species distribution models

Top PDF species distribution models:

Accounting for biotic interactions through alpha-diversity constraints in stacked species distribution models

Accounting for biotic interactions through alpha-diversity constraints in stacked species distribution models

Kuemmerlen, M., Stoll, S., Haase, P. and Kunin, W. E. (2017), Accounting for biotic interactions through alpha-diversity constraints in stacked species distribution models. Methods Ecol Evol. doi:10.1111/2041-210X.12731, which has been published in final form at https://doi.org/10.1111/2041-210X.12731. This article may be used for non-commercial purposes in accordance with the Wiley Terms and Conditions for Self-Archiving.

10 Read more

Cross-validation of species distribution models: removing spatial sorting bias and calibration with a null model

Cross-validation of species distribution models: removing spatial sorting bias and calibration with a null model

Abstract. Species distribution models are usually evaluated with cross-validation. In this procedure evaluation statistics are computed from model predictions for sites of presence and absence that were not used to train (fit) the model. Using data for 226 species, from six regions, and two species distribution modeling algorithms (Bioclim and MaxEnt), I show that this procedure is highly sensitive to ‘‘spatial sorting bias’’: the difference between the geographic distance from testing-presence to training-presence sites and the geographic distance from testing-absence (or testing-background) to training-presence sites. I propose the use of pairwise distance sampling to remove this bias, and the use of a null model that only considers the geographic distance to training sites to calibrate cross-validation results for remaining bias. Model evaluation results (AUC) were strongly inflated: the null model performed better than MaxEnt for 45% and better than Bioclim for 67% of the species. Spatial sorting bias and area under the receiver–operator curve (AUC) values increased when using partitioned presence data and random-absence data instead of independently obtained presence–absence testing data from systematic surveys. Pairwise distance sampling removed spatial sorting bias, yielding null models with an AUC close to 0.5, such that AUC was the same as null model calibrated AUC (cAUC). This adjustment strongly decreased AUC values and changed the ranking among species. Cross-validation results for different species are only comparable after removal of spatial sorting bias and/or calibration with an appropriate null model.
Show more

10 Read more

Accounting for biotic interactions through alpha-diversity constraints in stacked species distribution models

Accounting for biotic interactions through alpha-diversity constraints in stacked species distribution models

Kuemmerlen, M., Stoll, S., Haase, P. and Kunin, W. E. (2017), Accounting for biotic interactions through alpha-diversity constraints in stacked species distribution models. Methods Ecol Evol. doi:10.1111/2041-210X.12731, which has been published in final form at https://doi.org/10.1111/2041-210X.12731. This article may be used for non-commercial purposes in accordance with the Wiley Terms and Conditions for Self-Archiving.

29 Read more

What's All the Fuzz About? Effects of FIA Spatial Precision on the Performance of Species Distribution Models.

What's All the Fuzz About? Effects of FIA Spatial Precision on the Performance of Species Distribution Models.

Species distribution models (SDMs) relate species occurrence or abundance at known locations with the environmental and/or spatial characteristics of those locations (Elith & Leathwick 2009). They are used for aquatic or terrestrial plants or animals. SDMs have a broad spectrum of implications, including predicting species occurrence (Iverson & Prasad 1998) and species abundance (Fei & Steiner 2007, Harris 1999, Iverson & Prasad 1998) at unmeasured locations, the potential spread of an exotic species into a new spatial domain (Morin 2003), and potential shifts in species distributions under projected climate change scenarios (Iverson & Prasad 2002, Woodall et al. 2009, Gibson 2013, Serra-Diaz et al. 2016). To effectively model species distributions, it is important to consider what environmental factors constrain the species. Environmental variables may include precipitation,
Show more

32 Read more

Predicting the distributions of under-recorded Odonata using species distribution models

Predicting the distributions of under-recorded Odonata using species distribution models

Species distribution models (SDMs) use recorded species presence and the conditions associated with those records to infer unrecorded presence where conditions are similar (Elith & Leathwick, 2009). Aquatic insects, and the Odonata (dragonflies and damselflies) in particular, should be strongly influenced by water-energy variables. Odonata exhibit relationships between various aspects of their biology and temperature (Hassall & Thompson, 2008) and these suggest a tropical evolutionary history (Pritchard & Leggott, 1987), with temperate species possessing adaptations to cooler climates which permit the colonisation of habitat that is available during interglacial periods. Odonates rely strongly and nearly universally on persistent water bodies for breeding and larval habitats. Two previous studies have investigated the use of species distribution models for odonate distributions. The first investigated the distributions of 160 species of South African Odonata using the BIOCLIM method of distribution modelling due to its ease of implementation within GIS (Finch et al., 2006). However, this yielded over-predictions of generalists and species with distributional outliers. Finch et al. recommend the use of probabilistic modelling methods instead of BIOLCLIM for more accurate modelling. The second study used only two species, Schistolobos boliviensis (Daigle, 2007) and Tuberculobasis inversa (Selys, 1876) and was restricted to the tropics (De Almeida et al., in press). No studies have yet been conducted on North American Odonata, although this approach has been advocated previously as it may provide assistance in searching for previously unrecorded species in new areas (Samways in Bried & Mazzacano, 2010).
Show more

22 Read more

Prevalence, statistical thresholds, and accuracy assessment for species distribution models

Prevalence, statistical thresholds, and accuracy assessment for species distribution models

Received: 16 July 2012 – Revised: 19 March 2013 – Accepted: 22 April 2013 – Published: 13 May 2013 Abstract. For species distribution models, species frequency is termed prevalence and prevalence in samples should be similar to natural species prevalence, for unbiased samples. However, modelers commonly adjust sampling prevalence, producing a modeling prevalence that has a di ff erent frequency of occurrences than sampling prevalence. The separate e ff ects of (1) use of sampling prevalence compared to adjusted modeling prevalence and (2) modifications necessary in thresholds, which convert continuous probabilities to discrete presence or absence predictions, to account for prevalence, are unresolved issues. We examined e ff ects of prevalence and thresholds and two types of pseudoabsences on model accuracy. Use of sampling prevalence produced similar models compared to use of adjusted modeling prevalences. Mean correlation between pre- dicted probabilities of the least (0.33) and greatest modeling prevalence (0.83) was 0.86. Mean predicted prob- ability values increased with increasing prevalence; therefore, unlike constant thresholds, varying threshold to match prevalence values was e ff ective in holding true positive rate, true negative rate, and species prediction areas relatively constant for every modeling prevalence. The area under the curve (AUC) values appeared to be as informative as sensitivity and specificity, when using surveyed pseudoabsences as absent cases, but when the entire study area was coded, AUC values reflected the area of predicted presence as absent. Less frequent species had greater AUC values when pseudoabsences represented the study background. Modeling prevalence had a mild impact on species distribution models and accuracy assessment metrics when threshold varied with prevalence. Misinterpretation of AUC values is possible when AUC values are based on background absences, which correlate with frequency of species.
Show more

7 Read more

Climatic limitation of alien weeds in New Zealand: enhancing species distribution models with field data

Climatic limitation of alien weeds in New Zealand: enhancing species distribution models with field data

Correlative species distribution models (SDMs) are often used to quantify the potential ranges of alien species. Despite rising popularity, there is ongoing debate surrounding whether SDMs can predict non-equilibrium species, how well they capture underlying biological mechanisms versus drawing spurious correlations, and how realistic the ensuing projections are. There have been numerous calls to integrate SDMs with real-world performance data to validate and improve projections, but such studies remain rare. In this thesis, I investigated the potential distributions of three alien plant species, Aeonium arboreum, A. haworthii and Cotyledon orbiculata, in their introduced ranges of New Zealand. I used a combination of SDMs, observational and experimental approaches. I firstly developed correlative SDMs for the three species. Secondly, I quantified the species’ climatic limits in the study region of Banks Peninsula, New Zealand, using field transplant experiments and surveys. Finally, I combined the aforementioned plant performance data into a single climate-driven population model, which I used to test and enhance the original SDM projections. I found that the New Zealand distributions of all three species are climatically novel relative to their distributions elsewhere, and constitute shifts in their realized niches. Although SDMs indicated that much of New Zealand is climatically suitable, transplant experiments on Banks Peninsula confirmed that the climate of Banks Peninsula is limiting. In all three species, low growth rates, low germination, and high mortality at high elevations will limit spread. In contrast, surveys found little evidence of direct climatic limitation to fecundity within the species’ current distributions on Banks Peninsula. The final step of validating SDM projections against the population model revealed that the SDM performed
Show more

205 Read more

A comprehensive evaluation of predictive performance of 33 species distribution models at species and community levels

A comprehensive evaluation of predictive performance of 33 species distribution models at species and community levels

With inventory data on multiple species, one can additionally make assumptions about how the rela- tionship between environmental covariates and species occurrences is structured among species (Table 2, Fea- ture C). The widely used stacked species distribution models are first fit separately for each species, after which their predictions are combined. They thus assume that species respond individualistically to vari- ation in environmental conditions (Williams and Jack- son 2007, Guisan and Rahbek 2011). By comparison, the more recently developed joint species distribution models (JSDMs) represent the response of entire spe- cies assemblages to environmental variation, assuming, for example, that species with similar traits have simi- lar responses (Warton et al. 2015, Ovaskainen et al. 2017). In complex communities, it is difficult to pre- dict a priori the joint structure of species responses to environmental variation and thus one might assume that treating each species individually is more in line with our limited current understanding of community assembly. However, treating each species individually may come with a higher risk of overfitting, while bor- rowing information from other species may increase predictive performance if the species respond similarly enough to abiotic variation (Ovaskainen and Soininen 2011, Hui et al. 2013, Madon et al. 2013, Maguire et al. 2016). Intermediately common species may show more statistically reliable relationships with environ- mental variables than rare species with wide and scattered distributions (Segurado and Ara ujo 2004), so treating assemblages as a whole can in effect increase the statistical power of detecting true envi- ronment – species relationships for rarer species within communities (Ovaskainen and Soininen 2011, Hui et al. 2013).
Show more

24 Read more

Influence of artefacts in marine digital terrain models on habitat maps and species distribution models: a multiscale assessment

Influence of artefacts in marine digital terrain models on habitat maps and species distribution models: a multiscale assessment

Remote sensing techniques are currently the main methods providing elevation data used to produce Digital Terrain Models (DTM). Terrain attributes (e.g. slope, orientation, rugosity) derived from DTMs are commonly used as surrogates of spe- cies or habitat distribution in ecological studies. While DTMs’ errors are known to propagate to terrain attributes, their impact on ecological analyses is however rarely documented. This study assessed the impact of data acquisition artefacts on habitat maps and species distribution models. DTMs of German Bank (off Nova Scotia, Canada) at five different spatial scales were altered to artificially introduce different levels of common data acquisition artefacts. These data were used in 615 unsupervised classifications to map potential habitat types based on biophysical characteristics of the area, and in 615 supervised classifications (MaxEnt) to predict sea scallop distribution across the area. Differences between maps and models built from altered data and reference maps and models were assessed. Roll artefacts decreased map accuracy (up to 14% lower) and artificially increased models’ per- formances. Impacts from other types of artefacts were not consistent, either decreasing or increasing accuracy and performance measures. The spatial distribu- tion of habitats and spatial predictions of sea scallop distributions were always affected by data quality (i.e. artefacts), spatial scale of the data, and the selection of variables used in the classifications. This research demonstrates the importance of these three factors in building a study design, and highlights the need for error quantification protocols that can assist when maps and models are used in deci- sion-making, for instance in conservation and management.
Show more

15 Read more

Climate suitability for European ticks: assessing species distribution models against null models and projection under AR5 climate

Climate suitability for European ticks: assessing species distribution models against null models and projection under AR5 climate

The selection of appropriate predictor variables is crit- ical to the generation of robust and reliable SDMs [17]; it is therefore imperative to consider the ecology of the modelled species. Ticks are haematophagous and require one or more hosts to complete their life cycle, but con- siderable periods of their life are spent off-host (>95 % of life for most Ixodes ticks [18]), where environmental factors strongly influence their activity, demographic rates, and distribution [2, 19]. Key amongst these envir- onmental determinants is climate, as discussed by several in-depth reviews [2, 18, 20, 21]. Although micro- climate is a more direct influence on a tick than macro- climate, only data for the latter are available at this large spatial scale. Several authors have used macroclimatic temperature and rainfall variables to model tick distribu- tion (e.g. [22, 23]), although it is also important to in- clude a measure of water stress for ticks as this is not effectively represented by rainfall [24]. Saturation deficit quantifies the ‘drying power’ of the air [25], which can drive tick mortality through cuticular water loss [24, 26]. It has been suggested that saturation deficit better repre- sents the constraining influence of water stress than relative humidity does (see [27]) and it has emerged as a better predictor than relative humidity in a recent popu- lation model for I. ricinus [28]. Ticks are known to seek favourable relative humidity in microclimates [18] where they may actively absorb water [29]; however, available data on macroclimatic humidity is not representative of that in a tick ’ s microclimate [30]. As ticks are ectother- mal, temperature also affects development rates and ac- tivity [21, 31, 32], and temperature and water availability in concert can therefore influence geographic distribu- tions. The northern distribution of Ixodes ricinus in Eur- ope, for example, is thought to be limited by low temperatures, whilst high temperatures and saturation deficit govern its southern distributional limit (see [19]).
Show more

15 Read more

Accounting for preferential sampling in species distribution models

Accounting for preferential sampling in species distribution models

In studies of species distribution, collecting data on the species of interest is not a trivial problem ([Kery et al., 2010). With the excep‐ tion of a few studies (Thogmartin, Knutson, & Sauer, 2006), SDMs frequently rely on opportunistic data collection due to the high cost and time consuming nature of sampling data in the field, especially on a large spatial scale (Kery et al., 2010). Indeed, it is often infea‐ sible to collect data based on a well‐designed, randomized, and/or systematic sampling scheme to estimate the distribution of a specific species over the entire area of interest (Brotons, Herrando, & Pla, 2007). Hence, various types of opportunistic sampling schemes are commonly used. As an example, studies on sea mammals commonly resort to the affordable practice of sampling from recreational boats (so‐called platforms of opportunity), whose bearings are neither random nor systematic (Rodríguez, Brotons, Bustamante, & Seoane, 2007). Similarly, bird data are often derived from online databases such as eBird, which make available locations of birds sighted by bird‐ watchers, who tend to visit habitats suitable for interesting species (https://ebird.org/). Also, in the context of fishery ecology, fishery‐ dependent survey data are often derived from commercial fleets tend to be readily available for analysis. However, the fishing boats naturally tend to fish in locations where they expect a high concen‐ tration of their target species (Vasconcellos & Cochrane, 2005).
Show more

11 Read more

Sensitivity of fine scale species distribution models to locational uncertainty in occurrence data across multiple sample sizes

Sensitivity of fine scale species distribution models to locational uncertainty in occurrence data across multiple sample sizes

Although not widely recognized, observation data col- lected using modern positioning systems invariably contain locational uncertainty. For example, the current locational accuracy of most standard GPS units can be ~ 30 m (Frair et al. 2010). While this is small compared to those con- tained in digitized records, when these data sets are incor- porated into a fine-scale SDM framework, this minor locational error affects the accuracy of model predictions (Guisan et al. 2007). With technological advances in the collection of environmental data sets, SDMs are being built at increasingly finer resolutions, not more so than in the marine environment, where multibeam echosounders (MBESs), along with other techniques, are now capable of providing seafloor structure information at resolutions of < 2 m (Brown et al. 2011). Consequently, locational uncer- tainty continues to be problematic despite the development of improved positioning systems (Rigby, Pizarro & Wil- liams 2006). In a recent study, Rattray et al. (2014) quanti- fied the propagated error associated with each component of underwater camera positioning (a technique commonly used to collect observation data in marine ecosystems). They found a linear increase in location error with camera depth, equating to a 1 5 m horizontal error near the surface and 5 7 m error at a depth of 100 m. This suggests that the maximum error in location of a species observation may often exceed the resolution of the predictor data sets, and, thus, locational uncertainty remains an issue with data sets collected using modern positioning systems.
Show more

10 Read more

The influence of life history characteristics on flea (Siphonaptera) species distribution models

The influence of life history characteristics on flea (Siphonaptera) species distribution models

Preliminary climate and landscape variables were se- lected based on our knowledge of flea ecology, limiting candidate variables to only include predictors that are considered ecologically relevant to flea species (following [30–32]). All predictor variables and flea occurrence data were converted to Quarter Degree Grid Cell (QDGC) scale and cropped to the borders of South Africa. Five remotely-sensed climate-based variables (daytime land surface temperature (hereafter referred to as day temperature), Leaf Area Index (LAI), Normalised Difference Vegetation Index (NDVI), rainfall, water vapour, and soil characteristics) and one landscape fea- ture variable (Topography) were extracted from the NASA-NEO website (http://neo.sci.gsfc.nasa.gov/about/) as potential predictor variables (missing values were esti- mated as the average of contiguous cells). Climate is known to generally influence flea populations to a greater extent than host species, especially at regional and local scales [20, 33], with air temperature, rainfall and relative humidity being important for flea survival, see [1, 10, 33–36]. NDVI is widely used in arthropod vector distribution modelling and is a measure of pri- mary productivity (plant photosynthetic activity), and therefore can be considered as a proxy for general
Show more

10 Read more

Evaluation of Deep Species Distribution Models using Environment and Co-occurrences

Evaluation of Deep Species Distribution Models using Environment and Co-occurrences

This paper compared four main types of models aimed at predicting species dis- tribution: (i) a convolutional neural network trained on environmental variables extracted around the location of interest, (ii) a purely spatial model trained with a random forest, (iii) a co-occurrence based model aimed at predicting the like- lihood of presence of a given species thanks to the knowledge of the presence of other species, and (iv), two fusions models between the environmental CNN and the co-occurrences model, one late fusion of predictions and one learned jointly on the to inputs. Our study shows that the convolutional neural network model maintains a high score with unbiased environmental patches. Indeed, it achieved the best performance over the others GeoLifeCLEF 2018 submitted models. However the main contribution of our study is the new joint model on environment and co-occurrences that achieve good results, significantly better than the environmental CNN. This shows that there is useful information in co-occurrences and that this information is at least partly complementary to environmental information. Few studies currently use this co-occurrences in- formation. It would be interesting, in future work, to study more about how useful is the information in co-occurrences and how its complementarity with the environment can be explained.
Show more

15 Read more

Species distribution modelling: contrasting presence-only models with plot abundance data

Species distribution modelling: contrasting presence-only models with plot abundance data

Species distribution models (SDMs) are widely used in ecology and conservation. Presence-only SDMs such as MaxEnt frequently use natural history collections (NHCs) as occurrence data, given their huge numbers and accessibility. NHCs are often spatially biased which may generate inaccuracies in SDMs. Here, we test how the distribution of NHCs and MaxEnt predictions relates to a spatial abundance model, based on a large plot dataset for Amazonian tree species, using inverse distance weighting (IDW). We also propose a new pipeline to deal with inconsistencies in NHCs and to limit the area of occupancy of the species. We found a significant but weak positive relationship between the distribution of NHCs and IDW for 66% of the species. The relationship between SDMs and IDW was also significant but weakly positive for 95% of the species, and sensitivity for both analyses was high. Furthermore, the pipeline removed half of the NHCs records. Presence-only SDM applications should consider this limitation, especially for large biodiversity assessments projects, when they are automatically generated without subsequent checking. Our pipeline provides a conservative estimate of a species’ area of occupancy, within an area slightly larger than its extent of occurrence, compatible to e.g. IUCN red list assessments.
Show more

12 Read more

Species distribution modelling: contrasting presence only models with plot abundance data

Species distribution modelling: contrasting presence only models with plot abundance data

Species distribution models (SDMs) are widely used in ecology and conservation. Presence-only SDMs such as MaxEnt frequently use natural history collections (NHCs) as occurrence data, given their huge numbers and accessibility. NHCs are often spatially biased which may generate inaccuracies in SDMs. Here, we test how the distribution of NHCs and MaxEnt predictions relates to a spatial abundance model, based on a large plot dataset for Amazonian tree species, using inverse distance weighting (IDW). We also propose a new pipeline to deal with inconsistencies in NHCs and to limit the area of occupancy of the species. We found a significant but weak positive relationship between the distribution of NHCs and IDW for 66% of the species. The relationship between SDMs and IDW was also significant but weakly positive for 95% of the species, and sensitivity for both analyses was high. Furthermore, the pipeline removed half of the NHCs records. Presence-only SDM applications should consider this limitation, especially for large biodiversity assessments projects, when they are automatically generated without subsequent checking. Our pipeline provides a conservative estimate of a species’ area of occupancy, within an area slightly larger than its extent of occurrence, compatible to e.g. IUCN red list assessments.
Show more

12 Read more

Species Distribution Modeling

Species Distribution Modeling

The use of species distribution models (SDM) to map and monitor animal and plant distributions has become increasingly important in the context of awareness of environmental change and its ecological consequences. From their original inception as resource inventory and conservation mapping tools, SDM have evolved along with the increasing variety and availability of statistical methods, digital biological, and environmental data with which they are built in a geographic information system. Beyond predicting species distributions, these models have become an impor- tant and widely used decision-making tool for a variety of biogeographical applications, such as studying the effects of climate change, identifying potential protected areas, determining locations potentially susceptible to invasion, and mapping vector-borne disease spread and risk. This article outlines the steps involved in formulating an SDM and focuses on the conceptual and theoretical foundations on which it is based and identifies issues that have merited recent and will merit future research attention.
Show more

20 Read more

Species Distribution Modelling: Contrasting presence-only models with plot abundance data

Species Distribution Modelling: Contrasting presence-only models with plot abundance data

Species distribution models (SDMs) are widely used in ecology and conservation. Presence-only SDMs such as MaxEnt frequently use natural history collections (NHCs) as occurrence data, given their huge numbers and accessibility. NHCs are often spatially biased which may generate inaccuracies in SDMs. Here, we test how the distribution of NHCs and MaxEnt predictions relates to a spatial abundance model, based on a large plot dataset for Amazonian tree species, using inverse distance weighting (IDW). We also propose a new pipeline to deal with inconsistencies in NHCs and to limit the area of occupancy of the species. We found a significant but weak positive relationship between the distribution of NHCs and IDW for 66% of the species. The relationship between SDMs and IDW was also significant but weakly positive for 95% of the species, and sensitivity for both analyses was high. Furthermore, the pipeline removed half of the NHCs records. Presence-only SDM applications should consider this limitation, especially for large biodiversity assessments projects, when they are automatically generated without subsequent checking. Our pipeline provides a conservative estimate of a species’ area of occupancy, within an area slightly larger than its extent of occurrence, compatible to e.g. IUCN red list assessments.
Show more

12 Read more

The distribution of Gammarus species in estuaries. Part I.

The distribution of Gammarus species in estuaries. Part I.

Atypical features recorded were as follows: i uropod 3 inner ramus relatively short, between 80 and 9° % of segment I of outer ramus; ii hairs unusually dense, even more so than in the F[r]

52 Read more

Comparison of single distribution and mixture distribution models for modelling LGD

Comparison of single distribution and mixture distribution models for modelling LGD

Both accelerated failure time models and proportional hazards models (Cox regression model) are built for modelling both recovery rate and recovery amount. Here, the event of interest is debt write off, so the write-off debts are treated as uncensored; debts which were paid off or were still being paid are treated as censored. All the independent variables which are used in the linear regression model building are used here as well, and they are regrouped into dummy variables. Continuous variables were firstly cut into 10 to 15 bins to become 10 to 15 dummy variables, and put them into survival analysis model without any other characteristics. Observing coefficients of them from model output and bins with similar coefficients were binned together with their neighbours. The same method was used for nominal variables. Two continuous variables ‘default amount’ and ‘ratio of default amount to total loan’ were included in the models as their original forms as well.
Show more

19 Read more

Show all 10000 documents...