Comparing Ground Telemetry and Global Positioning System Methods to Determine Cougar Kill Rates

ABSTRACT We assessed whether use of 2 methods, intensive very high frequency (VHF) radiotelemetry and Global Positioning System (GPS) cluster sampling, yielded similar estimates of cougar (Puma concolor) kill rates in Yellowstone National Park, 1998–2005. We additionally determined biases (underestimation or overestimation of rates) resulting from each method. We used modeling to evaluate what characteristics of clusters best predicted a kill versus no kill and further evaluated which predictor(s) minimized effort and the number of missed kills. We conducted 16 VHF ground predation sequences resulting in 37 kill intervals (KIs) and 21 GPS sequences resulting in 84 KIs on 6 solitary adult females, 4 maternal females, and 5 adult males. Kill rates (days/kill and biomass [kg] killed/day) did not differ between VHF and GPS predation sampling methods for maternal females, solitary adult females, and adult males. Sixteen of 142 (11.3%) kills detected via GPS clusters were missed through VHF ground-based sampling, and the kill rate was underestimated by an average of 5.2 (95% CI = 3.8–6.6) days/kill over all cougar social classes. Five of 142 (3.5%) kills identified by GPS cluster sampling were incorrectly identified as the focal individual's kill from scavenging, and the kill rate was overestimated within the adult male social class by an average of 5.8 (95% CI = 3.0–8.5) days/ungulate kill. The number of nights (locations between 2000 hours and 0500 hours) a cougar spent at a cluster was the most efficient variable at predicting predation, minimizing the missed kills, and minimizing number of extra clusters that needed to be searched. In Yellowstone National Park, where competing carnivores displaced cougars from their kills, it was necessary to search extra sites where a kill may not have been present to ensure we did not miss small, ungulate prey kills or kills with displacement. Using predictions from models to assign unvisited clusters as no kill, small prey kill, or large prey kill can bias downward the number of kills a cougar made and bias upward kills made by competitors that displace cougars or scavenge cougar kills. Our findings emphasize that field visitation is crucial in determining displacement and scavenging events that can result in biases when using GPS cluster methods in multicarnivore systems.     

Discriminating grey wolf kill sites using GPS clusters

Characteristics of spatio-temporal clusters of locations from global positioning system (GPS)-collars have been used to distinguish kill sites of various predators. We deployed GPS collars on 9 grey wolves (Canis lupus) in the southwest area of Prince Albert National Park in central Saskatchewan, Canada, and used a GPS location clustering algorithm to identify kill sites of ungulate and other large-bodied prey during winter, December 2013–March 2017. We used logistic regression in a model-selection framework to determine if spatio-temporal and habitat characteristics of grey wolf GPS clusters could be used to reliably identify sites where wolves had killed prey. Global positioning system clusters were more likely to be wolf kill sites when they had a higher number of location fixes, did not begin within 300 m and 30 days of a previous cluster, did not begin within 1 km and 4 days of a previous cluster, began in the evening, had a high percentage of fixes occurring during the day, occurred farther from open habitat, and had both a high number of location fixes and a high percentage of fixes occurring during the day. Our results highlight the limits of using spatio-temporal clusters with a fix rate of 1/hour to discriminate wolf kill sites in systems dominated by deer (Odocoileus spp.) because of the associated short handling time with these prey.     

The generalized linear model for spatial data: assessing the effects of environmental covariates on population density in the field

With interest in spatial ecology growing, correlational field studies are likely to become increasingly important. Unfortunately, ecological field data often do not follow the assumptions of classical statistics, so techniques like the popular and powerful multiple linear regression and its variants are often unreliable, and results can be misleading. The generalized linear model (GLM) is a flexible extension of linear regression that has proved especially useful for discrete data. In this paper, the technique is adapted to accommodate spatially correlated, discrete data. Specifically, to demonstrate the approach, Japanese beetle grub [Popillia japonica Newman (Coleoptera, Scarabaeidae)] population density in the field is modeled as a function of soil organic matter content. The response variable (grub counts in small soil samples) was a spatially autocorrelated, discrete random variable. Three classes of GLMs of the association between soil organic matter content and grub density were compared: (i) regression (assuming normally distributed response variable), (ii) GLM assuming negative binomial counts, and (iii) GLM based on the assumption that the counts conformed to Taylor's power law (TPL). Because the grubs were distributed in patches rather than at random, models that explicitly accounted for the spatial autocorrelation of grub counts were constructed, and compared with models that assumed independent observations. The fitted values for the discrete GLMs [viz., (ii) and (iii)] differed noticeably from the fitted values from multiple regression; but fitted values among the negative binomial and TPL GLMs were virtually identical, regardless of whether the spatial covariance was incorporated into a model, whether a spherical or exponential variogram model was used, or whether variance function parameters were estimated over a large or small scale. However, P‐values for the overall significance of the models depended heavily on whether the GLM assumed a discrete or continuous response variable, and whether or not spatial autocorrelation in the response variable was accounted for. On average, P‐values were 45‐fold higher in the spatial GLMs than in the non‐spatial and 23‐fold higher in the discrete GLMs than in the continuous.     

Space–time clusters for early detection of grizzly bear predation

Accurate detection and classification of predation events is important to determine predation and consumption rates by predators. However, obtaining this information for large predators is constrained by the speed at which carcasses disappear and the cost of field data collection. To accurately detect predation events, researchers have used GPS collar technology combined with targeted site visits. However, kill sites are often investigated well after the predation event due to limited data retrieval options on GPS collars (VHF or UHF downloading) and to ensure crew safety when working with large predators. This can lead to missing information from small‐prey (including young ungulates) kill sites due to scavenging and general site deterioration (e.g., vegetation growth). We used a space–time permutation scan statistic (STPSS) clustering method (SaTScan) to detect predation events of grizzly bears (Ursus arctos) fitted with satellite transmitting GPS collars. We used generalized linear mixed models to verify predation events and the size of carcasses using spatiotemporal characteristics as predictors. STPSS uses a probability model to compare expected cluster size (space and time) with the observed size. We applied this method retrospectively to data from 2006 to 2007 to compare our method to random GPS site selection. In 2013–2014, we applied our detection method to visit sites one week after their occupation. Both datasets were collected in the same study area. Our approach detected 23 of 27 predation sites verified by visiting 464 random grizzly bear locations in 2006–2007, 187 of which were within space–time clusters and 277 outside. Predation site detection increased by 2.75 times (54 predation events of 335 visited clusters) using 2013–2014 data. Our GLMMs showed that cluster size and duration predicted predation events and carcass size with high sensitivity (0.72 and 0.94, respectively). Coupling GPS satellite technology with clusters using a program based on space–time probability models allows for prompt visits to predation sites. This enables accurate identification of the carcass size and increases fieldwork efficiency in predation studies.     

Building a mechanistic understanding of predation with GPS-based movement data

Quantifying kill rates and sources of variation in kill rates remains an important challenge in linking predators to their prey. We address current approaches to using global positioning system (GPS)-based movement data for quantifying key predation components of large carnivores. We review approaches to identify kill sites from GPS movement data as a means to estimate kill rates and address advantages of using GPS-based data over past approaches. Despite considerable progress, modelling the probability that a cluster of GPS points is a kill site is no substitute for field visits, but can guide our field efforts. Once kill sites are identified, time spent at a kill site (handling time) and time between kills (killing time) can be determined. We show how statistical models can be used to investigate the influence of factors such as animal characteristics (e.g. age, sex, group size) and landscape features on either handling time or killing efficiency. If we know the prey densities along paths to a kill, we can quantify the ‘attack success’ parameter in functional response models directly. Problems remain in incorporating the behavioural complexity derived from GPS movement paths into functional response models, particularly in multi-prey systems, but we believe that exploring the details of GPS movement data has put us on the right path.     

Evaluating Global Positioning System Telemetry Techniques for Estimating Cougar Predation Parameters

ABSTRACT Using clusters of locations obtained from Global Positioning System (GPS) telemetry collars to identify predation events may allow more efficient estimation of behavioral predation parameters for the study and management of large carnivore predator-prey systems. Applications of field- and model-based GPS telemetry cluster techniques, however, have met with mixed success. To further evaluate and refine these techniques for cougars (Puma concolor), we used data from visits to 1,735 GPS telemetry clusters, 637 of which were locations where cougars killed prey >8 kg in a multi-prey system in west-central Alberta. We tested 1) whether clusters were reliably created at kill locations, 2) the ability of logistic regression models to identify kill occurrence (prey >8 kg) and multinomial regression models to identify the prey species at a kill cluster, and 3) the duration of monitoring required to accurately estimate kill rate and prey composition. We found that GPS collars programmed to attempt location fixes every 3 hours consistently identified locations where prey >8 kg were handled, and cluster creation was robust to GPS location acquisition failures (poor collar fix success). The logistic regression model was capable of estimating cougar kill rate with a mean 5-fold cross validation error of <10%, provided the appropriate probability cutoff distinguishing kill clusters from non-kill clusters was selected. Logistic models also can be used to direct visits to clusters, reducing field efforts by as much as 25%, while still locating >95% of all kills. The multinomial model overpredicted occurrence of primary prey (deer) in the diet and underpredicted consumption of alternate prey (e.g., elk and moose) by as much as 100%. We conclude that a purely model-based approach should be used cautiously and that field visitation is required to obtain reliable information on species, sex, age, or condition of prey. Ultimately, we recommend a combined approach that involves using models to direct field visitation when estimating behavioral predation parameters. Regardless of the monitoring approach, long continuous monitoring periods (i.e., >100 days of a 180-day period) were necessary to reduce bias and imprecision in kill rate and prey composition estimates.     

Bias-reduced estimators of conditional odds ratios in matched case-control studies with unmatched confounding

We study bias-reduced estimators of exponentially transformed parameters in general linear models (GLMs) and show how they can be used to obtain bias-reduced conditional (or unconditional) odds ratios in matched case-control studies. Two options are considered and compared: the explicit approach and the implicit approach. The implicit approach is based on the modified score function where bias-reduced estimates are obtained by using iterative procedures to solve the modified score equations. The explicit approach is shown to be a one-step approximation of this iterative procedure. To apply these approaches for the conditional analysis of matched case-control studies, with potentially unmatched confounding and with several exposures, we utilize the relation between the conditional likelihood and the likelihood of the unconditional logit binomial GLM for matched pairs and Cox partial likelihood for matched sets with appropriately setup data. The properties of the estimators are evaluated by using a large Monte Carlo simulation study and an illustration of a real dataset is shown. Researchers reporting the results on the exponentiated scale should use bias-reduced estimators since otherwise the effects can be under or overestimated, where the magnitude of the bias is especially large in studies with smaller sample sizes.     

Joint regression analysis of correlated data using Gaussian copulas

Summary .  This article concerns a new joint modeling approach for correlated data analysis. Utilizing Gaussian copulas, we present a unified and flexible machinery to integrate separate one-dimensional generalized linear models (GLMs) into a joint regression analysis of continuous, discrete, and mixed correlated outcomes. This essentially leads to a multivariate analogue of the univariate GLM theory and hence an efficiency gain in the estimation of regression coefficients. The availability of joint probability models enables us to develop a full maximum likelihood inference. Numerical illustrations are focused on regression models for discrete correlated data, including multidimensional logistic regression models and a joint model for mixed normal and binary outcomes. In the simulation studies, the proposed copula-based joint model is compared to the popular generalized estimating equations, which is a moment-based estimating equation method to join univariate GLMs. Two real-world data examples are used in the illustration.     

Statistical Methods for Identifying Wolf Kill Sites Using Global Positioning System Locations

Abstract: Accurate estimates of kill rates remain a key limitation to addressing many predator—prey questions. Past approaches for identifying kill sites of large predators, such as wolves (Canis lupus), have been limited primarily to areas with abundant winter snowfall and have required intensive ground-tracking or aerial monitoring. More recently, attempts have been made to identify clusters of locations obtained using Global Positioning System (GPS) collars on predators to identify kill sites. However, because decision rules used in determining clusters have not been consistent across studies, results are not necessarily comparable. We illustrate a space—time clustering approach to statistically define clusters of wolf GPS locations that might be wolf kill sites, and we then use binary and multinomial logistic regression to model the probability of a cluster being a non—kill site, kill site of small-bodied prey species, or kill site of a large-bodied prey species. We evaluated our approach using field visits of kills and assessed the accuracy of the models using an independent dataset. The cluster-scan approach identified 42–100% of wolf-killed prey, and top logistic regression models correctly classified 100% of kills of large-bodied prey species, but 40% of small-bodied prey species were classified as nonkills. Although knowledge of prey distribution and vulnerability may help refine this approach, identifying small-bodied prey species will likely remain problematic without intensive field efforts. We recommend that our approach be utilized with the understanding that variation in prey body size and handling time by wolves will likely have implications for the success of both the cluster scan and logistic regression components of the technique. (JOURNAL OF WILDLIFE MANAGEMENT 72(3):798–807; 2008)     

The quantification of condensed tannins in African savanna tree species

We compared Quebracho with Sorghum tannin as standards for condensed tannin (CT) quantification in selected African savanna tree species in relation to the acid-butanol assay for CTs. Without exception, the use of Quebracho tannin as standard overestimated CTs, ranging from 0.7 to as much as 8.3 times. Sorghum tannin underestimated CTs by 0.26–0.79 times, except in one species where there was no difference in the CT concentration. Condensed tannins in African savanna trees showed qualitative and quantitative differences in chemical composition which may explain the variable reactivity in the acid-butanol assay. We propose the use of condensed tannins purified from the plant under investigation be used as standard since it will closely represent the CT structure and presumably chemical reactivity in the acid-butanol assay.     

Framing pictures: A conceptual framework to identify and correct for biases in detection probability of camera traps enabling multi‐species comparison

Obtaining reliable species observations is of great importance in animal ecology and wildlife conservation. An increasing number of studies use camera traps (CTs) to study wildlife communities, and an increasing effort is made to make better use and reuse of the large amounts of data that are produced. It is in these circumstances that it becomes paramount to correct for the species‐ and study‐specific variation in imperfect detection within CTs. We reviewed the literature and used our own experience to compile a list of factors that affect CT detection of animals. We did this within a conceptual framework of six distinct scales separating out the influences of (a) animal characteristics, (b) CT specifications, (c) CT set‐up protocols, and (d) environmental variables. We identified 40 factors that can potentially influence the detection of animals by CTs at these six scales. Many of these factors were related to only a few overarching parameters. Most of the animal characteristics scale with body mass and diet type, and most environmental characteristics differ with season or latitude such that remote sensing products like NDVI could be used as a proxy index to capture this variation. Factors that influence detection at the microsite and camera scales are probably the most important in determining CT detection of animals. The type of study and specific research question will determine which factors should be corrected. Corrections can be done by directly adjusting the CT metric of interest or by using covariates in a statistical framework. Our conceptual framework can be used to design better CT studies and help when analyzing CT data. Furthermore, it provides an overview of which factors should be reported in CT studies to make them repeatable, comparable, and their data reusable. This should greatly improve the possibilities for global scale analyses of (reused) CT data.     

Inter- and intra-specific gene-density-correlated radial chromosome territory arrangements are conserved in Old World monkeys

Recently it has been shown that the gene-density correlated radial distribution of human 18 and 19 homologous chromosome territories (CTs) is conserved in higher primates in spite of chromosomal rearrangements that occurred during evolution. However, these observations were limited to apes and New World monkey species. In order to provide further evidence for the evolutionary conservation of gene-density-correlated CT arrangements, we extended our previous study to Old World monkeys. They comprise the remaining species group to be analyzed in order to obtain a comprehensive overview of the nuclear topology of human 18 and 19 homologous CTs in higher primates. In the present study we investigated four lymphoblastoid cell lines from three species of Old World monkeys by three-dimensional fluorescence in situ hybridization (3D-FISH): two individuals of Japanese macaque (Macaca fuscata), crab-eating macaque (Macaca fascicularis), and an interspecies hybrid individual between African green monkey (Cercopithecus aethiops) and Patas monkey (Erythrocebus patas). Our data demonstrate that gene-poor human 18 homologous CTs are located preferentially close to the nuclear periphery, whereas gene-dense human 19 homologous CTs are oriented towards the nuclear center in all cell lines analyzed. The gene-density-correlated positioning of human 18 and 19 homologous CTs is evolutionarily conserved throughout all major higher primate lineages, despite chromosomal inversions, fusions, fissions or reciprocal translocations that occurred in the course of evolution in these species. This remarkable preservation of a gene-density-correlated chromatin arrangement gives further support for a functionally relevant higher-order chromatin architecture.     

PAAR‐Rhs proteins harbor various C‐terminal toxins to diversify the antibacterial pathways of type VI secretion systems

The type VI secretion system (T6SS) of bacteria plays a key role in competing for specific niches by the contact‐dependent killing of competitors. Recently, Rhs proteins with polymorphic C‐terminal toxin‐domains that inhibit or kill neighboring cells were identified. In this report, we identified a novel Rhs with an MPTase4 (Metallopeptidase‐4) domain (designated as Rhs‐CT1) that showed an antibacterial effect via T6SS in Escherichia coli. We managed to develop a specific strategy by matching the diagnostic domain‐architecture of Rhs‐CT1 (Rhs with an N‐terminal PAAR‐motif and a C‐terminal toxin domain) for effector retrieval and discovered a series of Rhs‐CTs in E. coli. Indeed, the screened Rhs‐CT3 with a REase‐3 (Restriction endonuclease‐3) domain also mediated interbacterial antagonism. Further analysis revealed that vgrG_O1 and eagR/DUF1795 (upstream of rhs‐ct) were required for the delivery of Rhs‐CTs, suggesting eagR as a potential T6SS chaperone. In addition to chaperoned Rhs‐CTs, neighborless Rhs‐CTs could be classified into a distinct family (Rhs‐Nb) sharing close evolutionary relationship with T6SS2‐Rhs (encoded in the T6SS2 cluster of E. coli). Notably, the Rhs‐Nb‐CT5 was confirmed bioinformatically and experimentally to mediate interbacterial antagonism via Hcp2B‐VgrG2 module. In a further retrieval analysis, we discovered various toxin/immunity pairs in extensive bacterial species that could be systematically classified into eight referential clans, suggesting that Rhs‐CTs greatly diversify the antibacterial pathways of T6SS.     

Generalized case-control sampling under generalized linear models

A generalized case-control (GCC) study, like the standard case-control study, leverages outcome-dependent sampling (ODS) to extend to nonbinary responses. We develop a novel, unifying approach for analyzing GCC study data using the recently developed semiparametric extension of the generalized linear model (GLM), which is substantially more robust to model misspecification than existing approaches based on parametric GLMs. For valid estimation and inference, we use a conditional likelihood to account for the biased sampling design. We describe analysis procedures for estimation and inference for the semiparametric GLM under a conditional likelihood, and we discuss problems with estimation and inference under a conditional likelihood when the response distribution is misspecified. We demonstrate the flexibility of our approach over existing ones through extensive simulation studies, and we apply the methodology to an analysis of the Asset and Health Dynamics Among the Oldest Old study, which motives our research. The proposed approach yields a simple yet versatile solution for handling ODS in a wide variety of possible response distributions and sampling schemes encountered in practice.     

The use of generalized linear models and generalized estimating equations in bioarchaeological studies

The current article explores whether the application of generalized linear models (GLM) and generalized estimating equations (GEE) can be used in place of conventional statistical analyses in the study of ordinal data that code an underlying continuous variable, like entheseal changes. The analysis of artificial data and ordinal data expressing entheseal changes in archaeological North African populations gave the following results. Parametric and nonparametric tests give convergent results particularly for P values <0.1, irrespective of whether the underlying variable is normally distributed or not under the condition that the samples involved in the tests exhibit approximately equal sizes. If this prerequisite is valid and provided that the samples are of equal variances, analysis of covariance may be adopted. GLM are not subject to constraints and give results that converge to those obtained from all nonparametric tests. Therefore, they can be used instead of traditional tests as they give the same amount of information as them, but with the advantage of allowing the study of the simultaneous impact of multiple predictors and their interactions and the modeling of the experimental data. However, GLM should be replaced by GEE for the study of bilateral asymmetry and in general when paired samples are tested, because GEE are appropriate for correlated data. Am J Phys Anthropol 153:473–483, 2014. © 2013 Wiley Periodicals, Inc.     

Modelling and predicting fungal distribution patterns using herbarium data

Aim The main aims of this study are: (1) to test if temperature and related parameters are the primary determinants of the regional distribution of macrofungi (as is commonly recognized for plants); (2) to test if the success of modelling fungal distribution patterns depends on species and distribution characteristics; and (3) to explore the potential of using herbarium data for modelling and predicting fungal species’ distributions. Location The study area, Norway, spans 58–71° N latitude and 4–32° E longitude, and embraces extensive ecological gradients in a small area. Methods The study is based on 1020 herbarium collections of nine selected species of macrofungi and a set of 75 environmental predictor variables, all recorded in a 5 × 5‐km grid covering Norway. Primarily, generalized linear model (GLM; logistic regression) analyses were used to identify the environmental variables that best accounted for the species’ recorded distributions in Norway. Second, Maxent analyses (using variables identified by GLM) were used to produce predictive potential distribution maps for these species. Results Variables relating to temperature and radiation were most frequently included in the GLMs, and between 24.8% and 59.8% of the variation in single‐species occurrence was accounted for. The fraction of variation explained by the GLMs ranged from 41.6% to 59.8% for species with restricted distributions, and from 24.8% to 39.3% for species with widespread/scattered and intermediate distributions. The two‐step procedure of GLM followed by Maxent gave predictions with very high values for the area under the curve (0.927–0.997), and maps of potential distribution were generally credible. Main conclusions We show that temperature is a key factor governing the distribution of macrofungi in Norway, indicating that fungi may respond strongly to global warming. We confirm that modelling success depends partly on species and distribution characteristics, notably on how the distribution relates to the extent of the study area. Our study demonstrates that the combination of GLM and Maxent may be a fruitful approach for biogeography. We conclude that herbarium data improve insight into factors that control the distributions of fungi, of particular value for research on fleshy fungi (mushrooms), which have largely cryptic life cycles.     

Testing Global Positioning System Telemetry to Study Wolf Predation on Deer Fawns

ABSTRACT We conducted a pilot study to test the usefulness of Global Positioning System (GPS) collars for investigating wolf (Canis lupus) predation on white-tailed deer (Odocoileus virginianus) fawns. Using GPS collars with short location-attempt intervals on 5 wolves and 5 deer during summers 2002–2004 in northeastern Minnesota, USA, demonstrated how this approach could provide new insights into wolf hunting behavior of fawns. For example, a wolf traveled ≥1.5–3.0 km and spent 20–22 hours in the immediate vicinity of known fawn kill sites and ≥0.7 km and 8.3 hours at scavenging sites. Wolf travel paths indicated that wolves intentionally traveled into deer summer ranges, traveled ≥0.7–4.2 km in such ranges, and spent <1–22 hours per visit. Each pair of 3 GPS-collared wolf pack members were located together for ≤6% of potential locations. From GPS collar data, we estimated that each deer summer range in a pack territory containing 5 wolves ≥1 year old and hunting individually would be visited by a wolf on average every 3–5 days. This approach holds great potential for investigating summer hunting behavior of wolves in areas where direct observation is impractical or impossible.     

Improvements on GPS Location Cluster Analysis for the Prediction of Large Carnivore Feeding Activities: Ground-Truth Detection Probability and Inclusion of Activity Sensor Measures

Animal space use studies using GPS collar technology are increasingly incorporating behavior based analysis of spatio-temporal data in order to expand inferences of resource use. GPS location cluster analysis is one such technique applied to large carnivores to identify the timing and location of feeding events. For logistical and financial reasons, researchers often implement predictive models for identifying these events. We present two separate improvements for predictive models that future practitioners can implement. Thus far, feeding prediction models have incorporated a small range of covariates, usually limited to spatio-temporal characteristics of the GPS data. Using GPS collared cougar (Puma concolor) we include activity sensor data as an additional covariate to increase prediction performance of feeding presence/absence. Integral to the predictive modeling of feeding events is a ground-truthing component, in which GPS location clusters are visited by human observers to confirm the presence or absence of feeding remains. Failing to account for sources of ground-truthing false-absences can bias the number of predicted feeding events to be low. Thus we account for some ground-truthing error sources directly in the model with covariates and when applying model predictions. Accounting for these errors resulted in a 10% increase in the number of clusters predicted to be feeding events. Using a double-observer design, we show that the ground-truthing false-absence rate is relatively low (4%) using a search delay of 2–60 days. Overall, we provide two separate improvements to the GPS cluster analysis techniques that can be expanded upon and implemented in future studies interested in identifying feeding behaviors of large carnivores.     

Generalized models vs. classification tree analysis: Predicting spatial distributions of plant species at different scales

Abstract. Statistical models of the realized niche of species are increasingly used, but systematic comparisons of alternative methods are still limited. In particular, only few studies have explored the effect of scale in model outputs. In this paper, we investigate the predictive ability of three statistical methods (generalized linear models, generalized additive models and classification tree analysis) using species distribution data at three scales: fine (Catalonia), intermediate (Portugal) and coarse (Europe). Four Mediterranean tree species were modelled for comparison. Variables selected by models were relatively consistent across scales and the predictive accuracy of models varied only slightly. However, there were slight differences in the performance of methods. Classification tree analysis had a lower accuracy than the generalized methods, especially at finer scales. The performance of generalized linear models also increased with scale. At the fine scale GLM with linear terms showed better accuracy than GLM with quadratic and polynomial terms. This is probably because distributions at finer scales represent a linear sub‐sample of entire realized niches of species. In contrast to GLM, the performance of GAM was constant across scales being more data‐oriented. The predictive accuracy of GAM was always at least equal to other techniques, suggesting that this modelling approach is more robust to variations of scale because it can deal with any response shape.     

Classification of black-pigmented anaerobes by pyrolysis mass spectrometry (PMS) and conventional tests (CTs)

Abstract Clinical (66: dental 53; vaginal 4; wound 9) and reference (5^∗) strains of pigmented Gram-negative anaerobic bacilli were examined in pyrolysis mass spectrometry (PMS) and conventional tests (CTs). The strains were identified in CTs as: Prevotella intermedia (48^∗); Pr. melaninogenica (1); Pr. corporis (7); Porphyromonas asaccharolytica (12^∗); P. endodontalis (1^∗) and P. gingivalis (2^∗). Numerical classification based on CTs resolved five clusters comprising strains identified as (I) Pr. corporis , (II) Pr. melaninogenica , (III) Pr. intermedia , (IV) P. gingivalis and (V) P. asaccharolytica and P. endodontalis . Numerical classification based on PMS showed a similar division, with decreasing homogeneity in the order Pr. intermedia, Pr. corporis, P. asaccharolytica , in agreement with the ordering of homogeneity for these species in CTs. PMS clusters corresponding the Porphyromonas spp. were clearly distinct from those of Prevotella spp. PMS and CT classifications disagreed on cluster membership for only six of the strains. PMS identification from blind challenge sets agreed with conventional identification for 64 of 67 strains.