Spatially explicit estimation of occupancy,detection probability and survey effort needed to inform conservation planning

Aim It is increasingly recognized the importance of accounting for imperfect detection in species distribution modelling and conservation planning. However, the integration of detectability into a spatially explicit frame has received little attention. We aim (1) to show how to develop distribution maps of both detection probability and survey effort required to reliably determine a species presence/absence and (2) to increase awareness of the spatial variation of detection error inherent in studies of species occurrence. Location North‐western Spain. Methods  We registered the presence/absence of the endangered Egyptian vulture (Neophron percnopterus) in 213 surveys performed in 40 of 104 territories once known to be occupied. We model simultaneously both detection probability and occurrence, using site occupancy modelling. With the resulting regression equations, we developed distribution maps of both detection probability and required sampling effort throughout the area. Results Of the studied territories, 72.5% were detected as occupied, but after accounting for imperfect detection, the proportion of sites truly occupied was 79%. Detectability decreased in territories with higher topographical irregularity and increased with both the time of day of the survey and the progress of the season. Spatial distribution of detectability showed a mainly north–south gradient following the distribution of slope in the area. The likelihood of occupancy increased with rockier, less forested surface and less topographical irregularity within the territory. A minimum of five surveys, on average, are needed to assess, with 95% probability, the occupancy status of a site, ranging from ≤ 3 to > 24 visits/territory depending on survey‐ and site‐specific features. Main conclusions Accounting for detectability and its sources of variation allows us to elaborate distribution maps of detectability‐based survey effort. These maps are useful tools to reliably assess (e.g. with 95% probability) occupancy status throughout a landscape and provide guidance for species conservation planning.     

Creating individual accessible area hypotheses improves stacked species distribution model performance

Aim

Stacked species distribution models (SDMs) are an important step towards estimating species richness, but frequently overpredict this metric and therefore erroneously predict which species comprise a given community. We test the idea that developing hypotheses about accessible area a priori can greatly improve model performance. By integrating dispersal ability via accessible area into SDM creation, we address an often‐overlooked facet of ecological niche modelling.

Innovation

By limiting the training and transference areas to theoretically accessible areas, we are creating more accurate SDMs on the basis of a taxon's explorable environments. This limitation of space and environment is a more accurate reflection of a taxon's true dispersal properties and more accurately reflects the geographical and environmental space to which a taxon is exposed. Here, we compare the predictive performance of stacked SDMs derived from spatially constrained and unconstrained training areas.

Main conclusions

Restricting a species’ training and transference areas to a theoretically accessible area greatly improves model performance. Stacked SDMs drawn from spatially restricted training areas predicted species richness and community composition more accurately than non‐restricted stacked SDMs. These accessible area‐based restrictions mimic true dispersal barriers to species and limit training areas to the suite of environments to those which a species is exposed to in nature. Furthermore, these restrictions serve to ‘clip’ predictions in geographical space, thus removing overpredictions in adjacent geographical regions where the species is known to be absent.     

Integrating detection probabilities in species distribution models of amphibians breeding in Mediterranean temporary ponds

Aim The imperfect detection of species may lead to erroneous conclusions about species–environment relationships. Accuracy in species detection usually requires temporal replication at sampling sites, a time‐consuming and costly monitoring scheme. Here, we applied a lower‐cost alternative based on a double‐sampling approach to incorporate the reliability of species detection into regression‐based species distribution modelling. Location Doñana National Park (south‐western Spain). Methods Using species‐specific monthly detection probabilities, we estimated the detection reliability as the probability of having detected the species given the species‐specific survey time. Such reliability estimates were used to account explicitly for data uncertainty by weighting each absence. We illustrated how this novel framework can be used to evaluate four competing hypotheses as to what constitutes primary environmental control of amphibian distribution: breeding habitat, aestivating habitat, spatial distribution of surrounding habitats and/or major ecosystems zonation. The study was conducted on six pond‐breeding amphibian species during a 4‐year period. Results Non‐detections should not be considered equivalent to real absences, as their reliability varied considerably. The occurrence of Hyla meridionalis and Triturus pygmaeus was related to a particular major ecosystem of the study area, where suitable habitat for these species seemed to be widely available. Characteristics of the breeding habitat (area and hydroperiod) were of high importance for the occurrence of Pelobates cultripes and Pleurodeles waltl. Terrestrial characteristics were the most important predictors of the occurrence of Discoglossus galganoi and Lissotriton boscai, along with spatial distribution of breeding habitats for the last species. Main conclusions We did not find a single best supported hypothesis valid for all species, which stresses the importance of multiscale and multifactor approaches. More importantly, this study shows that estimating the reliability of non‐detection records, an exercise that had been previously seen as a naïve goal in species distribution modelling, is feasible and could be promoted in future studies, at least in comparable systems.     

Assessing the effect of prevalence on the predictive performance of species distribution models using simulated data

Aim The proportion of sampled sites where a species is present is known as prevalence. Empirical studies have shown that prevalence can affect the predictive performance of species distribution models. This paper uses simulated species data to examine how prevalence and the form of species environmental dependence affect the assessment of the predictive performance of models. Methods Simulated species data were based on various functions of simulated environmental data with differing degrees of spatial correlation. Seven model performance measures – sensitivity, specificity, class‐average (CA), overall prediction success, kappa (κ), normalized mutual information (NMI) and area under the receiver operating characteristic curve (AUC) – were applied to species models fitted by three regression methods. The response of the performance measures to prevalence was then assessed. Three probability threshold selection methods used to convert fitted logistic model values to presence or absence were also assessed. Results The study shows that the extent to which prevalence affects model performance depends on the modelling technique and its degree of success in capturing dominant environmental determinants. It also depends on the statistic used to measure model performance and the probability threshold method. The response based on κ generally preferred models with medium prevalence. All performance measures were least affected by prevalence when the probability threshold was chosen to maximize predictive performance or was based directly on prevalence. In these cases, the responses based on AUC, CA and NMI generally preferred models with small or large prevalence. Main conclusions The effect of prevalence on the predictive performance of species distribution models has a methodological basis. Relevant factors include the success of the fitted distribution model in capturing the dominant environmental determinant, the model performance measure and the probability threshold selection method. The fixed probability threshold method yields a marked response of model performance to prevalence and is therefore not recommended. The study explains previous empirical results obtained with real data.     

Occupancy models including local and landscape variables are useful to assess the distribution of a salamander species at risk

Numerous amphibian species are at risk of extinction worldwide. Therefore, reliable estimations of the distribution and abundance of these species are necessary for their conservation. Generally, amphibians are difficult to detect in the wild, which compromises the accuracy of long-term population monitoring and management. Occupancy models are useful tools to assess how environmental variables, at a local and at a landscape scale, affect the distribution and abundance of organisms taking into account species imperfect detectability. In this study, we evaluated with an environmental multiscale approach the seasonal variation of the occupation area of the threatened salamander, Ambystoma ordinarium along its distribution range. We obtained readings in 60 streams of physicochemical variables associated with habitat quality and landscape features. We found that detection and occupation probability of A. ordinarium are seasonally associated with different environmental variables. During the dry season, detectability was positively associated with temperature and stream depth, whereas occupancy was positively associated with the proportion of crops in the landscape and stream elevation. In the rainy season, the detection probability was not explained by any variable considered, and occupancy was negatively associated with stream's electrical conductivity and dissolved oxygen. Based on the estimation of occupied sites, we showed that A. ordinarium presents a more restricted distribution range than previously projected. Therefore, our results reveal the importance of evaluating the accuracy of distribution estimates for the conservation of threatened species as A. ordinarium.     

Incorporating biotic interactions in the distribution models of African wild silk moths (Gonometa species,Lasiocampidae) using different representations of modelled host tree distributions

Biotic interactions influence species niches and may thus shape distributions. Nevertheless, species distribution modelling has traditionally relied exclusively on environmental factors to predict species distributions, while biotic interactions have only seldom been incorporated into models. This study tested the ability of incorporating biotic interactions, in the form of host plant distributions, to increase model performance for two host‐dependent lepidopterans of economic interest, namely the African silk moth species, Gonometa postica and Gonometa rufobrunnea (Lasiocampidae). Both species are dependent on a small number of host tree species for the completion of their life cycle. We thus expected the host plant distribution to be an important predictor of Gonometa distributions. Model performance of a species distribution model trained only on abiotic predictors was compared to four species distribution models that additionally incorporated biotic interactions in the form of four different representations of host plant distributions as predictors. We found that incorporating the moth–host plant interactions improved G. rufobrunnea model performance for all representations of host plant distribution, while for G. postica model performance only improved for one representation of host plant distribution. The best performing representation of host plant distribution differed for the two Gonometa species. While these results suggest that incorporating biotic interactions into species distribution models can improve model performance, there is inconsistency in which representation of the host tree distribution best improves predictions. Therefore, the ability of biotic interactions to improve species distribution models may be context‐specific, even for species which have obligatory interactions with other organisms.     

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.     

Using between-model comparisons to fine-tune linear models of species ranges

Aim To offer an objective approach to some of the problems associated with the development of logistic regression models: how to compare different models, determination of sample size adequacy, the influence of the ratio of positive to negative cells on model accuracy, and the appropriate scale at which the hypothesis of a non‐random distribution should be tested. Location Test data were taken from Southern Africa. Methods The approach relies mainly on the use of the AUC (Area under the Curve) statistic, based on ROC (threshold Receiver Operating Characteristic) plots, for between‐model comparisons. Data for the distribution of the bont tick Amblyomma hebraeum Koch (Acari: Ixodidae) are used to illustrate the methods. Results Methods for the estimation of minimum sample sizes and more accurate hypothesis‐testing are outlined. Logistic regression is robust to the assumption that uncollected cells can be scored as negative, provided that the sample size of cells scored as positive is adequate. The variation in temperature and rainfall at localities where A. hebraeum has been collected is significantly lower than expected from a random sample of points across the data set, suggesting that within‐site variation may be an important determinant of its distribution. Main conclusions Between‐model comparisons relying on AUCs can be used to enhance objectivity in the development and refinement of logistic regression models. Both between‐site and within‐site variability should be considered as potentially important factors determining species distributions.     

Latent Dirichlet Allocation reveals spatial and taxonomic structure in a DNA‐based census of soil biodiversity from a tropical forest

High‐throughput sequencing of amplicons from environmental DNA samples permits rapid, standardized and comprehensive biodiversity assessments. However, retrieving and interpreting the structure of such data sets requires efficient methods for dimensionality reduction. Latent Dirichlet Allocation (LDA) can be used to decompose environmental DNA samples into overlapping assemblages of co‐occurring taxa. It is a flexible model‐based method adapted to uneven sample sizes and to large and sparse data sets. Here, we compare LDA performance on abundance and occurrence data, and we quantify the robustness of the LDA decomposition by measuring its stability with respect to the algorithm's initialization. We then apply LDA to a survey of 1,131 soil DNA samples that were collected in a 12‐ha plot of primary tropical forest and amplified using standard primers for bacteria, protists, fungi and metazoans. The analysis reveals that bacteria, protists and fungi exhibit a strong spatial structure, which matches the topographical features of the plot, while metazoans do not, confirming that microbial diversity is primarily controlled by environmental variation at the studied scale. We conclude that LDA is a sensitive, robust and computationally efficient method to detect and interpret the structure of large DNA‐based biodiversity data sets. We finally discuss the possible future applications of this approach for the study of biodiversity.     

Refining predictions of climate change impacts on plant species distribution through the use of local statistics

Bioclimate envelope models are often used to predict changes in species distribution arising from changes in climate. These models are typically based on observed correlations between current species distribution and climate data. One limitation of this basic approach is that the relationship modelled is assumed to be constant in space; the analysis is global with the relationship assumed to be spatially stationary. Here, it is shown that by using a local regression analysis, which allows the relationship under study to vary in space, rather than conventional global regression analysis it is possible to increase the accuracy of bioclimate envelope modelling. This is demonstrated for the distribution of Spotted Meddick in Great Britain using data relating to three time periods, including predictions for the 2080s based on two climate change scenarios. Species distribution and climate data were available for two of the time periods studied and this allowed comparison of bioclimate envelope model outputs derived using the local and global regression analyses. For both time periods, the area under the receiver operating characteristics curve derived from the analysis based on local statistics was significantly higher than that from the conventional global analysis; the curve comparisons were also undertaken with an approach that recognised the dependent nature of the data sets compared. Marked differences in the future distribution of the species predicted from the local and global based analyses were evident and highlight a need for further consideration of local issues in modelling ecological variables.     

Soil nutritional factors improve models of plant species distribution: an illustration with Acer campestre (L.) in France

Aim  To estimate the relative importance of climate and soil nutritional variables for predicting the distribution of Acer campestre (L.) in French forests.
Location  France.
Methods  We used presence/absence information for A. campestre in 3286 forest plots scattered all over France, coupled with climatic and edaphic data. More than 150 climatic variables (temperature, precipitation, solar radiation, evapotranspiration, water balance) were obtained using a digital elevation model (DEM) and a geographical information system (GIS). Six direct soil variables (pH, C/N ratio, base saturation rate, concentrations of calcium, magnesium and potassium) were available from EcoPlant, a phytoecological data base for French forests. Using a forward stepwise logistic regression technique, we derived two distinct predictive models for A. campestre ; the first with climatic variables alone and the second with both climatic and edaphic variables.
Results  The distribution of A. campestre was poorly modelled when including only climatic variables. The inclusion of edaphic variables significantly improved the quality of predictions for this species, allowing prediction of patches of presence/absence within the study region.
Main conclusion  Soil nutritional variables may improve the performance of fine-scale (grain) plant species distribution models.