Correcting for bias in distribution modelling for rare species using citizen science data

Aim

To improve the accuracy of inferences on habitat associations and distribution patterns of rare species by combining machine‐learning, spatial filtering and resampling to address class imbalance and spatial bias of large volumes of citizen science data.

Innovation

Modelling rare species’ distributions is a pressing challenge for conservation and applied research. Often, a large number of surveys are required before enough detections occur to model distributions of rare species accurately, resulting in a data set with a high proportion of non‐detections (i.e. class imbalance). Citizen science data can provide a cost‐effective source of surveys but likely suffer from class imbalance. Citizen science data also suffer from spatial bias, likely from preferential sampling. To correct for class imbalance and spatial bias, we used spatial filtering to under‐sample the majority class (non‐detection) while maintaining all of the limited information from the minority class (detection). We investigated the use of spatial under‐sampling with randomForest models and compared it to common approaches used for imbalanced data, the synthetic minority oversampling technique (SMOTE), weighted random forest and balanced random forest models. Model accuracy was assessed using kappa, Brier score and AUC. We demonstrate the method by evaluating habitat associations and seasonal distribution patterns using citizen science data for a rare species, the tricoloured blackbird (Agelaius tricolor).

Main Conclusions

Spatial under‐sampling increased the accuracy of each model and outperformed the approach typically used to direct under‐sampling in the SMOTE algorithm. Our approach is the first to characterize winter distribution and movement of tricoloured blackbirds. Our results show that tricoloured blackbirds are positively associated with grassland, pasture and wetland habitats, and negatively associated with high elevations or evergreen forests during both winter and breeding seasons. The seasonal differences in distribution indicate that individuals move to the coast during the winter, as suggested by historical accounts.

     

Vulnerability of Australian tropical savanna birds to climate change

Assessments of species vulnerability to climate change should increase the effectiveness of interventions in the current decline in biodiversity. Species vulnerability to climate change is a consequence of their sensitivity and adaptive capacity, in combination with their exposure to climate change. We apply a vulnerability assessment framework to 243 bird species inhabiting the tropical savannas of northern Australia. We build on previous vulnerability studies by including detailed data for variables relating to species sensitivity to change (relative abundance, clutch size, sensitivity to fire and distribution area), species adaptive capacity (movement behaviour and dietary breadth) and proportional changes predicted for their geographic range (i.e. exposure to climate change). These are integrated to provide a ranking of vulnerability. Our analysis found that birds of Australian tropical savannas cluster together with high sensitivity, with a few wide‐ranging increasing species with very low sensitivity. Australian tropical savanna birds have a range of adaptive capacities, and the impact of climate change on these species is predicted to be substantial. Two already endangered species are among the most vulnerable. Species largely restricted to Cape York Peninsula (a geographically distinct region) had the greatest overall vulnerability; these species were, in general, sensitive due to small distributions, sensitivity to fire frequency and had a lower capacity for dispersal. It will be important for the future of Australian tropical savanna birds to mitigate ecological threats and maintain extensive areas of suitable habitat to facilitate species dispersal.     

Is color data from citizen science photographs reliable for biodiversity research?

1. Color research continuously demands better methods and larger sample sizes. Citizen science (CS) projects are producing an ever‐growing geo‐ and time‐referenced set of photographs of organisms. These datasets have the potential to make a huge contribution to color research, but the reliability of these data need to be tested before widespread implementation.
2. We compared the difference between color extracted from CS photographs with that of color extracted from controlled lighting conditions (i.e., the current gold standard in spectrometry) for both birds and plants. First, we tested the ability of CS photographs to quantify interspecific variability by assessing > 9,000 CS photographs of 537 Australian bird species with controlled museum spectrometry data. Second, we tested the ability of CS photographs to quantify intraspecific variability by measuring petal color data for two plant species using seven methods/sources with varying levels of control.
3. For interspecific questions, we found that by averaging out variability through a large sample size, CS photographs capture a large proportion of across species variation in plumage color within the visual part of the spectrum (R² = 0.68–0.71 for RGB space and 0.72–0.77 for CIE‐LAB space). Between 12 and 14 photographs per species are necessary to achieve this averaging effect for interspecific studies. Unsurprisingly, the CS photographs taken with commercial cameras failed to capture information in the UV part of the spectrum. For intraspecific questions, decreasing levels of control increase the color variation but averaging larger sample sizes can partially mitigate this, aside from particular issues related to saturation and irregularities in light capture.
4. CS photographs offer a very large sample size across space and time which offers statistical power for many color research questions. This study shows that CS photographs contain data that lines up closely with controlled measurements within the visual spectrum if the sample size is large enough, highlighting the potential of CS photographs for both interspecific and intraspecific ecological or biological questions. With regard to analyzing color in CS photographs, we suggest, as a starting point, to measure multiple random points within the ROI of each photograph for both patterned and unpatterned patches and approach the recommended sample size of 12–14 photographs per species for interspecific studies. Overall, this study provides groundwork in analyzing the reliability of a novel method, which can propel the field of studying color forward.

     

Different facets of the same niche: Integrating citizen science and scientific survey data to predict biological invasion risk under multiple global change drivers

Citizen science initiatives have been increasingly used by researchers as a source of occurrence data to model the distribution of alien species. Since citizen science presence-only data suffer from some fundamental issues, efforts have been made to combine these data with those provided by scientifically structured surveys. Surprisingly, only a few studies proposing data integration evaluated the contribution of this process to the effective sampling of species' environmental niches and, consequently, its effect on model predictions on new time intervals. We relied on niche overlap analyses, machine learning classification algorithms and ecological niche models to compare the ability of data from citizen science and scientific surveys, along with their integration, in capturing the realized niche of 13 invasive alien species in Italy. Moreover, we assessed differences in current and future invasion risk predicted by each data set under multiple global change scenarios. We showed that data from citizen science and scientific surveys captured similar species niches though highlighting exclusive portions associated with clearly identifiable environmental conditions. In terrestrial species, citizen science data granted the highest gain in environmental space to the pooled niches, determining an increased future biological invasion risk. A few aquatic species modelled at the regional scale reported a net loss in the pooled niches compared to their scientific survey niches, suggesting that citizen science data may also lead to contraction in pooled niches. For these species, models predicted a lower future biological invasion risk. These findings indicate that citizen science data may represent a valuable contribution to predicting future spread of invasive alien species, especially within national-scale programmes. At the same time, citizen science data collected on species poorly known to citizen scientists, or in strictly local contexts, may strongly affect the niche quantification of these taxa and the prediction of their future biological invasion risk.     

Testing the utility of species distribution modelling using Random Forests for a species in decline

Habitat suitability estimates derived from species distribution models (SDMs) are increasingly used to guide management of threatened species. Poorly estimating species’ ranges can lead to underestimation of threatened status, undervaluing of remaining habitat and misdirection of conservation funding. We aimed to evaluate the utility of a SDM, similar to the models used to inform government regulation of habitat in our study region, in estimating the contemporary distribution of a threatened and declining species. We developed a presence‐only SDM for the endangered New Holland Mouse (Pseudomys novaehollandiae) across Victoria, Australia. We conducted extensive camera trap surveys across model‐predicted and expert‐selected areas to generate an independent data set for use in evaluating the model, determining confidence in absence data from non‐detection sites with occupancy and detectability modelling. We assessed the predictive capacity of the model at thresholds based on (1) sum of sensitivity and specificity (SSS), and (2) the lowest presence threshold (LPT; i.e. the lowest non‐zero model‐predicted habitat suitability value at which we detected the species). We detected P. novaehollandiae at 40 of 472 surveyed sites, with strong support for the species’ probable absence from non‐detection sites. Based on our post hoc optimised SSS threshold of the SDM, 25% of our detection sites were falsely predicted as non‐suitable habitat and 75% of sites predicted as suitable habitat did not contain the species at the time of our survey. One occupied site had a model‐predicted suitability value of zero, and at the LPT, 88% of sites predicted as suitable habitat did not contain the species at the time of our survey. Our findings demonstrate that application of generic SDMs in both regulatory and investment contexts should be tempered by considering their limitations and currency. Further, we recommend engaging species experts in the extrapolation and application of SDM outputs.     

Combining population genetics,species distribution modelling and field assessments to understand a species vulnerability to climate change

Climate change is recognized as a major threat to biodiversity. Multidisciplinary approaches that combine population genetics and species distribution modelling to assess these threats and recommend conservation actions are critical but rare. Combined, these methods provide independent verification and a more compelling case for developing conservation actions. This study integrates these data streams together with field assessments and spatial analyses to develop future genetic resource management recommendations. The study species was Callistemon teretifolius (Needle Bottlebrush), a shrub species endemic to the Mount Lofty and Flinders Ranges, South Australia, and potentially vulnerable to climate change. Chloroplast microsatellite and Amplified Fragment Length Polymorphism data were combined with species distribution modelling (MaxEnt), spatial analysis and field assessment to evaluate climate change vulnerability. Two major genetic groups were identified (Mount Lofty and Flinders Ranges). Populations in the Flinders Ranges, especially the Southern Flinders Ranges exhibited the highest genetic diversity, indicating a possible genetic refugium. Lower genetic diversity to the south in the Mount Lofty Ranges and north in the Gammon Ranges may be due to post‐glacial expansion into these areas from the Flinders Ranges or loss of alleles. Low levels of contemporary gene flow were identified, which suggests Callistemon teretifolius may have a limited capacity to respond to climate change through migration. Range restrictions were predicted for all future climates, especially in the north. It is likely that C. teretifolius will be adversely affected by climate change, due to limited gene flow, predicted range restriction and loss of suitable habitat. The Southern Flinders Ranges should be a priority for conservation because it contains the highest number of individuals and genetic diversity. We recommend monitoring and adaptive management involving restoration in the Southern Flinders Ranges, potentially incorporating genetic translocations from other areas to capture diversity, to assist C. teretifolius to adapt to climate change.     

Challenges and perspectives for species distribution modelling in the neotropics

The workshop 'Species distribution models: applications, challenges and perspectives' held at Belo Horizonte (Brazil), 29-30 August 2011, aimed to review the state-of-the-art in species distribution modelling (SDM) in the neotropical realm. It brought together researchers in ecology, evolution, biogeography and conservation, with different backgrounds and research interests. The application of SDM in the megadiverse neotropics-where data on species occurrences are scarce-presents several challenges, involving acknowledging the limitations imposed by data quality, including surveys as an integral part of SDM studies, and designing the analyses in accordance with the question investigated. Specific solutions were discussed, and a code of good practice in SDM studies and related field surveys was drafted.     

Applying climatically associated species pools to the modelling of compositional change in tropical montane forests

Aim Predictive species distribution modelling is a useful tool for extracting the maximum amount of information from biological collections and floristic inventories. However, in many tropical regions records are only available from a small number of sites. This can limit the application of predictive modelling, particularly in the case of rare and endangered species. We aim to address this problem by developing a methodology for defining and mapping species pools associated with climatic variables in order to investigate potential species turnover and regional species loss under climate change scenarios combined with anthropogenic disturbance. Location The study covered an area of 6800 km² in the highlands of Chiapas, southern Mexico. Methods We derived climatically associated species pools from floristic inventory data using multivariate analysis combined with spatially explicit discriminant analysis. We then produced predictive maps of the distribution of tree species pools using data derived from 451 inventory plots. After validating the predictive power of potential distributions against an independent historical data set consisting of 3105 botanical collections, we investigated potential changes in the distribution of tree species resulting from forest disturbance and climate change. Results Two species pools, associated with moist and cool climatic conditions, were identified as being particularly threatened by both climate change and ongoing anthropogenic disturbance. A change in climate consistent with low‐emission scenarios of general circulation models was shown to be sufficient to cause major changes in equilibrium forest composition within 50 years. The same species pools were also found to be suffering the fastest current rates of deforestation and internal forest disturbance. Disturbance and deforestation, in combination with climate change, threaten the regional distributions of five tree species listed as endangered by the IUCN. These include the endemic species Magnolia sharpii Miranda and Wimmeria montana Lundell. Eleven vulnerable species and 34 species requiring late successional conditions for their regeneration could also be threatened. Main conclusions Climatically associated species pools can be derived from floristic inventory data available for tropical regions using methods based on multivariate analysis even when data limitations prevent effective application of individual species modelling. Potential consequences of climate change and anthropogenic disturbance on the species diversity of montane tropical forests in our study region are clearly demonstrated by the method.     

Projected changes in distributions of Australian tropical savanna birds under climate change using three dispersal scenarios

Identifying the species most vulnerable to extinction as a result of climate change is a necessary first step in mitigating biodiversity decline. Species distribution modeling (SDM) is a commonly used tool to assess potential climate change impacts on distributions of species. We use SDMs to predict geographic ranges for 243 birds of Australian tropical savannas, and to project changes in species richness and ranges under a future climate scenario between 1990 and 2080. Realistic predictions require recognition of the variability in species capacity to track climatically suitable environments. Here we assess the effect of dispersal on model results by using three approaches: full dispersal, no dispersal and a partial-dispersal scenario permitting species to track climate change at a rate of 30 km per decade. As expected, the projected distributions and richness patterns are highly sensitive to the dispersal scenario. Projected future range sizes decreased for 66% of species if full dispersal was assumed, but for 89% of species when no dispersal was assumed. However, realistic future predictions should not assume a single dispersal scenario for all species and as such, we assigned each species to the most appropriate dispersal category based on individual mobility and habitat specificity; this permitted the best estimates of where species will be in the future. Under this "realistic" dispersal scenario, projected ranges sizes decreased for 67% of species but showed that migratory and tropical-endemic birds are predicted to benefit from climate change with increasing distributional area. Richness hotspots of tropical savanna birds are expected to move, increasing in southern savannas and southward along the east coast of Australia, but decreasing in the arid zone. Understanding the complexity of effects of climate change on species' range sizes by incorporating dispersal capacities is a crucial step toward developing adaptation policies for the conservation of vulnerable species.     

Detecting outliers in species distribution data

Aim

Species distribution data play a pivotal role in the study of ecology, evolution, biogeography and biodiversity conservation. Although large amounts of location data are available and accessible from public databases, data quality remains problematic. Of the potential sources of error, positional errors are critical for spatial applications, particularly where these errors place observations beyond the environmental or geographical range of species. These outliers need to be identified, checked and removed to improve data quality and minimize the impact on subsequent analyses. Manually checking all species records within large multispecies datasets is prohibitively costly. This work investigates algorithms that may assist in the efficient vetting of outliers in such large datasets.

Location

We used real, spatially explicit environmental data derived from the western part of Victoria, Australia, and simulated species distributions within this same region.

Methods

By adapting species distribution modelling (SDM), we developed a pseudo‐SDM approach for detecting outliers in species distribution data, which was implemented with random forest (RF) and support vector machine (SVM) resulting in two new methods: RF_pdSDM and SVM_pdSDM. Using virtual species, we compared eight existing multivariate outlier detection methods with these two new methods under various conditions.

Results

The two new methods based on the pseudo‐SDM approach had higher true skill statistic (TSS) values than other approaches, with TSS values always exceeding 0. More than 70% of the true outliers in datasets for species with a low and intermediate prevalence can be identified by checking 10% of the data points with the highest outlier scores.

Main conclusions

Pseudo‐SDM‐based methods were more effective than other outlier detection methods. However, this outlier detection procedure can only be considered as a screening tool, and putative outliers must be examined by experts to determine whether they are actual errors or important records within an inherently biased set of data.     

Long‐term changes in the migration phenology of UK breeding birds detected by large‐scale citizen science recording schemes

The timing of migration is one of the key life‐history parameters of migratory birds. It is expected to be under strong selection, to be sensitive to changing environmental conditions and to have implications for population dynamics. However, most phenological studies do not describe arrival and departure phenologies for a species in a way that is robust to potential biases, or that can be clearly related to breeding populations. This hampers our ability to understand more fully how climate change may affect species’ migratory strategies, their life histories and ultimately their population dynamics. Using generalized additive models (GAMs) and extensive large‐scale data collected in the UK over a 40‐year period, we present standardized measures of migration phenology for common migratory birds, and examine how the phenology of bird migration has changed in the UK since the 1960s. Arrival dates for 11 of 14 common migrants became significantly earlier, with six species advancing their arrival by more than 10 days. These comprised two species, Blackcap Sylvia atricapilla and Chiffchaff Phylloscopus collybita, which winter closest to Britain in southern Europe and the arid northern zone of Africa, Common Redstart Phoenicurus phoenicurus, which winters in the arid zone, and three hirundines (Sand Martin Riparia riparia, House Martin Delichon urbicum and Barn Swallow Hirundo rustica), which winter in different parts of Africa. Concurrently, departure dates became significantly later for four of the 14 species and included species that winter in southern Europe (Blackcap and Chiffchaff) and in humid zones of Africa (Garden Warbler Sylvia borin and Whinchat Saxicola rubetra). Common Swift Apus apus was the exception in departing significantly earlier. The net result of earlier arrival and later departure for most species was that length of stay has become significantly longer for nine of the 14 species. Species that have advanced their timing of arrival showed the most positive trends in abundance, in accordance with previous studies. Related in part to earlier arrival and the relationship above, we also show that species extending their stay in Great Britain have shown the most positive trends. Further applications of our modelling approach will provide opportunities for more robust tests of relationships between phenological change and population dynamics than have been possible previously.     

Climate change,genetic markers and species distribution modelling

Ecologists and biogeographers are currently expending great effort forecasting shifts in species geographical ranges that may result from climate change. However, these efforts are problematic because they have mostly relied on presence‐only data that ignore within‐species genetic diversity. Technological advances in high‐throughput sequencing have now made it cost‐effective to survey the genetic structure of populations sampled throughout the range of a species. These data can be used to delineate two or more genetic clusters within the species range, and to identify admixtures of individuals within genetic clusters that reflect different patterns of ancestry. Species distribution models (SDMs) applied to the presence and absence of genetic clusters should provide more realistic forecasts of geographical range shifts that take account of genetic variability. High‐throughput sequencing and spatially explicit models may be used to further refine these projections.     

Identifying and correcting spatial bias in opportunistic citizen science data for wild ungulates in Norway

Many publications make use of opportunistic data, such as citizen science observation data, to infer large‐scale properties of species’ distributions. However, the few publications that use opportunistic citizen science data to study animal ecology at a habitat level do so without accounting for spatial biases in opportunistic records or using methods that are difficult to generalize. In this study, we explore the biases that exist in opportunistic observations and suggest an approach to correct for them. We first examined the extent of the biases in opportunistic citizen science observations of three wild ungulate species in Norway by comparing them to data from GPS telemetry. We then quantified the extent of the biases by specifying a model of the biases. From the bias model, we sampled available locations within the species’ home range. Along with opportunistic observations, we used the corrected availability locations to estimate a resource selection function (RSF). We tested this method with simulations and empirical datasets for the three species. We compared the results of our correction method to RSFs obtained using opportunistic observations without correction and to RSFs using GPS‐telemetry data. Finally, we compared habitat suitability maps obtained using each of these models. Opportunistic observations are more affected by human access and visibility than locations derived from GPS telemetry. This has consequences for drawing inferences about species’ ecology. Models naïvely using opportunistic observations in habitat‐use studies can result in spurious inferences. However, sampling availability locations based on the spatial biases in opportunistic data improves the estimation of the species’ RSFs and predicted habitat suitability maps in some cases. This study highlights the challenges and opportunities of using opportunistic observations in habitat‐use studies. While our method is not foolproof it is a first step toward unlocking the potential of opportunistic citizen science data for habitat‐use studies.