Comparing stormwater quality and watershed typologies across the United States: A machine learning approach.

Watersheds continue to be urbanized across different regions of the United States, increasing the number of impaired waterbodies due to urban stormwater. Using machine learning techniques, this study examined how stormwater quality and watershed characteristics are related at a national scale and compared stormwater quality across watersheds in diverse climates. We analyzed a selection of data from the National Stormwater Quality Database (NSQD) comprising 1,881 stormwater samples taken from 182 watersheds in 26 metropolitan areas in the United States between 1992 and 2003. Using an ensemble clustering algorithm, the stormwater quality in these samples was classified into "stormwater signatures," defined as distinct combinations of 9 contaminants including metals (Pb, Zn, Cu), particulates (TSS, TDS), and nutrients (BOD, TP, TKN, NOx). Next, multinomial logistic regression was applied to the NSQD data now classified by signature and combined with climate, weather, land use, and imperviousness data obtained from multiple sources. The results yielded 5 stormwater signatures with distinct aquatic toxicity implications and relationships to climate, weather, land use, and imperviousness: Signature 1 ("Ecotoxic and Eutrophic"), defined by high median concentrations of contaminants, likely represents the first flush in moderate-to-high imperviousness watersheds; Signature 2 ("Reduced Nitrates") represents a wet season signature, particularly for dry climates; Signature 3 ("Potentially Eutrophic") represents the first flush in low imperviousness watersheds; Signature 4 ("Elevated Particulates and Metals") represents a wet season signature, particularly on warmer days; finally, Signature 5 ("Most Dilute") is primarily a regional signature associated with the warm, wet climate of the southeastern US. This study serves as a proof-of-concept demonstrating how machine learning techniques can be used to identify patterns in high-dimensional and highly variable data. Applied to stormwater quality, these techniques identify major patterns in stormwater quality across the United States using a stormwater signature approach, which examines how contaminants co-occur and under what climate, weather, land use, and impervious conditions. The findings point to dominant processes driving stormwater generation and inform watershed monitoring, green infrastructure planning, stormwater quality under climate change, and opportunities for public engagement.

Insights and approaches using deep learning to classify wildlife.

The implementation of intelligent software to identify and classify objects and individuals in visual fields is a technology of growing importance to operatives in many fields, including wildlife conservation and management. To non-experts, the methods can be abstruse and the results mystifying. Here, in the context of applying cutting edge methods to classify wildlife species from camera-trap data, we shed light on the methods themselves and types of features these methods extract to make efficient identifications and reliable classifications. The current state of the art is to employ convolutional neural networks (CNN) encoded within deep-learning algorithms. We outline these methods and present results obtained in training a CNN to classify 20 African wildlife species with an overall accuracy of 87.5% from a dataset containing 111,467 images. We demonstrate the application of a gradient-weighted class-activation-mapping (Grad-CAM) procedure to extract the most salient pixels in the final convolution layer. We show that these pixels highlight features in particular images that in some cases are similar to those used to train humans to identify these species. Further, we used mutual information methods to identify the neurons in the final convolution layer that consistently respond most strongly across a set of images of one particular species. We then interpret the features in the image where the strongest responses occur, and present dataset biases that were revealed by these extracted features. We also used hierarchical clustering of feature vectors (i.e., the state of the final fully-connected layer in the CNN) associated with each image to produce a visual similarity dendrogram of identified species. Finally, we evaluated the relative unfamiliarity of images that were not part of the training set when these images were one of the 20 species "known" to our CNN in contrast to images of the species that were "unknown" to our CNN.

Modeling epidemics: A primer and Numerus Model Builder implementation.

Epidemiological models are dominated by compartmental models, of which SIR formulations are the most commonly used. These formulations can be continuous or discrete (in either the state-variable values or time), deterministic or stochastic, or spatially homogeneous or heterogeneous, the latter often embracing a network formulation. Here we review the continuous and discrete deterministic and discrete stochastic formulations of the SIR dynamical systems models, and we outline how they can be easily and rapidly constructed using Numerus Model Builder, a graphically-driven coding platform. We also demonstrate how to extend these models to a metapopulation setting using NMB network and mapping tools.

Fear-based niche shifts in neotropical birds.

Predation is a strong ecological force that shapes animal communities through natural selection. Recent studies have shown the cascading effects of predation risk on ecosystems through changes in prey behavior. Minimizing predation risk may explain why multiple prey species associate together in space and time. For example, mixed-species flocks that have been widely documented from forest systems, often include birds that eavesdrop on sentinel species (alarm calling heterospecifics). Sentinel species may be pivotal in (1) allowing flocking species to forage in open areas within forests that otherwise incur high predation risk, and (2) influencing flock occurrence (the amount of time species spend with a flock). To test this, we conducted a short-term removal experiment in an Amazonian lowland rainforest to test whether flock habitat use and flock occurrence was influenced by sentinel presence. Antshrikes (genus Thamnomanes) act as sentinels in Amazonian mixed-species flocks by providing alarm calls widely used by other flock members. The alarm calls provide threat information about ambush predators such as hawks and falcons which attack in flight. We quantified home range behavior, the forest vegetation profile used by flocks, and the proportion occurrence of other flocking species, both before and after removal of antshrikes from flocks. We found that when sentinel species were removed, (1) flock members shifted habitat use to lower risk habitats with greater vegetation cover, and (2) species flock occurrence decreased. We conclude that eavesdropping on sentinel species may allow other species to expand their realized niche by allowing them to safely forage in high-risk habitats within the forest. In allowing species to use extended parts of the forest, sentinel species may influence overall biodiversity across a diverse landscape.

Parasite biodiversity faces extinction and redistribution in a changing climate.

Climate change is a well-documented driver of both wildlife extinction and disease emergence, but the negative impacts of climate change on parasite diversity are undocumented. We compiled the most comprehensive spatially explicit data set available for parasites, projected range shifts in a changing climate, and estimated extinction rates for eight major parasite clades. On the basis of 53,133 occurrences capturing the geographic ranges of 457 parasite species, conservative model projections suggest that 5 to 10% of these species are committed to extinction by 2070 from climate-driven habitat loss alone. We find no evidence that parasites with zoonotic potential have a significantly higher potential to gain range in a changing climate, but we do find that ectoparasites (especially ticks) fare disproportionately worse than endoparasites. Accounting for host-driven coextinctions, models predict that up to 30% of parasitic worms are committed to extinction, driven by a combination of direct and indirect pressures. Despite high local extinction rates, parasite richness could still increase by an order of magnitude in some places, because species successfully tracking climate change invade temperate ecosystems and replace native species with unpredictable ecological consequences.

Carnivore carcasses are avoided by carnivores.

Ecologists have traditionally focused on herbivore carcasses as study models in scavenging research. However, some observations of scavengers avoiding feeding on carnivore carrion suggest that different types of carrion may lead to differential pressures. Untested assumptions about carrion produced at different trophic levels could therefore lead ecologists to overlook important evolutionary processes and their ecological consequences. Our general goal was to investigate the use of mammalian carnivore carrion by vertebrate scavengers. In particular, we aimed to test the hypothesis that carnivore carcasses are avoided by other carnivores, especially at the intraspecific level, most likely to reduce exposure to parasitism. We take a three-pronged approach to study this principle by: (i) providing data from field experiments, (ii) carrying out evolutionary simulations of carnivore scavenging strategies under risks of parasitic infection, and (iii) conducting a literature-review to test two predictions regarding parasite life-history strategies. First, our field experiments showed that the mean number of species observed feeding at carcasses and the percentage of consumed carrion biomass were substantially higher at herbivore carcasses than at carnivore carcasses. This occurred even though the number of scavenger species visiting carcasses and the time needed by scavengers to detect carcasses were similar between both types of carcasses. In addition, we did not observe cannibalism. Second, our evolutionary simulations demonstrated that a risk of parasite transmission leads to the evolution of scavengers with generally low cannibalistic tendencies, and that the emergence of cannibalism-avoidance behaviour depends strongly on assumptions about parasite-based mortality rates. Third, our literature review indicated that parasite species potentially able to follow a carnivore-carnivore indirect cycle, as well as those transmitted via meat consumption, are rare in our study system. Our findings support the existence of a novel coevolutionary relation between carnivores and their parasites, and suggest that carnivore and herbivore carcasses play very different roles in food webs and ecosystems.