Sensitivity of Systematic Reserve Selection to Decisions about Scale, Biological Data, and Targets: Case Study from Southern British Columbia

Abstract:  The identification of conservation areas based on systematic reserve-selection algorithms requires decisions related to both spatial and ecological scale. These decisions may affect the distribution and number of sites considered priorities for conservation within a region. We explored the sensitivity of systematic reserve selection by altering values of three essential variables. We used a 1:20,000–scale terrestrial ecosystem map and habitat suitability data for 29 threatened vertebrate species in the Okanagan region of British Columbia, Canada. To these data we applied a reserve-selection algorithm to select conservation sites while altering selection unit size and shape, features of biodiversity (i.e., vertebrate species), and area conservation targets for each biodiversity feature. The spatial similarity, or percentage overlap, of selected sets of conservation sites identified (1) with different selection units was ≤40%, (2) with different biodiversity features was 59%, and (3) with different conservation targets was ≥94%. Because any selected set of sites is only one of many possible sets, we also compared the conservation value (irreplaceability) of all sites in the region for each variation of the data. The correlations of irreplaceability were weak for different selection units (0.23 ≤ r ≤ 0.67), strong for different biodiversity features ( r = 0.84), and mixed for different conservation targets ( r = 0.16; 0.16; 1.00). Because of the low congruence of selected sites and weak correlations of irreplaceability for different selection units, recommendations from studies that have been applied at only one spatial scale must be considered cautiously.     

Using abiotic variables to predict importance of sites for species representation

In systematic conservation planning, species distribution data for all sites in a planning area are used to prioritize each site in terms of the site's importance toward meeting the goal of species representation. But comprehensive species data are not available in most planning areas and would be expensive to acquire. As a shortcut, ecologists use surrogates, such as occurrences of birds or another well‐surveyed taxon, or land types defined from remotely sensed data, in the hope that sites that represent the surrogates also represent biodiversity. Unfortunately, surrogates have not performed reliably. We propose a new type of surrogate, predicted importance, that can be developed from species data for a q% subset of sites. With species data from this subset of sites, importance can be modeled as a function of abiotic variables available at no charge for all terrestrial areas on Earth. Predicted importance can then be used as a surrogate to prioritize all sites. We tested this surrogate with 8 sets of species data. For each data set, we used a q% subset of sites to model importance as a function of abiotic variables, used the resulting function to predict importance for all sites, and evaluated the number of species in the sites with highest predicted importance. Sites with the highest predicted importance represented species efficiently for all data sets when q = 25% and for 7 of 8 data sets when q = 20%. Predicted importance requires less survey effort than direct selection for species representation and meets representation goals well compared with other surrogates currently in use. This less expensive surrogate may be useful in those areas of the world that need it most, namely tropical regions with the highest biodiversity, greatest biodiversity loss, most severe lack of inventory data, and poorly developed protected area networks.     

Effects of sample survey design on the accuracy of classification tree models in species distribution models

We evaluated the effects of probabilistic (hereafter DESIGN) and non-probabilistic (PURPOSIVE) sample surveys on resultant classification tree models for predicting the presence of four lichen species in the Pacific Northwest, USA. Models derived from both survey forms were assessed using an independent data set (EVALUATION). Measures of accuracy as gauged by resubstitution rates were similar for each lichen species irrespective of the underlying sample survey form. Cross-validation estimates of prediction accuracies were lower than resubstitution accuracies for all species and both design types, and in all cases were closer to the true prediction accuracies based on the EVALUATION data set. We argue that greater emphasis should be placed on calculating and reporting cross-validation accuracy rates rather than simple resubstitution accuracy rates. Evaluation of the DESIGN and PURPOSIVE tree models on the EVALUATION data set shows significantly lower prediction accuracy for the PURPOSIVE tree models relative to the DESIGN models, indicating that non-probabilistic sample surveys may generate models with limited predictive capability. These differences were consistent across all four lichen species, with 11 of the 12 possible species and sample survey type comparisons having significantly lower accuracy rates. Some differences in accuracy were as large as 50%. The classification tree structures also differed considerably both among and within the modelled species, depending on the sample survey form. Overlap in the predictor variables selected by the DESIGN and PURPOSIVE tree models ranged from only 20% to 38%, indicating the classification trees fit the two evaluated survey forms on different sets of predictor variables. The magnitude of these differences in predictor variables throws doubt on ecological interpretation derived from prediction models based on non-probabilistic sample surveys.     

Exploration of spatial morphometry and socioeconomic variables in modelling urban land use change

The incorporation of land use (LU) data with socioeconomic data is a main issue in modelling. This is as a result of difference in data model and scale. This study proposed and tested the change–pattern approach, which allows the incorporation of these data sets in modelling LU change. Focusing on LU dynamics for a selected part of the Thames Gateway within the City of London, the approach tested two different methods of input selection for the modelling operations. Variables selected from these two methods serve as inputs into several neural networks tested in order to identify the direction of change for each of the LU types within the study area. The result shows that direction of LU change across the study area could be identified when spatial morphology of the area and socioeconomic variables are considered. Some classes of change could be identified fairly accurately using landscape metrics indicating level of fragmentation, extent of LU patches, shape complexity of LU patches in combination with some socioeconomic variables.     

Fine-scale habitat modeling of a top marine predator: do prey data improve predictive capacity?

Predators and prey assort themselves relative to each other, the availability of resources and refuges, and the temporal and spatial scale of their interaction. Predictive models of predator distributions often rely on these relationships by incorporating data on environmental variability and prey availability to determine predator habitat selection patterns. This approach to predictive modeling holds true in marine systems where observations of predators are logistically difficult, emphasizing the need for accurate models. In this paper, we ask whether including prey distribution data in fine-scale predictive models of bottlenose dolphin (Tursiops truncatus) habitat selection in Florida Bay, Florida, U.S.A., improves predictive capacity. Environmental characteristics are often used as predictor variables in habitat models of top marine predators with the assumption that they act as proxies of prey distribution. We examine the validity of this assumption by comparing the response of dolphin distribution and fish catch rates to the same environmental variables. Next, the predictive capacities of four models, with and without prey distribution data, are tested to determine whether dolphin habitat selection can be predicted without recourse to describing the distribution of their prey. The final analysis determines the accuracy of predictive maps of dolphin distribution produced by modeling areas of high fish catch based on significant environmental characteristics. We use spatial analysis and independent data sets to train and test the models. Our results indicate that, due to high habitat heterogeneity and the spatial variability of prey patches, fine-scale models of dolphin habitat selection in coastal habitats will be more successful if environmental variables are used as predictor variables of predator distributions rather than relying on prey data as explanatory variables. However, predictive modeling of prey distribution as the response variable based on environmental variability did produce high predictive performance of dolphin habitat selection, particularly foraging habitat.     

Identifying Conservation Areas on the Basis of Alternative Distribution Data Sets

Abstract:  Distribution data on biodiversity features is a major component of conservation planning that are often inaccurate; thus, the true distribution of each feature is commonly over- or underrepresented. The selection of distribution data sets may therefore lead to variability in the spatial configuration and size of proposed reserve networks and uncertainty regarding the extent to which these networks actually contain the biodiversity features they were identified to protect. Our goals were to investigate the impact on reserve selection of choosing different distribution data sets and to propose novel methods to minimize uncertainty about target attainment within reserves. To do so, we used common prioritization methods (richness mapping, systematic reserve design, and a novel approach that integrates multiple types of distribution data) and three types of data on the distribution of mammals (predicted distribution models, occurrence records, and a novel combination of the two) to simulate the establishment of regional biodiversity reserves for the state of Arizona (U.S.A.). Using the results of these simulations, we explored variability in reserve placement and size as a function of the distribution data set. Spatial overlap of reserve networks identified with only predicted distribution data or only occurrence distribution data never exceeded 16%. In pairwise comparisons between reserves created with all three types of distribution data, overlap never achieved 50%. The reserve size required to meet conservation targets also varied with the type of distribution data used and the conservation goal; the largest reserve system was 10 times the smallest. Our results highlight the impact of employing different types of distribution data and identify novel tools for application to existing distribution data sets that can minimize uncertainty about target attainment.     

Species distribution modelling: Does one size fit all? A phytogeographic analysis of Salix in Ontario

Empirical models for predicting the distribution of organisms from environmental data have often focused on principles of ecological niche theory. However, even at large scales, there is little agreement over how to represent the dimensions of a species’ niche. The performance of such models is greatly affected by the nature of species distributional and environmental data. Regional scale distribution models were developed for 30 willow species in Ontario to examine (i) the predictive ability of logistic regression analysis, and (ii) the effects of using different distributional and environmental data sets. Two original measures of model accuracy and over-prediction were employed and evaluated using independent data. Models based on unique combinations of monthly climate data predicted distributions most accurately for all species. Models based on a fixed set of variables, while generating the highest average probabilities of occurrence for certain species with limited ranges, resulted in the greatest under- and over-estimates of willow distributions. Comparisons of models demonstrated climatic patterns among willows of differing habit and habitat. The distribution of dwarf willow species, present only in the Ontario arctic, followed gradients of summer maximum temperatures. The distribution of the tree species in the southerly portions of the province followed gradients of fall and winter minimum temperatures. Regardless of distributional and environmental data input, no algorithm maximized model performance for all species. Individual species models require individual approaches; i.e., the variable selection technique, the set of environmental factors used as predictors, and the nature of species distributional data must be carefully matched to the intended application. An understanding of evolutionary processes enhances the meaningful interpretation of individual species models. Unless sampling bias and species prevalence can be accounted for, models based on collection point data are best used to guide field surveys. While inferred range data may be better suited to determine potential ecological niches, overestimation of species prevalence and environmental tolerance must be recognized. A combination of available distributional data types is recommended to best determine species niches, an important step in developing conservation strategies.     

Forward selection of explanatory variables

This paper proposes a new way of using forward selection of explanatory variables in regression or canonical redundancy analysis. The classical forward selection method presents two problems: a highly inflated Type I error and an overestimation of the amount of explained variance. Correcting these problems will greatly improve the performance of this very useful method in ecological modeling. To prevent the first problem, we propose a two-step procedure. First, a global test using all explanatory variables is carried out. If, and only if, the global test is significant, one can proceed with forward selection. To prevent overestimation of the explained variance, the forward selection has to be carried out with two stopping criteria: (1) the usual alpha significance level and (2) the adjusted coefficient of multiple determination (Ra(2)) calculated using all explanatory variables. When forward selection identifies a variable that brings one or the other criterion over the fixed threshold, that variable is rejected, and the procedure is stopped. This improved method is validated by simulations involving univariate and multivariate response data. An ecological example is presented using data from the Bryce Canyon National Park, Utah, U.S.A.     

Transferability of Species Distribution Models: a Functional Habitat Approach for Two Regionally Threatened Butterflies

Abstract:  Numerous models for predicting species distribution have been developed for conservation purposes. Most of them make use of environmental data (e.g., climate, topography, land use) at a coarse grid resolution (often kilometres). Such approaches are useful for conservation policy issues including reserve-network selection. The efficiency of predictive models for species distribution is usually tested on the area for which they were developed. Although highly interesting from the point of view of conservation efficiency, transferability of such models to independent areas is still under debate. We tested the transferability of habitat-based predictive distribution models for two regionally threatened butterflies, the green hairstreak ( Callophrys rubi ) and the grayling ( Hipparchia semele ), within and among three nature reserves in northeastern Belgium. We built predictive models based on spatially detailed maps of area-wide distribution and density of ecological resources. We used resources directly related to ecological functions (host plants, nectar sources, shelter, microclimate) rather than environmental surrogate variables. We obtained models that performed well with few resource variables. All models were transferable—although to different degrees—among the independent areas within the same broad geographical region. We argue that habitat models based on essential functional resources could transfer better in space than models that use indirect environmental variables. Because functional variables can easily be interpreted and even be directly affected by terrain managers, these models can be useful tools to guide species-adapted reserve management.     

Environmental diversity as a surrogate for species representation

Because many species have not been described and most species ranges have not been mapped, conservation planners often use surrogates for conservation planning, but evidence for surrogate effectiveness is weak. Surrogates are well‐mapped features such as soil types, landforms, occurrences of an easily observed taxon (discrete surrogates), and well‐mapped environmental conditions (continuous surrogate). In the context of reserve selection, the idea is that a set of sites selected to span diversity in the surrogate will efficiently represent most species. Environmental diversity (ED) is a rarely used surrogate that selects sites to efficiently span multivariate ordination space. Because it selects across continuous environmental space, ED should perform better than discrete surrogates (which necessarily ignore within‐bin and between‐bin heterogeneity). Despite this theoretical advantage, ED appears to have performed poorly in previous tests of its ability to identify 50 × 50 km cells that represented vertebrates in Western Europe. Using an improved implementation of ED, we retested ED on Western European birds, mammals, reptiles, amphibians, and combined terrestrial vertebrates. We also tested ED on data sets for plants of Zimbabwe, birds of Spain, and birds of Arizona (United States). Sites selected using ED represented European mammals no better than randomly selected cells, but they represented species in the other 7 data sets with 20% to 84% effectiveness. This far exceeds the performance in previous tests of ED, and exceeds the performance of most discrete surrogates. We believe ED performed poorly in previous tests because those tests considered only a few candidate explanatory variables and used suboptimal forms of ED's selection algorithm. We suggest future work on ED focus on analyses at finer grain sizes more relevant to conservation decisions, explore the effect of selecting the explanatory variables most associated with species turnover, and investigate whether nonclimate abiotic variables can provide useful surrogates in an ED framework.     

Estimation of total sediment load concentration obtained by experimental study using artificial neural networks

Estimation of sediment concentration in rivers is very important for water resources projects planning and managements. The sediment concentration is generally determined from the direct measurement of sediment concentration of river or from sediment transport equations. Direct measurement is very expensive and cannot be conducted for all river gauge stations. However, sediment transport equations do not agree with each other and require many detailed data on the flow and sediment characteristics. The main purpose of the study is to establish an effective model which includes nonlinear relations between dependent (total sediment load concentration) and independent (bed slope, flow discharge, and sediment particle size) variables. In the present study, by performing 60 experiments for various independent data, dependent variables were obtained, because of the complexity of the phenomena, as a soft computing method artificial neural networks (ANNs) which is the powerful tool for input–output mapping is used. However, ANN model was compared with total sediment transport equations. The results show that ANN model is found to be significantly superior to total sediment transport equations.     

Determining factors that influence the dispersal of a pelagic species: A comparison between artificial neural networks and evolutionary algorithms

Because of increasing transport and trade there is a growing threat of marine invasive species being introduced into regions where they do not presently occur. So that the impacts of such species can be mitigated, it is important to predict how individuals, particularly passive dispersers are transported and dispersed in the ocean as well as in coastal regions so that new incursions of potential invasive species are rapidly detected and origins identified. Such predictions also support strategic monitoring, containment and/or eradication programs. To determine factors influencing a passive disperser, around coastal New Zealand, data from the genus Physalia (Cnidaria: Siphonophora) were used. Oceanographic data on wave height and wind direction and records of occurrences of Physalia on swimming beaches throughout the summer season were used to create models using artificial neural networks (ANNs) and Na?ve Bayesian Classifier (NBC). First, however, redundant and irrelevant data were removed using feature selection of a subset of variables. Two methods for feature selection were compared, one based on the multilayer perceptron and another based on an evolutionary algorithm. The models indicated that New Zealand appears to have two independent systems driven by currents and oceanographic variables that are responsible for the redistribution of Physalia from north of New Zealand and from the Tasman Sea to their subsequent presence in coastal waters. One system is centred in the east coast of northern New Zealand and the other involves a dynamic system that encompasses four other regions on both coasts of the country. Interestingly, the models confirm, molecular data obtained from Physalia in a previous study that identified a similar distribution of systems around New Zealand coastal waters. Additionally, this study demonstrates that the modelling methods used could generate valid hypotheses from noisy and complicated data in a system about which there is little previous knowledge.     

K-select analysis: a new method to analyse habitat selection in radio-tracking studies

Two kinds of wildlife habitat studies can be distinguished in the literature: hindcasting and forecasting studies. Hindcasting studies aim to emphasize among a large set of habitat variables those that are of interest for the focus species, whereas forecasting studies are intended to predict habitat selection according to a small number of habitat variables for a given area. We provide here a new analytical tool which relies on the concept of ecological niche, the K-select analysis, for hindcasting studies of habitat selection by animals using radio-tracking data. Each habitat variable defines one dimension in the ecological space. For each animal, the difference between the vector of average available habitat conditions and the vector of average used conditions defines the marginality vector. Its size is proportional to the importance of habitat selection, and its direction indicates which variables are selected. By performing a non-centered principal component analysis of the table containing the coordinates of the marginality vectors of each animal (row) on the habitat variables (column), the K-select analysis returns a linear combination of habitat variables for which the average marginality is greatest. It is a synthesis of variables which contributes the most to the habitat selection. As with principal component analysis, the biological significance of the factorial axes is deduced from the loading of variables. An example is provided: habitat selection by wild boar is studied in a Mediterranean habitat using the K-select analysis. The numerous advantages of the analysis (a large number of variables that can be included, individual variability in habitat selection taken into account, a lack of too strict underlying hypotheses) make it a powerful approach in radio-tracking studies designed to identify habitat variables that are selected by animals.     

Applications of artificial neural networks for patterning and predicting aquatic insect species richness in running waters

Two artificial neural networks (ANNs), unsupervised and supervised learning algorithms, were applied to suggest practical approaches for the analysis of ecological data. Four major aquatic insect orders (Ephemeroptera, Plecoptera, Trichoptera, and Coleoptera, i.e. EPTC), and four environmental variables (elevation, stream order, distance from the source, and water temperature) were used to implement the models. The data were collected and measured at 155 sampling sites on streams of the Adour–Garonne drainage basin (South-western France). The modelling procedure was carried out following two steps. First, a self-organizing map (SOM), an unsupervised ANN, was applied to classify sampling sites using EPTC richness. Second, a backpropagation algorithm (BP), a supervised ANN, was applied to predict EPTC richness using a set of four environmental variables. The trained SOM classified sampling sites according to a gradient of EPTC richness, and the groups obtained corresponded to geographic regions of the drainage basin and characteristics of their environmental variables. The SOM showed its convenience to analyze relationships among sampling sites, biological attributes, and environmental variables. After accounting for the relationships in data sets, the BP used to predict the EPTC richness with a set of four environmental variables showed a high accuracy (r=0.91 and r=0.61 for training and test data sets respectively). The prediction of EPTC richness is thus a valuable tool to assess disturbances in given areas: by knowing what the EPTC richness should be, we can determine the degree to which disturbances have altered it. The results suggested that methodologies successively using two different neural networks are helpful to understand ecological data through ordination first, and then to predict target variables.     

A regression strategy for analyzing environmental data generated by spatio-temporal processes

Highly complex spatio-temporal environmental data sets are becoming common in ecology because of the increasing use of large-scale simulation models and automated data collection devices. The spatial and temporal dimensions present real and difficult challenges for the interpretation of these data. A particularly difficult problem is that the relationship among variables can vary in dramatically in response to environmental variation; consequently, a single model may not provide adequate fit. The temporal dimension presents both opportunities for improved prediction because explanatory variables sometimes exert delayed effects on response variables, and problems because variables are often serially correlated. This article presents a regression strategy for accommodating these problems and exploiting serial correlation. The strategy is illustrated by a case study of simulated net primary production (SNPP) that compares ocean-atmosphere indices to terrestrial climate variables as predictors of SNPP across the conterminous United States, and describes spatial variation in the relative importance of terrestrial climate variables towards predicting SNPP. We found that the relationship between ocean-atmosphere indices and SNPP varies substantially over the United States, and that there is evidence of a substantive link only in the western portions of the United States. Evidence of multi-year delays in the effect of terrestrial climate effects on SNPP were also found.     

Long-term variability in the structure of subtidal benthic communities in Puget Sound,Washington, USA

Data on benthic infauna from 4 permanent stations in Puget Sound off Seattle, USA, collected during 1963–1964, 1967, and 1969, revealed considerable stability in numbers of species and specimens and in diversity within stations among sampling dates. The species composition of the faunal assemblages also remained rather constant during the period of investigation, but the relative dominance among the numerically important species varied somewhat. Biomass data did not differ significantly in 1964 and 1969, but the 1967 data were considerably lower at all stations.     

Habitat selection models to account for seasonal persistence in radio telemetry data

Models for the analysis of habitat selection data incorporate covariates in an independent multinomial selections model (McCracken et al. 1998) Ramsey and Usner 2003 and an extension of that model to include a persistence parameter (2003). In both cases, all parameters are assumed to be fixed through time. Radio telemetry data collected for habitat selection studies typically consist of animal relocations through time, suggesting the need for an extension to these models. We use a Bayesian approach that allows for the habitat selection probabilities, persistence parameter, or both, to change with season. These extensions are particularly important when movement patterns are expected to differ seasonally and/or when availabilities of habitats change throughout the study period due to weather or migration. We implement and compare the models using radio telemetry data for westslope cutthroat trout in two streams in eastern Oregon.     

Biological benthic tools as indicators of coastal marine ecosystems health

Coastal marine ecosystems are increasingly subjected to environmental stress and degradation due to pollution. Several research programmes have addressed this problem and produced relevant data sets for specific areas, often including consistent sets of environmental and biological variables. The value of existing information gathered from these types of data can be largely increased by combining them into a common data set to determine globally applicable relationships. To perform this exercise, the Intergovernmental Oceanographic Commission (IOC) of UNESCO has recently formed the Ad hoc Study Group on Benthic Indicators (http://www.ioc.unesco.org/benthicindicators) with the aim of developing robust indicators of benthic health. In this paper, initial products and ongoing activities of this international initiative are described and discussed. An expansion of initial IOC/UNESCO research on benthic fauna-organic carbon relationships is also presented. As part of this follow-up research, the relationship between total organic carbon concentrations of sediment and abundance, biomass and species diversity of benthic macrofauna was evaluated using data sets from 2 different regions of the world comprising 3 different coastal marine environments. The ability of identifying threshold levels in selected variables that could serve as indicators of related adverse environmental conditions leading to stress in the benthos is envisaged within the frame of a larger joint analysis, carried out by the IOC/UNESCO Study Group on Benthic Indicators, of merged data sets from several coastal regions worldwide.     

Mapping species distributions in one dimension by non-homogeneous hidden Markov models: the case of freshwater pearl mussels in the River Dee

The investigation of species distributions in rivers involves data which are inherently sequential and unlikely to be fully independent. To take these characteristics into account, we develop a Bayesian hierarchical model for mapping the distribution of freshwater pearl mussels in the River Dee (Scotland). At the top of the hierarchy the likelihood is used to describe the sequence of sites in which mussels were observed or not. Given that false observations can occur, and that “not observed” means both that the species was not present and that it was not observed, a Markov prior is introduced at the second level of the hierarchy to represent the sequence of sites in which mussels are estimated to occur. The Markov prior allows modelling the spatial dependency between neighbouring sites. A third level in the hierarchy is given by the representation of the transition probabilities of the Markov chain in terms of site-specific explanatory variables, through a logistic regression. The selection of the explanatory variables which influence the Markov process is performed by means of a simulation-based procedure, in the complex case of association between covariates. Four features were found to be associated with reduced chance of finding a local mussel population: tributaries, bridges, dredging, and waste water treatment works. These results complement the results of a previous study, providing new evidence for the causes of the deterioration of a highly threatened species.