Habitat suitability models for the conservation of thermophilic grasshoppers and bush crickets—simple or complex?

One goal of conservation biology is the assessment of effects of land use change on species distribution. One approach for identifying the factors, which determine habitat suitability for a species are statistical habitat distribution models. These models are quantitative and can be used for predictions in management scenarios. However, they often have one major shortcoming, which is their complexity. This means that they need several, often costly-to-determine parameters for predictions of species occurrence. We first used habitat suitability models to investigate and determine habitat preferences of three different Orthoptera species. Second, we compared the predictive powers of simple habitat suitability models considering only the ‘habitat type’ as predictor with more complex models taking different habitat factors into account. We found that the habitat type is the most reliable and robust factor, which determines the occurrence of the species studied. Thus, analyses of habitat suitability can easily be carried out on the basis of existing vegetation maps for the conservation of the three species under study. Our results can serve as a basis for the estimation of spatio-temporal distribution and survival probabilities of the species studied and might also be valuable for other species living in dry grasslands.     

Species distribution models predict temporal but not spatial variation in forest growth

Bioclimate envelope models have been widely used to illustrate the discrepancy between current species distributions and their potential habitat under climate change. However, the realism and correct interpretation of such projections has been the subject of considerable discussion. Here, we investigate whether climate suitability predictions correlate to tree growth, measured in permanent inventory plots and inferred from tree‐ring records. We use the ensemble classifier RandomForest and species occurrence data from ~200,000 inventory plots to build species distribution models for four important European forestry species: Norway spruce, Scots pine, European beech, and pedunculate oak. We then correlate climate‐based habitat suitability with volume measurements from ~50‐year‐old stands, available from ~11,000 inventory plots. Secondly, habitat projections based on annual historical climate are compared with ring width from ~300 tree‐ring chronologies. Our working hypothesis is that habitat suitability projections from species distribution models should to some degree be associated with temporal or spatial variation in these growth records. We find that the habitat projections are uncorrelated with spatial growth records (inventory plot data), but they do predict interannual variation in tree‐ring width, with an average correlation of .22. Correlation coefficients for individual chronologies range from values as high as .82 or as low as ?.31. We conclude that tree responses to projected climate change are highly site‐specific and that local suitability of a species for reforestation is difficult to predict. That said, projected increase or decrease in climatic suitability may be interpreted as an average expectation of increased or reduced growth over larger geographic scales.     

Mountain landscapes offer few opportunities for high‐elevation tree species migration

Climate change is anticipated to alter plant species distributions. Regional context, notably the spatial complexity of climatic gradients, may influence species migration potential. While high‐elevation species may benefit from steep climate gradients in mountain regions, their persistence may be threatened by limited suitable habitat as land area decreases with elevation. To untangle these apparently contradictory predictions for mountainous regions, we evaluated the climatic suitability of four coniferous forest tree species of the western United States based on species distribution modeling (SDM) and examined changes in climatically suitable areas under predicted climate change. We used forest structural information relating to tree species dominance, productivity, and demography from an extensive forest inventory system to assess the strength of inferences made with a SDM approach. We found that tree species dominance, productivity, and recruitment were highest where climatic suitability (i.e., probability of species occurrence under certain climate conditions) was high, supporting the use of predicted climatic suitability in examining species risk to climate change. By predicting changes in climatic suitability over the next century, we found that climatic suitability will likely decline, both in areas currently occupied by each tree species and in nearby unoccupied areas to which species might migrate in the future. These trends were most dramatic for high elevation species. Climatic changes predicted over the next century will dramatically reduce climatically suitable areas for high‐elevation tree species while a lower elevation species, Pinus ponderosa, will be well positioned to shift upslope across the region. Reductions in suitable area for high‐elevation species imply that even unlimited migration would be insufficient to offset predicted habitat loss, underscoring the vulnerability of these high‐elevation species to climatic changes.     

A performance based consensus approach for predicting spatial extent of the Chinese windmill palm (Trachycarpus fortunei) in New Zealand under climate change

The predicted distribution of the Chinese Windmill Palm (Trachycarpus fortunei) was modelled using several algorithms with inputs consisting of occurrence information and bioclimatic datasets. A global species distribution model was developed and projected into New Zealand to provide a visualization of suitability for the species in current and future conditions. To ensure model robustness, occurrence data was checked for redundancy, spatial auto-correlation and the environmental variables checked for cross-correlation and collinearity. The final maps predicting suitability resulted from ensembling the predictions of all the algorithms. The resulting ensembled maps were weighted based on the evaluation parameters AUC, Kappa and TSS. When reclassified into low, medium and high suitability categories, results show an expansion of high suitability areas accompanied by a reduction of low suitability areas for the species. The centroids of the high suitability areas also exhibit a general movement towards the Southwest under future climate conditions. The range expansion was proportional with the representative values of emission trajectories RCPs (2.5, 4.5, 6.0 and 8.5) used in projecting into future conditions. The movement magnitude and direction of predicted high suitability area centroids for the palm supports the use of the plant as an indicator of climate change.     

Species–habitat associations and demographic rates of forest trees

Niche‐driven effects on demographic processes generated in response to habitat heterogeneity partly shape local distributions of species. Thus, tree distributions are commonly studied in relation to habitat conditions to understand how niche differentiation contributes to species coexistence in forest communities. Many such studies implicitly assume that local abundance reflects habitat suitability, and that abundance is relatively stable over time. We compared models based on abundance with those based on demographic performance for making inferences about habitat association for 287 tree species from three large dynamic plots located in tropical, subtropical and temperate forests. The correlation between the predictions of the abundance‐based models and the demography‐based models varied widely, with correlation coefficients ranging nearly from ?1 to 1.This suggests that the two types of models capture different information about species–habitat associations. Demography‐based models evaluate habitat quality by focusing on population processes and thus should be preferred for understanding responses of tree species to habitat conditions, especially when habitat conditions are changing and species–habitat interactions cannot be considered to be at equilibrium.     

Prospective sampling based on model ensembles improves the detection of rare species

Identifying the geographic distribution of populations is a basic, yet crucial step in many fundamental and applied ecological projects, as it provides key information on which many subsequent analyses depend. However, this task is often costly and time consuming, especially where rare species are concerned and where most sampling designs generally prove inefficient. At the same time, rare species are those for which distribution data are most needed for their conservation to be effective. To enhance fieldwork sampling, model‐based sampling (MBS) uses predictions from species distribution models: when looking for the species in areas of high habitat suitability, chances should be higher to find them. We thoroughly tested the efficiency of MBS by conducting an important survey in the Swiss Alps, assessing the detection rate of three rare and five common plant species. For each species, habitat suitability maps were produced following an ensemble modeling framework combining two spatial resolutions and two modeling techniques. We tested the efficiency of MBS and the accuracy of our models by sampling 240 sites in the field (30 sites×8 species). Across all species, the MBS approach proved to be effective. In particular, the MBS design strictly led to the discovery of six sites of presence of one rare plant, increasing chances to find this species from 0 to 50%. For common species, MBS doubled the new population discovery rates as compared to random sampling. Habitat suitability maps coming from the combination of four individual modeling methods predicted well the species' distribution and more accurately than the individual models. As a conclusion, using MBS for fieldwork could efficiently help in increasing our knowledge of rare species distribution. More generally, we recommend using habitat suitability models to support conservation plans.     

Climate change likely to reduce orchid bee abundance even in climatic suitable sites

Studies have tested whether model predictions based on species’ occurrence can predict the spatial pattern of population abundance. The relationship between predicted environmental suitability and population abundance varies in shape, strength and predictive power. However, little attention has been paid to the congruence in predictions of different models fed with occurrence or abundance data, in particular when comparing metrics of climate change impact. Here, we used the ecological niche modeling fit with presence–absence and abundance data of orchid bees to predict the effect of climate change on species and assembly level distribution patterns. In addition, we assessed whether predictions of presence–absence models can be used as a proxy to abundance patterns. We obtained georeferenced abundance data of orchid bees (Hymenoptera: Apidae: Euglossina) in the Brazilian Atlantic Forest. Sampling method consisted in attracting male orchid bees to baits of at least five different aromatic compounds and collecting the individuals with entomological nets or bait traps. We limited abundance data to those obtained by similar standard sampling protocol to avoid bias in abundance estimation. We used boosted regression trees to model ecological niches and project them into six climate models and two Representative Concentration Pathways. We found that models based on species occurrences worked as a proxy for changes in population abundance when the output of the models were continuous; results were very different when outputs were discretized to binary predictions. We found an overall trend of diminishing abundance in the future, but a clear retention of climatically suitable sites too. Furthermore, geographic distance to gained climatic suitable areas can be very short, although it embraces great variation. Changes in species richness and turnover would be concentrated in western and southern Atlantic Forest. Our findings offer support to the ongoing debate of suitability–abundance models and can be used to support spatial conservation prioritization schemes and species triage in Atlantic Forest.     

A simple non-parametric GIS model for predicting species distribution: endemic birds in Bioko Island, West Africa

Species mapping is a useful conservation tool for predicting patterns of biological diversity, or identifying geographical areas of conservation significance. Mapping can also improve our understanding of the appropriateness of habitat areas for individual species. We outline a computer-based methodology, PREDICT, for the analysis of the habitat requirements of species in a combined GIS-statistical programming environment. The paper details the statistical background to the approach adopted, the program structure and input file information and then applies these techniques to bird data from Bioko Island, West Africa. It produces images and statistics that assess the potential of unstudied areas for wildlife for which presence/absence data and basic habitat information are available. Suitability for target species is determined within surveyed and non-surveyed squares by a form of weights of evidence. The program measures the degree of association between habitat factors and presence/absence of target species by means of ² tests. The overall suitability weighting of each square, as the sum of all individual habitat factor weightings, is finally displayed in maps depicting areas of highly suitable, suitable, unsuitable and highly unsuitable habitat. Statistical relations between vegetation, rainfall and landscape features on Bioko Island and the location of 9 endemic bird taxa are presented herein. Final confirmation of the accuracy of predictions of the studied bird taxa will ensue from future field observations. However, in a series of misclassification tests of the program, actual distribution detection rate was in excess of 90%. The use of PREDICT can guide investigations of little known species in remote areas and provide a practical solution to identify areas of high rare species diversity in need of conservation.     

Scenario-Led Habitat Modelling of Land Use Change Impacts on Key Species

Accurate predictions of the impacts of future land use change on species of conservation concern can help to inform policy-makers and improve conservation measures. If predictions are spatially explicit, predicted consequences of likely land use changes could be accessible to land managers at a scale relevant to their working landscape. We introduce a method, based on open source software, which integrates habitat suitability modelling with scenario-building, and illustrate its use by investigating the effects of alternative land use change scenarios on landscape suitability for black grouse Tetrao tetrix. Expert opinion was used to construct five near-future (twenty years) scenarios for the 800 km² study site in upland Scotland. For each scenario, the cover of different land use types was altered by 5–30% from 20 random starting locations and changes in habitat suitability assessed by projecting a MaxEnt suitability model onto each simulated landscape. A scenario converting grazed land to moorland and open forestry was the most beneficial for black grouse, and ‘increased grazing’ (the opposite conversion) the most detrimental. Positioning of new landscape blocks was shown to be important in some situations. Increasing the area of open-canopy forestry caused a proportional decrease in suitability, but suitability gains for the ‘reduced grazing’ scenario were nonlinear. ‘Scenario-led’ landscape simulation models can be applied in assessments of the impacts of land use change both on individual species and also on diversity and community measures, or ecosystem services. A next step would be to include landscape configuration more explicitly in the simulation models, both to make them more realistic, and to examine the effects of habitat placement more thoroughly. In this example, the recommended policy would be incentives on grazing reduction to benefit black grouse.     

Presence-absence versus presence-only modelling methods for predicting bird habitat suitability

Habitat suitability models can be generated using methods requiring information on species presence or species presence and absence. Knowledge of the predictive performance of such methods becomes a critical issue to establish their optimal scope of application for mapping current species distributions under different constraints. Here, we use breeding bird atlas data in Catalonia as a working example and attempt to analyse the relative performance of two methods: the Ecological Niche factor Analysis (ENFA) using presence data only and Generalised Linear Models (GLM) using presence/absence data. Models were run on a set of forest species with similar habitat requirements, but with varying occurrence rates (prevalence) and niche positions (marginality). Our results support the idea that GLM predictions are more accurate than those obtained with ENFA. This was particularly true when species were using available habitats proportionally to their suitability, making absence data reliable and useful to enhance model calibration. Species marginality in niche space was also correlated to predictive accuracy, i.e. species with less restricted ecological requirements were modelled less accurately than species with more restricted requirements. This pattern was irrespective of the method employed. Models for wide‐ranging and tolerant species were more sensitive to absence data, suggesting that presence/absence methods may be particularly important for predicting distributions of this type of species. We conclude that modellers should consider that species ecological characteristics are critical in determining the accuracy of models and that it is difficult to predict generalist species distributions accurately and this is independent of the method used. Being based on distinct approaches regarding adjustment to data and data quality, habitat distribution modelling methods cover different application areas, making it difficult to identify one that should be universally applicable. Our results suggest however, that if absence data is available, methods using this information should be preferably used in most situations.     

Is the Climate Right for Pleistocene Rewilding? Using Species Distribution Models to Extrapolate Climatic Suitability for Mammals across Continents

Species distribution models (SDMs) are increasingly used for extrapolation, or predicting suitable regions for species under new geographic or temporal scenarios. However, SDM predictions may be prone to errors if species are not at equilibrium with climatic conditions in the current range and if training samples are not representative. Here the controversial “Pleistocene rewilding” proposal was used as a novel example to address some of the challenges of extrapolating modeled species-climate relationships outside of current ranges. Climatic suitability for three proposed proxy species (Asian elephant, African cheetah and African lion) was extrapolated to the American southwest and Great Plains using Maxent, a machine-learning species distribution model. Similar models were fit for Oryx gazella, a species native to Africa that has naturalized in North America, to test model predictions. To overcome biases introduced by contracted modern ranges and limited occurrence data, random pseudo-presence points generated from modern and historical ranges were used for model training. For all species except the oryx, models of climatic suitability fit to training data from historical ranges produced larger areas of predicted suitability in North America than models fit to training data from modern ranges. Four naturalized oryx populations in the American southwest were correctly predicted with a generous model threshold, but none of these locations were predicted with a more stringent threshold. In general, the northern Great Plains had low climatic suitability for all focal species and scenarios considered, while portions of the southern Great Plains and American southwest had low to intermediate suitability for some species in some scenarios. The results suggest that the use of historical, in addition to modern, range information and randomly sampled pseudo-presence points may improve model accuracy. This has implications for modeling range shifts of organisms in response to climate change.     

Mathematical contributions to link biota with environment

Hutchinson's pioneering work on the niche concept, dating from 1957, inspired the development of many ecological models. The first proposals, BIOCLIM and HABITAT, were simple geometric approximations to the shape of the niche. Despite their simplicity, they combine two features that make them adequate for the purpose of exploring the niche: they fit a predefined shape to the empirical data; and produce binary or ordinal predictions rather than continuous predictions. Thus, both explicitly delineate a precise boundary for the niche. However, the two methods present some limitations: BIOCLIM assumes that the variables are independent in their action on the species; and HABITAT, although not having that limitation, only delineates the boundaries of the niches without distinguishing levels of suitability for the species. We propose, discuss and illustrate: (1) the use of depth functions to identify regions with distinct suitability inside the niche; and (2) a general framework to assess overlap of the niches of two species, which can be applied to predictions from models that decompose the niche into a finite number of measurable regions.     

Current and future distribution of a parasite with complex life cycle under global change scenarios: Echinococcus multilocularis in Europe

Global change is expected to have complex effects on the distribution and transmission patterns of zoonotic parasites. Modelling habitat suitability for parasites with complex life cycles is essential to further our understanding of how disease systems respond to environmental changes, and to make spatial predictions of their future distributions. However, the limited availability of high quality occurrence data with high spatial resolution often constrains these investigations. Using 449 reliable occurrence records for Echinococcus multilocularis from across Europe published over the last 35 years, we modelled habitat suitability for this parasite, the aetiological agent of alveolar echinococcosis, in order to describe its environmental niche, predict its current and future distribution under three global change scenarios, and quantify the probability of occurrence for each European country. Using a machine learning approach, we developed large-scale (25 × 25 km) species distribution models based on seven sets of predictors, each set representing a distinct biological hypothesis supported by current knowledge of the autecology of the parasite. The best-supported hypothesis included climatic, orographic and land-use/land-cover variables such as the temperature of the coldest quarter, forest cover, urban cover and the precipitation seasonality. Future projections suggested the appearance of highly suitable areas for E. multilocularis towards northern latitudes and in the whole Alpine region under all scenarios, while decreases in habitat suitability were predicted for central Europe. Our spatially explicit predictions of habitat suitability shed light on the complex responses of parasites to ongoing global changes.     

Predicting the spatial distribution of the blue-spotted maskray Neotrygon kuhlii (Myliobatiformes, Dasyatidae) on the Australian North and Northwest Shelf comparing two different methods of habitat modeling

Knowledge about distribution and habitat requirements of species is important for analyzing their role in marine ecosystems or establishing sanctuaries. However, knowledge is scarce especially in many chondrichthyan species. In this study, the spatial distribution of the stingray Neotrygon kuhlii on the Australian North and Northwest Shelf was predicted model-based for the first time. Predictions based on two different types of habitat suitability models, logistic regression and maximum entropy modeling. Catch data of N. kuhlii from Australian trawl surveys combined with randomly selected pseudo-absences were used for modeling together with data sets of several environmental variables. Both modeling methods yielded plausible and validated habitat suitability models containing water depth and salinity as significant independent variables. The model-based predictions of the probability of occurrence of N. kuhlii were similar for both methods and thus emphasized the goodness of the models. Following the predictions, N. kuhlii has its highest probability of occurrence in about 60 m water depth and at a salinity of about 35 PSU. The results indicate that both modeling methods are powerful tools to predict spatial distribution and habitat quality for marine fish species. Therefore, they are suitable for detecting possible distribution in areas with only few field records.     

Modeling habitat suitability and spread dynamics of two invasive rose species in protected areas of Mendoza,Argentina

Biological invasions are a main threat to biodiversity and natural resources, which calls for studies that identify the regions that present the greatest invasion risks. We assessed the potential distribution of two non-native rose species, Rosa canina and Rosa rubiginosa, in mountain environments in mid-western Argentina, using species distribution models and dynamic simulations. We first fitted the model for one protected area, Villavicencio Nature Reserve, and then we made predictions on the distribution of these species for other protected areas in the same region, where the presence of these species was observed but where there are no systematic surveys on their distribution. We also modeled the invasion dynamics of these species based on habitat suitability, considering the dispersal distance and the growth rate of the invaded area. High and very high suitability sites were detected in all the protected areas studied, suggesting high invasion risk in these protected areas. Our simulations of the spatio–temporal dynamics of the rose invasion in Villavicencio indicated that the spread depends strongly on the average seed dispersal distance, that the spread has been gradual since the rose introduction into the protected area, and that 150 years after the introduction even the areas identified as having low suitability are expected to have been invaded. This is the first study of this type for the region, where these invasive rose species are a serious problem. Taken together, our results may be useful to identify areas vulnerable to invasion and thus help generate effective preventive, monitoring, and control practices.     

Modelling potential Pleistocene habitat corridors between Afromontane forest regions

The unusually high floral and faunal similarity between the different regions of the Afromontane archipelago has been noted by biogeographers since the late 1800s. A possible explanation for this similarity is the spread of montane habitat into the intervening lowlands during the glacial periods of the Pleistocene, allowing biotic exchange between mountain ranges. In this study, we sought to infer the existence and most likely positions of these potential habitat corridors. We focused on sixteen Afromontane endemic tree, shrub, and bird species in the Cameroon Volcanic Line, East African Rift and Great Escarpment. Species were chosen based on distribution above 1200–1500 m in at least two of the major Afromontane regions. Ecological niche models were developed for each species in the present and projected to the mid-Holocene and the last glacial maximum (LGM). Models were thresholded to create binary maps of presence/absence and then summed across taxa to estimate potential LGM and mid-Holocene distributions. We found widespread climatic suitability for our montane taxa throughout the lowlands of Central Africa during the LGM, connecting all regions of the Afromontane archipelago except the Ethiopian Highlands and the Dahomey Gap. During the mid-Holocene, we noted more limited climatic suitability for fewer species in lowland areas. Although we set out to test predictions derived from alternatively hypothesized corridors, we instead found widespread climatic suitability connecting Afromontane regions across the entire Congo Basin for all species.

     

Local stream habitat variables predicted from catchment scale characteristics are useful for predicting fish distribution

South-east Queensland (Australia) streams were described by 21 local habitat variables that were chosen because of their potential association with fish distribution. An Assessment by a Nearest Neighbour Analysis (ANNA) model used large-scale variables that are robust to human influence to predict what the values of each of the 21 local habitat variables at each site would be without modification from human activity. The ANNA model used elevation, stream order, distance from source and longitude to predict the local habitat variables; other candidate predictor variables (mean rainfall, latitude and catchment area) were not found to be useful. The ANNA model was able to predict five of the 21 local habitat variables (average width, sand (%), cobble (%), rocks (%) and large woody debris) with an R ² of at least 0.2. The observed values of these five local habitat variables were used to model the distributions of individual fish species. The species distribution models were developed using logistic regression based on a subset of the data (some of the data were withheld for model validation) and a forward stepwise model selection procedure. There was no difference in predictive performance of fish distribution models for model predictions based on observed values and model predictions based on ANNA predicted values of local habitat variables in the withheld data (p-value = 0.85). Therefore, it is possible to predict the suitability of sites as habitat for given fish species using estimated (estimates based on large-scale variables) natural values of local habitat variables.     

Implications of Climate Change for Bird Conservation in the Southwestern U.S. under Three Alternative Futures

Future expected changes in climate and human activity threaten many riparian habitats, particularly in the southwestern U.S. Using Maximum Entropy (MaxEnt3.3.3) modeling, we characterized habitat relationships and generated spatial predictions of habitat suitability for the Lucy’s warbler (Oreothlypis luciae), the Southwestern willow flycatcher (Empidonax traillii extimus) and the Western yellow-billed cuckoo (Coccyzus americanus). Our goal was to provide site- and species-specific information that can be used by managers to identify areas for habitat conservation and/or restoration along the Rio Grande in New Mexico. We created models of suitable habitat for each species based on collection and survey samples and climate, biophysical, and vegetation data. We projected habitat suitability under future climates by applying these models to conditions generated from three climate models for 2030, 2060 and 2090. By comparing current and future distributions, we identified how habitats are likely to change as a result of changing climate and the consequences of those changes for these bird species. We also examined whether land ownership of high value sites shifts under changing climate conditions. Habitat suitability models performed well. Biophysical characteristics were more important that climate conditions for predicting habitat suitability with distance to water being the single most important predictor. Climate, though less important, was still influential and led to declines of suitable habitat of more than 60% by 2090. For all species, suitable habitat tended to shrink over time within the study area leaving a few core areas of high importance. Overall, climate changes will increase habitat fragmentation and reduce breeding habitat patch size. The best strategy for conserving bird species within the Rio Grande will include measures to maintain and restore critical habitat refugia. This study provides an example of a presence-only habitat model that can be used to inform the management of species at intermediate scales.     

How to account for habitat suitability in weed management programmes?

Designing efficient management strategies for already established invasive alien species is challenging. Here, we ask whether environmental suitability, as predicted by species distribution models, is a useful basis of cost-effective spatial prioritization in large-scale surveillance and eradication programmes. We do so by means of spatially and temporarily explicit simulations of the spread of a case study species (Ambrosia artemisiifolia L.) in Austria and southern Germany under different management regimes. We ran these simulations on a contiguous grid of the study area with each grid cell (~35 km²) characterized by a habitat suitability value derived from the predictions of a species distribution model. The management regimes differed in terms of (a) a minimum habitat suitability rank p (suitability threshold) used to separate cells for surveillance from those which are not controlled; and (b) the strategy for selecting cells for annual campaigns from the pool defined by p. According to the results (i.e., number of cells infested in 2050 as well as infested on average per year) the most efficient way to base surveillance on suitability is to define the temporal sequence of management according to the grid cells’ suitability ranks. Management success declines sharply when the suitability threshold is set too high, but only moderately when it is set too low. We conclude that accounting for environmental suitability is important for large-scale management programmes of invasive alien species and that species distribution models are hence useful tools for designing such programmes.     

Prevalence-adjusted optimisation of fuzzy habitat suitability models for aquatic invertebrate and fish species in New Zealand

For many years, habitat suitability models for aquatic species have been derived from ecological datasets by model optimisation. Previous research showed that optimisation of the predictive model performance did not necessarily lead to ecologically relevant models due to the impact of the dataset prevalence. Therefore, the adjusted average deviation was presented as a performance criterion that allowed incorporation of ecological relevance in the model optimisation process. This paper aims to analyse the relation between the adjusted average deviation (aAD) and the training set prevalence for three species in different New Zealand river systems: caddis flies Aoteapsyche spp., large brown trout Salmo trutta and rainbow trout Oncorhynchus mykiss. The aAD was implemented in a hill-climbing algorithm to optimise a fuzzy species distribution model for each species. Specifically, the hypotheses were tested that (1) similar relations between the aAD and the training set prevalence would be obtained, (2) training based on the aAD would lead to more accurate model predictions than training based on more frequently applied performance criteria such as CCI, and that (3) the final fuzzy model would produce a realistic model of habitat suitability. The approach in this paper may improve the transparency of the model training process and thus the insight into habitat suitability models. Consequently, this paper could lead to ecologically more relevant models and contribute to the implementation of these models in ecosystem management.