Performance Of Species Distribution Models Research Articles

Bioclimatic variables from precipitation and temperature records vs. remote sensing-based bioclimatic variables: Which side can perform better in species distribution modeling?

Bioclimatic variables are considered as an indispensable data type in species distribution modeling. Such variables are available from the WorldClim database for the entire earth surface and at various spatial resolutions. Moreover, convenient access to real-time satellite data and their products has recently created a new way to produce environmental variables. Therefore, in the present study, it was attempted to compare the performance of bioclimatic variables derived from precipitation and temperature instrumental records (scenario I) and variables derived from remote sensing data (scenario II) where both scenarios were from 2001 to 2017. The variables were employed to predict the distribution of Artemisia sieberi in central Iran through five Species Distribution Models (SDMs) such as Generalized Linear Model (GLM), Random Forest (RF), Classification Tree Analysis (CTA), Multivariate Adaptive Regression Splines (MARS), and Maximum Entropy (Maxent). The DEM layer was derived from 90-m Shuttle Radar Topography Mission (SRTM), 1-km MODIS land surface temperature and vegetation indices products, and downscaled PERSIANN-CDR precipitation data were employed as derivations of temperature and precipitation to produce bioclimatic variables for scenario II. The results obtained from independent sample t-test on AUCratio values derived from the correlative models showed that it had more satisfactory results when they were getting from the data of scenario II than the scenario I (p < .01). RF scored the highest partial AUC values (AUCratio) among the single models, and based on both scenarios, ensemble map was able to provide the most accurate predictions. There were also non-significant differences among the performance of RF, CTA and ensemble models under two scenarios (p < .01). Results emphasized the importance of bioclimatic variables derived from remote sensing to produce more up-to-date information and also to improve the predictive performance of SDMs. Finally, it was suggested that convenient access to reliable and up-to-date information can assist modelers to outline management practices well.

Testing whether ensemble modelling is advantageous for maximising predictive performance of species distribution models

Predictive performance is important to many applications of species distribution models (SDMs). The SDM ‘ensemble’ approach, which combines predictions across different modelling methods, is believed to improve predictive performance, and is used in many recent SDM studies. Here, we aim to compare the predictive performance of ensemble species distribution models to that of individual models, using a large presence–absence dataset of eucalypt tree species. To test model performance, we divided our dataset into calibration and evaluation folds using two spatial blocking strategies (checkerboard‐pattern and latitudinal slicing). We calibrated and cross‐validated all models within the calibration folds, using both repeated random division of data (a common approach) and spatial blocking. Ensembles were built using the software package ‘biomod2’, with standard (‘untuned’) settings. Boosted regression tree (BRT) models were also fitted to the same data, tuned according to published procedures. We then used evaluation folds to compare ensembles against both their component untuned individual models, and against the BRTs. We used area under the receiver‐operating characteristic curve (AUC) and log‐likelihood for assessing model performance. In all our tests, ensemble models performed well, but not consistently better than their component untuned individual models or tuned BRTs across all tests. Moreover, choosing untuned individual models with best cross‐validation performance also yielded good external performance, with blocked cross‐validation proving better suited for this choice, in this study, than repeated random cross‐validation. The latitudinal slice test was only possible for four species; this showed some individual models, and particularly the tuned one, performing better than ensembles. This study shows no particular benefit to using ensembles over individual tuned models. It also suggests that further robust testing of performance is required for situations where models are used to predict to distant places or environments.

Deciphering ecology from statistical artefacts: Competing influence of sample size, prevalence and habitat specialization on species distribution models and how small evaluation datasets can inflate metrics of performance

AbstractAimSample size and species characteristics, including prevalence and habitat specialization, can influence the predictive performance of species distribution models (SDMs). There is little agreement, however, on which metric of model performance to use. Here, we directly compare AUC and partial ROC as metrics of SDM performance through analyses on the effects of species traits and sample size on SDM performance.LocationThree counties dominated by agricultural lands and coniferous forest in Oregon's Willamette Valley and Coast Range ecoregions.MethodsWe systematically reduced a large avian point count dataset to alter sample sizes of 22 species of songbird. We used boosted regression trees to run SDMs for each species, quantified habitat specialization, and used mixed effects models to compare the influence of sample size, prevalence, and habitat specialization on SDM performance, calculated as AUC and partial ROC, across species. We calculated AUC and partial ROC with subset and independent evaluation data separately to more comprehensively investigate differences in metrics.ResultsWe found a positive quadratic effect of sample size and a strongly positive effect of habitat specialization on both metrics of model performance. We found a weak effect of prevalence on partial ROC and no effect in AUC. Contrary to expectations, when evaluated with a subset evaluation data, partial ROC was consistently highest in models with the smallest sample sizes. These small sample size models had correspondingly small sample sizes in subset evaluation datasets. Partial ROC evaluated with independent data and AUC evaluated with subset or independent data showed the expected positive correlation between sample size and model performance.Main ConclusionsWe found that small evaluation datasets can artificially inflate partial ROC. With literature recommended minimum SDM sample sizes as low as three, attention must be given to the effects of correspondingly low sample sizes in evaluation datasets.

Assessing the effect of fish size on species distribution model performance in southern Chilean rivers.

Despite its theoretical relationship, the effect of body size on the performance of species distribution models (SDM) has only been assessed in a few studies, and to date, the evidence shows unclear results. In this context, Chilean fishes provide an ideal case to evaluate this relationship due to their short size (fishes between 5 cm and 40 cm) and conservation status, providing evidence for species at the lower end of the worldwide fish size distribution and representing a relevant management tool for species conservation. We assessed the effect of body size on the performance of SDM in nine Chilean river fishes, considering the number of records, performance metrics, and predictor importance. The study was developed in the Bueno and Valdivia basins of southern Chile. We used a neural network modeling algorithm, training models with a cross-validation scheme. The effect of fish size on selected metrics was assessed using linear models and beta regressions. While no relationship between fish size and the number of presences was found, our results indicate that the model specificity increases with fish size. Additionally, the predictive importance of Riparian Vegetation and Within-Channel Structures variables decreases for larger species. Our results suggest that the relationship between the grain of the dataset and the home range of the species could bias SDM, leading in our case, to overprediction of absences. We also suggest that evolutionary adaptation to low slopes among Chilean fishes increases the relevance of riparian vegetation in the SDMs of smaller species. This study provides evidence on how species size may bias SDM, which could potentially be corrected by adjusting the model grain.

Employing inferences across scales: Integrating spatial data with different resolutions to enhance Maxent models

Challenges associated with developing species distribution models (SDMs) with high-resolution data (including from lidar) prompted our investigation into a complementary approach to enhance the performance of SDMs using spatial data with different resolutions. In our experiment we developed a model with Maxent (a presence-background SDM) with variables that had a 30-m resolution, and then used the output of the model to restrict the background sampling area for models developed with variables that had a 10-m resolution. According to common measures of model quality, this approach produced better models than both a model developed with the default Maxent background sampling area and a model developed using the conventional approach of resampling environmental data to a common spatial resolution. We then reviewed the ecological meaning of this approach and observed how model mechanics were impacted as restricting the background sampling areas led to background points that had a greater contrast with the presence points, and therefore different environmental characteristics than background points sampled from the default background sampling area.

LiDAR data as a proxy for light availability improve distribution modelling of woody species

Modern multifunctional forest management can profit from high-quality information on the potential distribution of woody species generated by species distribution models (SDMs). Forest structure is an important factor in determining the distribution of woody species in forests, for example because it affects light conditions within forest stands. Remotely sensed data from light detection and ranging (LiDAR) can capture this three-dimensional structure of forests, leading to the expectation that LiDAR-derived data should enhance the predictive performance of SDMs. We test if and how LiDAR-derived data increases the predictive performance of SDMs for light-demanding and shade-tolerant shrub and tree species in Swiss forests. Our analyses suggest that LiDAR-derived data generally increases predictive performance of models. However, the response to including LiDAR-derived data varies depending on plant functional type: the increase in predictive performance is largest for light-demanding shrubs, reduced for light-demanding trees, and is lost for shade-tolerant species. We further find that shade-tolerant and light-demanding species show opposing responses along the LiDAR-derived predictors. Our results suggest that LiDAR-derived data indeed capture some aspects of light availability in forests, and that including LiDAR-derived predictors in SDMs should be considered for light-demanding shrubs, but may be of less use for trees (especially if shade-tolerant). We conclude that improving SDMs and resulting maps by including LiDAR-derived predictors potentially helps to ameliorate multifunctional, biodiversity-friendly forest stand management.

An evaluation of stringent filtering to improve species distribution models from citizen science data

AbstractAimCitizen science data are increasingly used for modelling species distributions because they offer broad spatiotemporal coverage of local observations. However, such data are often collected without experimental design or set survey methods, raising the risk that bias and noise will compromise modelled predictions. We tested the ability of species distribution models (SDMs) built from these low‐structure citizen science data to match the quality of SDMs from systematically collected data and tested whether stringent data filtering improved predictions.LocationNortheastern USA.MethodsWe evaluated models built from a rapidly growing dataset of avian occurrences reported by birders—eBird—against models built from four independent, systematically collected datasets. We developed SDMs for 96 species using both data sources and compared their predictive abilities. We also tested whether culling eBird data by applying stringent data filters on survey effort or observer expertise improved predictions.ResultsWe found that SDMs built from low‐structure citizen science data matched or exceeded performance of SDMs from systematically collected datasets for 12%–31% of species ( = 22%), depending on the dataset. At least one culling option produced equivalent or better performance for 40%–70% of species ( = 49%). Data culling by restricting survey effort improved predictions more than restricting by observer expertise. The optimal effort restriction differed by dataset, and for three of the datasets was further informed by species traits.Main conclusionsSpecies distribution models developed using low‐structure citizen science data sometimes performed as well as those from systematic data. Culling generally improved models, but results were heterogeneous, prohibiting clear recommendations for how to cull. Our results indicate that the growing availability of citizen science data holds potential for creating high‐quality spatial predictions, but that time should be invested in determining how best to cull datasets and that one‐size‐fits‐all solutions beyond basic outlier filtering may be hard to find.

Scale effects on the performance of niche-based models of freshwater fish distributions: Local vs. upstream area influences

Niche-based species distribution models (SDMs) play a central role in studying species response to environmental change. Effective management and conservation plans for freshwater ecosystems require SDMs that accommodate hierarchical catchment ordering and provide clarity on the performance of such models across multiple scales. The scale-dependence components considered here are: (a) environment spatial structure, represented by hierarchical catchment ordering following the Strahler system; (b) analysis grain, that included 1st to 5th order catchments; and (c) response grain, the grain at which species respond most, represented by local and upstream catchment area effects. We used fish occurrence data from the Danube River Basin and various factors representing climate, land cover and anthropogenic pressures. Our results indicate that the choice of response grain – local vs. upstream area effects – and the choice of analysis grain, only marginally influence the performance of SDMs. Upstream effects tend to better predict fish distributions than corresponding local effects for anthropogenic and land cover factors, in particular for species sensitive to pollution. Key predictors and their relative importance are scale and species dependent. Consequently, choosing proper species dependent spatial scales and factors is imperative for effective river rehabilitation measures.

Incorporating biotic relationships improves species distribution models: Modeling the temporal influence of competition in conspecific nesting birds

AimComplex, biotic interactions are notably excluded from species distribution models (SDMs) as they are often difficult to quantify and accommodate in a traditional modeling framework, especially those with a temporal component. The territorial nature of breeding Cactus wren is well-documented and typically involves nest usurping (i.e., destruction) of conspecifics. Due to their similar nesting ecology, breeding Verdin are frequently the target of such behavior and are often forced to move or abandon nests. Using the Verdin/Cactus wren system as a case study, our goal was to evaluate the performance of SDMs that include only environmental predictors with SDMs that also include biotic relationships as predictors. LocationEast Foundation’s San Antonio Viejo Ranch in south Texas. MethodsWe built SDMs (MaxEnt, Boosted Regression Tree [BRT], and Random Forest [RF]) to project Verdin distribution during the early (April through mid-May), peak (mid-May through mid-June), and late (mid-June through mid-July) breeding periods using occurrence data collected during the 2015 and 2016 breeding seasons. We ran parallel analyses using relevant environmental features alone as predictors and then environmental features with observed Cactus wren density. ResultsRandom Forest (RF) produced the highest predictive performance SDMs for all three breeding periods (AUC = 0.81-0.99; TSS = 0.23-0.73). All models improved in predictive power (Δ AUC = 0.01-0.10) and model sensitivity (Δ TSS = 0.09-0.66) with the inclusion of Cactus wren density as a predictor of Verdin presence. Main conclusionsOur results indicate that SDM performance is improved by the inclusion of biotic relationships as predictors. Incorporating biotic interactions, as well as their temporal trends, is essential in efforts to monitor or conserve bird species with similar nesting ecologies. Further, modeling algorithms that can accommodate complex, non-linear relationships (e.g., Random Forest) should be preferred in SDM development and application.

Improving Species Distribution Modelling of freshwater invasive species for management applications.

Freshwater ecosystems rank among the most endangered ecosystems in the world and are under increasing threat from aquatic invasive species (AIS). Understanding the range expansion of AIS is key for mitigating their impacts. Most approaches rely on Species Distribution Models (SDMs) to predict the expansion of AIS, using mainly environmental variables, yet ignore the role of human activities in favouring the introduction and range expansion of AIS. In this study, we use five SDM algorithms (independently and in ensemble) and two accuracy measures (TSS, AUC), combined with a null modelling approach, to assess the predictive performance of the models and to quantify which predictors (environmental and anthropogenic from the native and introduced regions) best explain the distribution of nine freshwater invasive species (including fish, arthropods, molluscs, amphibians and reptiles) in a large island (Great Britain), and which species characteristics affect model performance. Our results show that the distribution of invasive species is difficult to predict by SDMs, even in cases when TSS and AUC model accuracy values are high. Our study strongly advocates the use of null models for testing SDMs performance and the inclusion of information from the native area and a variety of both human-related and environmental predictors for a more accurate modelling of the range expansion of AIS. Otherwise, models that only include climatic variables, or rely only on standard accuracy measures or a single algorithm, might result in mismanagement of AIS.

Additive Multivariate Gaussian Processes for Joint Species Distribution Modeling with Heterogeneous Data

Species distribution models (SDM) are a key tool in ecology, conservation and management of natural resources. Two key components of the state-of-the-art SDMs are the description for species distribution response along environmental covariates and the spatial random effect. Joint species distribution models (JSDMs) additionally include interspecific correlations which have been shown to improve their descriptive and predictive performance compared to single species models. Current JSDMs are restricted to hierarchical generalized linear modeling framework. These parametric models have trouble in explaining changes in abundance due, e.g., highly non-linear physical tolerance limits which is particularly important when predicting species distribution in new areas or under scenarios of environmental change. On the other hand, semi-parametric response functions have been shown to improve the predictive performance of SDMs in these tasks in single species models. Here, we propose JSDMs where the responses to environmental covariates are modeled with additive multivariate Gaussian processes coded as linear models of coregionalization. These allow inference for wide range of functional forms and interspecific correlations between the responses. We propose also an efficient approach for inference with Laplace approximation and parameterization of the interspecific covariance matrices on the euclidean space. We demonstrate the benefits of our model with two small scale examples and one real world case study. We use cross-validation to compare the proposed model to analogous semi-parametric single species models and parametric single and joint species models in interpolation and extrapolation tasks. The proposed model outperforms the alternative models in all cases. We also show that the proposed model can be seen as an extension of the current state-of-the-art JSDMs to semi-parametric models.

Scale effects on the performance of niche-based models of freshwater fish distributions

Niche-based species distribution models (SDMs) have become an essential tool in conservation and restoration planning. Given the current threats to freshwater biodiversity, it is of fundamental importance to address scale effects on the performance of niche-based SDMs of freshwater species’ distributions. The scale effects are addressed here in the context of hierarchical catchment ordering, considered as counterpart to coarsening grain-size by increasing grid-cell size. We combine fish occurrence data from the Danube River Basin, the hierarchical catchment ordering and multiple environmental factors representing topographic, climatic and anthropogenic effects to model fish occurrence probability across multiple scales. We focus on 1st to 5th order catchments. The spatial scale (hierarchical catchment order) only marginally influences the mean performance of SDMs, however the uncertainty of the estimates increases with scale. Key predictors and their relative importance are scale and species dependent. Our findings have useful implications for choosing proper species dependent spatial scales for river rehabilitation measures, and for conservation planning in areas where fine grain species data are unavailable.

How do species and data characteristics affect species distribution models and when to use environmental filtering?

ABSTRACT Species distribution models (SDMs) are widely used in ecology and conservation. However, their performance is known to be affected by a variety of factors related to species occurrence characteristics. In this study, we used a virtual species approach to overcome the difficulties associated with testing of combined effects of those factors on performance of presence-only SDMs when using real data. We focused on the individual and combined roles of factors related to response variable (i.e. sample size, sampling bias, environmental filtering, species prevalence, and species response to environmental gradients). Results suggest that environmental filtering is not necessarily helpful and should not be performed blindly, without evidence of bias in species occurrences. The more gradual the species response to environmental gradients is, the greater is the model sensitivity to an inappropriate use of environmental filtering, although this sensitivity decreases with higher species prevalence. Results show that SDMs are affected to the greatest degree by the species response to environmental gradients, species prevalence, and sample size. Models’ accuracy decreased with sample size below 300 presences. Furthermore, a high level of interactions among individual factors was observed. Ignoring the combined effects of factors may lead to misleading outcomes and conclusions.

Assessing effects of genetic, environmental, and biotic gradients in species distribution modelling

AbstractTo develop more reliable marine species distribution models (SDMs), we examine how genetic, climatic, and biotic interaction gradients give rise to prediction error in marine SDM. Genetic lineages with distinct ecological requirements spanning genetic gradients have yet to be treated separately in marine SDM, which are often constrained to modeling the potential distribution of one biological unit (e.g. lineage or species) at a time. By comparing SDM performance for the whole species or where observation and predictions were partitioned among geographically discontinuous genetic lineages, we first identified the appropriate biological unit for modeling sea scallop. Prediction errors, in particular contiguous omissions at the northern range margins were effectively halved in genetic lineage SDM (Total error=15%) verses whole species SDM. Remaining SDM prediction error was strongly associated with: i) Sharp climatic gradients (abrupt and persistent spatial shifts in limiting temperatures) found within continental shelf breaks and bottom channels. ii) A biotic gradient in the predation of sea scallop juveniles by the sand star within the Hudson Shelf USA. Our findings highlight how the accuracy of marine SDM is dependent on capturing the appropriate biological unit for modeling (e.g. lineages rather than species) and adequately resolving limiting abiotic and biotic interaction gradients.

How much is enough? Influence of number of presence observations on the performance of species distribution models

Abstract Species distribution modeling (SDM) can be useful for many applied purposes, e.g., mapping and monitoring of rare and endangered species. Sparse presence data are a recurrent, major obstacle to precise modeling of species distributions. Thus, knowing the minimum number of presences required to obtain reliable distribution models is of fundamental importance for applied use of SDM. This study uses a novel approach to assess the critical sample size (CSS) sufficient for an accurate prediction of species distributions with Maximum Entropy Modeling (MaxEnt). Large presence datasets for thirty insect species, ranging from generalists to specialists regarding their responses to main bioclimatic gradients, were used to produce reference distribution models. Models based on replicated subsamples of different size drawn randomly from the full dataset were compared to the reference model using the index of vector similarity distribution models. Models based on replicated subsamples of different size drawn randomly from the full dataset were compared to the reference model using the index of vector similarity (IVS). Two thresholds for IVS were determined based on comparison of nine reference models to random null models. The threshold values correspond to 0.95 and 0.99 probability that a model outperforms a random null model in terms of similarity to the reference dataset. For 90% of the species, clearly nonrandom models were obtained with less than 10 presence observations, and for 97% of the species with less than 15 presence observations. We conclude that the number of presence observations required to produce nonrandom models is generally low and, accordingly, that even sparse datasets may be useful for distribution modelling.

Effects of simulated observation errors on the performance of species distribution models

AbstractAimSpecies distribution information is essential under increasing global changes, and models can be used to acquire such information but they can be affected by different errors/bias. Here, we evaluated the degree to which errors in species data (false presences–absences) affect model predictions and how this is reflected in commonly used evaluation metrics.LocationWestern Swiss Alps.MethodsUsing 100 virtual species and different sampling methods, we created observation datasets of different sizes (100–400–1,600) and added increasing levels of errors (creating false positives or negatives; from 0% to 50%). These degraded datasets were used to fit models using generalized linear model, random forest and boosted regression trees. Model fit (ability to reproduce calibration data) and predictive success (ability to predict the true distribution) were measured on probabilistic/binary outcomes using Kappa, TSS, MaxKappa, MaxTSS and Somers'D (rescaled AUC).ResultsThe interpretation of models’ performance depended on the data and metrics used to evaluate them, with conclusions differing whether model fit, or predictive success were measured. Added errors reduced model performance, with effects expectedly decreasing as sample size increased. Model performance was more affected by false positives than by false negatives. Models with different techniques were differently affected by errors: models with high fit presenting lower predictive success (RFs), and vice versa (GLMs). High evaluation metrics could still be obtained with 30% error added, indicating that some metrics (Somers'D) might not be sensitive enough to detect data degradation.Main conclusionsOur findings highlight the need to reconsider the interpretation scale of some commonly used evaluation metrics: Kappa seems more realistic than Somers'D/AUC or TSS. High fits were obtained with high levels of error added, showing that RF overfits the data. When collecting occurrence databases, it is advisory to reduce the rate of false positives (or increase sample sizes) rather than false negatives.

BlockCV: An r package for generating spatially or environmentally separated folds for k‐fold cross‐validation of species distribution models

Abstract When applied to structured data, conventional random cross‐validation techniques can lead to underestimation of prediction error, and may result in inappropriate model selection. We present the r package blockCV, a new toolbox for cross‐validation of species distribution modelling. Although it has been developed with species distribution modelling in mind, it can be used for any spatial modelling. The package can generate spatially or environmentally separated folds. It includes tools to measure spatial autocorrelation ranges in candidate covariates, providing the user with insights into the spatial structure in these data. It also offers interactive graphical capabilities for creating spatial blocks and exploring data folds. Package blockCV enables modellers to more easily implement a range of evaluation approaches. It will help the modelling community learn more about the impacts of evaluation approaches on our understanding of predictive performance of species distribution models.

Effects of environmental data temporal resolution on the performance of species distribution models

Abstract Assessing how environmental changes affect the distribution and dynamics of animal populations is becoming increasingly important, enabling ecologists to better predict the effects of global warming, biodiversity reduction or habitat degradation. In this context, the use of remotely sensed and modeled environmental variables play a key role by providing spatially contiguous predictors for species distribution models. Nevertheless, little attention has been paid to how the temporal scale of the environmental predictors could affect predictions. Here, we investigate how different levels of temporal integration of environmental variables influence model performance. We present the case study of the distribution of the three most discarded species of the bottom trawl mixed-species fishery in the Cantabrian Sea from 2004 to 2008. We find clear improvements in model prediction performance when using environmental data aggregated by season, while the estimation of the relationship between species and their habitats is better represented by time-specific data. Our results also support the hypothesis that medium and lived species are more sensitive to seasonal environmental conditions, in contrast to results found for long-lived marine mammals and seabirds, indicating that their distributions are better predicted by long-term mean conditions.

Inclusion of edaphic predictors for enhancement of models to determine distribution of soil-transmitted helminths: the case of Zimbabwe

BackgroundReliable mapping of soil-transmitted helminth (STH) parasites requires rigorous statistical and machine learning algorithms capable of integrating the combined influence of several determinants to predict distributions. This study tested whether combining edaphic predictors with relevant environmental predictors improves model performance when predicting the distribution of STH, Ascaris lumbricoides and hookworms at a national scale in Zimbabwe.MethodsGeo-referenced parasitological data obtained from a 2010/2011 national survey indicating a confirmed presence or absence of STH among school children aged 10–15 years was used to calibrate ten species distribution models (SDMs). The performance of SDMs calibrated with a set of environmental and edaphic variables was compared to that of SDMs calibrated with environmental variables only. Model performance was evaluated using the true skill statistic and receiver operating characteristic curve.ResultsResults show a significant improvement in model performance for both A. lumbricoides and hookworms for all ten SDMs after edaphic variables were combined with environmental variables in the modelling of the geographical distribution of the two STHs at national scale. Using the top three performing models, a consensus prediction was developed to generate the first continuous maps of the potential distribution of the two STHs in Zimbabwe.ConclusionsThe findings from this study demonstrate significant model improvement if relevant edaphic variables are included in model calibration resulting in more accurate mapping of STH. The results also provide spatially-explicit information to aid targeted control of STHs in Zimbabwe and other countries with STH burden.

Effects of models and spatial resolutions on the species distribution model performance

Species distribution models (SDMs) forecast the probability of species presence based on their statistical relationships with habitat parameters. Many studies have compared SDM models and their accuracies, but few have surveyed the effects of the different parameters on SDMs’ performance. We tested the effects of spatial resolution and the model on SDMs’ performance. Ten models were utilized to build SDMs across different spatial resolutions for Hyaena hyaena in Iran. Our results showed that SDM models indicated a firm response to spatial resolution. Our results also revealed that SDMs’ performance strongly depended on the model and MaxEnt performed best across all spatial resolutions. This study indicated SDMs accuracy had not descending trend when spatial resolution raised, but all models showed a significant declining trend from 4 to 32 km.