Performance Of Species Distribution Models Research Articles

Building use-inspired species distribution models: Using multiple data types to examine and improve model performance.

Species distribution models (SDMs) are becoming an important tool for marine conservation and management. Yet while there is an increasing diversity and volume of marine biodiversity data for training SDMs, little practical guidance is available on how to leverage distinct data types to build robust models. We explored the effect of different data types on the fit, performance and predictive ability of SDMs by comparing models trained with four data types for a heavily exploited pelagic fish, the blue shark (Prionace glauca), in the Northwest Atlantic: two fishery dependent (conventional mark-recapture tags, fisheries observer records) and two fishery independent (satellite-linked electronic tags, pop-up archival tags). We found that all four data types can result in robust models, but differences among spatial predictions highlighted the need to consider ecological realism in model selection and interpretation regardless of data type. Differences among models were primarily attributed to biases in how each data type, and the associated representation of absences, sampled the environment and summarized the resulting species distributions. Outputs from model ensembles and a model trained on all pooled data both proved effective for combining inferences across data types and provided more ecologically realistic predictions than individual models. Our results provide valuable guidance for practitioners developing SDMs. With increasing access to diverse data sources, future work should further develop truly integrative modeling approaches that can explicitly leverage the strengths of individual data types while statistically accounting for limitations, such as sampling biases.

Causal inference and large‐scale expert validation shed light on the drivers of SDM accuracy and variance

AbstractAimTo develop a causal understanding of the drivers of Species distribution model (SDM) performance.LocationUnited Kingdom (UK).MethodsWe measured the accuracy and variance of SDMs fitted for 518 species of invertebrate and plant in the UK. Our measure of variance reflects variation among replicate model fits, and taxon experts assessed model accuracy. Using directed acyclic graphs, we developed a causal model depicting plausible effects of explanatory variables (e.g. species' prevalence, sample size) on SDM accuracy and variance and quantified those effects using a multilevel piecewise path model.ResultsAccording to our model, sample size and niche completeness (proportion of a species' niche covered by sampling) directly affect SDM accuracy and variance. Prevalence and range completeness have indirect effects mediated by sample size. Challenging conventional wisdom, we found that the effect of prevalence on SDM accuracy is positive. This reflects the facts that sample size has a positive effect on accuracy and larger sample sizes are possible for widespread species. It is possible, however, that the omission of an unobserved confounder biased this effect. Previous studies, which reported negative correlations between prevalence and SDM accuracy, conditioned on sample size.Main conclusionsOur model explicates the causal basis of previously reported correlations between SDM performance and species/data characteristics. It also suggests that niche completeness has similarly large effects on SDM accuracy and variance as sample size. Analysts should consider niche completeness, or proxies thereof, in addition to sample size when deciding whether modelling is worthwhile.

Data stochasticity and model parametrisation impact the performance of species distribution models: insights from a simulation study

Species distribution models (SDM) are widely used to describe and explain how species relate to their environment and predict their spatial distributions. As such, they are the cornerstone of most of spatial planning efforts worldwide. SDM can be implemented with a wide array of data types (presence-only, presence-absence, count...), which can either be point- or areal-based, and use a wide array of environmental conditions as predictor variables. The choice of the sampling type as well as the resolution of environmental conditions to be used are recognized as of crucial importance, yet we lack any quantification of the effects these decisions may have on SDM reliability. In the present work, we fill this gap with an unprecedented simulation procedure. We simulated 100 possible distributions of two different virtual species in two different regions. Species distribution were modelled using either segment- or areal-based sampling and five different spatial resolutions of environmental conditions. The SDM performances were inspected by statistical metrics, model composition, shapes of relationships and prediction quality. We provided clear evidence of stochasticity in the modelling process (particularly in the shapes of relationships): two dataset from the same survey, species and region could yield different results. Sampling type had stronger effects than spatial resolution on the final model relevance. The effect of coarsening the resolution was directly related to the resistance of the spatial features to changes of scale: SDM failed to adequately identify spatial distributions when the spatial features targeted by the species were diluted by resolution coarsening. These results have important implications for the SDM community, backing up some commonly accepted choices, but also by highlighting some up-to-now unexpected features of SDM (stochasticity). As a whole, this work calls for carefully weighted decisions in implementing models, and for caution in interpreting results.

Species Distribution Models: the delicate balance between signal and noise

A recommendation of: Charlotte Lambert, Auriane Virgili Data stochasticity and model parametrisation impact the performance of species distribution models: insights from a simulation study https://doi.org/10.1101/2023.01.17.524386

The Invasion History of Diaphorina citri (Hemiptera: Liviidae) in Puerto Rico: Past, Present, and Future Perspectives.

The Asian citrus psyllid, Diaphorina citri Kuwayama, 1908 (Hemiptera: Liviidae), is a phloem-feeding insect pest of Citrus L. ( Sapindales: Rutaceae), Murraya J. Koenig ex L. (Sapindales: Rutaceae), and other Rutaceae, native to the Indian subcontinent and transmits the citrus greening pathogens (huanglongbing). We herein evaluate the occurrence of D. citri in Puerto Rico at a fine scale and the potential invasion history of the Caribbean from the native range. We found that the genetic diversity of D. citri in Puerto Rico, using the mitochondrial genetic marker CO1, indicated that the Caribbean haplotype is unique and more related to Southwest Asia haplotypes than other groups. Results then suggest that the invasion of the Caribbean did not occur from the United States. Also, we found that isothermality was the abiotic variable that influenced D. citri species distribution model performance the most in current and future climate change scenarios. Given the data available from the Global Biodiversity Information Facility (GBIF) for present D.citri reported occurrence, future projections suggest that climate change might promote an expansion of the pest to other Caribbean islands and Latin American countries. Still, it might elicit a range contraction across the Cordillera Central in Puerto Rico, possibly promoting citriculture. Regarding pest management strategies, we must consider a holistic approach to understanding D. citri invasion and resilience in novel habitats by including broader and more robust genetic population structure studies and if its current distribution reflects projected models across the neotropics.

MaxEnt brings comparable results when the input data are being completed; Model parameterization of four species distribution models.

Species distribution models (SDMs) are practical tools to assess the habitat suitability of species with numerous applications in environmental management and conservation planning. The manipulation of the input data to deal with their spatial bias is one of the advantageous methods to enhance the performance of SDMs. However, the development of a model parameterization approach covering different SDMs to achieve well-performing models has rarely been implemented. We integrated input data manipulation and model tuning for four commonly-used SDMs: generalized linear model (GLM), gradient boosted model (GBM), random forest (RF), and maximum entropy (MaxEnt), and compared their predictive performance to model geographically imbalanced-biased data of a rare species complex of mountain vipers. Models were tuned up based on a range of model-specific parameters considering two background selection methods: random and background weighting schemes. The performance of the fine-tuned models was assessed based on recently identified localities of the species. The results indicated that although the fine-tuned version of all models shows great performance in predicting training data (AUC > 0.9 and TSS > 0.5), they produce different results in classifying out-of-bag data. The GBM and RF with higher sensitivity of training data showed more different performances. The GLM, despite having high predictive performance for test data, showed lower specificity. It was only the MaxEnt model that showed high predictive performance and comparable results for identifying test data in both random and background weighting procedures. Our results highlight that while GBM and RF are prone to overfitting training data and GLM over-predict nonsampled areas MaxEnt is capable of producing results that are both predictable (extrapolative) and complex (interpolative). We discuss the assumptions of each model and conclude that MaxEnt could be considered as a practical method to cope with imbalanced-biased data in species distribution modeling approaches.

Improving species distribution models for dominant trees in climate data-poor forests using high-resolution remote sensing

Species distribution models (SDMs) are useful for predictive and explanatory purposes, allowing biologists to identify how human and environmental factors influence distributions of plants and animals. Lack of high-resolution climatic variables is one of the challenges for accurately predicting distributions of organisms at local or landscape scales. This study used SDMs to predict four dominant tree species distributions in northern Iran temperate forests (400km2 area) using high-resolution Sentinel-2 data (20 m) and topographic variables. We divided the explanatory variables into four datasets with increasing complexity of Sentinel-2 data and modelled distributions using four statistical and machine learning algorithms: random forest (RF), generalized boosting model (GBM), generalized linear model (GLM), and generalized additive model (GAM). Our results suggested differences in the predictive performance of the four algorithms. We found the most complex dataset, including topographic variables, Sentinel-2 bands, and vegetation and soil indices, gave the best fit for the four tree species, improving the accuracy of models for the different species between 5 and 16%. We then selected the most complex dataset to produce an ensemble model of the modeling algorithms where evaluation criteria were varied for tree species. Our result showed that the performance of SDMs improved using different satellite remote sensing data including raw bands, topographic variables and indices in the Hyrcanian forest, northern Iran. Elevation was a more significant variable than Sentinel bands and Sentinel vegetation indices variables for predicting the tree species distributions. With the Hyrcanian Forest included in this study region of northern Iran recently declared a UNESCO World Heritage site, a key result of our improved species distribution maps for these four dominant tree species is to support conservation management of forest biodiversity in this region.

Testing the assumption of environmental equilibrium in an invasive plant species over a 130 year history

Invasive plants are an increasing threat to global biodiversity. Effective management depends on accurate predictions of their spread. However, modelling the geographic distribution of invasive species, particularly with correlative species distribution models (SDMs), is challenging. SDMs assume that species are in equilibrium with their environment (i.e. they occur in all suitable environments); this assumption is likely to be violated for a species that is actively invading new environments. This assumption is rarely assessed, and when violated can have consequences for model reliability. Using the invasive vine Vincetoxicum rossicum, we tested the hypotheses that: 1) invasive species' distribution in environmental and geographic space increase to a plateau over time; 2) this plateau is a useful proxy for equilibrium distribution, a key assumption underlying SDMs. We compare V. rossicum's expansion in environmental and geographic space between historical and current time periods and infer equilibrium when its distribution has remained stable for an extended period. We also compare the performance of SDMs from historical time periods in predicting the current geographic distribution of V. rossicum. We found that V. rossicum has reached equilibrium in environmental space, but is still expanding its geographic distribution. SDM performance was poor in the first 30 years following introduction, but improved as V. rossicum approached environmental equilibrium. Our findings demonstrate the power of including temporal dynamics and the need to consider environmental and geographic equilibrium separately when modelling the distribution of invasive species. In light of our findings, we address shortcomings of the current approach to defining an equilibrium distribution and present a new perspective for reconciling the potentially confounding influence of dispersal limitation when assessing equilibrium distribution.

Incorporating satellite remote sensing for improving potential habitat simulation of Prosopis cineraria (L.) Druce in United Arab Emirates

Predicting the suitable habitat of a species is one of the adequate and enhancing approach in biodiversity conservation planning and implementation. Species habitat distribution is closely linked with environmental factors and bioclimatic variables used as prediction variables in species distribution models (SDMs). Recent application of satellite remote sensing data and the bioclimatic variables has created an advanced way to improve the SDMs performance. In this study, MaxEnt was used to predict species habitat distribution. Our objectives are to assess the application of satellite remote sensing data in predicting the potential habitat suitability of the arid plant species. We have selected Prosopis cineraria (L.) Druce (Ghaf), the national tree of the UAE. We have chosen 33 environmental variables along with 90 species occurrences and for the final model simulation we have used three modeling scenarios. MaxEnt results showed that the model simulation with all key variables has substantially improved the potential habitat suitability prediction with a mean Area Under receiver operating Curve (AUC) value of 0.984, indicating a better predictive accuracy in the integration of satellite remote sensing data, edaphic variables and topographic parameters. Model results showed that the spatial proportions of the potential habitat suitability in the UAE consisted of high (2%), medium (6%) and low (9%) habitat suitability classes. The MaxEnt results revealed that precipitation of the coldest quarter (32.5%), NDVI (12.1%) and elevation (8%), had significant contribution to the potential habitat distribution of UAE Ghaf trees. Cold season precipitation is the most significant climate constraint on the habitat distribution. The most habitat suitability of Ghaf tree in the UAE was within a certain range of NDVI, LST, elevation, aspect, and precipitation patterns. This paper finding could be useful for environmental managers for implementing reforestation not only in UAE but also for the entire middle eastern regions with similar hyper arid conditions.

Positional errors in species distribution modelling are not overcome by the coarser grains of analysis

Abstract The performance of species distribution models (SDMs) is known to be affected by analysis grain and positional error of species occurrences. Coarsening of the analysis grain has been suggested to compensate for positional errors. Nevertheless, this way of dealing with positional errors has never been thoroughly tested. With increasing use of fine‐scale environmental data in SDMs, it is important to test this assumption. Models using fine‐scale environmental data are more likely to be negatively affected by positional error as the inaccurate occurrences might easier end up in unsuitable environment. This can result in inappropriate conservation actions. Here, we examined the trade‐offs between positional error and analysis grain and provide recommendations for best practice. We generated narrow niche virtual species using environmental variables derived from LiDAR point clouds at 5 × 5 m fine‐scale. We simulated the positional error in the range of 5 m to 99 m and evaluated the effects of several spatial grains in the range of 5 m to 500 m. In total, we assessed 49 combinations of positional accuracy and analysis grain. We used three modelling techniques (MaxEnt, BRT and GLM) and evaluated their discrimination ability, niche overlap with virtual species and change in realized niche. We found that model performance decreased with increasing positional error in species occurrences and coarsening of the analysis grain. Most importantly, we showed that coarsening the analysis grain to compensate for positional error did not improve model performance. Our results reject coarsening of the analysis grain as a solution to address the negative effects of positional error on model performance. We recommend fitting models with the finest possible analysis grain and as close to the response grain as possible even when available species occurrences suffer from positional errors. If there are significant positional errors in species occurrences, users are unlikely to benefit from making additional efforts to obtain higher resolution environmental data unless they also minimize the positional errors of species occurrences. Our findings are also applicable to coarse analysis grain, especially for fragmented habitats, and for species with narrow niche breadth.

Modelling the distribution of rare invertebrates by correcting class imbalance and spatial bias

AbstractAimSoil arthropods are important decomposers and nutrient cyclers, but are poorly represented on national and international conservation Red Lists. Opportunistic biological records for soil invertebrates are sparse, and contain few observations of rare species but a relatively large number of non‐detection observations (a problem known as class imbalance). Robinson et al. (Diversity and Distributions, 24, 460) proposed a method for under‐sampling non‐detection data using a spatial grid to improve class balance and spatial bias in bird data. For taxa that are less intensively sampled, datasets are smaller, which poses a challenge because under‐sampling data removes information. We tested whether spatially stratified under‐sampling improved prediction performance of species distribution models for millipedes, for which large datasets are not available. We also tested whether using environmental predictor variables provided additional information beyond what is captured by spatial position for predicting species distributions.LocationIsland of Ireland.MethodsWe tested the spatially stratified under‐sampling method of Robinson et al. (Diversity and Distributions, 24, 460) by using biological records to train species distribution models of rare millipedes.ResultsUsing spatially stratified under‐sampled data improved species distribution model sensitivity (true positive rate) but decreased model specificity (true negative rate). The spatial pattern of under‐sampling affected model performance. Training data that was under‐sampled in a spatially stratified way sometimes produced worse models than did data that was under‐sampled in an unstratified way. Geographic coordinates were as good as or better than environmental variables for predicting distributions of one out of six species.Main ConclusionsSpatially stratified under‐sampling improved prediction performance of species distribution models for rare millipedes. Spatially stratified under‐sampling was most effective for rarer species, although unstratified under‐sampling was sometimes more effective. The good prediction performance of models using geographic coordinates is promising for modelling distributions of poorly studied species for which little is known about ecological or physiological determinants of occurrence.

Do Two Different Approaches to the Season in Modeling Affect the Predicted Distribution of Fish? A Case Study for Decapterus maruadsi in the Offshore Waters of Southern Zhejiang, China

The relationships between environmental factors and fish density are closely related, and species distribution models (SDMs) have been widely used in exploring these relationships and predicting the spatial distribution of fishery resources. When exploring the prediction of the spatial distribution of species in different seasons, the method of choosing the appropriate approach to the season will help to improve the predictive performance of the model. Based on data collected from 2015 to 2020 during a survey off southern Zhejiang, the Tweedie-GAM was used to establish the relationship between the density of Decapterus maruadsi and environmental factors at different modeling approaches. The results showed that water temperature, salinity and depth were the main factors influencing D. maruadsi, and they operated through different mechanisms and even resulted in opposite trends of density in different seasons. Spatially, the two modeling approaches also differed in predicting the spatial distribution of D. maruadsi, with the seasonal model showing a higher density trend in inshore waters than in offshore waters in spring but showing the opposite trend in summer and autumn, which was more consistent with the actual spatial distribution of the resource. By analyzing the effects of two different approaches on the prediction of fishery resources, this study aims to provide research ideas and references for improving the predictive performance of SDMs.

A comparison in species distribution model performance of succulents using key species and subsets of environmental predictors.

Identifying the environmental drivers of the global distribution of succulent plants using the Crassulacean acid metabolism pathway of photosynthesis has previously been investigated through ensemble‐modeling of species delimiting the realized niche of the natural succulent biome. An alternative approach, which may provide further insight into the fundamental niche of succulent plants in the absence of dispersal limitation, is to model the distribution of selected species that are globally widespread and have become naturalized far beyond their native habitats. This could be of interest, for example, in defining areas that may be suitable for cultivation of alternative crops resilient to future climate change. We therefore explored the performance of climate‐only species distribution models (SDMs) in predicting the drivers and distribution of two widespread CAM plants, Opuntia ficus‐indica and Euphorbia tirucalli. Using two different algorithms and five predictor sets, we created distribution models for these exemplar species and produced an updated map of global inter‐annual rainfall predictability. No single predictor set produced markedly more accurate models, with the basic bioclim‐only predictor set marginally out‐performing combinations with additional predictors. Minimum temperature of the coldest month was the single most important variable in determining spatial distribution, but additional predictors such as precipitation and inter‐annual precipitation variability were also important in explaining the differences in spatial predictions between SDMs. When compared against previous projections, an a posteriori approach correctly does not predict distributions in areas of ecophysiological tolerance yet known absence (e.g., due to biotic competition). An updated map of inter‐annual rainfall predictability has successfully identified regions known to be depauperate in succulent plants. High model performance metrics suggest that the majority of potentially suitable regions for these species are predicted by these models with a limited number of climate predictors, and there is no benefit in expanding model complexity and increasing the potential for overfitting.

Effects of input data sources on species distribution model predictions across species with different distributional ranges

AbstractAimA major source of uncertainty in the application of species distribution models (SDMs) is related to input data quality. Citizen‐collected species occurrence data are often used for fitting SDMs when data from standardized and expert‐supported surveys are unavailable. Macroclimate variables are much more commonly used as predictors in SDMs than other sources coming from remote sensing data. Here, we assess the effects of using different data sources (in both response and predictor variables) on SDM performance across a wide range of species with contrasting distributional ranges.LocationIberian Peninsula.TaxonBirds.MethodsA SDM ensemble‐forecasting approach was implemented using bird data from two different data sources: the eBird project and Atlases. We fitted SDMs with three predictor types: macroclimate, remotely sensed ecosystem functional attributes (EFAs) and their combination. Species were grouped in four range size classes. We assessed the uncertainty of model predictions by different evaluation metrics. Generalized linear mixed‐effects models tested the effect on model performance of input data source across distributional range sizes while accounting for different accuracy metrics. Pairwise comparisons between range projections were used to assess their spatial similarity.ResultsData source, size class, predictor and accuracy metric showed significant effects on SDM performance. eBird‐based models outperformed those built with Atlas data for less widespread species. Climate predictors yielded models with the best performance, especially when combined with EFAs. However, the predictor contribution was consistent across bird datasets, being mostly driven by the species range.Main ConclusionsOur models demonstrated the usefulness and complementarity of different input data sources when modelling species distribution across different distributional ranges. These findings highlight the need to integrate different data sources to improve the model predictions at regional scale. Our framework also underlines that model uncertainty should be examined more exhaustively at early stages of the modelling process.

High resolution species distribution and abundance models cannot predict separate shrub datasets in adjacent Arctic fjords

AbstractAimImproving species distribution models (SDMs) and species abundance models (SAMs) of woody shrubs is critical for predicting biodiversity changes in the Arctic, which is experiencing especially high warming rates. Yet, it remains relatively unexplored if SDMs and SAMs can explain local scale patterns. We aim to identify predictor differences for the distribution versus abundance of two widespread Arctic shrub species with high resolution models and to compare validation approaches to assess the models’ predictive abilities.LocationNuup Kangerlua (NK) and Kangerluarsunnguaq (K), two adjacent fjords in Southwest Greenland.MethodsWe conducted two separate field surveys in either fjord to construct high resolution (~90 m) SDMs and SAMs forBetula nanaandSalix glauca, analysing the predictive influences of local scale climate, topography and soil moisture indicators. We then alternatively trained and validated models in either NK or K fjord and compared these results with the common split‐sample validation approach. Finally, we assessed if including local scale biotic predictors improves SAM performance.ResultsTemperature extremes and precipitation best predicted the distributions of both species, whereas insolation and soil moisture indicators best predicted abundances. Compared to split‐sample validation, both SDM and SAM performance was substantially reduced with separate survey validation. Regardless of validation approach, models performed poor to moderately well, and including local scale biotic parameters improved SAM performance.Main conclusionsSubstantial differences in model performance between validation approaches highlight the usefulness of using a separate survey for validating model predictive performance. We discuss various factors that might have caused poor model performance, such as not capturing all relevant predictors or enough local scale heterogeneity in predictor or response variables. We emphasise the need to include predictors relevant at the spatial scale of study, such as local scale biotic interactions, for improved predictions at high spatial resolution.

Predicting plant species distributions using climate‐based model ensembles with corresponding measures of congruence and uncertainty

AbstractAimThe increasing availability of regional and global climate data presents an opportunity to build better ecological models; however, it is not always clear which climate dataset is most appropriate. The aim of this study was to better understand the impacts that alternative climate datasets have on the modelled distribution of plant species, and to develop systematic approaches to enhancing their use in species distribution models (SDMs).LocationVictoria, southeast Australia and the Himalayan Kingdom of Bhutan.MethodsWe compared the statistical performance of SDMs for 38 plant species in Victoria and 12 plant species in Bhutan with multiple algorithms using globally and regionally calibrated climate datasets. Individual models were compared against one another and as SDM ensembles to explore the potential for alternative predictions to improve statistical performance. We develop two new spatially continuous metrics that support the interpretation of ensemble predictions by characterizing the per‐pixel congruence and variability of contributing models.ResultsThere was no clear consensus on which climate dataset performed best across all species in either study region. On average, multi‐model ensembles (across the same species with different climate data) increased AUC/TSS/Kappa/OA by up to 0.02/0.03/0.03/0.02 in Victoria and 0.06/0.11/0.11/0.05 in Bhutan. Ensembles performed better than most single models in both Victoria (AUC = 85%; TSS = 68%) and Bhutan (AUC = 86%; TSS = 69%). SDM ensembles using models fitted with alternative algorithms and/or climate datasets each provided a significant improvement over single model runs.Main conclusionsOur results demonstrate that SDM ensembles, built using alternative models of the same climate variables, can quantify model congruence and identify regions of the highest uncertainty while mitigating the risk of erroneous predictions. Algorithm selection is known to be a large source of error for SDMs, and our results demonstrate that climate dataset selection can be a comparably significant source of uncertainty.

Species profiles support recommendations for quality filtering of opportunistic citizen science data

Opportunistic citizen science data are commonly filtered in an attempt to improve their applicability for relating species occurrences with environmental variables. Recommendations on when and how to filter, however, have remained relatively general and associations between species traits and filtering recommendations are sparse. We collected six traits (body size, detectability, classification error rate, familiarity, reporting probability and range size) of 52 birds, 25 butterflies and 14 dragonflies. Both absolute (values not rescaled) and relative traits (values rescaled per taxonomic group) were linked to filter effects, i.e. the impact on three different measures of species distribution model performance caused by applying three different quality filters, for different degrees of sample size reduction. First, we applied multiple regressions that predicted the filter effects by either absolute (including taxonomic group) or relative traits. Second, a principal component and clustering analysis were performed to define five species profiles based on species traits that were retained after a multiple regression model selection. The analysis of the profiles indicated the relative importance of species traits and revealed new insights into the association of species traits with changes in model performance after data quality filtering. Both taxonomic group (more than absolute traits) and relative species traits (mainly classification error rate, range size and familiarity) defined the impact of data quality filtering on model performance and we discourage the selection of a quality filtering strategy based on one single species trait. Results further confirmed the importance of considering the goal of the study (i.e. increasing model discrimination capacity, sensitivity or specificity) as well as the change in sample size caused by stringent filtering. The general species knowledge amongst citizen scientists (importance of observer experience), together with the mechanism of record verification in an opportunistic data platform (importance of verifiable metadata) have the largest potential for enhancing the quality of opportunistic records.

Maximizing species distribution model performance when using historical occurrences and variables of varying persistency

AbstractOccurrence data used to build species distribution models often include historical records from locations in which the species no longer exists. When these records are paired with contemporary environmental values that no longer represent the conditions the species experienced, the model creates false associations that hurt predictive performance. The extent of mismatching increases with the number of historical occurrences and with inclusion of environmental variables that are prone to change over time. Indeed, the mismatch between occurrence data and contemporaneous environmental variables is a common dilemma when modeling rare or cryptic species, especially those of conservation concern that were once more abundant. Herein, we assess (1) the impact of historical occurrences on model performance across three sets of environmental variables of increasing persistency and (2) the performance of models built using selected‐historical occurrences from locations that showed evidence of limited environmental change over time. Concepts are tested on federally listed flatwoods salamanders, reflecting real‐world conservation management efforts. We predicted that, compared to other occurrence sets, (1) historical occurrences would perform best with environmental variables that were more persistent, (2) recent occurrences would perform best when the environmental variables were more impersistent, and that (3) our selected‐historical occurrences would perform best with a combination of persistent and impersistent variables. Our results showed the expected inversion of model performance of recent and historical occurrences across environmental variables of increasing persistency when evaluated by correct predictions. However, the inversion was not seen in area under the curve performance, in which historical occurrences outperformed recent occurrence models across all variable sets. Selected‐historical occurrences did not notably improve performance over all‐historical occurrences in any metric or variable set. To maximize utility and performance, modelers could acknowledge potential trade‐offs from inclusion of historical occurrences and consider number and age of recent and historical occurrences available, the persistency of environmental variables considered, and how their conservation goals are reflected in model design and evaluation, particularly with respect to sensitivity versus specificity. Our study lends support for inclusion of historical occurrences, with the potential exception of mostly impersistent variables when sensitivity is the highest priority.

A comparison of remotely sensed environmental predictors for avian distributions

With greater accessibility and processing power from online platforms, summaries of remotely sensed data are increasingly used in species distribution models (SDMs). Comparisons of the predictive power of these environmental variables could inform SDMs moving forward. We evaluated the performance of freely available Landsat data as predictor sets for SDMs. Our objectives were to (1) compare the performance of single season SDMs built on mean values of raw spectral bands, Tasseled Cap transformations, and eight different indices, including NDVI, (2) evaluate the performance gain with the addition of standard deviation, textural metrics, and additional seasons, and (3) compare the performance of SDMs built on these continuous spectral predictor sets to SDMs built on classified land cover data (e.g., percent forest cover). We used statewide point counts to build multi-scale SDMs for 13 avian species across Oregon, USA. We compared the performance of SDMs built on each predictor set based on our objectives. Of the Landsat-derived predictor sets, SDMs built on raw spectral bands had the highest overall performance with nearly equivalent performance in Tasseled-Cap models. While performance gains from standard deviations, textural metrics, and additional seasons were minimal in raw-band and Tasseled-Cap models, gains were appreciable in single-index models. Classified land cover models performed equivalently to raw band models. When predictive performance is paramount, means of raw Landsat bands are strong predictors for avian SDMs. When parsimonious variables are essential, SDMs of single indices (e.g., NDVI) greatly benefit from additional information, such as standard deviation.

Assessing the uncertainty arising from standard land‐cover mapping procedures when modelling species distributions

AbstractAimHabitat loss and degradation is one of the main threats to biodiversity worldwide. Ecological modelling community has been increasingly incorporating habitat changes in the species distribution models (SDMs). However, the effect that the uncertainty arising from the standard procedures of land‐use/land‐cover (LULC) mapping can have on SDMs has been overlooked. Remote sensing offers a great source of information to assess the habitat changes at different spatiotemporal scales. However, despite the great variety of satellite imagery preprocessing and supervised classification procedures currently available, their effects on SDMs remain largely unknown.LocationBaixa Limia–Serra do Xurés Natural Park, NW Spain.MethodsWe developed SDMs for 27 bird species with 4 modelling algorithms using 12 sets of the LULC predictors derived from the combination of 3 image preprocessing levels and 4 supervised classification algorithms.ResultsThe preprocessing level, classification method and modelling algorithm had a significant impact on the SDM performance and on the spatial and temporal predictions, with the considerable variation between species. Even if the effects of satellite image processing and classification procedures on the performance and transferability of SDMs were lower than that of the modelling algorithms themselves, the spatial and temporal predictions of SDMs were overall strongly affected by these LULC mapping procedures, sometimes overriding the effect of modelling algorithms.Main conclusionsWe argue that the uncertainty arising from the standard LULC mapping procedures when predicting species distribution dynamics should receive the same level of attention as the variability due to the different types of climate simulations and models often used in the ecological research. We also highlight the need to report accuracy of the satellite‐based LULC maps for each LULC class since some of these classes can be of key importance for certain types of the ecological applications.