Analysis of Few-Shot Techniques for Fungal Plant Disease Classification and Evaluation of Clustering Capabilities Over Real Datasets.

Plant fungal diseases are one of the most important causes of crop yield losses. Therefore, plant disease identification algorithms have been seen as a useful tool to detect them at early stages to mitigate their effects. Although deep-learning based algorithms can achieve high detection accuracies, they require large and manually annotated image datasets that is not always accessible, specially for rare and new diseases. This study focuses on the development of a plant disease detection algorithm and strategy requiring few plant images (Few-shot learning algorithm). We extend previous work by using a novel challenging dataset containing more than 100,000 images. This dataset includes images of leaves, panicles and stems of five different crops (barley, corn, rape seed, rice, and wheat) for a total of 17 different diseases, where each disease is shown at different disease stages. In this study, we propose a deep metric learning based method to extract latent space representations from plant diseases with just few images by means of a Siamese network and triplet loss function. This enhances previous methods that require a support dataset containing a high number of annotated images to perform metric learning and few-shot classification. The proposed method was compared over a traditional network that was trained with the cross-entropy loss function. Exhaustive experiments have been performed for validating and measuring the benefits of metric learning techniques over classical methods. Results show that the features extracted by the metric learning based approach present better discriminative and clustering properties. Davis-Bouldin index and Silhouette score values have shown that triplet loss network improves the clustering properties with respect to the categorical-cross entropy loss. Overall, triplet loss approach improves the DB index value by 22.7% and Silhouette score value by 166.7% compared to the categorical cross-entropy loss model. Moreover, the F-score parameter obtained from the Siamese network with the triplet loss performs better than classical approaches when there are few images for training, obtaining a 6% improvement in the F-score mean value. Siamese networks with triplet loss have improved the ability to learn different plant diseases using few images of each class. These networks based on metric learning techniques improve clustering and classification results over traditional categorical cross-entropy loss networks for plant disease identification.

A Workflow for the Performance of the Differential Ovarian Follicle Count Using Deep Neuronal Networks.

In order to automate the counting of ovarian follicles required in multigeneration reproductive studies performed in the rat according to Organization for Economic Co-operation and Development guidelines 443 and 416, the application of deep neural networks was tested. The manual evaluation of the differential ovarian follicle count is a tedious and time-consuming task that requires highly trained personnel. In this regard, deep learning outputs provide overlay pictures for a more detailed documentation, together with an increased reproducibility of the counts. To facilitate the planned good laboratory practice (GLP) validation a workflow was set up using MLFlow to make all steps from generating of scans, training of the neural network, uploading of study images to the neural network, generation and storage of the results in a compliant manner controllable and reproducible. PyTorch was used as main framework to build the Faster region-based convolutional neural network for the training. We compared the performances of different depths of ResNet models with specific regard to the sensitivity, specificity, accuracy of the models. In this paper, we describe all steps from data labeling, training of networks, and the performance metrics chosen to evaluate different network architectures. We also make recommendation on steps, which should be taken into consideration when GLP validation is aimed for.

Life-history trade-offs mediate 'personality' variation in two colour morphs of the pea aphid, Acyrthosiphon pisum.

Life-history trade-offs are considered a major driving force in the emergence of consistent behavioural differences (personality variation); but empirical tests are scarce. We investigated links between a personality trait (escape response), life-history and state variables (growth rate, size and age at first reproduction, age-dependent reproductive rates, lifetime reproductive success, life span) in red and green colour morphs of clonal pea aphids, Acyrthosiphon pisum. Escape response (dropping/non-dropping off a plant upon a predatory attack) was measured repeatedly to classify individuals as consistent droppers, consistent nondroppers or inconsistents. Red morphs experienced stronger trade-offs between early reproduction and life span than green morphs; and red consistent (non)droppers had highest lifetime reproductive success. Red droppers followed a risk-averse life-history strategy (high late reproduction), red nondroppers a risk-prone strategy (high early reproduction), while reproductive rates were equivalent for all green behavioural types and red inconsistents. This suggests that red morphs suffer the highest costs of dropping (they are most conspicuous to predators), which 'equivalates' fitness payoffs to both risk-takers (red non-droppers) and risk-averse red droppers. The strong trade-off also means that committing to a particular lifestyle (being consistent) maximises fitness. Our study suggests that life-history trade-offs likely mediate personality variation but effects might depend on interactions with other organismal characteristics (here: colour morph).

Discontinuity in the responses of ecosystem processes and multifunctionality to altered soil community composition.

Ecosystem management policies increasingly emphasize provision of multiple, as opposed to single, ecosystem services. Management for such "multifunctionality" has stimulated research into the role that biodiversity plays in providing desired rates of multiple ecosystem processes. Positive effects of biodiversity on indices of multifunctionality are consistently found, primarily because species that are redundant for one ecosystem process under a given set of environmental conditions play a distinct role under different conditions or in the provision of another ecosystem process. Here we show that the positive effects of diversity (specifically community composition) on multifunctionality indices can also arise from a statistical fallacy analogous to Simpson's paradox (where aggregating data obscures causal relationships). We manipulated soil faunal community composition in combination with nitrogen fertilization of model grassland ecosystems and repeatedly measured five ecosystem processes related to plant productivity, carbon storage, and nutrient turnover. We calculated three common multifunctionality indices based on these processes and found that the functional complexity of the soil communities had a consistent positive effect on the indices. However, only two of the five ecosystem processes also responded positively to increasing complexity, whereas the other three responded neutrally or negatively. Furthermore, none of the individual processes responded to both the complexity and the nitrogen manipulations in a manner consistent with the indices. Our data show that multifunctionality indices can obscure relationships that exist between communities and key ecosystem processes, leading us to question their use in advancing theoretical understanding--and in management decisions--about how biodiversity is related to the provision of multiple ecosystem services.

Invertebrate herbivory increases along an experimental gradient of grassland plant diversity.

Plant diversity is a key driver of ecosystem functioning best documented for its influence on plant productivity. The strength and direction of plant diversity effects on species interactions across trophic levels are less clear. For example, with respect to the interactions between herbivorous invertebrates and plants, a number of competing hypotheses have been proposed that predict either increasing or decreasing community herbivory with increasing plant species richness. We investigated foliar herbivory rates and consumed leaf biomass along an experimental grassland plant diversity gradient in year eight after establishment. The gradient ranged from one to 60 plant species and manipulated also functional group richness (from one to four functional groups-legumes, grasses, small herbs, and tall herbs) and plant community composition. Measurements in monocultures of each plant species showed that functional groups differed in the quantity and quality of herbivory damage they experienced, with legumes being more damaged than grasses or non-legume herbs. In mixed plant communities, herbivory increased with plant diversity and the presence of two key plant functional groups in mixtures had a positive (legumes) and a negative (grasses) effect on levels of herbivory. Further, plant community biomass had a strong positive impact on consumed leaf biomass, but little effect on herbivory rates. Our results contribute detailed data from a well-established biodiversity experiment to a growing body of evidence suggesting that an increase of herbivory with increasing plant diversity is the rule rather than an exception. Considering documented effects of herbivory on other ecosystem functions and the increase of herbivory with plant diversity, levels of herbivory damage might not only be a result, but also a trigger within the diversity-productivity relationship.

The nitrite-oxidizing community in activated sludge from a municipal wastewater treatment plant determined by fatty acid methyl ester-stable isotope probing.

Metabolically-active autotrophic nitrite oxidizers from activated sludge were labeled with (13)C-bicarbonate under exposure to different temperatures and nitrite concentrations. The labeled samples were characterized by FAME-SIP (fatty acid methyl ester-stable isotope probing). The compound cis-11-palmitoleic acid, which is the major lipid of the most abundant nitrite oxidizer in activated sludge, Candidatus Nitrospira defluvii, showed (13)C-incorporation in all samples exposed to 3 mM nitrite. Subsequently, the lipid cis-7-palmitoleic acid was labeled, and it indicated the activity of a nitrite oxidizer that was different from the known Nitrospira taxa in activated sludge. The highest incorporation of cis-7-palmitoleic acid label was found after incubation with a nitrite concentration of 0.3 mM at 17 and 22°C. While activity of Nitrobacter populations could not be detected by the FAME-SIP approach, an unknown nitrite oxidizer with the major lipid cis-9 isomer of palmitoleic acid exhibited (13)C-incorporation at 28°C with 30 mM nitrite. These results indicated flexibility of nitrite-oxidizing guilds in a complex community responding to different conditions. Labeled lipids so far not described for activated sludge-associated nitrifiers indicated the presence of unknown nitrite oxidizers in this habitat. The FAME-SIP-based information can be used to define appropriate conditions for the enrichment of nitrite-oxidizing guilds from complex samples.

Relevance and diversity of Nitrospira populations in biofilters of brackish RAS.

Lithoautotrophic nitrite-oxidizing bacterial populations from moving-bed biofilters of brackish recirculation aquaculture systems (RAS; shrimp and barramundi) were tested for their metabolic activity and phylogenetic diversity. Samples from the biofilters were labeled with (13)C-bicarbonate and supplemented with nitrite at concentrations of 0.3, 3 and 10 mM, and incubated at 17 and 28°C, respectively. The biofilm material was analyzed by fatty acid methyl ester - stable isotope probing (FAME-SIP). High portions of up to 45% of Nitrospira-related labeled lipid markers were found confirming that Nitrospira is the major autotrophic nitrite oxidizer in these brackish systems with high nitrogen loads. Other nitrite-oxidizing bacteria such as Nitrobacter or Nitrotoga were functionally not relevant in the investigated biofilters. Nitrospira-related 16S rRNA gene sequences were obtained from the samples with 10 mM nitrite and analyzed by a cloning approach. Sequence studies revealed four different phylogenetic clusters within the marine sublineage IV of Nitrospira, though most sequences clustered with the type strain of Nitrospira marina and with a strain isolated from a marine RAS. Three lipids dominated the whole fatty acid profiles of nitrite-oxidizing marine and brackish enrichments of Nitrospira sublineage IV organisms. The membranes included two marker lipids (16∶1 cis7 and 16∶1 cis11) combined with the non-specific acid 16∶0 as major compounds and confirmed these marker lipids as characteristic for sublineage IV species. The predominant labeling of these characteristic fatty acids and the phylogenetic sequence analyses of the marine Nitrospira sublineage IV identified organisms of this sublineage as main autotrophic nitrite-oxidizers in the investigated brackish biofilter systems.

Personality variation in a clonal insect: the pea aphid, Acyrthosiphon pisum.

Individuals are often consistent in their behavior but vary from each other in the level of behavior shown. Despite burgeoning interest in such animal personality variation, studies on invertebrates are scarce, and studies on clonal invertebrates nonexistent. This is surprising given the obvious advantages of using invertebrates/clones to tackle the crucial question why such consistent behavioral differences exist. Here we show that individuals of clonal pea aphids exhibit consistent behavioral differences in their escape responses to a predator attack (dropping vs. nondropping off a plant). However, behavior was not repeatable at the clonal level. Genetically identical clones expressed various phenotypes but different clones produced different proportions of each phenotype (dropper, nondropper, and inconsistent). Manipulations of early environmental conditions had little qualitative impact on such patterns. We discuss the importance of our findings for future studies of the evolutionary and ecological consequences of personality variation.

Black and white--does melanin change the bulk carbon and nitrogen isotope values of feathers?

Bird feathers are employed in a wide range of carbon and nitrogen isotope studies relating to diet and migration. Feathers are chemically inert with respect to carbon and nitrogen, after synthesis. It has always been assumed that feathers show isotope values characteristic of keratin, a fibrous structural protein from which they are formed. Little attention has been paid to other components of feathers such as melanin or carotenoids. Melanin is synthesized from tyrosine, which is depleted in both (13)C and (15)N. We compared isotope values of coeval black and white feathers in four different species. Black feather parts were in all cases significantly depleted in (13)C relative to white feather parts but in most species no clear trend was discernable for (15)N. We suggest that additional evaluation may be required to characterize the carbon and nitrogen isotope contribution of feather pigments like carotenoids. Care should be taken in future stable isotope studies when comparing differently coloured feathers.

Impacts of invasive plant species on riparian plant assemblages: interactions with elevated atmospheric carbon dioxide and nitrogen deposition.

Resource competition is commonly invoked to explain negative effects of invasive plants on native plant abundance. If invasives out-compete natives, global changes that elevate resource availability may interact with invasives to exacerbate impacts on native communities. Indeed, evidence is accumulating that elevated CO(2) and N deposition decrease native biomass and simultaneously increase invasive biomass. However, superior competitive ability, and a relative increase in the magnitude of invasive impacts under elevated resource availability, remain to be definitively proven. Using model, multi-species, multi-individual riparian plant communities, where planting density was maintained by replacement of native with exotic individuals, we conducted a greenhouse, competition experiment using native (to the UK) and invaded communities exposed to ambient and elevated CO(2) (CO(2) experiment) or N availability (N experiment). We tested two hypotheses: (1) invasives are superior competitors to natives at ambient atmospheric CO(2) and N deposition; (2) negative effects of invasives on natives are exacerbated under elevated CO(2) or N availability. Our results provide some support for the first hypothesis: in the CO(2) experiment native biomass was significantly lower in invaded communities. In the N experiment, native biomass was unaffected by the presence of exotics but other characteristics (e.g. root:shoot ratios) were altered. Differences in light availability between the experiments may have modified the effects of the invasives on the native assemblages but our design did not permit us to determine this definitively. The hypothesis that elevated CO(2) and N availability benefit invasives at the expense of natives was not supported by our results. This may be explained either because the invasives showed minor responses to the resource manipulations or because native and exotic species were differentially limited by CO(2) and N. Our results confirm the expectation that invasives alter the characteristics of native assemblages but lead us to question whether elevated resource availability will magnify these effects.

Decoupling the direct and indirect effects of nitrogen deposition on ecosystem function.

Elevated nitrogen (N) inputs into terrestrial ecosystems are causing major changes to the composition and functioning of ecosystems. Understanding these changes is challenging because there are complex interactions between 'direct' effects of N on plant physiology and soil biogeochemistry, and 'indirect' effects caused by changes in plant species composition. By planting high N and low N plant community compositions into high and low N deposition model terrestrial ecosystems we experimentally decoupled direct and indirect effects and quantified their contribution to changes in carbon, N and water cycling. Our results show that direct effects on plant growth dominate ecosystem response to N deposition, although long-term carbon storage is reduced under high N plant-species composition. These findings suggest that direct effects of N deposition on ecosystem function could be relatively strong in comparison with the indirect effects of plant community change.