The ghost of unbalanced species distribution data in geographical model predictions

Unbalanced samples are considered a drawback in predictive modelling of species' potential habitats, and a prevalence of 0.5 has been extensively recommended. We argue that unbalanced species distribution data are not such a problem from a statistical point of view, and that good models can be obtained provided that the right predictors and cut-off to convert probabilities into presence/absence are chosen. The effects of unbalanced prevalence should not be confused with those of low-quality data affected by false absences, low sample size, or unrepresentativeness of the environmental and spatial gradient. Finally, we point out the necessity of greater research effort aimed at improving both the quality of training data sets, and the processes of validating and testing of models.     

Predicting and mapping the spatial distribution of Chamaecyparis formosensis in central Taiwan in a GIS with species distribution models

Analyzing the relationship between species and environment is always a focal question of ecological research. In recent years species distribution models (SDMs) has been widely used to predict the spatial distribution of species. SDMs are numerical tools that combine observations and species occurrence or abundance with environmental variables to predict the spatial distribution of species across landscapes, sometimes requiring extrapolation in space and time. Chamaecyparis formosensis (Taiwan red cypress, TRCs) is a coniferous species endemic to Taiwan, where it natural grows in the central mountains at moderate to high altitudes of 800–2800 m, and most stands in the range of 1500–2150 m. It is threatened by habitat loss and over-cutting for its valuable timber. To preserve TRCs species and achieve sustainable use of biological resources, we choose TRCs as a target for the study to predict its distribution in central Taiwan.The pure forests of TRCs in the study area were mainly located in Pachsienshan (P), Shouchentashan (S) and Baigou Mountain (B) in central Taiwan, and the distribution data were originally obtained by The Third Survey of Forest Resources and Land Use in Taiwan. Elevation, slope, aspect, and three vegetation indices were derived from both SPOT-5 satellite images and DEM. GIS technique was used to overlay those factors. Discriminant analysis (DA), decision tree (DT) and maximum entropy (MAXENT), three commonly used SDMs, were applied based on above-mentioned six variables to predict the suitable habitat of TRCs, and to evaluate which the best model is in terms of accuracy and efficiency. Three experiment designs (ED1, ED2 and ED3) with different combinations of samples were used for model building and validation. The 200 target samples were collected from the site P–B, B–S and P–S for model building under ED1, ED2 and ED3 respectively, while the 100 samples were collected from the site S, P and B for model validation. All experiment designs had same 1350 background samples. The results showed that the overall accuracy and kappa coefficient of DT (96%, 0.88) was higher than that of MAXENT (91%, 0.70), and their accuracies were better than that of DA (84%, 0.58). All the three models were highly efficient in implementation of model construction and evaluation, while the DT model was difficult for generating the entire predicted map of potential habitat due to its complex conditional sentence. Vegetation indices derived from SPOT-5 satellite images could not improve model accuracy because of its insufficiency of spectral resolution and spatial resolution. High spatial resolution and spectral resolution remotely sensed imagery should be used in our future research to improve model performance and reliability.     

New frameworks for species surrogacy in monitoring highly variable coastal ecosystems: Applying the BestAgg approach to Mediterranean coastal lagoons

Benthic invertebrates are good indicators of aquatic ecosystem health. Yet, environmental monitoring and assessment of community changes in relation to both natural and human sources of disturbance involves considerable efforts for sample processing and time-consuming identifications of organisms, which make challenging large-scale and continuous monitoring programs required under the current regulatory frameworks at European scale. The use of higher taxa (e.g. families) as surrogates for species is a mainstream approach to reduce cost and time associated to fine taxonomic resolution in environmental studies, especially concerning macro-invertebrate communities. However, this approach of ‘taxonomic sufficiency’ simply relies on the static grouping of organisms in taxa belonging to a single higher taxonomic level irrespective of their ecological relevance or difficulties in their taxonomic identification, leading to unnecessary losses of taxonomic detail or ecological information. A new approach, namely the Best Practicable Aggregation of Species (BestAgg), has been recently developed as an alternative procedure for species surrogacy. BestAgg is based on aggregating species in the minimum number of surrogates sufficient to depict species-level community patterns consistently, while capitalizing on ecological information. Although the approach has been successfully applied to marine and freshwater invertebrate assemblages, its effectiveness in transitional water systems, where the complex and highly variable environmental conditions may affect the performance of surrogacy methods, still remain untested. Here, we applied the BestAgg approach to quantifying spatio-temporal patterns of variability of macro-invertebrate assemblages from Mediterranean coastal lagoons (Northern Adriatic Sea). Surrogates were defined using species-level data from a representative lagoon system, which served as pilot study. Then, they were used to analyze macro-invertebrate assemblages in two independent lagoons in the same geographic area. Results showed that BestAgg surrogates were effective in depicting multivariate patterns of macro-invertebrate assemblages as at species level over and beyond potential variations among the investigated lagoons. The use of families, following a more classic approach based on taxonomic sufficiency, also provided results comparable to those obtained using species. However, with respect to families, BestAgg surrogates allowed an estimated timesaving of 40% higher while still retaining an equivalent amount of information on species-level patterns. More importantly, BestAgg allowed exploiting different criteria of species aggregation, leading to a set of surrogates more aligned with ecological/functional characteristics of organisms, suggesting that BestAgg approach may provide a fresh perspective for optimizing trade-offs between pragmatism and the need for relevant ecological information in environmental assessment and monitoring.     

Understanding the conservation challenges and needs of culturally significant plant species through Indigenous Knowledge and species distribution models

Indigenous People in the Klamath River Basin have cared for and utilized ecosystems and component resources since time immemorial, proactively conserving species through continuous use and stewardship. Though many culturally significant plants are still tended and used by Indigenous people, many species are also experiencing prolonged stress from colonial forest management practices and environmental change. By integrating western and Indigenous ways of knowing, as part of a participatory and collaborative research and extension project, we present an approach to informing the conservation of four culturally significant plants (tanoak, evergreen huckleberry, beargrass, and iris) and understanding the influence of bioclimatic factors and stress on Indigenous people’s relationships with plants and the broader forest ecosystem. Mixed methods and ways of knowing generate a detailed assessment of each case study species that presence only species distribution models cannot supply alone. In this study we use MAXENT to model species distributions of our four study species and the flexible coding method in NVivo for qualitative interview and focus group data. Using species distribution models and 127 interviews and focus groups with cultural practitioners, we found significant shifts in huckleberry harvesting times, beargrass and iris cultural use quality, and tanoak acorn availability that must be addressed for the long-term vitality of these species and interconnected cultures and people. Tribes have generations of knowledge, experience, and connection to land that can help inform how to combat stressors and enhance productivity of forest foods and fibers and the health of forest ecosystems.     

Ecdysteroids from Vitex species: distribution and compilation of their 13C-NMR spectral data

Iridoids and ecdysteroids are found in some genera of the family Verbenaceae. In such cases, they are used as chemotaxonomic markers for the difficult task of taxonomic identification by using morphological characteristics of plants belonging to this family. The present work describes the distribution of ecdysteroids in plants from the genus Vitex from a review of previous work on seventeen Vitex species. In addition, (13)C-NMR data of the main ecdysteroids found in this genus are described. This study attempted to summarize previous research on ecdysteroids distribution in Vitex species with the addition of (13)C-NMR analysis to further refine the characterization of these compounds in the Verbenaceae family.     

Influences of temporal independence of data on modelling species distributions

Modelling species distributions has been widely used to understand present and future potential distributions of species, and can provide adaptation and mitigation information as references for conservation and management under climate change. However, various methods of data splitting to develop and validate functions of the models do not get enough attention, which may mislead the interpretation of predicted results. We used the Taiwanese endemic birds to test the influences of temporal independence of datasets on model performance and prediction. Training and testing data were considered to be independent if they were collected during different survey periods (1993–2004 and 2009–2010). The results indicated no significant differences of six model performance measures (AUC, kappa, TSS, accuracy, sensitivity, and specificity) among the combinations of training and testing datasets. Both species- and grid cell-based assessments differed significantly between predictions by the annual and pooled training data. We also found an average of 85.8% similarity for species presences and absences in different survey periods. The remaining dissimilarity was mostly caused by species observed in the late survey period but not in the early one. The method of data splitting, yielding training and testing data, is critical for resulting model species distributions. Even if similar model performance exists, different methods can lead to different species distributional maps. More attention needs to be given to this issue, especially when amplifying these models to project species distributions in a changing world.     

物种分布模型的发展及评价方法

物种分布模型已被广泛地应用于以保护区规划、气候变化对物种分布的影响等为目的的研究。回顾了已经得到广泛应用的多种物种分布模型,总结了评价模型性能的方法。基于物种分布模型的发展和应用以及性能评价中尚存在的问题,本文认为:在物种分布模型中集成样本选择模块能够避免模型预测过程中的过度拟合及欠拟合,增加变量选择模块可评估和降低变量之间自相关性的影响,增加生物因子以及将物种对环境的适应性机制(及扩散行为特征)和潜在分布模型进行结合,是提高模型预测性能的可行方案;在模型性能的评价方面,采用赤池信息量可对模型的预测性能进行客观评价。相关建议可为物种分布建模提供参考。     

Detecting outliers in species distribution data: Some caveats and clarifications on a virtual species study

Liu et al. (2018) used a virtual species approach to test the effects of outliers on species distribution models. In their simulations, they applied a threshold value over the simulated suitabilities to generate the species distributions, suggesting that using a probabilistic simulation approach would have been more complex and yield the same results. Here, we argue that using a probabilistic approach is not necessarily more complex and may significantly change results. Although the threshold approach may be justified under limited circumstances, the probabilistic approach has multiple advantages. First, it is in line with ecological theory, which largely assumes non‐threshold responses. Second, it is more general, as it includes the threshold as a limiting case. Third, it allows a better separation of the relevant intervening factors that influence model performance. Therefore, we argue that the probabilistic simulation approach should be used as a general standard in virtual species studies.     

How to resolve the SLOSS debate: Lessons from species-diversity models

The SLOSS debate - whether a single large reserve will conserve more species than several small - of the 1970s and 1980s never came to a resolution. The first rule of reserve design states that one large reserve will conserve the most species, a rule which has been heavily contested. Empirical data seem to undermine the reliance on general rules, indicating that the best strategy varies from case to case. Modeling has also been deployed in this debate. We may divide the modeling approaches to the SLOSS enigma into dynamic and static approaches. Dynamic approaches, covered by the fields of island equilibrium theory of island biogeography and metapopulation theory, look at immigration, emigration, and extinction. Static approaches, such as the one in this paper, illustrate how several factors affect the number of reserves that will save the most species.This article approaches the effect of different factors by the application of species-diversity models. These models combine species-area curves for two or more reserves, correcting for the species overlap between them. Such models generate several predictions on how different factors affect the optimal number of reserves. The main predictions are: Fewer and larger reserves are favored by increased species overlap between reserves, by faster growth in number of species with reserve area increase, by higher minimum-area requirements, by spatial aggregation and by uneven species abundances. The effect of increased distance between smaller reserves depends on the two counteracting factors: decreased species density caused by isolation (which enhances minimum-area effect) and decreased overlap between isolates. The first decreases the optimal number of reserves; the second increases the optimal number. The effect of total reserve-system area depends both on the shape of the species-area curve and on whether overlap between reserves changes with scale.The approach to modeling presented here has several implications for conservational strategies. It illustrates well how the SLOSS enigma can be reduced to a question of the shape of the species-area curve that is expected or generated from reserves of different sizes and a question of overlap between isolates (or reserves).     

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

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

Considerations regarding species distribution models for forest insects

1. Species distribution models, or SDMs, have become important decision support tools by answering fundamental questions about where species, including invasive species, are likely to survive and thrive based on environmental conditions.
2. For an inexperienced modeller or model reviewer, the terminology and technical aspects of SDMs can be overwhelming, and even well-trained modellers can struggle to understand the implications of various modelling choices.
3. Here, I outline some key considerations with respect to SDMs, focusing on their application to forest insects. Foremost, I assert that a model should be developed and evaluated with attention to relationships between an insect and its hosts, as those relationships determine much about the places the insect may occupy.
4. In my view, the most successful models are constructed carefully and incorporate honest assessments of their limitations, sources of error and uncertainty, and the degree of linkage between the model and the real-world circumstances it is meant to portray.

     

Transferable species distribution modelling: Comparative performance of Generalised Functional Response models

Predictive species distribution models (SDMs) are becoming increasingly important in ecology, in the light of rapid environmental change. However, the predictions of most current SDMs are specific to the habitat composition of the environments in which they were fitted. This may limit SDM predictive power because species may respond differently to a given habitat depending on the availability of all habitats in their environment, a phenomenon known as a functional response in resource selection. The Generalised Functional Response (GFR) framework captures this dependence by formulating the SDM coefficients as functions of habitat availability. The original GFR implementation used global polynomial functions of habitat availability to describe the functional responses. In this study, we develop several refinements of this approach and compare their predictive performance using two simulated and two real datasets. We first use local radial basis functions (RBF), a more flexible approach than global polynomials, to represent the habitat selection coefficients, and balance bias with precision via regularization to prevent overfitting. Second, we use the RBF-GFR and GFR models in combination with the classification and regression tree CART, which has more flexibility and better predictive powers for non-linear modelling. As further extensions, we use random forests (RFs) and extreme gradient boosting (XGBoost), ensemble approaches that consistently lead to variance reduction in generalization error. We find that the different methods are ranked consistently across the datasets for out-of-data prediction. The traditional stationary approach to SDMs and the GFR model consistently perform at the bottom of the ranking (simple SDMs underfit, and polynomial GFRs overfit the data). The best methods in our list provide non-negligible improvements in predictive performance, in some cases taking the out-of-sample R2 from 0.3 up to 0.7 across datasets. At times of rapid environmental change and spatial non-stationarity ignoring the effects of functional responses on SDMs, results in two different types of prediction bias (under-prediction or mis-positioning of distribution hotspots). However, not all functional response models perform equally well. The more volatile polynomial GFR models can generate biases through over-prediction. Our results indicate that there are consistently robust GFR approaches that achieve impressive gains in transferability across very different datasets.     

Global patterns of fern species diversity: An evaluation of fern data in GBIF

Despite that several studies have shown that data derived from species lists generated from distribution occurrence records in the Global Biodiversity Information Facility (GBIF) are not appropriate for those ecological and biogeographic studies that require high sampling completeness, because species lists derived from GBIF are generally very incomplete, Suissa et al. (2021) generated fern species lists based on data with GBIF for 100 km × 100 km grid cells across the world, and used the data to determine fern diversity hotspots and species richness–climate relationships. We conduct an evaluation on the completeness of fern species lists derived from GBIF at the grid–cell scale and at a larger spatial scale, and determine whether fern data derived from GBIF are appropriate for studies on the relations of species composition and richness with climatic variables. We show that species sampling completeness of GBIF is low (<40%) for most of the grid cells examined, and such low sampling completeness can substantially bias the investigation of geographic and ecological patterns of species diversity and the identification of diversity hotspots. We conclude that fern species lists derived from GBIF are generally very incomplete across a wide range of spatial scales, and are not appropriate for studies that require data derived from species lists in high completeness. We present a map showing global patterns of fern species diversity based on complete or nearly complete regional fern species lists.     

Bunching up the background betters bias in species distribution models

Sets of presence records used to model species’ distributions typically consist of observations collected opportunistically rather than systematically. As a result, sampling probability is geographically uneven, which may confound the model's characterization of the species’ distribution. Modelers frequently address sampling bias by manipulating training data: either subsampling presence data or creating a similar spatial bias in non‐presence background data. We tested a new method, which we call ‘background thickening’, in the latter category. Background thickening entails concentrating background locations around presence locations in proportion to presence location density. We compared background thickening to two established sampling bias correction methods – target group background selection and presence thinning – using simulated data and data from a case study. In the case study, background thickening and presence thinning performed similarly well, both producing better model discrimination than target group background selection, and better model calibration than models without correction. In the simulation, background thickening performed better than presence thinning when the number of simulated presence locations was low, and vice versa. We discuss drawbacks to target group background selection, why background thickening and presence thinning are conservative but robust sampling bias correction methods, and why background thickening is better than presence thinning for small sample sizes. Particularly, background thickening is advantageous for treating sampling bias when data are scarce because it avoids discarding presence records.     

Analysis of regional species distribution models based on radio-telemetry datasets from multiple small-scale studies

The identification of core habitat areas and resulting prediction maps are vital tools for land managers. Often, agencies have large datasets from multiple studies over time that could be combined for a more informed and complete picture of a species. Colorado Parks and Wildlife has a large database for greater sage-grouse (Centrocercus urophasianus) including 11 radio-telemetry studies completed over 12 years (1997–2008) across northwestern Colorado. We divided the 49,470-km² study area into 1-km² grids with the number of sage-grouse locations in each grid cell that contained at least 1 location counted as the response variable. We used a generalized linear mixed model (GLMM) using land cover variables as fixed effects and individual birds and populations as random effects to predict greater sage-grouse location counts during breeding, summer, and winter seasons. The mixed effects model enabled us to model correlations that may exist in grouped data (e.g., correlations among individuals and populations). We found only individual groupings accounted for variation in the summer and breeding seasons, but not the winter season. The breeding and summer seasonal models predicted sage-grouse presence in the currently delineated populations for Colorado, but we found little evidence supporting a winter season model. According to our models, about 50% of the study area in Colorado is considered highly or moderately suitable habitat in both the breeding and summer seasons. As oil and gas development and other landscape changes occur in this portion of Colorado, knowledge of where management actions can be accomplished or possible restoration can occur becomes more critical. These seasonal models provide data-driven, distribution maps that managers and biologists can use for identification and exploration when investigating greater sage-grouse issues across the Colorado range. Using historic data for future decisions on species management while accounting for issues found from combining datasets allows land managers the flexibility to use all information available. © 2013 The Wildlife Society.