Field validation supports novel niche modeling strategies in a cryptic endangered amphibian

Many studies employ ecological niche models (ENMs) to predict species’ occurrences in undersampled regions, generally without field confirmation. Here, we use field surveys to test the relative utility of four potential refinements to the standard ENM approach: 1) altering model complexity based on AIC_c, 2) selecting background points from a biologically informed region, 3) using target‐group background to account for sampling bias in existing localities, and 4) using many rangewide localities (global model) versus fewer proximal localities (local model) to construct geographically restricted range predictions. We used Maxent to predict new localities for the California tiger salamander Ambystoma californiense, an endangered species that often goes undocumented due to its cryptic lifestyle. We followed this with a field survey of 260 previously unsampled potential breeding sites in Solano County, CA and used the resulting presence/absence data to compare all factorial combinations of the four model refinements using a new application of the Kruskal–Wallis test for ENM outputs. Our field surveys led to the discovery of 81 previously undocumented breeding localities for the California tiger salamander and demonstrated that ENMs could be significantly improved by utilizing target‐group background to account for spatial sampling bias and local models to focus model output on the subregion of the range being surveyed. Our results clearly demonstrate the potential for local models to outperform global models, and we recommend supplementing traditional Maxent global models that utilize all known localities with local models, particularly when species occupy geographically structured, heterogeneous habitat types. We also recommend using target‐group background since the improvement we observed when including it in our models was significant and very similar to that documented by previous studies. Most importantly, we emphasize the importance of field verification to enable rigorous statistical comparisons among models.     

Transferability and model evaluation in ecological niche modeling: a comparison of GARP and Maxent

We compared predictive success in two common algorithms for modeling species' ecological niches, GARP and Maxent, in a situation that challenged the algorithms to be general – that is, to be able to predict the species' distributions in broad unsampled regions, here termed transferability. The results were strikingly different between the two algorithms – Maxent models reconstructed the overall distributions of the species at low thresholds, but higher predictive levels of Maxent predictions reflected overfitting to the input data; GARP models, on the other hand, succeeded in anticipating most of the species' distributional potential, at the cost of increased (apparent, at least) commission error. Receiver operating characteristic (ROC) tests were weak in discerning models able to predict into broad unsampled areas from those that were not. Such transferability is clearly a novel challenge for modeling algorithms, and requires different qualities than does predicting within densely sampled landscapes – in this case, Maxent was transferable only at very low thresholds, and biases and gaps in input data may frequently affect results based on higher Maxent thresholds, requiring careful interpretation of model results.     

Predicting species distributions from small numbers of occurrence records: a test case using cryptic geckos in Madagascar

Aim Techniques that predict species potential distributions by combining observed occurrence records with environmental variables show much potential for application across a range of biogeographical analyses. Some of the most promising applications relate to species for which occurrence records are scarce, due to cryptic habits, locally restricted distributions or low sampling effort. However, the minimum sample sizes required to yield useful predictions remain difficult to determine. Here we developed and tested a novel jackknife validation approach to assess the ability to predict species occurrence when fewer than 25 occurrence records are available. Location Madagascar. Methods Models were developed and evaluated for 13 species of secretive leaf‐tailed geckos (Uroplatus spp.) that are endemic to Madagascar, for which available sample sizes range from 4 to 23 occurrence localities (at 1 km² grid resolution). Predictions were based on 20 environmental data layers and were generated using two modelling approaches: a method based on the principle of maximum entropy (Maxent) and a genetic algorithm (GARP). Results We found high success rates and statistical significance in jackknife tests with sample sizes as low as five when the Maxent model was applied. Results for GARP at very low sample sizes (less than c. 10) were less good. When sample sizes were experimentally reduced for those species with the most records, variability among predictions using different combinations of localities demonstrated that models were greatly influenced by exactly which observations were included. Main conclusions We emphasize that models developed using this approach with small sample sizes should be interpreted as identifying regions that have similar environmental conditions to where the species is known to occur, and not as predicting actual limits to the range of a species. The jackknife validation approach proposed here enables assessment of the predictive ability of models built using very small sample sizes, although use of this test with larger sample sizes may lead to overoptimistic estimates of predictive power. Our analyses demonstrate that geographical predictions developed from small numbers of occurrence records may be of great value, for example in targeting field surveys to accelerate the discovery of unknown populations and species.     

Modeling Spatial Trends in Estimated Species Richness using Breeding Bird Survey Data: A Valuable Tool in Biodiversity Assessment

We used data from the French breeding bird survey to estimate local bird species richness within sampled sites, using capture–recapture models. We investigated the possible effects of habitat structure and composition (landscape fragmentation, habitat cover and diversity) on estimated species richness at a local scale, and used the identified trends to help with modeling species richness at a large spatial scale. We performed geostatistical analyses based on spatial autocorrelation – cokriging models – to interpolate estimated species richness over the entire country, providing an opportunity to predict species-rich areas. We further compared species richness obtained with this method to species and rarity richness obtained using a national atlas of breeding birds. Estimated species richness was higher in species richness hotspots identified by the atlas. Combining informations on rare species from Atlas and species richness estimates from sound sampling based schemes should help with identifying species-rich areas for various taxa and locating biodiversity hotspots to be protected as high conservation value areas, especially in temperate zones where diversity hotspots are likely to match centers of high species richness because of very few centers of true endemicity.     

Effects of sample standardization on mean species detectabilities and estimates of relative differences in species richness among assemblages

Ecological surveys provide the basic information needed to estimate differences in species richness among assemblages. Comparable estimates of the differences in richness between assemblages require equal mean species detectabilities across assemblages. However, mean species detectabilities are often unknown, typically low, and potentially different from one assemblage to another. As a result, inferences regarding differences in species richness among assemblages can be biased. We evaluated how well three methods used to produce comparable estimates of species richness achieved equal mean species detectabilities across diverse assemblages: rarefaction, statistical estimators, and standardization of sampling effort on mean taxonomic similarity among replicate samples (MRS). We used simulated assemblages to mimic a wide range of species-occurrence distributions and species richness to compare the performance of these three methods. Inferences regarding differences in species richness based on rarefaction were highly biased when richness estimates were compared among assemblages with distinctly different species-occurrence distributions. Statistical estimators only marginally reduced this bias. Standardization on MRS yielded the most comparable estimates of differences in species richness. These findings have important implications for our understanding of species-richness patterns, inferences drawn from biological monitoring data, and planning for biodiversity conservation.     

An evaluation of a GARP model as an approach to predicting the spatial distribution of non-vagile invertebrate species

One of the primary goals of any systematic, taxonomic or biodiversity study is the characterization of species distributions. While museum collection data are important for ascertaining distributional ranges, they are often biased or incomplete. The Genetic Algorithm for Rule-set Prediction (GARP) is an ecological niche modelling method based on a genetic algorithm that has been argued to provide an accurate assessment of the spatial distribution of organisms that have dispersal capabilities. The primary objective of this study is to evaluate the accuracy of a GARP model to predict the spatial distribution of a non-invasive, non-vagile invertebrate whose full distributional range was unknown. A GARP predictive model based on seven environmental parameters and 42 locations known from historical museum records for species of the trapdoor spider genus Promyrmekiaphila was produced and subsequently used as a guide for ground truthing the model. The GARP model was neither a significant nor an accurate predictor of spider localities and was outperformed by more simplistic BIOCLIM and GLM models. The isolated nature of Promyrmekiaphila populations mandates that environmental layers and their respective resolutions are carefully chosen for model production. Our results strongly indicate that, for modelling the spatial distribution of low vagility organisms, one should employ a modelling method whose results are more conducive to interpretation than models produced by a 'black box' algorithm such as GARP.     

Modelling ecological niches from low numbers of occurrences: assessment of the conservation status of poorly known viverrids (Mammalia, Carnivora) across two continents

The conservation of poorly known species is difficult because of incomplete knowledge on their biology and distribution. We studied the contribution of two ecological niche modelling tools, the Genetic Algorithm for Rule-set Prediction (GARP) and maximum entropy (Maxent), in assessing potential ranges and distributional connectivity among 12 of the least known African and Asian viverrids. The level of agreement between GARP and Maxent predictions was low when < 15 occurrences were available, probably indicating a minimum number below that necessary to obtain models with good predictive power. Unexpectedly, our results suggested that Maxent extrapolated more than GARP in the context of small sample sizes. Predictions were overlapped with current land use and location of protected areas to estimate the conservation status of each species. Our analyses yielded range predictions generally contradicting with extents of occurrence established by the IUCN. We evidenced a high level of disturbance within predicted distributions in West and East Africa, Sumatra, and South-East Asia, and identified within West African degraded lowlands four relatively preserved areas that might be of prime importance for the conservation of rainforest taxa. Knowing whether these species of viverrids may survive in degraded or alternative habitats is of crucial importance for further conservation planning. The level of coverage of species suitable ranges by existing and proposed IUCN reserves was low, and we recommend that the total surface of protected areas be substantially increased on both continents.     

The challenge of modeling niches and distributions for data‐poor species: a comprehensive approach to model complexity

Models of species ecological niches and geographic distributions now represent a widely used tool in ecology, evolution, and biogeography. However, the very common situation of species with few available occurrence localities presents major challenges for such modeling techniques, in particular regarding model complexity and evaluation. Here, we summarize the state of the field regarding these issues and provide a worked example using the technique Maxent for a small mammal endemic to Madagascar (the nesomyine rodent Eliurus majori). Two relevant model‐selection approaches exist in the literature (information criteria, specifically AICc; and performance predicting withheld data, via a jackknife), but AICc is not strictly applicable to machine‐learning algorithms like Maxent. We compare models chosen under each selection approach with those corresponding to Maxent default settings, both with and without spatial filtering of occurrence records to reduce the effects of sampling bias. Both selection approaches chose simpler models than those made using default settings. Furthermore, the approaches converged on a similar answer when sampling bias was taken into account, but differed markedly with the unfiltered occurrence data. Specifically, for that dataset, the models selected by AICc had substantially fewer parameters than those identified by performance on withheld data. Based on our knowledge of the study species, models chosen under both AICc and withheld‐data‐selection showed higher ecological plausibility when combined with spatial filtering. The results for this species intimate that AICc may consistently select models with fewer parameters and be more robust to sampling bias. To test these hypotheses and reach general conclusions, comprehensive research should be undertaken with a wide variety of real and simulated species. Meanwhile, we recommend that researchers assess the critical yet underappreciated issue of model complexity both via information criteria and performance on withheld data, comparing the results between the two approaches and taking into account ecological plausibility.     

Biased data reduce efficiency and effectiveness of conservation reserve networks

Complementarity-based reserve selection algorithms efficiently prioritize sites for biodiversity conservation, but they are data-intensive and most regions lack accurate distribution maps for the majority of species. We explored implications of basing conservation planning decisions on incomplete and biased data using occurrence records of the plant family Proteaceae in South Africa. Treating this high-quality database as 'complete', we introduced three realistic sampling biases characteristic of biodiversity databases: a detectability sampling bias and two forms of roads sampling bias. We then compared reserve networks constructed using complete, biased, and randomly sampled data. All forms of biased sampling performed worse than both the complete data set and equal-effort random sampling. Biased sampling failed to detect a median of 1-5% of species, and resulted in reserve networks that were 9-17% larger than those designed with complete data. Spatial congruence and the correlation of irreplaceability scores between reserve networks selected with biased and complete data were low. Thus, reserve networks based on biased data require more area to protect fewer species and identify different locations than those selected with randomly sampled or complete data.     

Modeling and mapping the current and future climatic-niche of endangered Himalayan musk deer

Identification of geographical space enveloped by suitable climatic conditions (i.e., climatic niche) that support species survival over space and time is crucial in conservation biogeography. Numerous algorithms (e.g., Maxent, GARP) with increasing accuracy have been devised and are being employed to overcome the challenges of forecasting climatic niche of species with incomplete information. The current study was conducted to map the distribution of current and future climatic niche of endangered Himalayan musk deer, a species endemic to Asia. Maxent and GARP modeling algorithms were individually employed to forecast current and future climatic niche of the species using randomly collected occurrence records of the species and bioclimatic variables with 30″ resolution from ‘WorldClim’ datasets. Both the modeling processes performed optimally with regard to AUC and TSS values and forecasted an increase/expansion of climatically-suitable geographical space in the future. A final climatic niche distribution map was produced by combining the binary maps generated from each of the processes to produce a relatively realistic and potentially accurate distribution of climatic niche of the species over space and time. Conservation of forecasted suitable geographical space is recommended and future survey efforts for potentially unexplored populations of the species in the forecasted suitable area are suggested.     

Database records as a surrogate for sampling effort provide higher species richness estimations

The compilation of all the available taxonomic and distributional information on the species present in a territory frequently generates a biased picture of the distribution of biodiversity due to the uneven distribution of the sampling effort performed. Thus, quality protocol assessments such as those proposed by Hortal et al. (Conservation Biology 21:853–863, 2007) must be done before using this kind of information for basic and applied purposes. The discrimination of localities that can be considered relatively well-surveyed from those not surveyed enough is a key first step in this protocol and can be attained by the previous definition of a sampling effort surrogate and the calculation of survey completeness using different estimators. Recently it has been suggested that records from exhaustive databases can be used as a sampling-effort surrogate to recognize probable well-surveyed localities. In this paper, we use an Iberian dung beetle database to identify the 50 × 50 km UTM cells that appear to be reliably inventoried, using both data derived from standardized sampling protocols and database records as a surrogate for sampling effort. Observed and predicted species richness values in the shared cells defined as well-surveyed by both methods suggest that the use of database records provides higher species richness values, which are proportionally greater in the richest localities by the inclusion of rare species.     

Ecological niche modeling re‐examined: A case study with the Darwin's fox

Many previous studies have attempted to assess ecological niche modeling performance using receiver operating characteristic (ROC) approaches, even though diverse problems with this metric have been pointed out in the literature. We explored different evaluation metrics based on independent testing data using the Darwin's Fox (Lycalopex fulvipes) as a detailed case in point. Six ecological niche models (ENMs; generalized linear models, boosted regression trees, Maxent, GARP, multivariable kernel density estimation, and NicheA) were explored and tested using six evaluation metrics (partial ROC, Akaike information criterion, omission rate, cumulative binomial probability), including two novel metrics to quantify model extrapolation versus interpolation (E‐space index I) and extent of extrapolation versus Jaccard similarity (E‐space index II). Different ENMs showed diverse and mixed performance, depending on the evaluation metric used. Because ENMs performed differently according to the evaluation metric employed, model selection should be based on the data available, assumptions necessary, and the particular research question. The typical ROC AUC evaluation approach should be discontinued when only presence data are available, and evaluations in environmental dimensions should be adopted as part of the toolkit of ENM researchers. Our results suggest that selecting Maxent ENM based solely on previous reports of its performance is a questionable practice. Instead, model comparisons, including diverse algorithms and parameterizations, should be the sine qua non for every study using ecological niche modeling. ENM evaluations should be developed using metrics that assess desired model characteristics instead of single measurement of fit between model and data. The metrics proposed herein that assess model performance in environmental space (i.e., E‐space indices I and II) may complement current methods for ENM evaluation.     

Using higher taxa as surrogates of species-level data in three Portuguese protected areas: a case study on Spheciformes (Hymenoptera)

Protected areas are the focus of most conservation efforts worldwide. Despite vast amount of investment in protected areas, biodiversity loss continues. This has led to increasing efforts to develop measures to assess the effectiveness of protected areas. The reliability of these measures depends on the quality of the information collected. However, because the resources available for the collection of information are limited, several strategies have been developed to reduce the resources necessary. In this study the combination of two resource reduction approaches—bioindicator and higher-taxa—is proposed. Spheciformes have been found to be useful as biodiversity, ecological and environmental indicators. Identification to the species level is usually very costly, but the use of genus-level information has been suggested. Tribe- and genus-level data for Spheciformes were assessed for their ability to predict the number of species independently of other variables—sampling area, geographic location, vegetation type, disturbance regime, and sampling effort—at three Portuguese protected areas. Tribe and genus-level data were found to be good indicators, with genus being the more reliable taxonomic level. Sampling effort was the only external variable that affected the relationship between species and higher-taxa richness. Genus-level data were also found to be useful for ranking sites according to richness or composition, and for determining richness-based and rarity-based complementary sets of sites for conservation. Using genus richness as a surrogate for species richness seems a promising approach for monitoring and contributing to the establishment of protected areas in Portugal and the entire Mediterranean region.     

Improved species‐occurrence predictions in data‐poor regions: using large‐scale data and bias correction with down‐weighted Poisson regression and Maxent

Species distribution modelling (SDM) has become an essential method in ecology and conservation. In the absence of survey data, the majority of SDMs are calibrated with opportunistic presence‐only data, incurring substantial sampling bias. We address the challenge of correcting for sampling bias in the data‐sparse situations. We modelled the relative intensity of bat records in their entire range using three modelling algorithms under the point‐process modelling framework (GLMs with subset selection, GLMs fitted with an elastic‐net penalty, and Maxent). To correct for sampling bias, we applied model‐based bias correction by incorporating spatial information on site accessibility or sampling efforts. We evaluated the effect of bias correction on the models’ predictive performance (AUC and TSS), calculated on spatial‐block cross‐validation and a holdout data set. When evaluated with independent, but also sampling‐biased test data, correction for sampling bias led to improved predictions. The predictive performance of the three modelling algorithms was very similar. Elastic‐net models have intermediate performance, with slight advantage for GLMs on cross‐validation and Maxent on hold‐out evaluation. Model‐based bias correction is very useful in data‐sparse situations, where detailed data are not available to apply other bias correction methods. However, bias correction success depends on how well the selected bias variables describe the sources of bias. In this study, accessibility covariates described bias in our data better than the effort covariate, and their use led to larger changes in predictive performance. Objectively evaluating bias correction requires bias‐free presence–absence test data, and without them the real improvement for describing a species’ environmental niche cannot be assessed.     

Predicting species distributions in poorly-studied landscapes

Conservationists are increasingly relying on distribution models to predict where species are likely to occur, especially in poorly-surveyed but biodiverse areas. Modeling is challenging in these cases because locality data necessary for model formation are often scarce and spatially imprecise. To identify methods best suited to modeling in these conditions, we compared the success of three algorithms (Maxent, Mahalanobis Typicalities and Random Forests) at predicting distributions of eight bird and eight mammal species endemic to the eastern slopes of the central Andes. We selected study species to have a range of locality sample sizes representative of the data available for endemic species of this region and also that vary in their distribution characteristics. We found that for species that are known from moderate numbers (N = 38–94) of localities, the three methods performed similarly for species with restricted distributions but Maxent and Random Forests yielded better results for species with wider distributions. For species with small numbers of sample localities (N = 5–21), Maxent produced the most consistently successful results, followed by Random Forests and then Mahalanobis Typicalities. Because evaluation statistics for models derived from few localities can be suspect due to the poor spatial representation of the evaluation data, we corroborated these results with review by scientists familiar with the species in the field. Overall, Maxent appears to be the most capable method for modeling distributions of Andean bird and mammal species because of the consistency of results in varying conditions, although the other methods have strengths in certain situations. Electronic supplementary material  The online version of this article (doi:) contains supplementary material, which is available to authorized users.     

The effect of sample size and species characteristics on performance of different species distribution modeling methods

Species distribution models should provide conservation practioners with estimates of the spatial distributions of species requiring attention. These species are often rare and have limited known occurrences, posing challenges for creating accurate species distribution models. We tested four modeling methods (Bioclim, Domain, GARP, and Maxent) across 18 species with different levels of ecological specialization using six different sample size treatments and three different evaluation measures. Our assessment revealed that Maxent was the most capable of the four modeling methods in producing useful results with sample sizes as small as 5, 10 and 25 occurrences. The other methods compensated reasonably well (Domain and GARP) to poorly (Bioclim) when presented with datasets of small sample sizes. We show that multiple evaluation measures are necessary to determine accuracy of models produced with presence-only data. Further, we found that accuracy of models is greater for species with small geographic ranges and limited environmental tolerance, ecological characteristics of many rare species. Our results indicate that reasonable models can be made for some rare species, a result that should encourage conservationists to add distribution modeling to their toolbox.     

Woody plants and the prediction of climate-change impacts on bird diversity

Current methods of assessing climate-induced shifts of species distributions rarely account for species interactions and usually ignore potential differences in response times of interacting taxa to climate change. Here, we used species-richness data from 1005 breeding bird and 1417 woody plant species in Kenya and employed model-averaged coefficients from regression models and median climatic forecasts assembled across 15 climate-change scenarios to predict bird species richness under climate change. Forecasts assuming an instantaneous response of woody plants and birds to climate change suggested increases in future bird species richness across most of Kenya whereas forecasts assuming strongly lagged woody plant responses to climate change indicated a reversed trend, i.e. reduced bird species richness. Uncertainties in predictions of future bird species richness were geographically structured, mainly owing to uncertainties in projected precipitation changes. We conclude that assessments of future species responses to climate change are very sensitive to current uncertainties in regional climate-change projections, and to the inclusion or not of time-lagged interacting taxa. We expect even stronger effects for more specialized plant–animal associations. Given the slow response time of woody plant distributions to climate change, current estimates of future biodiversity of many animal taxa may be both biased and too optimistic.     

气候变化背景下紫楠在中国的适宜分布区模拟

为了解气候对紫楠(Phoebe sheareri)分布的影响,应用Maxent和GARP模型模拟了紫楠在当前气候下的中国适宜分布区,分析了影响其分布的主要环境因子,并预测了未来气候情境下其分布区的变化。结果表明,紫楠适宜分布在长江中下游及以南的各省区。影响紫楠分布的主要环境因子有年降雨量、最干季均温、降雨的季节性、相对湿度和6-8月的日照时数,这5个因子的累积贡献率达84.3%。在未来气候情境下,广东、云南、广西和海南等地区的适生区面积会显著锐减,而陕西中部、河南南部、安徽东部和江苏北部适生区面积会大幅度增加。因此,在未来气候变化背景下,紫楠的适宜分布区有向北扩张的趋势。     

Characterising and predicting benthic biodiversity for conservation planning in deepwater environments

Understanding patterns of biodiversity in deep sea systems is increasingly important because human activities are extending further into these areas. However, obtaining data is difficult, limiting the ability of science to inform management decisions. We have used three different methods of quantifying biodiversity to describe patterns of biodiversity in an area that includes two marine reserves in deep water off southern Australia. We used biological data collected during a recent survey, combined with extensive physical data to model, predict and map three different attributes of biodiversity: distributions of common species, beta diversity and rank abundance distributions (RAD). The distribution of each of eight common species was unique, although all the species respond to a depth-correlated physical gradient. Changes in composition (beta diversity) were large, even between sites with very similar environmental conditions. Composition at any one site was highly uncertain, and the suite of species changed dramatically both across and down slope. In contrast, the distributions of the RAD components of biodiversity (community abundance, richness, and evenness) were relatively smooth across the study area, suggesting that assemblage structure (i.e. the distribution of abundances of species) is limited, irrespective of species composition. Seamounts had similar biodiversity based on metrics of species presence, beta diversity, total abundance, richness and evenness to the adjacent continental slope in the same depth ranges. These analyses suggest that conservation objectives need to clearly identify which aspects of biodiversity are valued, and employ an appropriate suite of methods to address these aspects, to ensure that conservation goals are met.     

Spatial sampling bias and model complexity in stream‐based species distribution models: A case study of Paddlefish (Polyodon spathula) in the Arkansas River basin,USA

Leveraging existing presence records and geospatial datasets, species distribution modeling has been widely applied to informing species conservation and restoration efforts. Maxent is one of the most popular modeling algorithms, yet recent research has demonstrated Maxent models are vulnerable to prediction errors related to spatial sampling bias and model complexity. Despite elevated rates of biodiversity imperilment in stream ecosystems, the application of Maxent models to stream networks has lagged, as has the availability of tools to address potential sources of error and calculate model evaluation metrics when modeling in nonraster environments (such as stream networks). Herein, we use Maxent and customized R code to estimate the potential distribution of paddlefish (Polyodon spathula) at a stream‐segment level within the Arkansas River basin, USA, while accounting for potential spatial sampling bias and model complexity. Filtering the presence data appeared to adequately remove an eastward, large‐river sampling bias that was evident within the unfiltered presence dataset. In particular, our novel riverscape filter provided a repeatable means of obtaining a relatively even coverage of presence data among watersheds and streams of varying sizes. The greatest differences in estimated distributions were observed among models constructed with default versus AIC_C‐selected parameterization. Although all models had similarly high performance and evaluation metrics, the AIC_C‐selected models were more inclusive of westward‐situated and smaller, headwater streams. Overall, our results solidified the importance of accounting for model complexity and spatial sampling bias in SDMs constructed within stream networks and provided a roadmap for future paddlefish restoration efforts in the study area.