “niche model” has been shown to accurately predict degree distributions of empirical food webs, however, the niche model-generated food webs were referenced against empirical food webs that had their species grouped together based on their taxonomic and/or trophic relationships (aggregated food webs). Here, we explore the effects of species aggregation on the ability of the niche model to predict the total- (sum of prey and predator links per node), in- (number of predator links per node), and out- (number of prey links per node) degree distributions of empirical food webs by examining two food webs that can be aggregated at different levels of resolution. The results showed that (1) the cumulative total- and out-degree distributions were consistent with the niche model predictions when the species were aggregated, (2) when the species were disaggregated (i.e., higher resolution), there were mixed conclusions with regards to the niche model's ability to predict total- and out-degree distributions, (3) the model's ability to predict the in-degree distributions of the two food webs was generally inadequate. Although it has been argued that universal functional form based on the niche model could describe the degree distribution patterns of empirical food webs, we believe there are some limitations to the model's ability to accurately predict the structural properties of food webs.     

Systematic mistakes are likely in bounded optimal decision-making systems

Systematic mistakes can be distinguished from other types of mistakes in that they are repeatable and predictable within a given organism, and are not due to uncertainty or lack of information. Here we provide a mathematical definition for the concept of systematic mistakes, which captures the way this concept has been used in the behavioral sciences. We also provide an analytical model of information processing networks that are made of large numbers of components, in analogy to the brain being made of a large number of neurons. We show that, for almost all behavioral tasks, and for a wide range of limitations on the computational complexity of the decision-making network, the best possible decision-makers will make systematic mistakes. This result, together with available empirical evidence, suggests that violations of rationality in humans and animals are consistent with natural selection, as the latter operates under constraints.     

High proportion of smaller ranged hummingbird species coincides with ecological specialization across the Americas

Ecological communities that experience stable climate conditions have been speculated to preserve more specialized interspecific associations and have higher proportions of smaller ranged species (SRS). Thus, areas with disproportionally large numbers of SRS are expected to coincide geographically with a high degree of community-level ecological specialization, but this suggestion remains poorly supported with empirical evidence. Here, we analysed data for hummingbird resource specialization, range size, contemporary climate, and Late Quaternary climate stability for 46 hummingbird–plant mutualistic networks distributed across the Americas, representing 130 hummingbird species (ca 40% of all hummingbird species). We demonstrate a positive relationship between the proportion of SRS of hummingbirds and community-level specialization, i.e. the division of the floral niche among coexisting hummingbird species. This relationship remained strong even when accounting for climate, furthermore, the effect of SRS on specialization was far stronger than the effect of specialization on SRS, suggesting that climate largely influences specialization through species'' range-size dynamics. Irrespective of the exact mechanism involved, our results indicate that communities consisting of higher proportions of SRS may be vulnerable to disturbance not only because of their small geographical ranges, but also because of their high degree of specialization.     

Climatic control of dispersal–ecological specialization trade‐offs: a metacommunity process at the heart of the latitudinal diversity gradient?

We outline the potentially important role of dispersal in linking diversity patterns at different spatial and temporal scales, and the resulting potential to link hypotheses explaining macroscale patterns of diversity. We do this by proposing a possible mechanism linking climate to diversity patterns: we argue that climate, via effects of continuity of habitat availability in space and time, mediates a dispersal–ecological specialization trade‐off at the metacommunity level that leads to latitudinal trends in dispersal ability, ecological specialization, range sizes, speciation and species richness, ultimately driving the latitudinal diversity gradient. This trade‐off constitutes a possible mechanism for the strong macroscale correlation between climate and species richness that is consistent with recent ideas about niche conservatism and gradient lengths, as well as other leading hypotheses. We present an overview of predictions derived from our ideas. Of these, some have already been tested and supported and others are still open to debate or need testing. Together they provide a unique set of predictions that allows falsification.     

Forage selection by mule deer: does niche breadth increase with population density?

Effects of population density of mule deer Odocoileus hemionus on forage selection were investigated by comparing diet characteristics of two subpopulations of deer in southern California, USA, that differed in population density during winter. Quality of diet for deer, as indexed by faecal crude protein, was higher at the low-density site than at the high-density site in winter, when deer densities were different. Quality of diet was similar in summer when both areas had comparable densities of deer. Both outcomes are consistent with predictions from density-dependent selection of diets by deer. Dietary niche breadth, however, differed in a manner opposite to predictions of niche theory based on diet selection under an ideal-free distribution. During winter, when differences in density between the two study sites were pronounced, niche breadth along the dietary axis in the low-density area was twice that of the high-density site. Generalist herbivores feeding primarily on low-quality browse at high population density in winter would be expected to increase their dietary breadth by feeding on additional species of plants as they depleted their food supply. Mule deer in our study, however, decreased the breadth of their dietary niche as population density increased. We hypothesize that by rapidly eliminating high-quality forages from an area by heavy grazing, deer at higher population densities narrowed their dietary niche. Theoretical models for changes in niche dimensions, including the ideal-free distribution, need to consider such empirical outcomes.     

The latitudinal gradient in niche breadth: concepts and evidence

We examine Robert MacArthur's hypothesis that niche breadth is positively associated with latitude (the latitude-niche breadth hypothesis). This idea has been influential and long standing, yet no studies have evaluated its generality or the validity of its assumptions. We review the theoretical arguments suggesting a positive relationship between niche breadth and latitude. We also use available evidence to evaluate the assumptions and predictions of MacArthur's latitude-niche breadth hypothesis. We find that neither the assumptions nor the predictions of the hypothesis are supported by data. We propose an alternative hypothesis linking latitude with niche breadth. Unlike previous ideas, our conceptual framework does not require equilibrial assumptions and is based on recently uncovered patterns of species interactions.     

Consequences of host adaptation for performance of vesicular stomatitis virus in novel thermal environments

Although it is widely assumed that the selective advantage of niche specialization drives species biodiversity, some theory suggests that generalists are favored over specialists when environments change unexpectedly. But this idea is rarely tested empirically, and its relevance is unknown for microparasites such as RNA viruses. Due to their small genome sizes pleiotropy is not uncommon in RNA viruses. Therefore, the genetic architectures underlying generalist traits may be indirectly molded by selection to better prepare generalist organisms for growth in new environments. Previously, vesicular stomatitis viruses were evolved to specialize on a single host, or to generalize on multiple hosts. Here we test whether virus generalists arising in the context of host adaptation also perform differently than specialists when viruses grow at novel temperatures. We compared thermal reaction norms of performance, within and among groups of viral specialists and generalists. Results showed that host adaptation was consequential for some fitness traits at novel temperatures due to modification of pleiotropic viral genes. Contrary to theoretical predictions, host generalists were selectively disadvantaged at extreme cool and warm environments. Multi-host adaptation may compromise the evolved thermostability of viral proteins, creating a cost of host generalization when viruses replicate at extreme temperatures.     

The reality and importance of founder speciation in evolution

A founder event occurs when a new population is established from a small number of individuals drawn from a large ancestral population. Mayr proposed that genetic drift in an isolated founder population could alter the selective forces in an epistatic system, an observation supported by recent studies. Carson argued that a period of relaxed selection could occur when a founder population is in an open ecological niche, allowing rapid population growth after the founder event. Selectable genetic variation can actually increase during this founder-flush phase due to recombination, enhanced survival of advantageous mutations, and the conversion of non-additive genetic variance into additive variance in an epistatic system, another empirically confirmed prediction. Templeton combined the theories of Mayr and Carson with population genetic models to predict the conditions under which founder events can contribute to speciation, and these predictions are strongly confirmed by the empirical literature. Much of the criticism of founder speciation is based upon equating founder speciation to an adaptive peak shift opposed by selection. However, Mayr, Carson and Templeton all modeled a positive interaction of selection and drift, and Templeton showed that founder speciation is incompatible with peak-shift conditions. Although rare, founder speciation can have a disproportionate importance in adaptive innovation and radiation, and examples are given to show that "rare" does not mean "unimportant" in evolution. Founder speciation also interacts with other speciation mechanisms such that a speciation event is not a one-dimensional process due to either selection alone or drift alone.     

Niche breadth,rarity and ecological characteristics within a regional flora spanning large environmental gradients

Aim Species specialization, which plays a fundamental role in niche differentiation and species coexistence, is a key biological trait in relation to the responses of populations to changing environments. Species with a limited niche breadth are considered to experience a higher risk of extinction than generalist species. This work aims to measure the degree of specialization in the regional flora of the French Alps and test whether species specialization is related to species rarity and ecological characteristics. Location This study was conducted in the French Alps region, which encompasses a large elevational gradient over a relatively limited area (26,000 km²). Methods Specialization was estimated for approximately 1200 plant species found in the region. Given the inherent difficulty of pinpointing the critical environmental niche axes for each individual species, we used a co‐occurrence‐based index to estimate species niche breadths (specialization index). This comprehensive measurement included crucial undetermined limiting niche factors, acting on both local and regional scales, and related to both biotic and abiotic interactions. The specialization index for each species was then related to a selection of plant typologies such as Grime strategies and Raunkiaer life‐forms, and to two measurements of plant rarity, namely regional area of occupancy and local abundance. Results Specialist species were mainly found in specific and harsh environments such as wetlands, cold alpine habitats and dry heathlands. These species were usually geographically restricted but relatively dominant in their local communities. Although none of the selected traits were sufficient predictors of specialization, pure competitors were over‐represented amongst generalist species, whereas stress‐tolerant species tended to be more specialized. Main conclusions Our results suggest that co‐occurrence‐based indices of niche breadth are a satisfactory method for inferring plant specialization using large species samples across very heterogeneous environments. Our results are an empirical validation of the tolerance–dominance trade‐off and also provide interesting insights into the long‐standing question of which biological properties characterize species with narrow niche breadth that are potentially threatened by global changes in the environment.     

Genetically informed ecological niche models improve climate change predictions

We examined the hypothesis that ecological niche models (ENMs) more accurately predict species distributions when they incorporate information on population genetic structure, and concomitantly, local adaptation. Local adaptation is common in species that span a range of environmental gradients (e.g., soils and climate). Moreover, common garden studies have demonstrated a covariance between neutral markers and functional traits associated with a species’ ability to adapt to environmental change. We therefore predicted that genetically distinct populations would respond differently to climate change, resulting in predicted distributions with little overlap. To test whether genetic information improves our ability to predict a species’ niche space, we created genetically informed ecological niche models (gENMs) using Populus fremontii (Salicaceae), a widespread tree species in which prior common garden experiments demonstrate strong evidence for local adaptation. Four major findings emerged: (i) gENMs predicted population occurrences with up to 12‐fold greater accuracy than models without genetic information; (ii) tests of niche similarity revealed that three ecotypes, identified on the basis of neutral genetic markers and locally adapted populations, are associated with differences in climate; (iii) our forecasts indicate that ongoing climate change will likely shift these ecotypes further apart in geographic space, resulting in greater niche divergence; (iv) ecotypes that currently exhibit the largest geographic distribution and niche breadth appear to be buffered the most from climate change. As diverse agents of selection shape genetic variability and structure within species, we argue that gENMs will lead to more accurate predictions of species distributions under climate change.     

Tuned inhibition in perceptual decision-making circuits can explain seemingly suboptimal confidence behavior

Current dominant views hold that perceptual confidence reflects the probability that a decision is correct. Although these views have enjoyed some empirical support, recent behavioral results indicate that confidence and the probability of being correct can be dissociated. An alternative hypothesis suggests that confidence instead reflects the magnitude of evidence in favor of a decision while being relatively insensitive to the evidence opposing the decision. We considered how this alternative hypothesis might be biologically instantiated by developing a simple neural network model incorporating a known property of sensory neurons: tuned inhibition. The key idea of the model is that the level of inhibition that each accumulator unit receives from units with the opposite tuning preference, i.e. its inhibition ‘tuning’, dictates its contribution to perceptual decisions versus confidence judgments, such that units with higher tuned inhibition (computing relative evidence for different perceptual interpretations) determine perceptual discrimination decisions, and units with lower tuned inhibition (computing absolute evidence) determine confidence. We demonstrate that this biologically plausible model can account for several counterintuitive findings reported in the literature where confidence and decision accuracy dissociate. By comparing model fits, we further demonstrate that a full complement of behavioral data across several previously published experimental results—including accuracy, reaction time, mean confidence, and metacognitive sensitivity—is best accounted for when confidence is computed from units without, rather than units with, tuned inhibition. Finally, we discuss predictions of our results and model for future neurobiological studies. These findings suggest that the brain has developed and implements this alternative, heuristic theory of perceptual confidence computation by relying on the diversity of neural resources available.     

An agent-based model of group decision making in baboons

We present an agent-based model of the key activities of a troop of chacma baboons (Papio hamadryas ursinus) based on the data collected at De Hoop Nature Reserve in South Africa. We analyse the predictions of the model in terms of how well it is able to duplicate the observed activity patterns of the animals and the relationship between the parameters that control the agent's decision procedure and the model's predictions. At the current stage of model development, we are able to show that across a wide range of decision parameter values, the baboons are able to achieve their energetic and social time requirements. The simulation results also show that decisions concerning movement (group action selection) have the greatest influence on the outcomes. Those cases where the model's predictions fail to agree with the observed activity patterns have highlighted key elements that were missing from the field data, and that would need to be collected in subsequent fieldwork. Based on our experience, we believe group decision making is a fertile field for future research, and agent-based modelling offers considerable scope for understanding group action selection.     

Global patterns of specialization and coexistence in bird assemblages

Aim Increased specialization has been hypothesized to facilitate local coexistence and thus high species richness, but empirical evaluations of the richness–specialization relationships have been relatively scant. Here, we provide a first assessment of this relationship for terrestrial bird assemblages at global extent and from fine to coarse grains. Location World‐wide. Methods We use two indices of specialization that describe species‐level resource use: diet and habitat specialization. The relationship between richness and mean assemblage‐level specialization was independently assessed at realm, biome‐realm, 12,100 km² equal‐area grid cells and fine‐grained scales. To identify assemblages that are diverse relative to environmental conditions we: (1) applied quantile regressions, (2) statistically accounted for other environmental variables which may constrain richness, and (3) parsed the data according to the residuals of a model relating species richness to the environmental variables. Results Assemblage species richness increases with both measures of specialization at all scales. Statistically, richness appears constrained by levels of specialization, with the highest richness values only found in specialized assemblages. Richness is positively associated with specialization even after accounting for gradients in resource availability. Net primary productivity and assemblage specialization have complementary statistical effects on assemblage species richness. Contrary to expectations based on niche partitioning of local resources, the relationship between specialization and richness is steep even at coarse scales. Main conclusions The results demonstrate that for an entire clade, totalling > 9000 species, specialization and species richness are related, at least for diverse assemblages. The strong patterns observed across scales suggest that this relationship does not solely originate from (1) limits on coexistence in present‐day assemblages, or (2) increased specialization in richer assemblages imposed by species’ abilities to partition ecological space. Instead, regional‐scale influences on the species pool may determine much of the observed relationship between richness and specialization. Although causal attribution is not straightforward, these findings support the idea that, for the scale of our analysis, specialization may be related to the past origination of high‐diversity assemblages, rather than their contemporary assembly.     

Comparing alternative models to empirical data: cognitive models of western scrub-jay foraging behavior

Animals often select one item from a set of candidates, as when choosing a foraging site or mate, and are expected to possess accurate and efficient rules for acquiring information and making decisions. Little is known, however, about the decision rules animals use. We compare patterns of information sampling by western scrub-jays (Aphelocoma californica) when choosing a nut with three decision rules: best of n (BN), flexible threshold (FT), and comparative Bayes (CB). First, we use a null hypothesis testing approach and find that the CB decision rule, in which individuals use past experiences to make nonrandom assessment and choice decisions, produces patterns of behavior that more closely correspond to observed patterns of nut sampling in scrub-jays than the other two rules. This approach does not allow us to quantify how much better CB is at predicting scrub-jay behavior than the other decision rules. Second, we use a model selection approach that uses Akaike Information Criteria to quantify how well alternative models approximate observed data. We find that the CB rule is much more likely to produce the observed patterns of scrub-jay behavior than the other rules. This result provides some of the best empirical evidence of the use of Bayesian information updating by a nonhuman animal.     

The importance of information on relatives for the prediction of genomic breeding values and the implications for the makeup of reference data sets in livestock breeding schemes

Background

The theory of genomic selection is based on the prediction of the effects of genetic markers in linkage disequilibrium with quantitative trait loci. However, genomic selection also relies on relationships between individuals to accurately predict genetic value. This study aimed to examine the importance of information on relatives versus that of unrelated or more distantly related individuals on the estimation of genomic breeding values.

Methods

Simulated and real data were used to examine the effects of various degrees of relationship on the accuracy of genomic selection. Genomic Best Linear Unbiased Prediction (gBLUP) was compared to two pedigree based BLUP methods, one with a shallow one generation pedigree and the other with a deep ten generation pedigree. The accuracy of estimated breeding values for different groups of selection candidates that had varying degrees of relationships to a reference data set of 1750 animals was investigated.

Results

The gBLUP method predicted breeding values more accurately than BLUP. The most accurate breeding values were estimated using gBLUP for closely related animals. Similarly, the pedigree based BLUP methods were also accurate for closely related animals, however when the pedigree based BLUP methods were used to predict unrelated animals, the accuracy was close to zero. In contrast, gBLUP breeding values, for animals that had no pedigree relationship with animals in the reference data set, allowed substantial accuracy.

Conclusions

An animal''s relationship to the reference data set is an important factor for the accuracy of genomic predictions. Animals that share a close relationship to the reference data set had the highest accuracy from genomic predictions. However a baseline accuracy that is driven by the reference data set size and the overall population effective population size enables gBLUP to estimate a breeding value for unrelated animals within a population (breed), using information previously ignored by pedigree based BLUP methods.