Random forests for classification in ecology

Classification procedures are some of the most widely used statistical methods in ecology. Random forests (RF) is a new and powerful statistical classifier that is well established in other disciplines but is relatively unknown in ecology. Advantages of RF compared to other statistical classifiers include (1) very high classification accuracy; (2) a novel method of determining variable importance; (3) ability to model complex interactions among predictor variables; (4) flexibility to perform several types of statistical data analysis, including regression, classification, survival analysis, and unsupervised learning; and (5) an algorithm for imputing missing values. We compared the accuracies of RF and four other commonly used statistical classifiers using data on invasive plant species presence in Lava Beds National Monument, California, USA, rare lichen species presence in the Pacific Northwest, USA, and nest sites for cavity nesting birds in the Uinta Mountains, Utah, USA. We observed high classification accuracy in all applications as measured by cross-validation and, in the case of the lichen data, by independent test data, when comparing RF to other common classification methods. We also observed that the variables that RF identified as most important for classifying invasive plant species coincided with expectations based on the literature.     

Guiding principles for rewilding

There has been much recent interest in the concept of rewilding as a tool for nature conservation, but also confusion over the idea, which has limited its utility. We developed a unifying definition and 10 guiding principles for rewilding through a survey of 59 rewilding experts, a summary of key organizations’ rewilding visions, and workshops involving over 100 participants from around the world. The guiding principles convey that rewilding exits on a continuum of scale, connectivity, and level of human influence and aims to restore ecosystem structure and functions to achieve a self-sustaining autonomous nature. These principles clarify the concept of rewilding and improve its effectiveness as a tool to achieve global conservation targets, including those of the UN Decade on Ecosystem Restoration and post-2020 Global Biodiversity Framework. Finally, we suggest differences in rewilding perspectives lie largely in the extent to which it is seen as achievable and in specific interventions. An understanding of the context of rewilding projects is the key to success, and careful site-specific interpretations will help achieve the aims of rewilding.     

Effect of scale on trait predictors of species responses to agriculture

Species persistence in human‐altered landscapes can depend on factors operating at multiple spatial scales. To understand anthropogenic impacts on biodiversity, it is useful to examine relationships between species traits and their responses to land‐use change. A key knowledge gap concerns whether these relationships vary depending on the scale of response under consideration. We examined how local‐ and large‐scale habitat variables influence the occupancy dynamics of a bird community in cloud forest zones in the Colombian Chocó‐Andes. Using data collected across a continuum of forest and agriculture, we examined which traits best predict species responses to local variation in farmland and which traits best predict species responses to isolation from contiguous forest. Global range size was a strong predictor of species responses to agriculture at both scales; widespread species were less likely to decline as local habitat cover decreased and as distance from forest increased. Habitat specialization was a strong predictor of species responses only at the local scale. Open‐habitat species were particularly likely to increase as pasture increased, but they were relatively insensitive to variation in distance to forest. Foraging plasticity and flocking behavior were strong predictors of species responses to distance from forest, but not their responses to local habitat. Species with lower plasticity in foraging behaviors and obligate flock‐following species were more likely to decline as distance from contiguous forest increased. For species exhibiting these latter traits, persistence in tropical landscapes may depend on the protection of larger contiguous blocks of forest, rather than the integration of smaller‐scale woodland areas within farmland. Species listed as threatened or near threatened on the International Union for Conservation of Nature Red List were also more likely to decline in response to both local habitat quality and isolation from forest relative to least‐concern species, underlining the importance of contiguous forests for threatened taxa.     

Consequential cognition: Exploring how attribution theory sheds new light on the firm‐level consequences of product recalls

It is unclear why some firms suffer greater negative consequences than others following a product recall. To shed light on this question, we extend attribution theory to the firm level to explore how consumers engage in an attributional process following product recalls that shapes their responses to firms. Integrating attribution theory and the demand side theoretical perspective, we assert that consumer judgments of responsibility toward manufacturers are shaped by causal data regarding the locus of causality (i.e., manufacturer or supplier) and controllability (i.e., prior knowledge or awareness) of a recalled product's potential for harm. We then examine the impact of product recall characteristics on judgments of responsibility and firm‐level outcomes using an experimental test involving responses from 320 subjects. Our findings suggest that judgments of responsibility are attributed to the manufacturing firm more when consumers are given causal information indicating that the firm is the source of, or is aware of, a product's defects. The results also indicate that judgments of responsibility can have costly firm‐level consequences in the form of reputational damage, diminished consumer purchase intentions, and increased legal damage recommendations. We discuss theoretical contributions, practical implications, and opportunities for further research.     

Assimilation efficiency of a temperate-zone intertidal fish (Cebidichthys violaceus) fed diets of macroalgae

We studied assimilation efficiencies of the temperate-zone intertidal fish Cebidichthys violaceus (Girard, 1854) fed in the laboratory on each of the following species of macroalgae: Spongomorpha coalita (Chlorophyta), Ulva lobata (Chlorophyta), Iridaea flaccida (Rhodophyta) and Porphyra perforata (Rhodophyta). Together, these 4 algae make up over 75% of the natural summer diet of C. violaceus. Assimilation efficiency was calculated by proximate organic analysis of food and feces; the amount of ash in food and feces was used as a standard. Depending on the algal species, the fish assimilated 43 to 81% of the protein, 21 to 44% of the lipid, 45 to 62% of the carbohydrate and 31 to 52% of all three classes of organic material combined. These data are the first results showing that a temperate-zone marine fish can assimilate macroalgal constituents. Protein, carbohydrate and total organic material were absorbed more efficiently from rhodophytes than from chlorophytes. Conversely, lipid was absorbed more efficiently from chlorophytes than from rhodophytes. These results are compared with previous work showing that C. violaceus in nature eats more chlorophytes than rhodophytes, but in laboratory preference tests prefers rhodophytes to chlorophytes.     

Arsenic and heavy metal mobility in iron oxide-amended contaminated soils as evaluated by short- and long-term leaching tests

Several iron-bearing additives were evaluated for their effectiveness in the attenuation of arsenic (As) in various contaminated soils. These were selected for their known or potential ability to adsorb As anions, thus changing the speciation of As in a soil system. Three soils with different sources of As contamination were investigated (canal dredgings, coal fly ash deposits, and low-level alkali waste). The amendments used were goethite (alpha-FeOOH), iron grit, iron (II) and (III) sulphates (plus lime), and lime, applied to the soils at a rate of 1% w/w. A series of leachate extraction tests (UKEA, ASTM and modified Dutch column leaching test) were conducted on the equilibrated amended soils. These were used to firstly evaluate the potential of the amendments as immobilising agents, and secondly to compare the short- and long-term durability of their effects. Column tests demonstrated the efficiency of iron oxides over the longer time scale; these treatments significantly reduced concentrations of arsenic in leachates from all treated soils. Amended soils were also observed to contain higher levels of lead (Pb) and cadmium (Cd) in their leachates, signifying that certain Fe-oxides potentially increased heavy metal mobility in treated soils. The conclusions were that whilst Fe-oxides may be used as effective in situ amendments to attenuate As in soils, their effects on other trace elements, such as Pb and Cd, require careful consideration.     

Can adsorption isotherms predict sediment bioavailability?

The adsorption and desorption of 2,4-dichlorophenol (DCP) and pentachlorophenol (PCP) were studied for a range of synthetic particles, a dimethylditallowammonium exchanged clay and a natural sediment. The synthetic particles were Dowex 1X8400, Toyopearl Phenyl 650M and Toyopearl SP 650M. The bioaccumulation of the DCP and PCP from these particles was then studied using the oligochaete, Lumbriculus variegatus. There is a correlation between contaminant-particle interactions, as determined from adsorption and desorption isotherms, and bioaccumulation. Bioaccumulation by L. variegatus was found to be highest from the systems where differences in the classification of adsorption and desorption isotherms were observed.     

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.     

Locational Errors in Maps from Environmental Surveys and Their Implications for Information Extraction

1.Information on all maps, including those which depict characters of the natural environment, is subject to locational error. 2.Locational errors on field-sketched maps are attributable to inaccuracies both in the base map itself and in the method of locating and recording features onto the base map. 3.Two main sources of error associated with field sketch mapping were quantified using the River Corridor Survey maps as an example. First, channel widths are routinely exaggerated on these maps to allow recording of in-channel features. Distortions of river bank length of the order of 25% were shown to arise from a four-fold exaggeration of the width of a sinuous channel. Second, discrepancies in feature location of the order of 20% were found along a 500 m river stretch, although errors could be reduced to approximately 2% if undistorted, large-scale base maps were used. 4.An abundance scale appropriate to the inherent locational inaccuracy was devised to extract semi-quantitative information from River Corridor Surveys recorded on distorted base maps. Application of this scale to 29 River Corridor Survey maps of a 14·5 km river corridor produced a useful representation of environmental gradients within the corridor. 5.With the increasing potential to integrate and analyse environmental information within large computer databases including Geographic Information Systems, it is essential that the level of accuracy is recognised and quantified so that the errors can be allowed for in the development of methods of extracting information from the maps.