Gradient forests: calculating importance gradients on physical predictors

In ecological analyses of species and community distributions there is interest in the nature of their responses to environmental gradients and in identifying the most important environmental variables, which may be used for predicting patterns of biodiversity. Methods such as random forests already exist to assess predictor importance for individual species and to indicate where along gradients abundance changes. However, there is a need to extend these methods to whole assemblages, to establish where along the range of these gradients the important compositional changes occur, and to identify any important thresholds or change points. We develop such a method, called "gradient forest," which is an extension of the random forest approach. By synthesizing the cross-validated R2 and accuracy importance measures from univariate random forest analyses across multiple species, sampling devices, and surveys, gradient forest obtains a monotonic function of each predictor that represents the compositional turnover along the gradient of the predictor. When applied to a synthetic data set, the method correctly identified the important predictors and delineated where the compositional change points occurred along these gradients. Application of gradient forest to a real data set from part of the Great Barrier Reef identified mud fraction of the sediment as the most important predictor, with highest compositional turnover occurring at mud fraction values around 25%, and provided similar information for other predictors. Such refined information allows for more accurate capturing of biodiversity patterns for the purposes of bioregionalization, delineation of protected areas, or designing of biodiversity surveys.     

Balancing the edge effects budget: bay scallop settlement and loss along a seagrass edge

Edge effects are a dominant subject in landscape ecology literature, yet they are highly variable and poorly understood. Often, the literature suggests simple models for edge effects-positive (enhancement at the edge), negative (enhancement at the interior), or no effect (neutral)--on a variety of metrics, including abundance, diversity, and mortality. In the marine realm, much of this work has focused on fragmented seagrass habitats due to their importance for a variety of commercially important species. In this study, the settlement, recruitment, and survival of bay scallops was investigated across a variety of seagrass patch treatments. By simultaneously collecting settlers (those viable larvae available to settle and metamorphose) and recruits (those settlers that survive some period of time, in this case, 6 weeks) on the same collectors, we were able to demonstrate a "balance" between positive and negative edge effects, resulting in a net neutral effect. Scallop settlement was significantly enhanced along seagrass edges, regardless of patch type while survival was elevated within patch interiors. However, recruitment (the net result of settlement and post-settlement loss) did not vary significantly from edge to center, representing a neutral effect. Further, results suggest that post-settlement loss, most likely due to predation, appears to be the dominant mechanism structuring scallop abundance, not patterns in settlement. These data illustrate the complexity of edge effects, and suggest that the metric used to investigate the effect (be it abundance, survival, or other metrics) can often influence the magnitude and direction of the perceived effect. Traditionally, high predation along a habitat edge would have indicated an "ecological trap" for the species in question; however, this study demonstrates that, at the population level, an ecological trap may not exist.     

Multi-decadal drought and amplified moisture variability drove rapid forest community change in a humid region

Climate variability, particularly the frequency of extreme events, is likely to increase in the coming decades, with poorly understood consequences for terrestrial ecosystems. Hydroclimatic variations of the Medieval Climate Anomaly (MCA) provide a setting for studying ecological responses to recent climate variability at magnitudes and timescales comparable to expectations of coming centuries. We examined forest response to the MCA in the humid western Great Lakes region of North America, using proxy records of vegetation, fire, and hydroclimate. Multi-decadal moisture variability during the MCA was associated with a widespread, episodic decline in Fagus grandifolia (beech) populations. Spatial patterns of drought and forest changes were coherent, with beech declining only in areas where proxy-climate records indicate that severe MCA droughts occurred. The occurrence of widespread, drought-induced ecological changes in the Great Lakes region indicates that ecosystems in humid regions are vulnerable to rapid changes in drought magnitude and frequency.     

Assessment of monitoring power for highly mobile vertebrates

Monitoring of population trends is a critical component of conservation management, and development of practical methods remains a priority, particularly for species that challenge more standard approaches. We used field-parameterized simulation models to examine the effects of different errors on monitoring power and compared alternative methods used with two species of threatened pteropodids (flying-foxes), Pteropus conspicillatus and P. poliocephalus, whose mobility violates assumptions of closure on short and long timescales. The influence of three errors on time to 80% statistical power was assessed using a Monte Carlo approach. The errors were: (1) failure to count all animals at a roost, (2) errors associated with enumeration, and (3) variability in the proportion of the population counted due to the movement of individuals between roosts. Even with perfect accuracy and precision for these errors only marginal improvements in power accrued (-1%), with one exception. Improving certainty in the proportion of the population being counted reduced time to detection of a decline by over 6 yr (43%) for fly-out counts and almost 10 yr (71%) for walk-through counts. This error derives from the movement of animals between known and unknown roost sites, violating assumptions of population closure, and because it applies to the entire population, it dominates all other sources of error. Similar errors will accrue in monitoring of a wide variety of highly mobile species and will also result from population redistribution under climate change. The greatest improvements in monitoring performance of highly mobile species accrue through an improved understanding of the proportion of the population being counted, and consequently monitoring of such species must be done at the scale of the species or population range, not at the local level.     

Eradication via destratification: whole-lake mixing to selectively remove rainbow smelt, a cold-water invasive species

Invasive species can have severe effects on aquatic ecosystems. After invasions occur, eradication should be considered whenever the potential loss of ecosystem services outweighs the cost of the eradication method. Here we evaluate the possibility of destratifying Crystal Lake, Wisconsin, USA, to eradicate the invasive fish rainbow smelt (Osmerus mordax). We modeled the effects of three destratification scenarios (non-, low-, and high-mixing) using both physical and biological models. Field observations were used to calibrate the models. Water temperatures estimated from 18 unique DYRESM simulations were used in a bioenergetics model to estimate growth of five age classes of rainbow smelt under normal and destratified conditions. Our simulations indicate that destratification can eliminate optimal rainbow smelt thermal habitat resulting in mortality. Destratified lake temperatures also surpassed several physiological critical temperatures. Bioenergetics simulations predicted a weight loss of 45-55% in yearling and adult rainbow smelt. We found that destratification is potentially effective for eradicating cold-water species in temperate lakes.     

Genetic structure among coastal tailed frog populations at Mount St. Helens is moderated by post-disturbance management

Catastrophic disturbances often provide "natural laboratories" that allow for greater understanding of ecological processes and response of natural populations. The 1980 eruption of the Mount St. Helens volcano in Washington, USA, provided a unique opportunity to test biotic effects of a large-scale stochastic disturbance, as well as the influence of post-disturbance management. Despite severe alteration of nearly 600 km2 of habitat, coastal tailed frogs (Ascaphus truei) were found within a portion of the blast area five years after eruption. We investigated the genetic source of recolonization within the blast area and tested whether post-eruption salvage logging and subsequent tree planting influenced tailed frog movement patterns. Our results support widespread recolonization across the blast area from multiple sources, as all sites are grouped into one genetic cluster. Landscape genetic models suggest that gene flow through the unmanaged portion of the blast area is influenced only by distance between sites and the frost-free period (r2 = 0.74). In contrast, gene flow pathways within the blast area where salvage logging and replanting occurred post-eruption are strongly limited (r2 = 0.83) by the physiologically important variables of heat load and precipitation. These data suggest that the lack of understory and coarse wood (downed and standing dead tree boles) refugia in salvaged areas may leave frogs more susceptible to desiccation and mortality than those frogs moving through the naturally regenerated area. Simulated populations based on the landscape genetic models show an increase in the inbreeding coefficient in the managed area relative to the unmanaged blast area. In sum, we show surprising resilience of an amphibian species to a catastrophic disturbance, and we suggest that, at least for this species, naturally regenerating habitat may better maintain long-term genetic diversity of populations than actively managed habitat.     

Downscaling species occupancy from coarse spatial scales

The measurement and prediction of species' populations at different spatial scales is crucial to spatial ecology as well as conservation biology. An efficient yet challenging goal to achieve such population estimates consists of recording empirical species' presence and absence at a specific regional scale and then trying to predict occupancies at finer scales. So far the majority of the methods have been based on particular species' distributional features deemed to be crucial for downscaling occupancy. However, only a minority of them have dealt explicitly with specific spatial features. Here we employ a wide class of spatial point processes, the shot noise Cox processes (SNCP), to model species occupancies at different spatial scales and show that species' spatial aggregation is crucial for predicting population estimates at fine scales starting from coarser ones. These models are formulated in continuous space and locate points regardless of the arbitrary resolution that one employs to study the spatial pattern. We compare the performances of nine models, calibrated at regional scales and demonstrate that a very simple class of SNCP, the Thomas process, is able to outperform other published models in predicting occupancies down to areas four orders of magnitude smaller than the ones employed for the parameterization. We conclude by explaining the ability of the approach to infer spatially explicit information from spatially implicit measures, the potential of the framework to combine niche and spatial models, and the possibility of reversing the method to allow upscaling.     

First-instar monarch larval growth and survival on milkweeds in southern California: effects of latex, leaf hairs and cardenolides

Growth rate and survival of first-instar larvae of Danaus plexippus, a milkweed specialist, depended on milkweed species, and was related to the amount of latex produced from wounds, leaf cardenolide concentrations and the presence of leaf hairs. Larval growth was more rapid and survival was higher on leaves of Asclepias californica with experimentally reduced latex, and this species has characteristically high latex, low- to mid-range cardenolide concentrations, and very hirsute leaves. Similarly, growth was higher on reduced latex leaves of both A. eriocarpa (a high latex/high cardenolide, hirsute species) and A. erosa (glabrous fleshy leaves, high latex/high cardenolides). There were no differences in either survival or growth rate between larvae on reduced latex or control leaves of the low latex/low cardenolide A. fascicularis with soft glabrous leaves and both survival and growth rate were higher on this species than the other species tested. Larval growth rates on leaves with reduced latex were similar among ten milkweed species tested to date but differed from growth rates on intact leaves suggesting that latex and possibly included cardenolides are both important in first-instar monarch larval growth, development and survival. We show for a range of ecologically important milkweeds that experiments on cut plant material (no latex outflow) lead to higher growth rates compared to intact plants. Such laboratory assays based on detached leaves will be misleading if the objective is to determine the impact of treatments such as Bt-maize pollen on monarchs on field plants.     

Use of DNA barcoding to reveal species composition of convenience seafood

Increased education of consumers can be an effective tool for conservation of commercially harvested marine species when product labeling is accurate and allows an informed choice. However, generic labeling (e.g., as white fish or surimi) and mislabeling of seafood prevents this and may erode consumer confidence in seafood product labels in general. We used DNA barcoding to identify the species composition of two types of convenience seafood (i.e., products processed for ease of consumption): fish fingers (long pieces of fish covered with bread crumbs or batter, n = 241) and seafood sticks (long pieces of cooked fish, n = 30). In products labeled as either white fish or surimi, four teleost species were present. Less than 1.5% of fish fingers with species-specific information were mislabeled. Results of other studies show substantially more mislabeling (e.g., >25%) of teleost products, which likely reflects the lower economic gains associated with mislabeling of convenience seafood compared with whole fillets. In addition to species identification, seafood product labels should be required to contain information about, for example, harvesting practices, and our data indicate that consumers can have reasonable confidence in the accuracy of the labels of convenience seafood and thus select brands on the basis of information about current fisheries practice.