Bee species perform distinct foraging behaviors that are best described by different movement models.

In insect-pollinated plants, the foraging behavior of pollinators affects their pattern of movement. If distinct bee species vary in their foraging behaviors, different models may best describe their movement. In this study, we quantified and compared the fine scale movement of three bee species foraging on patches of Medicago sativa. Bee movement was described using distances and directions traveled between consecutive racemes. Bumble bees and honey bees traveled shorter distances after visiting many flowers on a raceme, while the distance traveled by leafcutting bees was independent of flower number. Transition matrices and vectors were calculated for bumble bees and honey bees to reflect their directionality of movement within foraging bouts; leafcutting bees were as likely to move in any direction. Bee species varied in their foraging behaviors, and for each bee species, we tested four movement models that differed in how distances and directions were selected, and identified the model that best explained the movement data. The fine-scale, within-patch movement of bees could not always be explained by a random movement model, and a general model of movement could not be applied to all bee species.

Bee species visiting Medicago sativa differ in pollen deposition curves with consequences for gene flow.

PREMISE: Pollinator foraging behavior can influence pollen dispersal and gene flow. In many plant species a pollinator trips a flower by applying pressure to release its sexual organs. We propose that differences in tripping rate among grooming pollinators could generate distinct pollen deposition curves, the pattern of pollen deposition over successive flowers visited. This study compares the pollen deposition curves of two grooming pollinators, a social bumble bee and a solitary leafcutting bee, with distinct tripping rates on Medicago sativa flowers. We predict a steeper deposition curve for pollen moved by leafcutting bees, the pollinator with the higher tripping rate. METHODS: Medicago sativa plants carrying a gene (GUS) whose product is easily detected by staining, were used as pollen donors. After visiting the GUS plants, a bee was released on a linear array of conventional M. sativa plants. The number of GUS pollen grains deposited over successive flowers visited or over cumulative distances was examined. Distinct mixed effect Poisson regression models, illustrating different rates of decay in pollen deposition, were fitted to the pollen data for each bee species. RESULTS: Pollen decay was steeper for leafcutting bees relative to bumble bees for both models of flowers visited and cumulative distance, as predicted by their higher tripping rate. CONCLUSIONS: This is the first report of a difference in pollen deposition curves between two bee species, both grooming pollinators. Such differences could lead to distinct impacts of bee species on gene flow, genetic differentiation, introgression, and ultimately speciation.

Patch selection by bumble bees navigating discontinuous landscapes.

Pollen and nectar resources are unevenly distributed over space and bees must make routing decisions when navigating patchy resources. Determining the patch selection process used by bees is crucial to understanding bee foraging over discontinuous landscapes. To elucidate this process, we developed four distinct probability models of bee movement where the size and the distance to the patch determined the attractiveness of a patch. A field experiment with a center patch and four peripheral patches of two distinct sizes and distances from the center was set up in two configurations. Empirical transition probabilities from the center to each peripheral patch were obtained at two sites and two years. The best model was identified by comparing observed and predicted transition probabilities, where predicted values were obtained by incorporating the spatial dimensions of the field experiment into each model's mathematical expression. Bumble bees used both patch size and isolation distance when selecting a patch and could assess the total amount of resources available in a patch. Bumble bees prefer large, nearby patches. This information will facilitate the development of a predictive framework to the study of bee movement and of models that predict the movement of genetically engineered pollen in bee-pollinated crops.

Vegetation productivity summarized by the Dynamic Habitat Indices explains broad-scale patterns of moose abundance across Russia.

Identifying the factors that determine habitat suitability and hence patterns of wildlife abundances over broad spatial scales is important for conservation. Ecosystem productivity is a key aspect of habitat suitability, especially for large mammals. Our goals were to a) explain patterns of moose (Alces alces) abundance across Russia based on remotely sensed measures of vegetation productivity using Dynamic Habitat Indices (DHIs), and b) examine if patterns of moose abundance and productivity differed before and after the collapse of the Soviet Union. We evaluated the utility of the DHIs using multiple regression models predicting moose abundance by administrative regions. Univariate models of the individual DHIs had lower predictive power than all three combined. The three DHIs together with environmental variables, explained 79% of variation in moose abundance. Interestingly, the predictive power of the models was highest for the 1980s, and decreased for the two subsequent decades. We speculate that the lower predictive power of our environmental variables in the later decades may be due to increasing human influence on moose densities. Overall, we were able to explain patterns in moose abundance in Russia well, which can inform wildlife managers on the long-term patterns of habitat use of the species.

Linking the foraging behavior of three bee species to pollen dispersal and gene flow.

Foraging behaviors that impact gene flow can guide the design of pollinator strategies to mitigate gene flow. Reduced gene flow is expected to minimize the impact of genetically engineered (GE) crops on feral and natural populations and to facilitate the coexistence of different agricultural markets. The goal of this study is to link foraging behavior to gene flow and identify behaviors that can help predict gene flow for different bee species. To reach this goal, we first examined and compared the foraging behaviors of three distinct bee species, the European honey bee, Apis mellifera L., the common eastern bumble bee, Bombus impatiens Cr., and the alfalfa leafcutting bee, Megachile rotundata F., foraging on Medicago sativa flowers. Each foraging behavior investigated differed among bee species. Both social bees exhibited directionality of movement and had similar residence, in contrast to the random movement and shorter residence of the solitary bee. Tripping rate and net distance traveled differed among the three bee species. We ranked each behavior among bee species and used the relative ranking as gene flow predictor before testing the predictions against empirical gene flow data. Tripping rate and net distance traveled, but not residence, predicted relative gene dispersal among bee species. Linking specific behaviors to gene flow provides mechanisms to explain differences in gene flow among bee species and guides the development of management practices to reduce gene flow. Although developed in one system, the approach developed here can be generalized to different plant/pollinator systems.

Comparing Genome-Wide Association Study Results from Different Measurements of an Underlying Phenotype.

Increasing popularity of high-throughput phenotyping technologies, such as image-based phenotyping, offer novel ways for quantifying plant growth and morphology. These new methods can be more or less accurate and precise than traditional, manual measurements. Many large-scale phenotyping efforts are conducted to enable genome-wide association studies (GWAS), but it is unclear exactly how alternative methods of phenotyping will affect GWAS results. In this study we simulate phenotypes that are controlled by the same set of causal loci but have differing heritability, similar to two different measurements of the same morphological character. We then perform GWAS with the simulated traits and create receiver operating characteristic (ROC) curves from the results. The areas under the ROC curves (AUCs) provide a metric that allows direct comparisons of GWAS results from different simulated traits. We use this framework to evaluate the effects of heritability and the number of causative loci on the AUCs of simulated traits; we also test the differences between AUCs of traits with differing heritability. We find that both increasing the number of causative loci and decreasing the heritability reduce a trait's AUC. We also find that when two traits are controlled by a greater number of causative loci, they are more likely to have significantly different AUCs as the difference between their heritabilities increases. When simulation results are applied to measures of tassel morphology, we find no significant difference between AUCs from GWAS using manual and image-based measurements of typical maize tassel characters. This finding indicates that both measurement methods have similar ability to identify genetic associations. These results provide a framework for deciding between competing phenotyping strategies when the ultimate goal is to generate and use phenotype-genotype associations from GWAS.

Floral traits influencing plant attractiveness to three bee species: Consequences for plant reproductive success.

PREMISE OF THE STUDY: The ability to attract pollinators is crucial to plants that rely on insects for pollination. We contrasted the roles of floral display size and flower color in attracting three bee species and determined the relationships between plant attractiveness (number of pollinator visits) and seed set for each bee species. METHODS: We recorded pollinator visits to plants, measured plant traits, and quantified plant reproductive success. A zero-inflated Poisson regression model indicated plant traits associated with pollinator attraction. It identified traits that increased the number of bee visits and traits that increased the probability of a plant not receiving any visits. Different components of floral display size were examined and two models of flower color contrasted. Relationships between plant attractiveness and seed set were determined using regression analyses. KEY RESULTS: Plants with more racemes received more bee visits from all three bee species. Plants with few racemes were more likely not to receive any bee visits. The role of flower color varied with bee species and was influenced by the choice of the flower color model. Increasing bee visits increased seed set for all three bee species, with the steepest slope for leafcutting bees, followed by bumble bees, and finally honey bees. CONCLUSIONS: Floral display size influenced pollinator attraction more consistently than flower color. The same plant traits affected the probability of not being visited and the number of pollinator visits received. The impact of plant attractiveness on female reproductive success varied, together with pollinator effectiveness, by pollinator species.

Factors related to building loss due to wildfires in the conterminous United States.

Wildfire is globally an important ecological disturbance affecting biochemical cycles and vegetation composition, but also puts people and their homes at risk. Suppressing wildfires has detrimental ecological effects and can promote larger and more intense wildfires when fuels accumulate, which increases the threat to buildings in the wildland-urban interface (WUI). Yet, when wildfires occur, typically only a small proportion of the buildings within the fire perimeter are lost, and the question is what determines which buildings burn. Our goal was to examine which factors are related to building loss when a wildfire occurs throughout the United States. We were particularly interested in the relative roles of vegetation, topography, and the spatial arrangement of buildings, and how their respective roles vary among ecoregions. We analyzed all fires that occurred within the conterminous United States from 2000 to 2010 and digitized which buildings were lost and which survived according to Google Earth historical imagery. We modeled the occurrence as well as the percentage of buildings lost within clusters using logistic and linear regression. Overall, variables related to topography and the spatial arrangement of buildings were more frequently present in the best 20 regression models than vegetation-related variables. In other words, specific locations in the landscape have a higher fire risk, and certain development patterns can exacerbate that risk. Fire policies and prevention efforts focused on vegetation management are important, but insufficient to solve current wildfire problems. Furthermore, the factors associated with building loss varied considerably among ecoregions suggesting that fire policy applied uniformly across the United States will not work equally well in all regions and that efforts to adapt communities to wildfires must be regionally tailored.

Glycomacropeptide for nutritional management of phenylketonuria: a randomized, controlled, crossover trial.

BACKGROUND: To prevent cognitive impairment, phenylketonuria requires lifelong management of blood phenylalanine (Phe) concentration with a low-Phe diet. The diet restricts intake of Phe from natural proteins in combination with traditional amino acid medical foods (AA-MFs) or glycomacropeptide medical foods (GMP-MFs) that contain primarily intact protein and a small amount of Phe. OBJECTIVE: We investigated the efficacy and safety of a low-Phe diet combined with GMP-MFs or AA-MFs providing the same quantity of protein equivalents in free-living subjects with phenylketonuria. DESIGN: This 2-stage, randomized crossover trial included 30 early-treated phenylketonuria subjects (aged 15-49 y), 20 with classical and 10 with variant phenylketonuria. Subjects consumed, in random order for 3 wk each, their usual low-Phe diet combined with AA-MFs or GMP-MFs. The treatments were separated by a 3-wk washout with AA-MFs. Fasting plasma amino acid profiles, blood Phe concentrations, food records, and neuropsychological tests were obtained. RESULTS: The frequency of medical food intake was higher with GMP-MFs than with AA-MFs. Subjects rated GMP-MFs as more acceptable than AA-MFs and noted improved gastrointestinal symptoms and less hunger with GMP-MFs. ANCOVA indicated no significant mean ± SE increase in plasma Phe (62 ± 40 µmol/L, P = 0.136), despite a significant increase in Phe intake from GMP-MFs (88 ± 6 mg Phe/d, P = 0.026). AA-MFs decreased plasma Phe (-85 ± 40 µmol/L, P = 0.044) with stable Phe intake. Blood concentrations of Phe across time were not significantly different (AA-MFs = 444 ± 34 µmol/L, GMP-MFs = 497 ± 34 µmol/L), suggesting similar Phe control. Results of the Behavior Rating Inventory of Executive Function were not significantly different. CONCLUSIONS: GMP-MFs provide a safe and acceptable option for the nutritional management of phenylketonuria. The greater acceptability and fewer side effects noted with GMP-MFs than with AA-MFs may enhance dietary adherence for individuals with phenylketonuria. This trial was registered at www.clinicaltrials.gov as NCT01428258.

Clinical relevance of the discrepancy in phenylalanine concentrations analyzed using tandem mass spectrometry compared with ion-exchange chromatography in phenylketonuria.

INTRODUCTION: Metabolic control of phenylketonuria (PKU) and compliance with the low-phenylalanine (phe) diet are frequently assessed by measuring blood phe concentrations in dried blood spots (DBS) collected by patients instead of plasma phe concentrations. OBJECTIVE: Our objective was to investigate the difference in blood phe concentrations in DBS collected by subjects and analyzed using either a validated newborn screening tandem mass spectrometry (MS/MS) protocol or ion-exchange chromatography (IEC) compared to plasma phe concentrations obtained simultaneously and analyzed using IEC. DESIGN: Three to four fasting blood samples were obtained from 29 subjects with PKU, ages 15-49 years. Capillary blood was spotted on filter paper by each subject and the DBS analyzed using both MS/MS and IEC. Plasma was isolated from venous blood and analyzed using IEC. RESULTS: Blood phe concentrations in DBS analyzed using MS/MS are 28% ± 1% (n = 110, p < 0.0001) lower than plasma phe concentrations analyzed using IEC resulting in a blood phe concentration of 514 ± 23 µmol/L and a plasma phe concentration of 731 ± 32 µmol/L (mean ± SEM). This discrepancy is larger when plasma phe is > 600 µmol/L. Due to the large variability across subjects of 13.2%, a calibration factor to adjust blood phe concentrations is not recommended. Analysis of DBS using IEC reduced the discrepancy to 15 ± 2% lower phe concentrations compared to plasma analyzed using IEC (n = 38, p = 0.0001). This suggests that a major contributor to the discrepancy in phe concentrations is the analytical method. CONCLUSION: Use of DBS analyzed using MS/MS to monitor blood phe concentrations in individuals with PKU yields significantly lower phe levels compared to plasma phe levels analyzed using IEC. Optimization of current testing methodologies for measuring phe in DBS, along with patient education regarding the appropriate technique for spotting blood on filter paper is needed to improve the accuracy of using DBS to measure phe concentrations in PKU management.

Spatial scan statistics for detection of multiple clusters with arbitrary shapes.

In applying scan statistics for public health research, it would be valuable to develop a detection method for multiple clusters that accommodates spatial correlation and covariate effects in an integrated model. In this article, we connect the concepts of the likelihood ratio (LR) scan statistic and the quasi-likelihood (QL) scan statistic to provide a series of detection procedures sufficiently flexible to apply to clusters of arbitrary shape. First, we use an independent scan model for detection of clusters and then a variogram tool to examine the existence of spatial correlation and regional variation based on residuals of the independent scan model. When the estimate of regional variation is significantly different from zero, a mixed QL estimating equation is developed to estimate coefficients of geographic clusters and covariates. We use the Benjamini-Hochberg procedure (1995) to find a threshold for p-values to address the multiple testing problem. A quasi-deviance criterion is used to regroup the estimated clusters to find geographic clusters with arbitrary shapes. We conduct simulations to compare the performance of the proposed method with other scan statistics. For illustration, the method is applied to enterovirus data from Taiwan.

Are the numbers adding up? Exploiting discrepancies among complementary population models.

Large carnivores are difficult to monitor because they tend to be sparsely distributed, sensitive to human activity, and associated with complex life histories. Consequently, understanding population trend and viability requires conservationists to cope with uncertainty and bias in population data. Joint analysis of combined data sets using multiple models (i.e., integrated population model) can improve inference about mechanisms (e.g., habitat heterogeneity and food distribution) affecting population dynamics. However, unobserved or unobservable processes can also introduce bias and can be difficult to quantify. We developed a Bayesian hierarchical modeling approach for inference on an integrated population model that reconciles annual population counts with recruitment and survival data (i.e., demographic processes). Our modeling framework is flexible and enables a realistic form of population dynamics by fitting separate density-dependent responses for each demographic process. Discrepancies estimated from shared parameters among different model components represent unobserved additions (i.e., recruitment or immigration) or removals (i.e., death or emigration) when annual population counts are reliable. In a case study of gray wolves in Wisconsin (1980-2011), concordant with policy changes, we estimated that a discrepancy of 0% (1980-1995), -2% (1996-2002), and 4% (2003-2011) in the annual mortality rate was needed to explain annual growth rate. Additional mortality in 2003-2011 may reflect density-dependent mechanisms, changes in illegal killing with shifts in wolf management, and nonindependent censoring in survival data. Integrated population models provide insights into unobserved or unobservable processes by quantifying discrepancies among data sets. Our modeling approach is generalizable to many population analysis needs and allows for identifying dynamic differences due to external drivers, such as management or policy changes.

Stable isotope turnover and half-life in animal tissues: a literature synthesis.

Stable isotopes of carbon, nitrogen, and sulfur are used as ecological tracers for a variety of applications, such as studies of animal migrations, energy sources, and food web pathways. Yet uncertainty relating to the time period integrated by isotopic measurement of animal tissues can confound the interpretation of isotopic data. There have been a large number of experimental isotopic diet shift studies aimed at quantifying animal tissue isotopic turnover rate λ (%·day(-1), often expressed as isotopic half-life, ln(2)/λ, days). Yet no studies have evaluated or summarized the many individual half-life estimates in an effort to both seek broad-scale patterns and characterize the degree of variability. Here, we collect previously published half-life estimates, examine how half-life is related to body size, and test for tissue- and taxa-varying allometric relationships. Half-life generally increases with animal body mass, and is longer in muscle and blood compared to plasma and internal organs. Half-life was longest in ecotherms, followed by mammals, and finally birds. For ectotherms, different taxa-tissue combinations had similar allometric slopes that generally matched predictions of metabolic theory. Half-life for ectotherms can be approximated as: ln (half-life) = 0.22*ln (body mass) + group-specific intercept; n = 261, p<0.0001, r2 = 0.63. For endothermic groups, relationships with body mass were weak and model slopes and intercepts were heterogeneous. While isotopic half-life can be approximated using simple allometric relationships for some taxa and tissue types, there is also a high degree of unexplained variation in our models. Our study highlights several strong and general patterns, though accurate prediction of isotopic half-life from readily available variables such as animal body mass remains elusive.

Modelling avian biodiversity using raw, unclassified satellite imagery.

Applications of remote sensing for biodiversity conservation typically rely on image classifications that do not capture variability within coarse land cover classes. Here, we compare two measures derived from unclassified remotely sensed data, a measure of habitat heterogeneity and a measure of habitat composition, for explaining bird species richness and the spatial distribution of 10 species in a semi-arid landscape of New Mexico. We surveyed bird abundance from 1996 to 1998 at 42 plots located in the McGregor Range of Fort Bliss Army Reserve. Normalized Difference Vegetation Index values of two May 1997 Landsat scenes were the basis for among-pixel habitat heterogeneity (image texture), and we used the raw imagery to decompose each pixel into different habitat components (spectral mixture analysis). We used model averaging to relate measures of avian biodiversity to measures of image texture and spectral mixture analysis fractions. Measures of habitat heterogeneity, particularly angular second moment and standard deviation, provide higher explanatory power for bird species richness and the abundance of most species than measures of habitat composition. Using image texture, alone or in combination with other classified imagery-based approaches, for monitoring statuses and trends in biological diversity can greatly improve conservation efforts and habitat management.

Commonly rare and rarely common: comparing population abundance of invasive and native aquatic species.

Invasive species are leading drivers of environmental change. Their impacts are often linked to their population size, but surprisingly little is known about how frequently they achieve high abundances. A nearly universal pattern in ecology is that species are rare in most locations and abundant in a few, generating right-skewed abundance distributions. Here, we use abundance data from over 24,000 populations of 17 invasive and 104 native aquatic species to test whether invasive species differ from native counterparts in statistical patterns of abundance across multiple sites. Invasive species on average reached significantly higher densities than native species and exhibited significantly higher variance. However, invasive and native species did not differ in terms of coefficient of variation, skewness, or kurtosis. Abundance distributions of all species were highly right skewed (skewness>0), meaning both invasive and native species occurred at low densities in most locations where they were present. The average abundance of invasive and native species was 6% and 2%, respectively, of the maximum abundance observed within a taxonomic group. The biological significance of the differences between invasive and native species depends on species-specific relationships between abundance and impact. Recognition of cross-site heterogeneity in population densities brings a new dimension to invasive species management, and may help to refine optimal prevention, containment, control, and eradication strategies.

Human and biophysical influences on fire occurrence in the United States.

National-scale analyses of fire occurrence are needed to prioritize fire policy and management activities across the United States. However, the drivers of national-scale patterns of fire occurrence are not well understood, and how the relative importance of human or biophysical factors varies across the country is unclear. Our research goal was to model the drivers of fire occurrence within ecoregions across the conterminous United States. We used generalized linear models to compare the relative influence of human, vegetation, climate, and topographic variables on fire occurrence in the United States, as measured by MODIS active fire detections collected between 2000 and 2006. We constructed models for all fires and for large fires only and generated predictive maps to quantify fire occurrence probabilities. Areas with high fire occurrence probabilities were widespread in the Southeast, and localized in the Mountain West, particularly in southern California, Arizona, and New Mexico. Probabilities for large-fire occurrence were generally lower, but hot spots existed in the western and south-central United States The probability of fire occurrence is a critical component of fire risk assessments, in addition to vegetation type, fire behavior, and the values at risk. Many of the hot spots we identified have extensive development in the wildland--urban interface and are near large metropolitan areas. Our results demonstrated that human variables were important predictors of both all fires and large fires and frequently exhibited nonlinear relationships. However, vegetation, climate, and topography were also significant variables in most ecoregions. If recent housing growth trends and fire occurrence patterns continue, these areas will continue to challenge policies and management efforts seeking to balance the risks generated by wildfires with the ecological benefits of fire.

Land-cover change and avian diversity in the conterminous United States.

Changes in land use and land cover have affected and will continue to affect biological diversity worldwide. Yet, understanding the spatially extensive effects of land-cover change has been challenging because data that are consistent over space and time are lacking. We used the U.S. National Land Cover Dataset Land Cover Change Retrofit Product and North American Breeding Bird Survey data to examine land-cover change and its associations with diversity of birds with principally terrestrial life cycles (landbirds) in the conterminous United States. We used mixed-effects models and model selection to rank associations by ecoregion. Land cover in 3.22% of the area considered in our analyses changed from 1992 to 2001, and changes in species richness and abundance of birds were strongly associated with land-cover changes. Changes in species richness and abundance were primarily associated with changes in nondominant types of land cover, yet in many ecoregions different types of land cover were associated with species richness than were associated with abundance. Conversion of natural land cover to anthropogenic land cover was more strongly associated with changes in bird species richness and abundance than persistence of natural land cover in nearly all ecoregions and different covariates were most strongly associated with species richness than with abundance in 11 of 17 ecoregions. Loss of grassland and shrubland affected bird species richness and abundance in forested ecoregions. Loss of wetland was associated with bird abundance in forested ecoregions. Our findings highlight the value of understanding changes in nondominant land cover types and their association with bird diversity in the United States.

Glycomacropeptide, a low-phenylalanine protein isolated from cheese whey, supports growth and attenuates metabolic stress in the murine model of phenylketonuria.

Phenylketonuria (PKU) is caused by a mutation in the phenylalanine (phe) hydroxylase gene and requires a low-phe diet plus amino acid (AA) formula to prevent cognitive impairment. Glycomacropeptide (GMP) contains minimal phe and provides a palatable alternative to AA formula. Our objective was to compare growth, body composition, and energy balance in Pah(enu2) (PKU) and wild-type mice fed low-phe GMP, low-phe AA, or high-phe casein diets from 3-23 wk of age. The 2 × 2 × 3 design included main effects of genotype, sex, and diet. Fat and lean mass were assessed by dual-energy X-ray absorptiometry, and acute energy balance was assessed by indirect calorimetry. PKU mice showed growth and lean mass similar to wild-type littermates fed the GMP or AA diets; however, they exhibited a 3-15% increase in energy expenditure, as reflected in oxygen consumption, and a 3-30% increase in food intake. The GMP diet significantly reduced energy expenditure, food intake, and plasma phe concentration in PKU mice compared with the casein diet. The high-phe casein diet or the low-phe AA diet induced metabolic stress in PKU mice, as reflected in increased energy expenditure and intake of food and water, increased renal and spleen mass, and elevated plasma cytokine concentrations consistent with systemic inflammation. The low-phe GMP diet significantly attenuated these adverse effects. Moreover, total fat mass, %body fat, and the respiratory exchange ratio (CO(2) produced/O(2) consumed) were significantly lower in PKU mice fed GMP compared with AA diets. In summary, GMP provides a physiological source of low-phe dietary protein that promotes growth and attenuates the metabolic stress induced by a high-phe casein or low-phe AA diet in PKU mice.

An evaluation of prior influence on the predictive ability of Bayesian model averaging.

Model averaging is gaining popularity among ecologists for making inference and predictions. Methods for combining models include Bayesian model averaging (BMA) and Akaike's Information Criterion (AIC) model averaging. BMA can be implemented with different prior model weights, including the Kullback-Leibler prior associated with AIC model averaging, but it is unclear how the prior model weight affects model results in a predictive context. Here, we implemented BMA using the Bayesian Information Criterion (BIC) approximation to Bayes factors for building predictive models of bird abundance and occurrence in the Chihuahuan Desert of New Mexico. We examined how model predictive ability differed across four prior model weights, and how averaged coefficient estimates, standard errors and coefficients' posterior probabilities varied for 16 bird species. We also compared the predictive ability of BMA models to a best single-model approach. Overall, Occam's prior of parsimony provided the best predictive models. In general, the Kullback-Leibler prior, however, favored complex models of lower predictive ability. BMA performed better than a best single-model approach independently of the prior model weight for 6 out of 16 species. For 6 other species, the choice of the prior model weight affected whether BMA was better than the best single-model approach. Our results demonstrate that parsimonious priors may be favorable over priors that favor complexity for making predictions. The approach we present has direct applications in ecology for better predicting patterns of species' abundance and occurrence.

Terrestrial, benthic, and pelagic resource use in lakes: results from a three-isotope Bayesian mixing model.

Fluxes of organic matter across habitat boundaries are common in food webs. These fluxes may strongly influence community dynamics, depending on the extent to which they are used by consumers. Yet understanding of basal resource use by consumers is limited, because describing trophic pathways in complex food webs is difficult. We quantified resource use for zooplankton, zoobenthos, and fishes in four low-productivity lakes, using a Bayesian mixing model and measurements of hydrogen, carbon, and nitrogen stable isotope ratios. Multiple sources of uncertainty were explicitly incorporated into the model. As a result, posterior estimates of resource use were often broad distributions; nevertheless, clear patterns were evident. Zooplankton relied on terrestrial and pelagic primary production, while zoobenthos and fishes relied on terrestrial and benthic primary production. Across all consumer groups terrestrial reliance tended to be higher, and benthic reliance lower, in lakes where light penetration was low due to inputs of terrestrial dissolved organic carbon. These results support and refine an emerging consensus that terrestrial and benthic support of lake food webs can be substantial, and they imply that changes in the relative availability of basal resources drive the strength of cross-habitat trophic connections.