Accurate image-based identification of macroinvertebrate specimens using deep learning-How much training data is needed?

Image-based methods for species identification offer cost-efficient solutions for biomonitoring. This is particularly relevant for invertebrate studies, where bulk samples often represent insurmountable workloads for sorting, identifying, and counting individual specimens. On the other hand, image-based classification using deep learning tools have strict requirements for the amount of training data, which is often a limiting factor. Here, we examine how classification accuracy increases with the amount of training data using the BIODISCOVER imaging system constructed for image-based classification and biomass estimation of invertebrate specimens. We use a balanced dataset of 60 specimens of each of 16 taxa of freshwater macroinvertebrates to systematically quantify how classification performance of a convolutional neural network (CNN) increases for individual taxa and the overall community as the number of specimens used for training is increased. We show a striking 99.2% classification accuracy when the CNN (EfficientNet-B6) is trained on 50 specimens of each taxon, and also how the lower classification accuracy of models trained on less data is particularly evident for morphologically similar species placed within the same taxonomic order. Even with as little as 15 specimens used for training, classification accuracy reached 97%. Our results add to a recent body of literature showing the huge potential of image-based methods and deep learning for specimen-based research, and furthermore offers a perspective to future automatized approaches for deriving ecological data from bulk arthropod samples.

Vulnerability of Aquatic Insect Species to Insecticides, Depending on Their Flight Period and Adult Life Span.

Effects of insecticides on terrestrial adult life stages of otherwise aquatic insects, such as mayflies (Ephemeroptera), stoneflies (Plecoptera), and caddisflies (Trichoptera), are largely unknown. In the present study, a risk model was used to pinpoint the species most likely to experience effects due to spray drift exposure during the adult life stage. Using data from an earlier case study with lambda-cyhalothrin, 6 species with different life cycle traits were used to explore how life cycle characteristics may influence vulnerability. In addition, we performed a generic calculation of the potential effect on the terrestrial life stages of 53 species (including 47 species with unknown sensitivity). Our approach incorporated temporal and spatial distribution of both the insect and the insecticide, creating different exposure conditions among species due to variation in the relative proportion of the populations present at the time of insecticide spraying. The Ephemeroptera species represented were least vulnerable due to their extremely short adult life span and relatively short flight period. Based on their life cycle characteristics, Plecoptera and Trichoptera species were more vulnerable. These vulnerable species segregated into 2 distinct groups; one with a long adult life span to emergent period ratio and another with a high overlap between emergent period and spraying season. We therefore recommend that future ecotoxicological tests be done on species with these life cycle characteristics. Environ Toxicol Chem 2021;40:1778-1787. © 2021 SETAC.

Terrestrial adult stages of freshwater insects are sensitive to insecticides.

Terrestrial adult stages of freshwater insects may be exposed to pesticides by wind drift, over-spray, contact or feeding. However, studies addressing insecticide effects on freshwater invertebrates focus primarily on the impact of pesticides reaching the streams and potentially harming the aquatic juvenile stages. This is also reflected in the current risk assessment procedures, which do not include testing of adult freshwater insects. In order to assess the potential impact of insecticides on adult stages of freshwater insects, we exposed six common species to the insecticides Karate (lambda-cyhalothrin) and Confidor (imidacloprid). Dose-response relations were established, and LD50 estimates were compared to those of the honey bee, Apis mellifera L. (Hymenoptera: Apidae), which is the standard terrestrial test insect when pesticides are evaluated prior to commercial release. Generally, the tested species were more sensitive to the studied insecticides than the honey bee. In order to examine whether the sensitivity of adult stages of freshwater insects corresponds with the sensitivity of the juvenile stages of the same species, the ranking of the two life stages with respect to the toxicity of Karate was compared, revealing some correspondence, but also some dissimilarities. Our results strongly indicate that terrestrial adult stages of aquatic insects are not adequately protected by current risk assessment procedures.

Effects on hawthorn the year after simulated spray drift.

This paper describes the effect of a herbicide applied at levels consistent with off target movement on hawthorn the year following exposure. In the experiment, metsulfuron-methyl was applied in five dosages to individual trees in seven different hawthorn hedgerows. Spraying was conducted both at the bud stage and at the early flowering. Five endpoints (i.e. leaves, buds, flowers, green berries and mature berries) were sampled and counted. All were significantly reduced with increasing exposure. Present day risk assessment of effects on non-target plants is therefore likely to overlook significant effects on perennial non-target plants in the spray drift zone due to the focus on results from short-term laboratory test studies. The significance of the present study is underlined by the fact that the effects observed were significant, even though other influential factors such as herbivory and differential pollination were not eliminated and that experiments were conducted in multiple locations.

Metsulfuron spray drift reduces fruit yield of hawthorn (Crataegus monogyna L.).

This study was carried out to investigate whether spray drift of metsulfuron has a potential to negatively affect hawthorn (Crataegus monogyna) hedgerows near agricultural fields. For this purpose four doses of metsulfuron ranging from 5% to 40% of the field dose (4 g metsulfuron per hectare) were sprayed on trees in seven different hawthorn hedgerows. The actual deposition on the leaves was measured by means of a tracer (glycine). Spraying was conducted both at the bud stage and at early flowering. Leaves, flowers, green berries and mature berries were harvested and the number and weight of each were measured. The spraying at the bud stage caused a highly significant reduction in number and dry weight of berries, whereas it had no effects on leaf and flower production. The berry reduction was close to 100% at actual depositions relevant for spray drift under normal conditions. Spraying at early flowering also significantly reduced berries although the effect was smaller than for the spraying at bud stage. The early flower stage spraying caused no reduction in number and size of leaves. The possible ecological consequence is that metsulfuron spray drift from agricultural fields has a potential to reduce the amount of berries available for frugivorous birds in nearby hedgerows. A potential need for regulatory measures to reduce herbicide spray drift to hedgerows situated near agricultural fields with herbicide use is also indicated.

Toxicity of eight polycyclic aromatic compounds to red clover (Trifolium pratense), ryegrass (Lolium perenne), and mustard (Sinapsis alba).

The effect of eight polycyclic aromatic compounds (PACs) on the seed emergence and early life-stage growth of three terrestrial plants (Sinapsis alba, Trifolium pratense and Lolium perenne) were studied in a greenhouse, using a Danish agricultural soil with an organic carbon content of 1.6%. After three weeks of exposure, seed emergence and seedling weight (fresh weight and dry weight) were determined. Exposure concentrations were verified with chemical analysis. The substances tested were four polycyclic aromatic hydrocarbons (fluoranthene, pyrene, phenanthrene and fluorene), the N-, S-, and O-substituted analogues of fluorene (carbazole, dibenzothiophene and dibenzofuran, respectively), and the quinoline representative acridine. Seedling growth was a far more sensitive endpoint than seed emergence for all substances. Concentrations estimated to give a 20% reduction of seedling fresh weight (EC20-values) ranged from 36 to 290 mgkg(-1) for carbazole, 43 to 93 mgkg(-1) for dibenzofuran, 37 to 110 mgkg(-1) for dibenzothiophene, 140 to 650 mgkg(-1) for fluoranthene, 55 to 380 mgkg(-1) for fluorene, 37 to 300 mgkg(-1) for phenanthrene, and 49 to 1300 mgkg(-1) for pyrene. For acridine, no toxicity was observed within the concentration range tested (1-1000 mgkg(-1)). As illustrated by the EC20-values, there was a rather large difference in sensitivity between the species, and T. pratense was the most sensitive of the species tested.

Comparison of the toxicity of dimethoate and cypermethrin in the laboratory and the field when applying the same bioassay.

The present paper studies the relationship between laboratory- and field-based survival data of beetle larvae sprayed with either dimethoate or cypermethrin. The comparison for dimethoate was based on an earlier constructed pesticide-effect model derived from greenhouse data. To predict field effects, the model was supplied with field data for temperatures, the insecticide concentrations on the plants, and the control mortality rate of the beetles. The same model structure was unable to predict greenhouse mortality from cypermethrin treatment, for which reason the differences between laboratory and field for this compound are discussed on the basis of direct observations. For best comparison, the same bioassay method was used in both the greenhouse and the field in the present study. Tests were based on Gastrophysa polygoni larvae dwelling on the undersides of leaves of Fallopia convolvulus plants. Dietary uptake was regarded as the main source of exposure. The simulation results for dimethoate underestimated the field mortality even after accounting for the higher control mortality in the field in the model. It was shown that an additional increase of the toxicant-induced death rate by more than a factor of five was required for the model to fit the field data. This indicated a relatively severe impact of dimethoate under field conditions. For cypermethrin, the effects in the field increased with time, pointing at increasing exposure of and/or effects on the larvae. Contributions of the present results toward the bridging of the gap between laboratory and field are discussed.

Global solar radiation as the factor controlling induction of diapause in the pod midge (Dasyneura brassicae Winn.).

In this paper we examine the hypothesis that diapause induction in the polyvoltine pod midge Dasyneura brassicae Winn. (Diptera: Cecidomyiidae) is controlled by cumulative global solar radiation during the larval stage. The correlation between field observations of relative diapause for 3 years and four environmental factors (daylength, temperature, cumulative daylight time, and cumulative global solar radiation) was investigated. The cumulative global solar radiation during the larval stage clearly showed the strongest correlation with the observed relative diapause. It was estimated that larvae which cumulated less than 71.1 kWh m-2 entered diapause. Compared to former theories on diapause induction in insects, this hypothesis explains how temperature and light can function together and, furthermore, it removes the requirement for biological clock involvement in diapause induction in insects.