No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
The LOKI underwater imaging system and an automatic identification model for the detection of zooplankton taxa in the Arctic Ocean
Abstract
We deployed the Lightframe On-sight Keyspecies Investigation
(LOKI) system, a novel underwater imaging system providing
cutting-edge imaging quality, in the Canadian Arctic during fall
2013. A Random Forests machine learning model was built to
automatically identify zooplankton in LOKI images. The model
successfully distinguished between 114 different categories of
zooplankton and particles. The high resolution taxonomical tree included
many species, stages, as well as sub-groups based on animal
orientation or condition in images. Results from a machine learning
regression model of prosome length (R^2 = 0.97) were used
as a key predictor in the automatic identification model. Model internal
validation of the automatic identification model on test data
demonstrated that the model performed with overall high accuracy
(86%) and specificity (86%). This was confirmed by confusion
matrices for external testing results, based on automatic identifications
for 2 complete stations. For station 101, from which images
had also been used for training, accuracy and specificity were 85%.
For station 126, from which images had not been used to train the
model, accuracy and specificity were 81%. Further comparisons between
model results and microscope identifications of zooplankton
in samples from the two test stations were in good agreement
for most taxa. LOKI’s image quality makes it possible to build accurate
automatic identification models of very high taxonomic detail,
which will play a critical role in future studies of zooplankton dynamics
and zooplankton coupling with other trophic levels.
... • Evaluation and discoveries: Examples of these include the evaluation of corals and their inhabitants [16], seabed analysis and mapping (photo-mosaicking) [17], object classification and discovery [18], [19], plant identification [20], the automatic recognition of fish [14], [21], lobster [22], [23], plankton [24], and other species, as well as tracking and direction finding [25]. • Monitoring and management: Examples of these include environmental monitoring (e.g. ...
... • • Shrimp Detection by SLIC [23] • Plant Detection by Gabor Wavelet [20] • Plant Recognition by Gabor Wavelet [153] • Plankton Recognition by MSPA [24], [204] • Coral Detection by Clustering [205] • Object Detection by SLIC [18] • Object Detection by Gabor Wavelet [17] • Object Detection by Clustering [19] • General Underwater Clustering by FCM [206] • General Underwater Clustering by RG [207] • Coral Recognition by MSPA [134] • Coral Recognition by RG [208] Math Calculus (based on solving a Differential or Integral Equation) ...
... • Grey level co-occurrence matrices .. properties (contrast, correlation, energy, .. and homogeneity) [22], [24], [25], [134], [139], [204], [210], [215], [217] ...
The Internet of Underwater Things (IoUT) is an emerging communication ecosystem developed for connecting underwater objects in maritime and underwater environments. The IoUT technology is intricately linked with intelligent boats and ships, smart shores and oceans, automatic marine transportations, positioning and navigation, underwater exploration, disaster prediction and prevention, as well as with intelligent monitoring and security. The IoUT has an influence at various scales ranging from a small scientific observatory, to a midsized harbor, and to covering global oceanic trade. The network architecture of IoUT is intrinsically heterogeneous and should be sufficiently resilient to operate in harsh environments. This creates major challenges in terms of underwater communications, whilst relying on limited energy resources. Additionally, the volume, velocity, and variety of data produced by sensors, hydrophones, and cameras in IoUT is enormous, giving rise to the concept of Big Marine Data (BMD), which has its own processing challenges. Hence, conventional data processing techniques will falter, and bespoke Machine Learning (ML) solutions have to be employed for automatically learning the specific BMD behavior and features facilitating knowledge extraction and decision support. The motivation of this paper is to comprehensively survey the IoUT, BMD, and their synthesis. It also aims for exploring the nexus of BMD with ML. We set out from underwater data collection and then discuss the family of IoUT data communication techniques with an emphasis on the state-of-the-art research challenges. We then review the suite of ML solutions suitable for BMD handling and analytics. We treat the subject deductively from an educational perspective, critically appraising the material surveyed.
... • Evaluation and discoveries: Examples of these include the evaluation of corals and their inhabitants [16], seabed analysis and mapping (photo-mosaicking) [17], object classification and discovery [18], [19], plant identification [20], the automatic recognition of fish [14], [21], lobster [22], [23], plankton [24], and other species, as well as tracking and direction finding [25]. • Monitoring and management: Examples of these include environmental monitoring (e.g. ...
... • • Shrimp Detection by SLIC [23] • Plant Detection by Gabor Wavelet [20] • Plant Recognition by Gabor Wavelet [153] • Plankton Recognition by MSPA [24], [204] • Coral Detection by Clustering [205] • Object Detection by SLIC [18] • Object Detection by Gabor Wavelet [17] • Object Detection by Clustering [19] • General Underwater Clustering by FCM [206] • General Underwater Clustering by RG [207] • Coral Recognition by MSPA [134] • Coral Recognition by RG [208] Math Calculus (based on solving a Differential or Integral Equation) ...
... • Grey level co-occurrence matrices .. properties (contrast, correlation, energy, .. and homogeneity) [22], [24], [25], [134], [139], [204], [210], [215], [217] ...
The Internet of Underwater Things (IoUT) is an emerging communication ecosystem developed for connecting underwater objects in maritime and underwater environments. The IoUT technology is intricately linked with intelligent boats and ships, smart shores and oceans, automatic marine trans-portations, positioning and navigation, underwater exploration, disaster prediction and prevention, as well as with intelligent monitoring and security. The IoUT has an influence at various scales ranging from a small scientific observatory, to a mid-sized harbor, and to covering global oceanic trade. The network architecture of IoUT is intrinsically heterogeneous and should be sufficiently resilient to operate in harsh environments. This creates major challenges in terms of underwater communications, whilst relying on limited energy resources. Additionally, the volume, velocity, and variety of data produced by sensors, hydrophones, and cameras in IoUT is enormous, giving rise to the concept of Big Marine Data (BMD), which has its own processing challenges. Hence, conventional data processing techniques will falter, and bespoke Machine Learning (ML) solutions have to be employed for automatically learning the specific BMD behavior and features facilitating knowledge extraction and decision support. The motivation of this paper is to comprehensively survey the IoUT, BMD, and their synthesis. It also aims for exploring the nexus of BMD with ML. We set out from underwater data collection and then discuss the family of IoUT data communication techniques with an emphasis on the state-of-the-art research challenges. We then review the suite of ML solutions suitable for BMD handling and analytics. We treat the subject deductively from an educational perspective, critically appraising the material surveyed. Accordingly, the reader will become familiar with the pivotal issues of IoUT and BMD processing, whilst gaining an insight into the state-of-the-art applications, tools, and techniques. Finally, we analyze of the architectural challenges of the IoUT, followed by proposing a range of promising direction for research and innovation in the broad areas of IoUT and BMD. Our hope is to inspire researchers, engineers, data scientists, and governmental bodies to further progress the field, to develop new tools and techniques, as well as to make informed decisions and set regulations related to the maritime and underwater environments around the world.
... The Lightframe Onsight Key-species Investigation (LOKI) optical profiler (Schultz et al., 2010;Schmid et al., 2016) captures the fine-scale vertical distribution of mesozooplankton. The optical resolution of images is sufficient for either visual (Schulz et al., 2010;Hirche et al., 2014) or computer-assisted (Schmid et al., 2016(Schmid et al., , 2018 taxonomic identification of zooplankton organisms. ...
... The Lightframe Onsight Key-species Investigation (LOKI) optical profiler (Schultz et al., 2010;Schmid et al., 2016) captures the fine-scale vertical distribution of mesozooplankton. The optical resolution of images is sufficient for either visual (Schulz et al., 2010;Hirche et al., 2014) or computer-assisted (Schmid et al., 2016(Schmid et al., , 2018 taxonomic identification of zooplankton organisms. Computerassisted identification has the advantage that large quantities of plankton images (tens to hundreds of millions) can be identified efficiently (Luo et al., 2018) when human expert identification is no longer practical. ...
... In the present study, the LOKI profiler was deployed from 10 m above the seafloor to the surface at six stations in northern Baffin Bay to investigate the fine-scale vertical co-distribution (1 m resolution) of Calanus hyperboreus, C. glacialis and chlorophyll a at different hours of the day, at the end of the grazing season in August. Zooplankton images were identified automatically by a classifier developed using machine learning (Schmid et al., 2016). The observed distributions of Calanus in relation to its microalgal food at different light intensities were then compared to the predictions of the Predator Avoidance Hypothesis. ...
Studying the distribution of zooplankton in relation to their prey and predators is challenging, especially in situ. Recent developments in underwater imaging enable such fine-scale research. We deployed the Lightframe On-sight Keyspecies Investigation (LOKI) image profiler to study the fine-scale (1 m) vertical distribution of the copepods Calanus hyperboreus and C. glacialis in relation to the subsurface chlorophyll maximum (SCM) at the end of the grazing season in August in the North Water and Nares Strait (Canadian Arctic). The vertical distribution of both species was generally consistent with the predictions of the Predator Avoidance Hypothesis. In the absence of a significant SCM, both copepods remained at depth during the night. In the presence of a significant SCM, copepods remained at depth in daytime and a fraction of the population migrated in the SCM at night. All three profiles where the numerically dominant copepodite stages C4 and C5 of the two species grazed in the SCM at night presented the same intriguing pattern: the abundance of C. hyperboreus peaked in the core of the SCM while that of C. glacialis peaked just above and below the core SCM. These distributions of the same-stage congeners in the SCMs were significantly different. Lipid fullness of copepod individuals was significantly higher in C. hyperboreus in the core SCM than in C. glacialis above and below the core SCM. Foraging interference resulting in the exclusion from the core SCM of the smaller C. glacialis by the larger C. hyperboreus could explain this vertical partitioning of the actively grazing copepodite stages of the two species. Alternatively, specific preferences for microalgal and/or microzooplankton food hypothetically occupying different layers in the SCM could explain the observed partitioning. Investigating the observed fine-scale co-distributions further will enable researchers to better predict potential climate change effects on these important Arctic congeners.
... Underwater imaging systems that take photographs of individual zooplankton in situ are able to overcome the low vertical resolution of nets (Hirche et al., 2014;Luo et al., 2014), and are therefore useful for studying copepod lipids. This study uses the Lightframe On-sight Keyspecies Investigation (LOKI) system (Schulz et al., 2010;Schmid et al., 2016). Images were identified using a copepod stagespecific automatic zooplankton identification model (Schmid et al., 2016) based on machine learning. ...
... This study uses the Lightframe On-sight Keyspecies Investigation (LOKI) system (Schulz et al., 2010;Schmid et al., 2016). Images were identified using a copepod stagespecific automatic zooplankton identification model (Schmid et al., 2016) based on machine learning. The appearance of copepod lipid sacs was investigated in hundreds of LOKI images of individuals of three copepod species that dominate the biomass of mesozooplankton in the Arctic Ocean: the diapausing species C. hyperboreus and C. glacialis, and the non-diapausing Metridia longa. ...
... For photosynthetically active radiation (E PAR ), we used data collected by the ArcticNet main CTD (Biospherical QCP-2300). Detailed information on LOKI design and setup are presented in Schulz et al. (2010) and Schmid et al. (2016). ...
Copepod lipids fuel the Arctic marine ecosystem, but information on the fine-scale distribution of copepods and lipids is nonexistent. This study investigated the fine-scale (1 m) vertical distribution of the copepods Calanus hyperboreus, Calanus glacialis and Metridia longa during a Lagrangian drift in the North Water Polynya using the Lightframe On-sight Keyspecies Investigation (LOKI) imaging system. A copepod species- and stage-specific automatic identification model based on machine learning, a subcategory of artificial intelligence, was used to identify images taken by LOKI. Lipids were measured from images of copepods taken over the whole water column (1m resolution). Diel vertical migration (DVM) in all three species was detected. In C. hyperboreus and C. glacialis C4-females as well as M. longa C5-females lipid load of deep copepod individuals was significantly higher than that of shallower individuals. Vertical distribution profiles
and individual lipid loads suggested that individuals with lower lipid load continued DVM, while others with high lipid load ceased migrating, remaining at depth. Calanus hyperboreus individuals seemed to migrate to diapause at lower lipid fullness (50%) than C. glacialis (60%). A bioenergetics model showed that Calanus females had enough lipids to diapause for over a year, highlighting the significant lipid overhead they use for capital breeding.
KEYWORDS: copepod lipids; DVM; diapause; fine scale vertical distribution; underwater imaging; machine learning; automatic zooplankton identification model; North Water Polynya; Arctic Ocean
... The LOKI was deployed as part of a 5-Net Vertical Sampler (5NVS; Fig. 7). The 5NVS is a sampling setup on which up to 5 different nets can be deployed (Schmid 2016). During LOKI deployment, integrated sensors (Fig. 8) record environmental data. ...
... In the process, the outline of the zooplankter in the image is detected and parameters such as surface area, circularity, convexity and homogeneity measured (see The final selection of images was then quality controlled by using ADOBE Bridge and Photoshop. In the process, smears on the camera lens or light reflections on the borders of the Schmid et al. 2016). ...
... The resolution obtained with the LOKI has proved insufficient to discern C. glacialis C2 from C. hyperboreus C1 and C. glacialis C3 from C. hyperboreus C2 and therefore these taxa were pooled into two groups (Schmid et al. 2016). This leads currently to difficulties when interpreting the observed pattern of these mixed groups. ...
In the Arctic Ocean copepods transfer energy produced by autotrophs to higher trophic levels. Net tows have long been used as the main zooplankton sampling tool but can hardly indicate the abundance and distribution of these animals on a fine-scale level. The LOKI (Lightframe On-sight Keyspecies Investigation) is an in-situ optical underwater imaging device capable of recording the vertical distribution of zooplankton at a scale of ~30-60 cm. Using this device, the fine-scale distribution of the key herbivorous copepods Calanus glacialis and C. hyperboreus as well as that of the omnivore Metridia longa was analyzed. It was investigated how the fine-scale vertical distribution of copepods was coupled to fluorescence of microalgae during the phytoplankton bloom and how that pattern differed between eastern and western North Water Polynya. Eastern and western NOW showed substantial differences in the fluorescence of microalgae which seemed to influence the fine-scale copepod distribution. Whereas copepod abundance and fluorescence seemed almost vertically separated on the Canadian side of the polynya, the abundance of the same taxa showed a strong vertical coupling with fluorescence on the Greenland side. The results suggest a strong influence of physical parameters in the NOW (e.g. temperature and upwelling), likely governed by the West Greenland Current and Baffin Island Current, on microalgae and hence fine-scale zooplankton distribution.
... To address this issue, several studies have turned to automatic or semiautomatic identification of the organisms (e.g. Ashjian et al., 2005;Rolke & Lenz, 1984;Schmid et al., 2016). There are several programs for (semi-) automatic identification available, such as ImageJ (Schneider et al., 2012), the Video Annotation and Reference System (VARS) (Gomes-Pereira et al., 2016), the web based Ecotaxa (Picheral et al., 2017), and the recently developed MorphoCluster (Schröder et al., 2020). ...
... Comparison of taxon richness (A), Shannon diversity index (B), Pielou's evenness (C) andSimpson diversity index (D) between MultiNet (red) and Lightframe On-sight Key species Investigation (LOKI)(Schmid et al., 2016) (blue) along a Barents Sea to Arctic Ocean transect in summer 2017. Each dot represents a station. ...
Analysis of the biodiversity and abundance of key zooplankton species along a Barents Sea into Arctic Ocean transect, using in-situ imaging and semi-automatic identification.
... Next, we evaluate the performance of our algorithms on an image dataset extracted from the Woods Hole Oceanographic Institution (WHOI) plankton database 6 . Machine learning methods are becoming a popular way to characterize and classify plankton [7][8][9][10][11][12][13][14] . ...
... Geometric features include area, eccentricity, rectangularity and other morphological descriptors, that have been used to distinguish plankton by shape and size 17 . The invariant Hu 31 (7) and Zernike moments 32 (25) are widely used in shape representation, recognition and reconstruction. Texture based features encode the structural diversity of plankton. ...
The acquisition of increasingly large plankton digital image datasets requires automatic methods of recognition and classification. As data size and collection speed increases, manual annotation and database representation are often bottlenecks for utilization of machine learning algorithms for taxonomic classification of plankton species in field studies. In this paper we present a novel set of algorithms to perform accurate detection and classification of plankton species with minimal supervision. Our algorithms approach the performance of existing supervised machine learning algorithms when tested on a plankton dataset generated from a custom-built lensless digital device. Similar results are obtained on a larger image dataset obtained from the Woods Hole Oceanographic Institution. Additionally, we introduce a new algorithm to perform anomaly detection on unclassified samples. Here an anomaly is defined as a significant deviation from the established classification. Our algorithms are designed to provide a new way to monitor the environment with a class of rapid online intelligent detectors.
... For the semi-transparent Daphnia, the organs in the body were able to be clearly presented (Figure 14b). In this work, the tests were carried out in simulated conditions in the laboratory which has few noises comparing with in situ observation, such as the results from LOKI system [11]. In the further, our developed imaging system will be deployed in coastal area. ...
... Comparison with LOKI. (a) Images captured by LOKI[11]; (b) images captured by the proposed system. ...
A cost-effective and low-power-consumption underwater microscopic imaging system was developed to capture high-resolution zooplankton images in real-time. In this work, dark-field imaging was adopted to reduce backscattering and background noise. To produce an accurate illumination, a novel illumination optimization scheme for the light-emitting diode (LED) array was proposed and applied to design a lighting system for the underwater optical imaging of zooplankton. A multiple objective genetic algorithm was utilized to find the best location of the LED array, which resulted in the specific illumination level and most homogeneous irradiance in the target area. The zooplankton imaging system developed with the optimal configuration of LEDs was tested with Daphnia magna under laboratory conditions. The maximal field of view was 16 mm × 13 mm and the optical resolution was 15 μm. The experimental results showed that the imaging system developed could capture high-resolution and high-definition images of Daphnia. Subsequently, Daphnia individuals were accurately segmented and their geometrical characters were measured by using a classical image processing algorithm. This work provides a cost-effective zooplankton measuring system based on an optimization illumination configuration of an LED array, which has a great potential for minimizing the investment and operating costs associated with long-term in situ monitoring of the physiological state and population conditions of zooplankton.
... Although MSPA is typically applied in landscape connectivity assessments (e.g. Clerici and Vogt, 2013;Carlier and Moran, 2019), its capacity to automatically identify focal species has been recently demonstrated for zooplankton (Schmid et al., 2016). Its potential has not yet been explored in the context of individual, vascular plant species. ...
... tiff format) into an MSPA-compliant image dataset, and ii. Perform batch MSPA using the Guidos Toolbox 'batch' feature; � The potential to apply machine learning processes using MSPA (previously demonstrated using an automatic zooplankton identification model in Schmid et al., 2016) presents similar opportunities for automated plant quantification models; � The use of Otsu's method (Otsu, 1979) as an objective method of automatic optimal thresholding selection may also improve and speed up monochrome image preparation in GIS; � The option to divide 'Core' into user-selected size sub-classes of small/medium/large in MSPA could provide the ability to differentiate stages of leaf-growth or indeed differentiate species based on mature leaf size; � Although the focus of this study was to quantify the cover of a distinctly ovate leaf, MPSA could be used to estimate cover values of other species with lobed leaf morphologies (e.g. Quercus spp.) by examining the 'Edge' to 'Core' ratio of MSPA object classes within an image; � It is possible to quantify linear vegetation such as Poaceae spp. ...
The most commonly used method for measuring vegetation cover is visual estimation, which is highly subjective, potentially leading to measurement errors. This poses serious implications to the assessment and continued management of plant species cover, for example in the control of invasive plant species. Morphological analysis of digital imagery has, to date, been primarily applied in the classification of landscape features. Our novel application of morphological image analysis provides an objective method for detection and accurate cover assessment of an invasive alien plant species (IAS), giving reduced measurement errors when compared to visual estimation. Importantly, this method is entirely based on free software. Guidos Toolbox is a collection of generic raster image processing routines, including Morphological Spatial Pattern Analysis (MSPA), which classifies and quantifies features according to shape. MSPA was employed in this study to detect and quantify cover of invasive Petasites pyrenaicus (Winter heliotrope) in digital images of 1 m � 1 m plots. Its efficacy was compared to that of two other methods-GIS Digitisation (used as an accurate baseline) and Visual Estimation (standard method). We tested the limit of MSPA usability on images of varying complexity, i.e. "simple", intermediate" or "complex", depending on presence/absence of other vascular plant species and the species richness of plot. Our results show good agreement between all three methods. MSPA measurement of P. pyrenaicus cover was most closely aligned with the GIS Digitisation (concordance correlation coefficients of 0.966). Visual Estimation was less closely aligned with GIS Digitisation (concordance correlation coefficients of 0.888). However, image complexity resulted in differing levels of agreement; with the closest agreement being achieved between MSPA and GIS Digitisation when used on images of lower and higher complexity. MSPA consistently provides higher accuracy and precision for P. pyrenaicus cover measurement than the standard Visual Estimation method. Our methodology is applicable to a range of focal vegetation species, both herbaceous and graminoid. Future application of MSPA for larger-scale surveying and monitoring via remote sensing is discussed, potentially reducing resource demands and increasing cover measurement consistency and accuracy. We recommend this method forms part of vegetation management toolkits for not only environmental managers, but for anyone concerned with plant cover assessment, from agricultural systems to sustainable resource use.
... Despite the advantages of in-situ imaging systems, their usage remains limited. Imaging gear is expensive relative to traditional plankton nets, and they often collect vast amounts of data (gigabyte to 10's of terabytes per cruise, translating into millions to billions of plankton images), which either have to be analysed and classified manually 29,39 , or automated using machine and deep learning [40][41][42] . Only recently have algorithmic approaches become sufficiently accurate, and graphics processing units (GPUs) powerful enough, to tackle this task. ...
... Organismal data derived from In-situ Ichthyoplankton Imaging System (ISIIS) imagery (numbers 1-34, all measured in ind. m −3 ), variables derived from an Acoustic Doppler Current Profiler (ADCP; numbers[35][36][37][38], and in-situ environmental data from ISIIS sensors (numbers[39][40][41]. *only used in larval fish models. ...
Eddies can enhance primary as well as secondary production, creating a diverse meso- and sub-mesoscale seascape at the eddy front which can affect the aggregation of plankton and particles. Due to the coarse resolution provided by sampling with plankton nets, our knowledge of plankton distributions at these edges is limited. We used a towed, undulating underwater imaging system to investigate the physical and biological drivers of zoo- and ichthyoplankton aggregations at the edge of a decaying mesoscale eddy (ME) in the Straits of Florida. Using a sparse Convolutional Neural Network we identified 132 million images of plankton. Larval fish and Oithona spp. copepod concentrations were significantly higher in the eddy water mass, compared to the Florida Current water mass, only four days before the ME's dissipation. Larval fish and Oithona distributions were tightly coupled, indicating potential predator-prey interactions. Larval fishes are known predators of Oithona, however, Random Forests models showed that Oithona spp. and larval fish concentrations were primarily driven by variables signifying the physical footprint of the ME, such as current speed and direction. These results suggest that eddy-related advection leads to largely passive overlap between predator and prey, a positive, energy-efficient outcome for predators at the expense of prey.
... maxillipeds, antennae) of different type and evolution stages in zooplankton image data. The results were then ingested into a machine-learning algorithm for automatic identification of key zooplankton taxa in the Canadian Arctic (Schmid, Aubry, Grigor, & Fortier, 2016). ...
... For example, the MSPA algorithm has already been applied in a wide variety of application fields. Figure 2 shows how MSPA can detect a structural connecting pathway in a maze puzzle, or helps at microscale in an automatic zooplankton identification model (Schmid et al., 2016). In addition to structural patterns, the same methodology describes functional movement patterns of squirrels . ...
The increased availability of mapped environmental data calls for better tools to analyze the spatial characteristics and information contained in those maps. Publicly available, user-friendly and universal tools are needed to foster the interdisciplinary development and application of methodologies for the extraction of image object information properties contained in digital raster maps. That is the overarching goal of GuidosToolbox, which is a set of customized, thematically grouped raster image analysis methodologies provided in a graphical user interface and for all popular operating systems. The Toolbox contains a wide selection of dedicated algorithms and tools, which are complemented by batch-processing and pre- and post-processing routines, all designed to objectively describe and quantify various spatial properties of image objects in digital raster data. While first developed for the analysis of remote sensing data in environmental applications, the Toolbox now provides a generic framework that is applicable to image analysis at any scale and for any kind of digital raster data.
... Individual variability has long been recognized as a key property of plankton ecology since population dynamics and trophic interactions (that are of primary interest for marine ecologists) are emerging properties of individual characteristics and behaviors (Båmstedt, 1988). Modern experimental and in situ observation methods are providing increasingly detailed and abundant individuallevel data (e.g., Schmid et al., 2016), while current numerical approaches allow for testing how and how well individual-based models can effectively represent emerging properties at higher organizational levels (Neuheimer et al., 2010;Morozov et al., 2013). ...
... The effect of turbulence may only become important for denser eggs rising very slowly toward the surface or directly after the spawning if it occurs in a thin layer pattern, hence producing a high concentration gradient. This effect could be studied more efficiently with new in situ sampling devices such as the LOKI underwater imaging system that can provide highly resolved vertical distribution of adult females C. hyperboreus, their eggs, a whole suite of potential other intraguild predators beyond M. longa and the physical properties of the water column as well (Schmid et al., 2016). From a biological point of view, egg density had an overwhelming impact on egg vertical distribution patterns (Figure 7). ...
Communities of large copepods form an essential hub of matter and energy fluxes in Arctic marine food webs. Intraguild predation on eggs and early larval stages occurs among the different species of those communities and it has been hypothesized to impact its structure and function. In order to better understand the interactions between dominant copepod species in the Arctic, we conducted laboratory experiments that quantified intraguild predation between the conspicuous and omnivorous Metridia longa and the dominant Calanus hyperboreus. We recorded individual egg ingestion rates for several conditions of temperature, egg concentration, and alternative food presence. In each of these experiments, at least some females ingested eggs but individual ingestion rates were highly variable. The global mean ingestion rate of M. longa on C. hyperboreus eggs was 5.8 eggs ind −1 d −1 , or an estimated 37% of M. longa daily metabolic need. Among the different factors tested and the various individual traits considered (prosome length, condition index), only the egg concentration had a significant and positive effect on ingestion rates. We further explored the potential ecological impacts of intraguild predation in a simple 1D numerical model of C. hyperboreus eggs vertical distribution in the Amundsen Gulf. Our modeling results showed an asymmetric relationship in that M. longa has little potential impact on the recruitment of C. hyperboreus (<3% egg standing stock removed by IGP at most) whereas the eggs intercepted by the former can account for a significant portion of its metabolic requirement during winter (up to a third).
... Finally, photographic systems have become the preferred method for the observation of in situ particles, most likely because the resulting images are similar to those obtained from traditional microscopic analyses (Giering et al. 2020). There are numerous commercially available or custom-made devices available from various oceanic research groups such as the Video Plankton Recorder (VPR) (Davis et al. 2005) or the Continuous Particle Imaging and Classification system (CPICS) (Giering et al. 2020) for colored images, and the Underwater Vision Profiler (UVP) (Picheral et al. 2010(Picheral et al. , 2022, the In Situ Ichthyoplankton Imaging System (ISIIS) (Giering et al. 2020), the Shadowed Image Particle Profiling and Evaluation (SIPPER) (Samson et al. 2004), the profiling underwater camera system KIELVISION (Taucher et al. 2018), ParCa system (Ratmeyer and Wefer 1996), or the Lightframe On-Sight Keyspecies Investigation (LOKI) (Schmid et al. 2016) for monochrome images. Their advantages mainly result in their deployment from ships or on autonomous platforms and can deliver large datasets covering spatially and temporally distribution of particles. ...
Marine particles are key to the cycling of major elements on Earth and play an important role in the balance of nutrients in the ocean. Three main categories of marine particles link the different parts of the open ocean by shaping carbon distribution: (i) Sinking; (ii) Suspended, and (iii) Ascending. Atmospheric carbon captured by phytoplankton in the surface water, is partly sequestered by sinking particles to the bottom of the ocean, having an important role in controlling global climate. Suspended particles represent a major substrate of organic carbon for heterotrophic microorganisms, being more likely to get remineralized. Ascending particles, depending on their content, point of origin, and ascending velocity, may lead to carbon remineralization in the upper layers of the ocean in closer proximity to the atmosphere. Marine particles are hotspots of microbial activity and thus heavily colonized by microorganisms whose dynamics play an important role in organic matter degradation, aggregation and sinking, thus, directly influencing the biological carbon pump efficiency. Microbiomes of marine particles differ depending on particle size, source, and age. Nevertheless, these factors are generally overlooked, and particles are mostly studied as "bulk" without considering the high heterogeneity between individual particles. This hinders our understanding of the carbon budget in the ocean and thus future predictions of climate change. In this review we discuss characteristics of known particle-types and associated sampling methods. We further identify gaps in knowledge and highlight the need to better understand the single particles ecosystem to improve upscaling rates to the global scale.
... Such systems acquire a huge amount of image data for which manual identification is impractical [7]. Hence, machine learning has nowadays become one of the most studied approaches for the characterization of plankton data [3,[8][9][10][11][12][13][14][15][16]. In particular, there has been a surge in interest towards models based on artificial neural networks (ANNs), due to their successes in big data problems and their high expressive power, specifically in the form of convolutional neural networks (CNNs) [17][18][19][20][21]. ...
... Such systems acquire a huge amount of image data for which manual identification is impractical [7]. Hence, machine learning has nowadays become one of the most studied approaches for the characterization of plankton data [3,[8][9][10][11][12][13][14][15][16]. In particular, there has been a surge in interest towards models based on artificial neural networks (ANNs), due to their successes in big data problems and their high expressive power, specifically in the form of convolutional neural networks (CNNs) [17][18][19][20][21]. ...
Monitoring plankton populations in situ is fundamental to preserve the aquatic ecosystem. Plankton microorganisms are in fact susceptible of minor environmental perturbations, that can reflect into consequent morphological and dynamical modifications. Nowadays, the availability of advanced automatic or semi-automatic acquisition systems has been allowing the production of an increasingly large amount of plankton image data. The adoption of machine learning algorithms to classify such data may be affected by the significant cost of manual annotation, due to both the huge quantity of acquired data and the numerosity of plankton species. To address these challenges, we propose an efficient unsupervised learning pipeline to provide accurate classification of plankton microorganisms. We build a set of image descriptors exploiting a two-step procedure. First, a Variational Autoencoder (VAE) is trained on features extracted by a pre-trained neural network. We then use the learnt latent space as image descriptor for clustering. We compare our method with state-of-the-art unsupervised approaches, where a set of pre-defined hand-crafted features is used for clustering of plankton images. The proposed pipeline outperforms the benchmark algorithms for all the plankton datasets included in our analysis, providing better image embedding properties.
... These techniques generate data comparable to those obtained by manual light microscopy, but in a high-throughput way. Still, some differences can be detected due to the differences in manual vs automatic classification, sample preservation vs in situ observations, and between sampled seawater volumes Jakobsen & Carstensen, 2011;Álvarez et al., 2014;Schmid et al., 2016;Haraguchi et al., 2017;Detmer et al., 2019;Hrycik et al., 2019;Kraft et al., 2021). ...
A major challenge in characterizing plankton communities is the collection, identification and quantification of samples in a time-efficient way. The classical manual microscopy counts are gradually being replaced by high throughput imaging and nucleic acid sequencing. DNA sequencing allows deep taxonomic resolution (including cryptic species) as well as high detection power (detecting rare species), while RNA provides insights on function and potential activity. However, these methods are affected by database limitations, PCR bias, and copy number variability across taxa. Recent developments in high-throughput imaging applied in situ or on collected samples (high-throughput microscopy, Underwater Vision Profiler, FlowCam, ZooScan, etc) has enabled a rapid enumeration of morphologically-distinguished plankton populations, estimates of biovolume/biomass, and provides additional valuable phenotypic information. Although machine learning classifiers generate encouraging results to classify marine plankton images in a time efficient way, there is still a need for large training datasets of manually annotated images. Here we provide workflow examples that couple nucleic acid sequencing with high-throughput imaging for a more complete and robust analysis of microbial communities. We also describe the publicly available and collaborative web application EcoTaxa, which offers tools for the rapid validation of plankton by specialists with the help of automatic recognition algorithms. Finally, we describe how the field is moving with citizen science programs, unmanned autonomous platforms with in situ sensors, and sequencing and digitalization of historical plankton samples.
... Advances in optical technology, combined with the increased desire to use non-lethal observation approaches, have driven the development of new sensors and instruments to document marine ecosystems (Bicknell et al., 2016), but these instruments generally require an external light source. For example, Optical probes such as the Underwater Vision Profiler (Picheral et al., 2010), the Laser-Optical Plankton Counter (Basedow et al., 2013;Herman & Harvey, 2006), the Video Plankton Recorder (Sainmont et al., 2014), and the Light frame On-sight Key species Investigation system (Schmid et al., 2016;Schulz et al., 2010) all use light sources and optical sensors to assess the vertical distribution and abundance of zooplankton. Researchers and the industry alike increasingly use high-definition (HD) video cameras or stereo-cameras mounted on trawls to document the catchability of different species or size classes of fish (Boldt et al., 2018;Underwood et al., 2020;Williams et al., 2010). ...
The globally widespread adoption of Artificial Light at Night (ALAN) began in the mid‐20th century. Yet, it is only in the last decade that a renewed research focus has emerged into its impacts on ecological and biological processes in the marine environment that are guided by natural intensities, moon phase, natural light and dark cycles and daily light spectra alterations. The field has diversified rapidly from one restricted to impacts on a handful of vertebrates, to one in which impacts have been quantified across a broad array of marine and coastal habitats and species. Here we review the current understanding of ALAN impacts in diverse marine ecosystems. The review presents the current state of knowledge across key marine and coastal ecosystems (sandy and rocky shores, coral reefs and pelagic) and taxa (birds and sea turtles), introducing how ALAN can mask seabirds and sea turtles navigation, cause changes in animals predation patterns and failure of coral spawning synchronization, as well as inhibition of zooplankton Diel Vertical Migration. Mitigation measures are recommended, however, while strategies for mitigation were easily identified, barriers to implementation are poorly understood. Finally, we point out knowledge gaps that if addressed would aid in the prediction and mitigation of ALAN impacts in the marine realm.
... Particle camera systems (Honjo et al., 1984;Asper, 1987;Ratmeyer and Wefer, 1996;Gorsky et al., 2000) have significantly improved during the past three decades. It is now possible to develop higher resolution camera systems capable of imaging smaller aggregates than previously possible (Picheral et al., 2010;Schmid et al., 2016;Giering et al., 2020). The advantage of particle cameras lies in their non-destructive nature of acquiring particle size-distribution and abundance, in situ. ...
Most studies on the potential impacts of deep-sea mining in the Clarion Clipperton Zone (CCZ) have largely focused on benthic ecosystems but ignore the pelagic environment. To model full-scale impacts, it is important to understand how sediment discharge might affect the pelagic zone as well. This study combines in situ optics, hydrography, and remote sensing to describe particle abundance and size distribution through the entire water column in the CCZ (German sector). CCZ surface waters were characterized as productive over the year. During the winter, we observed the formation of a sharp transition zone in Chla concentration, identifying the area as a productive transitional zone toward a more depleted ocean gyre. In the German sector, median particle size was small (± 77 μm), and large particles (>300 μm) were rare. By assessing particle flux attenuation, we could show that the presence of a thick oxygen minimum zone (OMZ) plays an essential role in export and transformation of settling aggregates, with strong diel variations. We suggest that the combination of small aggregate size, bottom currents and slow seafloor consolidation may explain the extremely low sedimentation rate in the CCZ. We conclude that sediment incorporations and ballasting effect on settling particulate matter represent the most significant hazard on midwater and benthic ecosystems.
... Particle camera systems (Honjo et al., 1984;Asper, 1987;Ratmeyer and Wefer, 1996;Gorsky et al., 2000) have significantly improved during the past three decades. It is now possible to develop higher resolution camera systems capable of imaging smaller aggregates than previously possible (Picheral et al., 2010;Schmid et al., 2016;Giering et al., 2020). The advantage of particle cameras lies in their non-destructive nature of acquiring particle size-distribution and abundance, in situ. ...
Most studies on the potential impacts of deep-sea mining in the Clarion Clipperton Zone (CCZ) have largely focused on benthic ecosystems but ignore the pelagic environment. To model full-scale impacts, it is important to understand how sediment discharge might affect the pelagic zone as well. This study combines in situ optics, hydrography, and remote sensing to describe particle abundance and size distribution through the entire water column in the CCZ (German sector). CCZ surface waters were characterized as productive over the year. During the winter, we observed the formation of a sharp transition zone in Chla concentration, identifying the area as a productive transitional zone toward a more depleted ocean gyre. In the German sector, median particle size was small (± 77 µm), and large particles (>300 µm) were rare. By assessing particle flux attenuation, we could show that the presence of a thick oxygen minimum zone (OMZ) plays an essential role in export and transformation of settling aggregates, with strong diel variations. We suggest that the combination of small aggregate size, bottom currents and slow seafloor consolidation may explain the extremely low sedimentation rate in the CCZ. We conclude that sediment incorporations and ballasting effect on settling particulate matter represent the most significant hazard on midwater and benthic ecosystems.
... Therefore, there is a great need to observe and monitor the plankton. Compared with the traditional observation method of manual net sampling and light microscopy examination, the rapid development of in situ optical imaging technology and artificial intelligence has greatly improved the tempo-spatial scales and the efficiency of plankton observation [3][4][5][6][7]. According to the variations in illumination, in situ plankton imager can be simply classified into the brightfield and the darkfield categories. ...
The application of in situ imaging technology has greatly promoted our understanding of marine plankton ecology. However, the out-of-focus blur, which is the main cause of in situ plankton image degradation, leads not only to subsequent recognition difficulty for both human and machine, but also to inaccurate measurement of seawater volume and hence plankton abundance quantification. Therefore, it is necessary to develop methods to evaluate the blur, reserve those high-quality in-focus
images for subsequent analysis, and remove blurred ones to cater for limited computation, storage and transmission bandwidth resources in a field deployed system. In this work, we report two algorithms exclusively developed for focusing evaluation of in situ darkfield plankton images that are based on handcrafted and automatic convolution neural network feature extraction, respectively. They complement each other in computational efficiency and evaluation performance, and are expected to be deployed in darkfield imaging systems under different field scenarios with different powering, computing, storage and networking conditions.
... The above mentioned approaches to automated identification of species or individuals are also being scaled to ecosystem scale and applied to diversity assessment, conservation, and resource management (Christin et al., 2019). Examples using techniques detailed in the previous section include detecting and estimat-ing abundance of zooplankton (Schmid et al., 2016) and detecting and counting sea turtles and whales using drone and satellite imagery (Gray et al., 2019;Guirado et al., 2019). Other uses combine digital imagery with LiDAR and other remote sensing or geospatial data for mapping of vegetation (Guo et al., 2020;Li et al., 2020b,c;Kislov and Korznikov, 2020;Korznikov et al., 2021), forest carbon stock (Asner et al., 2018), and the footprint of fishing across the world's oceans (Kroodsma et al., 2018). ...
Deep learning is driving recent advances behind many everyday technologies, including those relying on speech and image recognition, natural language processing, and autonomous driving. It is also gaining popularity in biology, where it has been used for automated species identification, environmental monitoring, behavioral studies, DNA sequencing, and population genetics and phylogenetics, among other applications. Deep learning relies on artificial neural networks for predictive modeling and excels at recognizing complex patterns. Operating within the machine learning paradigm, deep learning can be viewed as an alternative to likelihood-based inference methods. It has desirable properties, including good performance and scaling with increasing complexity, while posing unique challenges such as sensitivity to bias in input data. In this review we provide a gentle introduction to deep learning, review its applications in ecology and evolution, and discuss its limitations and efforts to overcome them. We also provide a practical primer for biologists interested in including deep learning in their toolkit and identify its possible future applications.
... To achieve this objective, we sought to quantify individual traits of copepods. A powerful way to capture such traits is by analyzing in situ images Schmid et al. 2016). Modern devices can image thousands of individuals in their immediate environment while simultaneously measuring physical and biological variables like temperature or fluorescence. ...
Imaging techniques are increasingly used in ecology studies, producing vast quantities of data. Inferring functional
traits from individual images can provide original insights on ecosystem processes. Morphological traits are,
as other functional traits, individual characteristics influencing an organism’s fitness. We measured them from in
situ image data to study an Arctic zooplankton community during sea ice break-up. Morphological descriptors
(e.g., area, lightness, complexity) were automatically measured on ~28,000 individual copepod images from a
high-resolution underwater camera deployed at more than 150 sampling sites across the ice-edge. A statisticallydefined
morphological space allowed synthesizing morphological information into interpretable and continuous
traits (size, opacity, and appendages visibility). This novel approach provides theoretical and methodological
advantages because it gives access to both inter- and intra-specific variability by automatically analyzing a large
dataset of individual images. The spatial distribution of morphological traits revealed that large copepods are associated
with ice-covered waters, while open waters host smaller individuals. In those ice-free waters, copepods also
seem to feed more actively, as suggested by the increased visibility of their appendages. These traits distributions
are likely explained by bottom-up control: high phytoplankton concentrations in the well-lit open waters encourages
individuals to actively feed and stimulates the development of small copepod stages. Furthermore, copepods
located at the ice edge were opaquer, presumably because of full guts or an increase in red pigmentation. Our morphological
trait-based approach revealed ecological patterns that would have been inaccessible otherwise, including
color and posture variations of copepods associated with ice-edge environments in Arctic ecosystems.
... Color correction is a method for estimating quality [17,47,[112][113][114], while other methods ruining information have little destruction and they are not very accurate. g) Classification and clustering Although learning in classification is preferred to clustering, but clustering is used in segmentation and even image quality improvement [115,116]. Deep learning theory (also known as hierarchical learning) is part of a broader category of machine learning methods based on learning data representations, as opposed to task-specific algorithms. This process is modeled using a deep graph with several processed layers, including linear and nonlinear converting layers [89,[117][118][119]. ...
In recent years, deep sea and ocean explorations have attracted more attention in the marine industry. Most of the marine vehicles, including robots, submarines, and ships, would be equipped with automatic imaging of deep sea layers. There is a reason which the quality of the images taken by the underwater devices is not optimal due to water properties and impurities. Consequently, water absorbs a series of colors, so processing gets more difficult. Scattering and absorption are related to underwater imaging light and are called light attenuation in water. The examination has previously shown that the emergence of some inherent limitations is due to the presence of artifacts and environmental noise in underwater images. As a result, it is hard to distinguish objects from their backgrounds in those images in a real-time system. This paper discusses the effect of the software and hardware parts for the underwater image, surveys the state-of-art different strategies and algorithms in underwater image enhancement, and measures the algorithm performance from various aspects. We also consider the important conducted studies on the field of quality enhancement in underwater images. We have analyzed the methods from five perspectives: (a) hardware and software tools, (b) a variety of underwater imaging techniques, (c) improving real-time image quality, (d) identifying specific objectives in underwater imaging, and (e) assessments. Finally, the advantages and disadvantages of the presented real/non-real-time image processing techniques are addressed to improve the quality of the underwater images. This systematic review provides an overview of the major underwater image algorithms and real/non-real-time processing.
... Finer-scale data can be collected on the scale of the individual through to sub-mesoscales (Luo et al. 2014, Schmid et al. 2018) using underwater imaging that can resolve not only the distributions of larval fishes, but also that of their patchy prey and predator fields. Several imaging systems are in existence today (e.g., Video Plankton Recorder (VPR), Davis et al. 2005; In Situ Ichthyoplankton Imaging System (ISIIS), Cowen and Guigand, 2008; Underwater Vision Profiler 5 (UVP5), Picheral et al. 2010; Lightframe On-sight Keyspecies Investigation (LOKI), Schmid et al. 2016), but their early usage has been limited, in part due to system costs and image analysis challenges associated with millions to billions of images. ...
The Northern California Current (NCC) is a complex, dynamic system experiencing distinctly different levels of upwelling and downwelling, ranging from intermittent upwelling in summer to downwelling in winter. In recent years, warm water anomalies along the Oregon coast have had significant effects on coastal plankton assemblages. To resolve some of the fine-scale responses to these conditions, we used a towed, undulating underwater imaging system to investigate fine-scale (1 m vertical) zoo-and ichthyoplankton distributions along a 57-km section parallel to the Newport Hydrographic Line encompassing the shelf, shelf break, and slope off the central coast of Oregon. A sparse Convolutional Neural Network was used to automate the identification of 52 million plankton images of 64 plankton taxa, ranging from protists to copepods, larval fishes, and gelatinous organisms. Taxa distributions were interpolated over the whole transect, providing unprecedented insight into their horizontal and vertical distributions and revealing seven broad patterns of distribution. Additional fine-scale distribution data enable examination of some of the physical and biological processes underlying these fine-scale distribution patterns, building upon the historical time series data that exists for this region and advancing our knowledge of planktonic processes in this productive region of the NCC.
... Illumination techniques are manifold with large differences in light source and direction. Particles can be illuminated from the front, one side (P-Cam; Lampitt and Iversen, unpublished), two sides (UVP; Picheral et al., 2010), all sides (LOKI; Schmid et al., 2016), or from the back (VPR, ISIIS; Davis et al., 2005). The light sources vary from simple scattered light (P-Cam) to laser sheets (SIPPER; Samson et al., 2004) and collimated LED beams (UVP, ISIIS; Table 1). ...
Optical particle measurements are emerging as an important technique for understanding the ocean carbon cycle, including contributions to estimates of their downward flux, which sequesters carbon dioxide (CO2) in the deep sea. Optical instruments can be used from ships or installed on autonomous platforms, delivering much greater spatial and temporal coverage of particles in the mesopelagic zone of the ocean than traditional techniques, such as sediment traps. Technologies to image particles have advanced greatly over the last two decades, but the quantitative translation of these immense datasets into biogeochemical properties remains a challenge. In particular, advances are needed to enable the optimal translation of imaged objects into carbon content and sinking velocities. In addition, different devices often measure different optical properties, leading to difficulties in comparing results. Here we provide a practical overview of the challenges and potential of using these instruments, as a step toward improvement and expansion of their applications.
... Machine learning methods are becoming a popular way to characterize and classify plankton [7]-67 [14]. A recent paper [15] CC-BY-NC 4.0 International license It is made available under a (which was not peer-reviewed) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. ...
The acquisition of increasingly large plankton digital image datasets requires automatic methods of recognition and classification. As data size and collection speed increases, manual annotation and database representation are often bottlenecks for utilization of machine learning algorithms for taxonomic classification of plankton species in field studies. In this paper we present a novel set of algorithms to perform accurate detection and classification of plankton species with minimal supervision. Our algorithms approach the performance of existing supervised machine learning algorithms when tested on a plankton dataset generated from a custom-built lensless digital device. Similar results are obtained on a larger image dataset obtained from the Woods Hole Oceanographic Institution. Our algorithms are designed to provide a new way to monitor the environment with a class of rapid online intelligent detectors.
Author Summary
Plankton are at the bottom of the aquatic food chain and marine phytoplankton are estimated to be responsible for over 50% of all global primary production [1] and play a fundamental role in climate regulation. Thus, changes in plankton ecology may have a profound impact on global climate, as well as deep social and economic consequences. It seems therefore paramount to collect and analyze real time plankton data to understand the relationship between the health of plankton and the health of the environment they live in. In this paper, we present a novel set of algorithms to perform accurate detection and classification of plankton species with minimal supervision. The proposed pipeline is designed to provide a new way to monitor the environment with a class of rapid online intelligent detectors.
... The other challenge posed by turbid waters is data processing; countless particles are imaged, and manual annotation of these images becomes a near-impossible task. Automated processing is being tested by some major research groups, and thus this major roadblock is diminishing (Benfield et al. 2007;Sosik and Olson 2007;Schmid et al. 2016;Orenstein and Beijbom 2017;Robinson et al. 2017;Luo et al. 2018). Very recently two new in situ imaging sensors became available, the Zoocam (Ohman et al. 2019), which is attached to the Zooglider, and the Continuous Particle Image Classification System (CPIC; www.coastaloceanvision.com), which can be mounted on a CTD frame. ...
Understanding plankton dynamics in marine ecosystems has been advanced using in situ molecular and imaging instrumentation. A range of research objectives have been addressed through high‐resolution autonomous sampling, from food web characterization to harmful algal bloom dynamics. When used together, molecular and imaging sensors can cover the full‐size range, genetic identity, and life stages of plankton. Here, we briefly review a selection of in situ instrumentation developed for the collection of molecular and imaging information on plankton communities in marine ecosystems. In addition, we interviewed a selection of instrumentation developers to determine if the transfer of sensor technology from marine to freshwater ecosystems is feasible and to describe the process of creating in situ sensors. Finally, we discuss the status of in situ molecular and imaging sensors in freshwater ecosystems and how some of the reviewed sensors could be used to address basic and applied research questions.
... Particularly Morphological Spatial Pattern Analysis (MSPA) and Network Connectivity Analysis (NCA) can help researchers to build hypothesis about this issue. The generic setup of MSPA has been used to identify and map forest patterns, both structural (Vogt et al., 2006) and functional (Vogt et al., 2007), to identify key connectors for habitat suitability (Saura et al., 2011), for riparian corridor conservation studies (Clerici & Vogt, 2013), or evaluation of the US green infrastructure (Wickham et al., 2010) up to classifying zooplankton ( Schmid et al., 2016). A unique feature of MSPA is the automatic detection of connecting pathways between core areas of image objects. ...
Background
Despite its wide distribution worldwide, only 4.6% of temperate grasslands are included within systems of protected areas. In Argentina, this situation is even more alarming: only 1.05% is protected. The study area (central area of the southern Salado River basin) has a large extent of grasslands of Paspalum quadrifarium (Pq) which has been target since the middle of the last century of a variety of agricultural management practices including fire burning for cattle grazing.
Methods
Five binary images of presence-absence data of Pq from a 42-year range (1974–2016) derived from a land cover change study were used as base data. Morphological Spatial Pattern Analysis (MSPA), Morphological Change Detection (MCD) and Network Connectivity Analysis (NCA) were performed to the data using Guidos Toolbox (GTB) for the estimation of habitat and connectivity dynamics of the Pq patches (fragments).
Results
A loss of the coverage area and habitat nuclei of this grassland was observed during the study period, with some temporal oscillation but no recovery to initial states. Additional drastic reduction in connectivity was also evident in resulting maps. The number of large Pq grassland fragments (>50 ha) decreased at beginning of the study period. Also, fragmentation measured as number of components (patches) was higher at the end of the study period. The Pq pajonal nuclei had their minimum representativeness in 2000, and recovered slightly in area in 2011, but with a significant percentage increase of smaller patches (=islets) and linear elements as bridges and branches. Large corridors (mainly edge of roads) could be observed at the end of the study period, while the total connectivity of the landscape pattern drops continuously. Statistics of links shows mean values decreasing from 1974 to 2016. On the other hand, maximum values of links decreased up to 19% in 2011, and recovered to a 54% of their original value in 2016.
Discussion
Pq fragmentation and habitat reduction could have an impact on the ecosystem functioning and the mobility of some species of native fauna. The connecting elements of the landscape were maintained and/or recovered in percentage in 2011 and 2016. This fact, although favoring the dispersion of the present diversity in the habitat nuclei could cause degradation by an edge effect. Part of the area has the potential to be taken as an area of research and as an example of livestock management, since it is the one that would most preserve the biodiversity of the Pq environment. On the methodological side, the use of a proved tool as GTB is useful for monitoring dynamics of a grassland-habitat fragmentation.
... The most promising methods use optical and imaging tools (Schmid, Aubry, Grigor, & Fortier, 2016), possibly in combination with biochemical or genetic analysis (e.g. Wagner, Durbin, & Buckley, 1998). ...
... Adapted fromHopcroft et al. (2008). Photographs of living holo-zooplankton captured in-situ by a zooplankton imager (Lightframe On-sight Key Species Investigation System;Schmid et al. 2016), during deployments in the Canadian Arctic (summer 2014). Typical physiology of a chaetognath belonging to the family Sagittidae. ...
Chapter 1: General Introduction
Chapter 2: Polar night ecology of a pelagic predator, the chaetognath Parasagitta elegans (published in Polar Biology)
Chapter 3: Growth and reproduction of the chaetognaths Eukrohnia hamata and Parasagitta elegans in the Canadian Arctic Ocean: capital breeding versus income breeding (published in Journal of Plankton Research)
Chapter 4: Feeding strategies of arctic chaetognaths: are they really “tigers of the plankton”? (undergoing revision for publication - results should not be taken as final)
Chapter 5: General Conclusions
... While OPCs are the most common in-situ optical instruments, the field is developing rapidly and there are a range of other systems which deserve to be mentioned. In particular, camera and imaging systems such as ZooScan (Laboratory only; Grosjean et al., 2004), FlowCam (Laboratory only; Sieracki et al., 1998), Zooplankton Visualization system (ZOOVIS; Trevorrow et al., 2005), Video Plankton Recorder (VPR; Davis et al., 2005), Lightframe On-sight Keyspecies Investigation (LOKI; Schmid et al., 2016), and the In Situ Ichthyoplankton Imaging System (ISIIS; Cowen and Guigand, 2008) have become more widespread. Additionally, increased effort has been invested in the identification of zooplankton from images (Zooniverse, www.planktonportal.org). ...
Zooplankton are the intermediate trophic level between phytoplankton and fish, and are an important component of carbon and nutrient cycles, accounting for a large proportion of the energy transfer to pelagic fishes and the deep ocean. Given zooplankton's importance, models need to adequately represent zooplankton dynamics. A major obstacle, though, is the lack of model assessment. Here we try and stimulate the assessment of zooplankton in models by filling three gaps. The first is that many zooplankton observationalists are unfamiliar with the biogeochemical, ecosystem, size-based and individual-based models that have zooplankton functional groups, so we describe their primary uses and how each typically represents zooplankton. The second gap is that many modelers are unaware of the zooplankton data that are available, and are unaccustomed to the different zooplankton sampling systems, so we describe the main sampling platforms and discuss their strengths and weaknesses for model assessment. Filling these gaps in our understanding of models and observations provides the necessary context to address the last gap—a blueprint for model assessment of zooplankton. We detail two ways that zooplankton biomass/abundance observations can be used to assess models: data wrangling that transforms observations to be more similar to model output; and observation models that transform model outputs to be more like observations. We hope that this review will encourage greater assessment of zooplankton in models and ultimately improve the representation of their dynamics.
The pelagic ecosystem of the Arctic Ocean is threatened by severe changes such as the reduction in sea‐ice coverage and increased inflow of warmer Atlantic water. The latter is already altering the zooplankton community, highlighting the need for monitoring studies. It is therefore essential to accelerate the taxonomic identification to speed up sample analysis, and to expand the analysis to biomass and size assessments, providing data for modeling efforts. Our case study in Fram Strait illustrates that image‐based analyses with the ZooScan provide abundance data and taxonomic resolutions that are comparable to microscopic analyses and are suitable for zooplankton monitoring purposes in the Arctic. We also show that image analysis allows to differentiate developmental stages of the key species Calanus spp. and Metridia longa and, thus, to study their population dynamics. Our results emphasize that older preserved samples can be successfully reanalyzed with ZooScan. To explore the applicability of image parameters for calculating total mesozooplankton and Calanus spp. biomasses, we used (1) conversion factors (CFs) translating wet mass to dry mass (DM), and (2) length–mass (LM) relationships. For Calanus spp., the calculated biomass values yielded similar results as direct DM measurements. Total mesozooplankton biomass ranged between 1.6 and 15 (LM) or 2.4 and 21 (CF) g DM m−2, respectively, which corresponds to previous studies in Fram Strait. Ultimately, a normalized biomass size spectra analysis provides 1st insights into the mesozooplankton size structure at different depths, revealing steep slopes in the linear fit in communities influenced by Atlantic water inflow.
Monitoring living plankton status has greatly developed owing to machine vision technology. One of these developments is the automatic video acquisition system, which can capture high-resolution information and record considerable details of a scene. However, existing optical sensors cannot usually obtain a whole image, in which all details of plankton are fully clear, due to the influence of many factors, such as depth-of-field limitation of lenses, movement of plankton, and large difference in plankton scale. The captured video also needs to be filtered to eliminate the redundancy before using for the sparsity of the sample. This procedure is time-consuming and costly. Therefore, in this article, we develop an end-to-end plankton database collection system that can directly generate complete, clear plankton images from video. First, the regions of interest of the plankton are extracted. Then, the same plankton which appears in successive frames is identified to reconstruct the clearest morphological and structure features through their detailed information. Experimental results indicate that the proposed system can effectively compress the original data. The proposed fusion method also outperforms the state-of-the-art methods, especially for images with anisotropic blur. Furthermore, this system can monitor the abundance and distribution of marine plankton with the help of an embedded computing platform.
In situ observations of pelagic fish and zooplankton with optical instruments usually rely on external light sources. However, artificial light may attract or repulse marine organisms, which results in biased measurements. It is often assumed that most pelagic organisms do not perceive the red part of the visible spectrum and that red light can be used for underwater optical measurements of biological processes. Using hull-mounted echosounders above an acoustic probe or a baited video camera, each equipped with light sources of different colours (white, blue and red), we demonstrate that pelagic organisms in Arctic and temperate regions strongly avoid artificial light, including visible red light (575–700 nm), from instruments lowered in the water column. The density of organisms decreased by up to 99% when exposed to artificial light and the distance of avoidance varied from 23 to 94 m from the light source, depending on colours, irradiance levels and, possibly, species communities. We conclude that observations from optical and acoustic instruments, including baited cameras, using light sources with broad spectral composition in the 400–700 nm wavelengths do not capture the real state of the ecosystem and that they cannot be used alone for reliable abundance estimates or behavioural studies.
The paper presents an underwater holographic sensor to study marine particles—a miniDHC digital holographic camera, which may be used as part of a hydrobiological probe for accompanying (background) measurements. The results of field measurements of plankton are given and interpreted, their verification is performed. Errors of measurements and classification of plankton particles are estimated. MiniDHC allows measurement of the following set of background data, which is confirmed by field tests: plankton concentration, average size and size dispersion of individuals, particle size distribution, including on major taxa, as well as water turbidity and suspension statistics. Version of constructing measuring systems based on modern carriers of operational oceanography for the purpose of ecological diagnostics of the world ocean using autochthonous plankton are discussed. The results of field measurements of plankton using miniDHC as part of a hydrobiological probe are presented and interpreted, and their verification is carried out. The results of comparing the data on the concentration of individual taxa obtained using miniDHC with the data obtained by the traditional method using plankton catching with a net showed a difference of no more than 23%. The article also contains recommendations for expanding the potential of miniDHC, its purpose indicators, and improving metrological characteristics.
Faced with global change, maintaining biodiversity and the proper functioning of ecosystems requires the implementation of appropriate monitoring and management tools. The aim of this thesis work is to analyse the spatio-temporal variability of the different facets of biodiversity (species diversity, functional diversity and isotopic diversity) and to study their complementarity in order to describe more exhaustively the long-term evolution of coastal benthic communities in response to different anthropogenic pressures. To this end, two datasets of long-term monitoring programs on fine sands benthic communities in the English Channel were used: one in the eastern Bay of Seine and one in the Bay of Morlaix. The study of the temporal changes of these communities showed very contrasted temporal dynamics. A relative stability of the community in the eastern Bay of Seine opposes the high variability of the community of the Bay of Morlaix marked by two abrupt changes. Partially congruent results between species diversity and functional diversity were reported; if the traits approach describes explicitly the consequences of structural changes on the global functioning of the ecosystem, it is sensitive to the properties of the indices. Furthermore, their values vary according to whether the species are weighted by densities or biomass, and then to their distribution within communities. The use of isotopic diversity indices has been tested to offer another approach to assess the functional variability on benthic communities centred on food webs.
Aquatic ecologists face challenges in identifying the general rules of the functioning of ecosystems. A common framework, including freshwater, marine, benthic, and pelagic ecologists, is needed to bridge communication gaps and foster knowledge sharing. This framework should transcend local specificities and taxonomy in order to provide a common ground and shareable tools to address common scientific challenges. Here, we advocate the use of functional trait-based approaches (FTBAs) for aquatic ecologists and propose concrete paths to go forward. Firstly, we propose to unify existing definitions in FTBAs to adopt a common language. Secondly, we list the numerous databases referencing functional traits for aquatic organisms. Thirdly, we present a synthesis on traditional as well as recent promising methods for the study of aquatic functional traits, including imaging and genomics. Finally, we conclude with a highlight on scientific challenges and promising venues for which FTBAs should foster opportunities for future research. By offering practical tools, our framework provides a clear path forward to the adoption of trait-based approaches in aquatic ecology. © 2020 The Authors. Limnology and Oceanography published by Wiley Periodicals LLC. on behalf of Association for the Sciences of Limnology and Oceanography.
The rapid growth of machine learning (ML) has resulted in an almost overwhelmingly large number of modelling techniques, demanding better elucidation of their strengths and weaknesses in applied contexts. Tree-based methods such as Random Forests (RF) and Boosted Regression Trees (BRT) are powerful ML approaches that make no assumptions about the functional forms of the relationship with predictors, are flexible in handling missing data, and can easily capture complex, non-linear interactions. As with many ML methods, however, RF and BRT are potentially vulnerable to overfitting and a subsequent loss of generalizability.
The rise of in situ plankton imaging systems, particularly high‐volume imagers such as the In Situ Ichthyoplankton Imaging System, has increased the need for fast processing and accurate classification tools that can identify a high diversity of organisms and nonliving particles of biological origin. Previous methods for automated classification have yielded moderate results that either can resolve few groups at high accuracy or many groups at relatively low accuracy. However, with the advent of new deep learning tools such as convolutional neural networks (CNNs), the automated identification of plankton images can be vastly improved. Here, we describe an image processing procedure that includes preprocessing, segmentation, classification, and postprocessing for the accurate identification of 108 classes of plankton using spatially sparse CNNs. Following a filtering process to remove images with low classification scores, a fully random evaluation of the classification showed that average precision was 84% and recall was 40% for all groups. Reliably classifying rare biological classes was difficult, so after excluding the 12 rarest taxa, classification accuracy for the remaining biological groups became > 90%. This method provides proof of concept for the effectiveness of an automated classification scheme using deep‐learning methods, which can be applied to a range of plankton or biological imaging systems, with the eventual application in a variety of ecological monitoring and fisheries management contexts.
In Arctic seas, primary production and the availability of food for zooplankton are strongly pulsed over the short productive summer. We tested the hypothesis that Eukrohnia hamata and Parasagitta elegans, two similar and sympatric arctic chaetognaths, partition resources through different reproductive strategies. The two species had similar natural longevities of around 2 years. Eukrohnia hamata, which occurred at epi- and meso-pelagic depths, spawned two distinct broods in autumn and spring. Offspring production coincided with drops in the frequency of E. hamata with visible lipid reserves, characteristic of capital breeders. Growth was positive from April to January and negative in February and March. Growth and maturation were similar for the two broods. Storage reserves contained in an oil vacuole may allow E. hamata to reproduce and grow outside the short production season. Parasagitta elegans produced one brood in summer–autumn during peak production in near-surface waters, characteristic of income breeders. In winter, P. elegans co-inhabited meso-pelagic waters with E. hamata, where it neither grew nor reproduced. As the Arctic warms, the development of an autumn phytoplankton bloom could favour the summer–autumn brood of P. elegans.
(Journal of Plankton Research is attributed as the original place of publication: https://doi.org/10.1093/plankt/fbx045)
ZOOMIE is an image treatment tool developed to ensure optimal quality for images collected with the Lightframe On-sight Keyspecies Investigation (LOKI) System, an underwater zooplankton camera system. ZOOMIE does that by identifying cases where multiple pictures of the same specimen have been taken (hereafter referred to as double images), a phenomenon that frequently occurs when imaging plankton in a constrained volume during vertical deployments. The process of identifying double pictures can be carried out manually but is very time consuming. By applying ZOOMIE, the time needed to identify double images is substantially reduced. It is essential to account for double images when representative distributions of images are wanted. ZOOMIE can automatically filter thousands of images based on previously extracted image parameters (e.g. area, mean grey pixel value, kurtosis; here extracted using the LOKI browser software (Isitec GmbH; http://www.isitec.de/start.htm)). The filtering is based on a set of rules that compares the image parameters of multiple images in order to detect double images and exclude them. The set of rules can be changed easily in the ZOOMIE scripts so that researchers can easily adapt the thresholds for finding double images necessary for their LOKI settings. After running the actual script to find double images, other scripts can be executed to automatically transfer images flagged for exclusion to a new folder. Finally, the results can be visualized on an internal homepage, using the actual images which are linked to the database. Here we can validate the outcome of the processing and we can manually adapt the outcome through dragging and dropping of images to verify if any images were wrongly allocated to a double image group. Although ZOOMIE was developed for LOKI images and the exclusion of double images, ZOOMIE could easily be adapted to handle other tasks requiring the handling and comparison of large numbers of images. See https://zenodo.org/ for software, files and citation information: http://dx.doi.org/10.5281/zenodo.17928
The importance of small planktonic copepods and their roles in pelagic marine food webs. Zoological Studies 43(2): 255-266. Small planktonic marine copepods (< 1 mm in length) are the most abundant metazoans on Earth. Included are adults and copepodites of calanoid genera such as Paracalanus, Clausocalanus, and Acartia; cyclopoid genera such as Oithona, Oncaea, and Corycaeus; planktonic harpacticoids of the genus Microsetella; and nauplii of almost all copepod species. Despite the abundance of small copepods, they have historically been undersampled due to the use of nets with meshes >200-333 mum. Recent studies have shown, however, that when appropriate net meshes of 100 pm or less are used, small copepods vastly exceed the abundance and sometimes the biomass of larger ones. Failure to adequately account for small copepods may cause serious underestimations of zooplankton abundance and biomass, the copepod grazing impact on phytoplankton primary production, zooplankton-mediated fluxes of chemicals and materials, and trophic interactions in the sea. The feeding ecology of small copepods is less well-known than that of adults of larger copepod species, such as members of the genus Calanus. Further, most feeding information for small copepods is for coastal genera such as Acartia, rather than for offshore taxa. Although it is generally assumed that small copepods, including nauplii, feed primarily upon small-sized phytoplankton cells, most such information comes from rearing or feeding studies on limited laboratory diets. There have been few examinations of actual copepod feeding on mixed diets of natural phytoplankton and microzooplankton found in the sea, but some of those have produced surprises. For instance, some species of Oithona and Paracalanus and even nauplii of Arctic Calanus spp. may feed primarily as predators upon heterotrophic protists, rather than as grazers of phytoplankton. Also, nauplii of various tropical copepod species have been shown to feed upon bacterioplankton. Thus, numerous basic questions remain as to the feeding ecology and grazing/predation impact of small copepods in the sea. Despite limited knowledge of what small copepods eat, it is clear that many higher-trophic-level consumers eat them. Numerous studies have shown that copepod nauplii, Oithona spp., and other small copepods are important prey of fish larvae and other planktivores. Small copepods exhibit a variety of reproductive strategies to compensate for losses to their populations due to predation. These include having high fecundity and growth rates, when not limited by insufficient food; having high reproduction and growth rates at warmer temperatures; having limited motion and low respiration rates, allowing the investment of more energy in reproduction; and having extended longevity to maximize lifetime reproductive output. Thus, small copepods are important links in marine food webs, serving as major grazers of phytoplankton, as components of the microbial loop, and as prey for ichthyoplankton and other larger pelagic carnivores. Our present inadequate understanding of the true abundance, biomass, trophic ecology, and role of small copepods in biogenic fluxes precludes proper understanding of the ecology of the sea.
The large oscillations of abiotic factors in the Arctic is critical in structuring its marine biota and the biodiversity of its indigenous populations and communities. The seasonal light cycle is modified by the sea ice cover, creating a situation dominated by phytoplankton blooms, that follow the receding ice edge, and in leads as the sea ice opens during the Arctic summer (Sakshaug 1997, 2003; Hegseth 1998; Falk-Petersen et al. 2000a; Engelsen et al. 2002). Blooms of phytoplankton propagate through Arctic waters (Zenkevich 1963) and carbon fixed through photosynthesis is rapidly converted into large, specialised lipid (marine fat) stores by the herbivorous Calanus species (Lee 1975; Sargent and Henderson 1986). These high-energy lipids are then rapidly transferred upwards through the food chain in large amounts (Falk-Petersen et al. 1990). The increase in lipid level from 10-20% of dry mass in phytoplankton to 50 - 70% in herbivorous zooplankton is probably one of the most fundamental specialisations in polar bioproduction. The lipid - based energy flux is one of the primary reasons for the large stocks of fish and mammals in Arctic waters. The importance of the diatom => Calanus food chain in the Arctic pelagic food has been demonstrated by Falk-Petersen et al. (1986; 2002) and Scott et al. (2002). A wide spectrum of predators from zooplankton to fish and sea birds has also been analysed by using fatty acid trophic markers in Arctic waters. In all of these studies the Calanus C20 and C22 lipid trophic markers were strikingly dominant, demonstrating the importance of the Calanus species in the Arctic pelagic ecosystem (Falk-Petersen et al. 2001, 2002, 2004; Dahl et al. 2003. The population size spectrum and energy content of the key Calanus species, being potential prey for zooplankton-eating fish and sea birds, is therefore instrumental in structuring the biodiversity of Arctic ecosystems. We believe that understanding the climate variability is a key to understand the biology of Arctic animals and the biodiversity of Arctic systems. In this paper we discuss how different climate regimes in the Nordic Seas can influence Calanus- based Arctic food chains.
In situ imaging of small particles and biota in the water column is commonly used by scientists. Scrutinising such objects by imaging confronts system designers and researchers with a number of technical challenges. Here physical and technical aspects of optical subsea technologies are discussed, including parameters for the identifi cation of suitable systems tailored to the specific scientific question at hand. The chapter tries to 'illuminate' why arbitrary combinations may not necessarily be ultimate guarantors for the required performance, points out common pitfalls, and provides an introduction to the necessary fundamentals of optics and design.
At ice edges in the Canadian High Arctic, seabirds and marine mammals eat arctic cod (Boreogadus saida) and to a lesser extent, zooplankton (calanoid copepods and Parathemisto) and ice-associated amphipods. Cod eat ice-associated amphipods, other ice-associated taxa (harpacticoid and cyclopoid copepods), and zooplankton. Calanoid copepods, Parathemisto, and the ice-associated amphipods studied (Onisimus glacialis, Apherusa glacialis, Gammarus wilkitzkü) all eat primarily diatom algae characteristic of the under-ice flora. From this information, a food web at the ice edge is constructed.
Key words: trophic relationships, arctic, ice edges, seabirds, marine mammals, cod, epontic community, zooplankton
Diel vertical migration (DVM) is a common behavior adopted by zooplankton species. DVM is a prominent adaptation for avoiding visual predation during daylight hours and still being able to feed on surface phytoplankton blooms during night. Here, we report on a DVM study using a Video Plankton Recorder (VPR), a tool that allows mapping of vertical zooplankton distributions with a far greater spatial resolution than conventional zooplankton nets. The study took place over a full day–night cycle in Disko Bay, Greenland, during the peak of the phytoplankton spring bloom. The sampling revealed a large abundance of copepods performing DVM (up during night and down during day). Migration behavior was expressed differently among the abundant groups with either a strong DVM (euphausiids), an absence of DVM (i.e., permanently deep; ostracods) or a marked DVM, driven by strong surface avoidance during the day and more variable depth preferences at night (Calanus spp.). The precise individual depth position provided by the VPR allowed us to conclude that the escape from surface waters during daytime reduces feeding opportunities but also lowers the risk of predation (by reducing the light exposure) and thereby is likely to influence both state (hunger, weight and stage) and survival. The results suggest that the copepods select day and night time habitats with similar light levels (~10−9 μmol photon s−1 m−2). Furthermore, Calanus spp. displayed state-dependent behavior, with DVM most apparent for smaller individuals, and a deeper residence depth for the larger individuals.
Computer-automated image capture and analysis techniques are increasingly used for the study of plankton. Unfortunately, because most of the automatic image processing packages recognize only numbers of pixels and grey scale values, although a great many parameters are automatically measured on each image (area, perimeter, average level of grey, etc. . . .), these do not directly equate to variables that appear in the biological literature. The relationships between prosome lengths and computer-generated characters were derived from images acquired using the ZooScan system for two of the most common copepod families off south–eastern Japan: Calanoida and Poecilostomatoida. The equivalent spherical diameter (ESD), a character automatically measured by the ZooScan-ZooProcess system, was found to be highly correlated with prosome length in the two orders, and a single equation could be established (regression coeecient r2 = 0.943) for all Calanoida excepting those of the family Eucalanidae: PL (mm)0.080.023*ESD (pixels).
Recent receding of the ice pack allows more sunlight to penetrate into the Arctic Ocean, enhancing productivity of a single annual phytoplankton bloom. Increasing river runoff may however enhance the yet pronounced upper-ocean stratification and prevent any significant wind-driven vertical mixing and upward supply of nutrients, counteracting the additional light available to phytoplankton. Vertical mixing of the upper-ocean is the key-process that will determine the fate of marine Arctic ecosystems. Here, we reveal an unexpected consequence of the Arctic ice loss: regions are now developing a second bloom in the fall, which coincides with delayed freeze-up and increased exposure of the sea surface to wind stress. This implies that wind-driven vertical mixing during fall is indeed significant, at least enough to promote further primary production. The Arctic Ocean seems to be experiencing a fundamental shift from a polar to a temperate mode, which is likely to alter the marine ecosystem.
The incorrect notion that kurtosis somehow measures “peakedness” (flatness, pointiness, or modality) of a distribution is remarkably persistent, despite attempts by statisticians to set the record straight. This article puts the notion to rest once and for all. Kurtosis tells you virtually nothing about the shape of the peak—its only unambiguous interpretation is in terms of tail extremity, that is, either existing outliers (for the sample kurtosis) or propensity to produce outliers (for the kurtosis of a probability distribution). To clarify this point, relevant literature is reviewed, counterexample distributions are given, and it is shown that the proportion of the kurtosis that is determined by the central μ ± σ range is usually quite small.
Plankton counting and analysis is essential in ecological study, yet scant literature exists as to the reliability of those counts and the consistency of the experts who make the counts. To assess how variable expert taxonomic identifications are, a set of six archived mesozooplankton samples from a series of Longhurst Hardy Plankton Recorder net hauls were counted by expert zooplankton analysts located at six marine laboratories. Sample identifications were repeated on two separate days with over 700 target specimens counted and identified on each day across the samples. Twenty percent of the analysts returned counts that varied by more than 10%. Thirty-three percent of analysts exhibited low identification consistencies, returning Intraclass Correlation Coefficient scores of less than 0.80. Statistical analyses of these data suggest that over 83% of the observed categorical count variance can be attributed to inconsistencies within analysts. We suggest this is the root cause of variation in expert specimen labelling consistency.
The climate in the Arctic is changing faster than in mid-latitudes. This is shown by increased temperatures, loss of summer sea ice, earlier snow melt, impacts on ecosystems, and increased economic access. Arctic sea ice volume has decreased by 75 % since the 1980s. Long-lasting global anthropogenic forcing from CO2 has increased over the previous decades and is anticipated to increase over the next decades. Temperature increases in response to greenhouse gases are amplified in the Arctic through feedback processes associated with shifts in albedo, ocean and land heat storage, and near-surface longwave radiation fluxes. Thus for the next few decades out to 2040, continuing environmental changes in the Arctic are very likely, and the appropriate response is to plan for adaptation to these changes. For example, it is very likely that the Arctic Ocean will become nearly seasonally sea ice free before 2050 and possibly within a decade or two, which in turn will further increase Arctic temperatures, economic access, and ecological shifts. Mitigation becomes an important option to reduce potential Arctic impacts in the second half of the 21st century. Using the most recent set of climate model projections (CMIP5), multi-model mean temperature projections show an Arctic-wide end of century increase of +13 ° C in late fall and +5 ° C in late spring for a business as usual emission scenario (RCP8.5) in contrast to +7 ° C in late fall and + 3° C in late spring if civilization follows a mitigation scenario (RCP4.5). Such temperature increases demonstrate the heightened sensitivity of the Arctic to greenhouse gas forcing.
When a sample of individual organisms is classified taxonomically with the possibility of classification error, the taxonomic counts of the classified individuals are biased estimates of the true counts in the sample. This note describes a simple method for correcting for this bias based on the classification probabilities of the classifier. The method is illustrated using some data from the Video Plankton Recorder.
A real-time segmentation of images requires features which are fast to calculate and a segmentation procedure which can fastly classify
pixels or regions with respect to these features. Automatic surveillance of a honeybee comb requires the segmentation of homogeneous
image regions. Such regions correspond to visible brood cells while the regions which are covered by a crowd of bees appear non-
homogeneous. In this paper a novel and efficient way to calculate an existing homogeneity feature is presented. Second, it is extended
to a novel feature and empirical results are provided which show improvements over the basic homogeneity value. Third, a simple but
effective segmentation procedure is presented and segmentation results are provided.
A two-vessel exercise was conducted over the southern flank of Georges Bank during the onset of vernal stratification in May 1992. The Video Plankton Recorder (VPR), a towed video system, was used to map out the fine-scale distributions of zooplankton to a depth of 70 m along a trackline which described a regular grid (3.5 × 4.5 km) in Lagrangian space. A second vessel following a parallel course conducted Multiple Opening/Closing Net and Environmental Sensing System (MOCNESS) sampling during the last section of the grid, which provided an opportunity to compare data from the two systems. Both the VPR and the MOCNESS provided similar data on the taxonomic composition of the plankton which was numerically dominated by copepods (Calanus, Pseudocalanus, Oithona), pteropods (Limacina) and larvaceans (Oikopleura). The absence of rare (
Optical imaging samplers are becoming widely used in plankton ecology, but image analysis methods have lagged behind image acquisition rates. Automated methods for analysis and recognition of plankton images have been developed, which are capable of real-time processing of incoming image data into major taxonomic groups. The limited accuracy of these methods can require significant manual post-processing to correct the automatically generated results, in order to obtain accurate estimates of plankton abundance patterns. We present here a dual-classification method in which each plankton image is first identified using a shaped-based feature set and a neural network classifier, and then a second time using a texture-based feature set and a support vector machine classifier. The plankton image is considered to belong to a given taxon only if the 2 identifications agree; otherwise it is labeled as unknown, This dual-classification method greatly reduces the false positive rate, and thus gives better abundance estimation in regions of low relative abundance. A confusion matrix is computed from a set of training images in order to determine the detection and false positives rates. These rates are used to correct abundances estimated from the automatic classification results. Aside from the manual sorting required to generate the initial training set of images, this dual-classification method is fully automatic and does not require subsequent manual correction of automatically sorted images. The resulting abundances agree closely with those obtained using manually sorted results. A set of images from a Video Plankton Recorder was used to evaluate this method and compare it with previously reported single-classifier results for major taxa.
This paper synthesizes the trophic dynamics of a Canadian arctic marine ecosystem in so far as it is known, using new data on primary production, zooFlankton, the bivalve Mya truncata, and arctic cod (Boreogadus saida), as well as literature values for marine mammals and seabirds. The 98 000 km reglon has a high rate of primary production relative to other parts of arctic Canada. About 60 g Cm-' are fixed annually, of which approximately 90% is contributed by phytoplankton, 10% by ice algae, and 1% by kelp. Phytoplankton production is twofold higher along the south coast of Comwallis Island than elsewhere in Barrow Strait. Four copepod species, of which Pseudocalanus acuspes is the most important energeti- cally, graze about one-third of the phytoplankton production. Bivalves maintain high biomass but low energy flow, acting as sedimenting agents. Arctic cod is a major component, with 125 000 tonnes bein consumed by marine mammals and 23 000 tonnes by seabirds annually. Our hydro- acoustic estimate for mean arctic cod density, 0.0022 fish.m', is probably too low, partly because we have been unable to quantify dense aggrega- tions of schooling fish. The ecological efficiency of ringed seal is near maximum, with 5% of ringed seal ingestion going to bears and man as seal flesh. The data on total kill and prey consumption in whales and birds is incomplete because they migrate out of the Lancaster Sound region in win- ter. The food chain is very long, with bears occupying the fifth trophic level; this is reflected by high biomagnification factors for persistent lipophilic pollutants such as PCBs. There are major data gaps for some zooplankton and most of the benthos, as well as for winter populations and energetics. This trophic analysis is therefore incomplete and efficiencies for entire trophic levels cannot be calculated.
Zooplankton is a key element in aquatic food webs. Rapid mapping of
abundance, combined with information on taxonomic and size composition
is necessary to understand ecosystem dynamics. Classical sampling with
towed plankton nets does not allow resolving fine scale distributions
along hydrographic gradients (e.g. fronts and clines) although such
structures determine community assemblages and trophic interactions.
Furthermore, sample analysis is labor intensive and time consuming. To
overcome these shortcomings, Lightframe On-sight Keyspecies
Investigation (LOKI), a new imaging device, was developed for sensing
spatial variability of plankton distribution on scales below the 1~m
level. Here, we give a brief description of the LOKI system and
demonstrate its potential for taxonomic identification using images of
various zooplankton taxa collected in the south east Pacific.
When Victor Hensen deployed the first
true plankton1 net in 1887, he and his
colleagues were attempting to answer
three fundamental questions: What
planktonic organisms are present in
the ocean? How many of each type are
present? How does the plankton’s composition
change over time? Although
answering these questions has remained
a central goal of oceanographers, the
sophisticated tools available to enumerate
planktonic organisms today offer
capabilities that Hensen probably could
never have imagined.
ZooScan with ZooProcess and Plankton Identifier (PkID) software is an integrated analysis system for acquisition and classification of digital zooplankton images from preserved zooplankton samples. Zooplankton samples are digitized by the ZooScan and processed by ZooProcess and PkID in order to detect, enumerate, measure and classify the digitized objects. Here we present a semi-automatic approach that entails automated classification of images followed by manual validation, which allows rapid and accurate classification of zooplankton and abiotic objects. We demonstrate this approach with a biweekly zooplankton time series from the Bay of Villefranche-sur-mer, France. The classification approach proposed here provides a practical compromise between a fully automatic method with varying degrees of bias and a manual but accurate classification of zooplankton. We also evaluate the appropriate number of images to include in digital learning sets and compare the accuracy of six classification algorithms. We evaluate the accuracy of the ZooScan for automated measurements of body size and present relationships between machine measures of size and C and N content of selected zooplankton taxa. We demonstrate that the ZooScan system can produce useful measures of zooplankton abundance, biomass and size spectra, for a variety of ecological studies.
In this article, we review evidence of how climate change has already resulted in clearly discernable changes in marine Arctic ecosystems. After defining the term ‘footprint’ and evaluating the availability of reliable baseline information we review the published literature to synthesize the footprints of climate change impacts in marine Arctic ecosystems reported as of mid-2009. We found a total of 51 reports of documented changes in Arctic marine biota in response to climate change. Among the responses evaluated were range shifts and changes in abundance, growth/condition, behaviour/phenology and community/regime shifts. Most reports concerned marine mammals, particularly polar bears, and fish. The number of well-documented changes in planktonic and benthic systems was surprisingly low. Evident losses of endemic species in the Arctic Ocean, and in ice algae production and associated community remained difficult to evaluate due to the lack of quantitative reports of its abundance and distribution. Very few footprints of climate change were reported in the literature from regions such as the wide Siberian shelf and the central Arctic Ocean due to the limited research effort made in these ecosystems. Despite the alarming nature of warming and its strong potential effects in the Arctic Ocean the research effort evaluating the impacts of climate change in this region is rather limited.
Over the last two decades, there has been an accelerating advancement of acoustic and optical plankton samplers, opening many opportunities for fine-scale studies of plankton distribution. To date, however, the imaging systems have been limited in the volume of water being sampled, thereby restricting their utility to quantifying highly abundant, small zooplankton like copepods, but not relatively rarer, larger ichthyo- and other meso-zooplankton (e.g., larval decapods, salps, pteropods, ctenophores, etc.). Here we describe an imaging system, In situ ichthyoplankton imaging system (ISIIS), that is capable of In situ (i.e., noninvasive) sampling of sufficiently large volumes of water at very high resolution, allowing quantitative measurement of these rare plankton, while at the same time also recording the smaller more abundant taxa. Capitalizing on state-of-the-art digital line scan cameras and high-throughput computer data transfer and storage, combined with shadow photographic lighting techniques, we have designed and built a towed system capable of imaging at 68-micron pixel resolution, yet with up to a 20-cm depth of field (with a 14-cm field of view). This system is coupled with various environmental sensors (e.g., CTD, fluorometer), enabling the evaluation of fine-scale, taxon-specific distributions in relation to environmental conditions. Field testing demonstrated high-resolution imagery of plankters, while quantitatively imaging >70 L s-1 continuously for a 78-min trial. © 2008, by the American Society of Limnology and Oceanography, Inc.
Jefferson T. Turner (2004) The importance of small planktonic copepods and their roles in pelagic marine food webs. Zoological Studies 43(2): 255-266. Small planktonic marine copepods (< 1 mm in length) are the most abundant metazoans on Earth. Included are adults and copepodites of calanoid genera such as Paracalanus, Clausocalanus, and Acartia; cyclopoid genera such as Oithona, Oncaea, and Corycaeus; plank-tonic harpacticoids of the genus Microsetella; and nauplii of almost all copepod species. Despite the abun-dance of small copepods, they have historically been undersampled due to the use of nets with meshes > 200-333 µm. Recent studies have shown, however, that when appropriate net meshes of 100 µm or less are used, small copepods vastly exceed the abundance and sometimes the biomass of larger ones. Failure to adequate-ly account for small copepods may cause serious underestimations of zooplankton abundance and biomass, the copepod grazing impact on phytoplankton primary production, zooplankton-mediated fluxes of chemicals and materials, and trophic interactions in the sea. The feeding ecology of small copepods is less well-known than that of adults of larger copepod species, such as members of the genus Calanus. Further, most feeding information for small copepods is for coastal genera such as Acartia, rather than for offshore taxa. Although it is generally assumed that small copepods, including nauplii, feed primarily upon small-sized phytoplankton cells, most such information comes from rearing or feeding studies on limited laboratory diets. There have been few examinations of actual copepod feeding on mixed diets of natural phytoplankton and microzooplank-ton found in the sea, but some of those have produced surprises. For instance, some species of Oithona and Paracalanus and even nauplii of Arctic Calanus spp. may feed primarily as predators upon heterotrophic pro-tists, rather than as grazers of phytoplankton. Also, nauplii of various tropical copepod species have been shown to feed upon bacterioplankton. Thus, numerous basic questions remain as to the feeding ecology and grazing/predation impact of small copepods in the sea. Despite limited knowledge of what small copepods eat, it is clear that many higher-trophic-level consumers eat them. Numerous studies have shown that copepod nauplii, Oithona spp., and other small copepods are important prey of fish larvae and other planktivores. Small copepods exhibit a variety of reproductive strategies to compensate for losses to their populations due to pre-dation. These include having high fecundity and growth rates, when not limited by insufficient food; having high reproduction and growth rates at warmer temperatures; having limited motion and low respiration rates, allow-ing the investment of more energy in reproduction; and having extended longevity to maximize lifetime repro-ductive output. Thus, small copepods are important links in marine food webs, serving as major grazers of phytoplankton, as components of the microbial loop, and as prey for ichthyoplankton and other larger pelagic carnivores. Our present inadequate understanding of the true abundance, biomass, trophic ecology, and role of small copepods in biogenic fluxes precludes proper understanding of the ecology of the sea. Small planktonic marine copepods (< 1 mm in length) are undoubtedly the most abundant metazoans on Earth. Included are adults and copepodites of calanoid genera such as Paracalanus, Pseudocalanus, Acartia, and Clausocalanus; cyclopoid genera such as Oithona, and Oncaea and Corycaeus; planktonic harpacticoids of the genus Microsetella; and nau-plii of almost all copepod species. Because the early stages of all copepods include nauplii, even copepods that are comparatively large as adults are small when young. 255 Zoological Studies 43(2): 255-266 (2004) 256 Despite their overwhelming abundance and pivotal position in marine food webs, there is still comparatively less knowledge of these small cope-pods than for larger calanoid taxa such as mem-bers of the genus Calanus (Marshall and Orr 1955, Tande and Miller 1996 2000). This is particularly true for the feeding and reproductive ecology of small copepods. Nonetheless, there is a substan-tial body of literature indicating that small cope-pods are important prey items for larval fish and other zooplanktivorous consumers. Accordingly, this paper reviews the abundance of small plank-tonic marine copepods, their feeding ecology, their role as prey for predators at higher trophic levels, and aspects of reproductive biology which allow sufficient reproductive success to counter preda-tion losses.
The Arctic bloom consists of two distinct categories of primary producers, ice algae growing within and on the underside of the sea ice, and phytoplankton growing in open waters. Long chain omega-3 fatty acids, a subgroup of polyunsaturated fatty acids (PUFAs) produced exclusively by these algae, are essential to all marine organisms for successful reproduction, growth, and development. During an extensive field study in the Arctic shelf seas, we followed the seasonal biomass development of ice algae and phytoplankton and their food quality in terms of their relative PUFA content. The first PUFA-peak occurred in late April during solid ice cover at the onset of the ice algal bloom, and the second PUFA-peak occurred in early July just after the ice break-up at the onset of the phytoplankton bloom. The reproduction and growth of the key Arctic grazer Calanus glacialis perfectly coincided with these two bloom events. Females of C. glacialis utilized the high-quality ice algal bloom to fuel early maturation and reproduction, whereas the resulting offspring had access to ample high-quality food during the phytoplankton bloom 2 months later. Reduction in sea ice thickness and coverage area will alter the current primary production regime due to earlier ice break-up and onset of the phytoplankton bloom. A potential mismatch between the two primary production peaks of high-quality food and the reproductive cycle of key Arctic grazers may have negative consequences for the entire lipid-driven Arctic marine ecosystem.
Rises in surface air temperature (SAT) in response to increasing concentrations of greenhouse gases (GHGs) are expected to be amplified in northern high latitudes, with warming most pronounced over the Arctic Ocean owing to the loss of sea ice. Observations document recent warming, but an enhanced Arctic Ocean signal is not readily evident. This disparity, combined with varying model projections of SAT change, and large variability in observed SAT over the 20th century, may lead one to question the concept of Arctic amplification. Disparity is greatly reduced, however, if one compares observed trajectories to near-future simulations (2010–2029), rather than to the doubled-CO2 or late 21st century conditions that are typically cited. These near-future simulations document a preconditioning phase of Arctic amplification, characterized by the initial retreat and thinning of sea ice, with imprints of low-frequency variability. Observations show these same basic features, but with SATs over the Arctic Ocean still largely constrained by the insulating effects of the ice cover and thermal inertia of the upper ocean. Given the general consistency with model projections, we are likely near the threshold when absorption of solar radiation during summer limits ice growth the following autumn and winter, initiating a feedback leading to a substantial increase in Arctic Ocean SATs.
Anthropogenic fragmentation of landscapes is known as a major reason for the loss of species in industrialized countries. Landscape fragmentation caused by roads, railway lines, extension of settlement areas, etc., further enhances the dispersion of pollutants and acoustic emissions and affects local climatic conditions, water balance, scenery, and land use. In this study, three new measures of fragmentation are introduced: degree of landscape division (D), splitting index (S), and effective mesh size (m). They characterize the anthropogenic penetration of landscapes from a geometric point of view and are calculated from the distribution function of the remaining patch sizes.First, D, S, and m are defined, their mathematical properties are discussed, and their reactions to the six fragmentation phases of perforation, incision, dissection, dissipation, shrinkage, and attrition are analysed. Then they are compared with five other known fragmentation indices with respect to nine suitability criteria such as intuitive interpretation, low sensivity to very small patches, monotonous reaction to different fragmentation phases, and detection of structural differences. Their ability to distinguish spatial patterns is illustrated by means of two series of model patterns. In particular, the effective mesh size (m), representing an intensive and area-proportionately additive measure, proves to be well suited for comparing the fragmentation of regions with differing total size.
The authors present evidence of the difficulties facing human taxonomists/ecologists in
identifying marine dinoflagellates. This is especially important for work on harmful algal blooms in
marine aquaculture. It is shown that it is difficult for people to categorise specimens from species with
significant morphological variation, perhaps with morphologies overlapping with those of other
species. Trained personnel can be expected to achieve 67 to 83% self-consistency and 43%
consensus between people in an expert taxonomic labelling task. Experts who are routinely engaged
in particular discriminations can return accuracies in the range of 84 to 95%. In general, neither
human nor machine can be expected to give highly accurate or repeatable labelling of specimens. It
is also shown that automation methods can perform as well as humans on these complex
categorisations.
In general, organisms are not distributed uniformly over space, consider for example, the distribution of primary production over the world’s oceans (Fig. 2.30). Moreover, such heterogeneity occurs at all spatial scales. We can use the size of a patch of higher abundance or the distances among such patches as a way to assess the scale of heterogeneity. There are significantly higher or lower rates of production in distances of 105–108 m within any linear transect in Fig. 2.30. Spatial variability
is also pervasive at smaller distances: the variation in concentration of dissolved CO2 in surface waters of the Gulf of Maine is evidence of spatially heterogeneous biological activity on a scale of 104−105 m, while the variability in concentration of chlorophyll
in St. Margaret’s Bay, Nova Scotia
, varies on a scale of about 104 m (Figs. 13.1a, b). Measurements of numbers of zooplankton m−2 off the California coast
show patches of tens of meters (Fig. 13.2); a careful examination of spatial variability of phytoplankton production shows significant patches in a scale of 10−1 m, most clearly in samples from the Gulf Stream
(Fig. 13.1d). Plankton
recorders (for example, Davis et al. 1992) provide data over spatial scales of microns to hundreds of meters, confirming the presence of aggregated distributions
of organisms at scales of 20 cm–200 m.
Many textbooks teach a rule of thumb stating that the mean is right of the median under right skew, and left of the median under left skew. This rule fails with surprising frequency. It can fail in multimodal distributions, or in distributions where one tail is long but the other is heavy. Most commonly, though, the rule fails in discrete distributions where the areas to the left and right of the median are not equal. Such distributions not only contradict the textbook relationship between mean, median, and skew, they also contradict the textbook interpretation of the median. We discuss ways to correct ideas about mean, median, and skew, while enhancing the desired intuition.
Bagging predictors is a method for generating multiple versions of a predictor and using these to get an aggregated predictor. The aggregation averages over the versions when predicting a numerical outcome and does a plurality vote when predicting a class. The multiple versions are formed by making bootstrap replicates of the learning set and using these as new learning sets. Tests on real and simulated data sets using classification and regression trees and subset selection in linear regression show that bagging can give substantial gains in accuracy. The vital element is the instability of the prediction method. If perturbing the learning set can cause significant changes in the predictor constructed, then bagging can improve accuracy.
The support-vector network is a new learning machine for two-group classification problems. The machine conceptually implements the following idea: input vectors are non-linearly mapped to a very high-dimension feature space. In this feature space a linear decision surface is constructed. Special properties of the decision surface ensures high generalization ability of the learning machine. The idea behind the support-vector network was previously implemented for the restricted case where the training data can be separated without errors. We here extend this result to non-separable training data. High generalization ability of support-vector networks utilizing polynomial input transformations is demonstrated. We also compare the performance of the support-vector network to various classical learning algorithms that all took part in a benchmark study of Optical Character Recognition.
High-resolution photomicrographs of phytoplankton cells and chains can now be acquired with imagingin-flow systems at rates that make manual identification impractical for many applications. To address the challenge for automated taxonomie identification of images generated by our custom-built submersible Imaging FlowCytobot, we developed an approach that relies on extraction of image features, which are then presented to a machine learning algorithm for classification. Our approach uses a combination of image feature types including size, shape, symmetry, and texture characteristics, plus orientation invariant moments, diffraction pattern sampling, and co-occurrence matrix statistics. Some of these features required preprocessing with image analysis techniques including edge detection after phase congruency calculations, morphological operations, boundary representation and simplification, and rotation. For the machine learning strategy, we developed an approach that combines a feature selection algorithm and use of a support vector machine specified with a rigorous parameter selection and training approach. After training, a 22-category classifier provides 88% overall accuracy for an independent test set, with individual category accuracies ranging from 68% to 99%. We demonstrate application of this classifier to a nearly uninterrupted 2-month time series of images acquired in Woods Hole Harbor, including use of statistical error correction to derive quantitative concentration estimates, which are shown to be unbiased with respect to manual estimates for random subsamples. Our approach, which provides taxonomically resolved estimates of phytoplankton abundance with fine temporal resolution (hours for many species), permits access to scales of variability from tidal to seasonal and longer. © 2007, by the American Society of Limnology and Oceanograpny, Inc.
Standard and imaging flow cytometry were used to examine the composition of phytoplankton assemblages in and around a massive under-ice bloom in the Chukchi Sea in 2011. In the core of this bloom, roughly 100 km northwest of Hanna Shoal, diatoms represented roughly 87% of the water column carbon-specific biomass of phytoplankton, while nanophytoplankton contributed ~9%. Picoeukaryotes were also observed in this bloom, as were phycoerythrin-containing cells consistent with Synechococcus spp., but picophytoplankton, dinoflagellates, and prymnesiophytes each represented only ~1% of the bloom's phytoplankton biomass. More broadly along this part of the Chukchi shelf, nanophytoplankton typically comprised a larger fraction of phytoplankton biomass in the water column, 22% on average but up to 82% at certain locations. Dinoflagellates and prymnesiophytes contributed at most 2% of water column biomass at any location and were most abundant in the deeper slope stations northeast of Hanna Shoal, east of the bloom. Picophytoplankton were most abundant in these deeper slope stations as well, and also in recently ice-free areas to the south around Hanna Shoal. These cell-derived estimates of phytoplankton carbon biomass, which were computed from imaging and standard cytometric observations of phytoplankton cell sizes and from published carbon:volume relationships, agree well with independent measurements of particulate organic carbon concentration from traditional biochemical assays.
Random forests are a combination of tree predictors such that each tree depends on the values of a random vector sampled independently and with the same distribution for all trees in the forest. The generalization error for forests converges a.s. to a limit as the number of trees in the forest becomes large. The generalization error of a forest of tree classifiers depends on the strength of the individual trees in the forest and the correlation between them. Using a random selection of features to split each node yields error rates that compare favorably to Adaboost (Y. Freund & R. Schapire, Machine Learning: Proceedings of the Thirteenth International conference, ***, 148–156), but are more robust with respect to noise. Internal estimates monitor error, strength, and correlation and these are used to show the response to increasing the number of features used in the splitting. Internal estimates are also used to measure variable importance. These ideas are also applicable to regression.
Loss of Arctic sea ice has accelerated recently, culminating in a 2007 summer minimum ice extent that was 23% below the previous low. To quantify the impact of this unprecedented loss of ice on marine primary production, we have coupled satellite-derived sea ice, SST, and chlorophyll to a primary production algorithm parameterized for Arctic waters. Annual primary production in the Arctic has increased yearly by an average of 27.5 Tg C yr−1 since 2003 and by 35 Tg C yr−1 between 2006 and 2007. 30% of this increase is attributable to decreased minimum summer ice extent and 70% to a longer phytoplankton growing season. Should these trends continue, additional loss of ice during Arctic spring could boost productivity >3-fold above 1998–2002 levels, potentially altering marine ecosystem structure and the degree of pelagic-benthic coupling. Changes in carbon export could in turn modify benthic denitrification on the vast continental shelves.
It is evident that organisms have aggregated, patchy distributions of abundance on a wide variety of space and time scales. This can easily be seen in the terrestrial and littoral environments. It has been more difficult to observe in the pelagic realm simply because we cannot see into the ocean. Thus we must depend upon sampling to gain an impression of the space-time scales of pattern in this habitat. Despite the difficulties, enough sampling of the right sort has now been done so that we can make some very general statements about the nature of pattern in the ocean, particularly with regard to pattern in the distribution and abundance of planktonic organisms. All the evidence indicates that plankton is patchy on a broad spectrum of scales. Because this aggregated spatial pattern is such a general phenomenon, there is little question of its ecological and evolutionary importance. Further, because we assume our samples represent a larger universe, patchiness strongly affects our efforts to obtain estimates of the abundance of organisms and our ability to detect significant spatial and temporal changes in abundance. It is therefore of great importance that we understand its nature, causes, and effects.
The development of quantitative zooplankton collecting systems began with Hensen (1887 Berichte der Kommssion wissenschaftlichen Untersuchung der deutschen Meere in Kiel5, 1–107; 1895 Ergebnisse der Plankton-Expedition der Humbolt-Stiftung. Kiel and Leipzig: Lipsius and Tischer). Non-opening closing nets, opening closing nets (mostly messenger based), high-speed samplers, and planktobenthos net systems all had their start in his era — the late 1800s and early 1900s. This was also an era in which many of the fundamental questions about the structure and dynamics of the plankton in the worlds oceans were first posed. Fewer new systems were introduced between 1912 and 1950 apparently due in part to the two World Wars. The continuous plankton recorder stands out as a truly innovative device developed during this period (Hardy 1926b Nature, London118, 630). Resurgence in development of mechanically-based instruments occurred during the 1950s and 1960s. A new lineage of high-speed samplers, the Gulf series, began in the 1950s and a number of variants were developed in the 1960s and 1970s. Net systems specifically designed to collect neuston first appeared in the late 1950s. During the 1960s, many focused field and experimental tank experiments were carried out to investigate the hydrodynamics of nets, and much of our knowledge concerning net design and construction criteria was developed. The advent of reliable electrical conducting cables and electrically-based control systems during this same period gave rise first to a variety of cod-end samplers and then to the precursors of the acoustically and electronically-controlled multi-net systems and environmental sensors, which appeared in the 1970s. The decade of the 1970s saw a succession of multi-net systems based both on the Bé multiple plankton sampler and on the Tucker trawl. The advent of the micro-computer stimulated and enabled the development of sophisticated control and data logging electronics for these systems in the 1980s. In the 1990s, acoustic and optical technologies gave rise to sensor systems that either complement multiple net systems or are deployed without nets. Multi-sensor systems with high data telemetry rates through electro-optical cable are now being deployed in towed bodies and on remotely operated vehicles. In the offing are new molecular technologies to identify species in situ, and realtime data analysis, image processing, and 3D/4D display. In the near future, it is likely that the use of multi-sensor systems deployed on autonomous vehicles will yield world wide coverage of the distribution and abundance of zooplankton.
Several sources of error associated with subsampling zooplankton were investigated, using five subsampling devices on “pure” (single-taxon) samples. Bias occurred in 8·8 % of these experiments. Coefficients of variation (CV) were generally similar and ranged from 4·8 to 30·5%, in agreement with previous research. For fish eggs the Stempel pipette was most precise and very fast, though it is often impractical for normal samples because of clogging. For the fish larvae and invertebrates the Folsom splitter was the most precise and the Huntsman Marine Laboratory (HML) beaker technique was the fastest. The HML beaker technique appeared to cause less damage to delicate organisms than the other devices examined. In five of six subsampling comparisons with the pure-sample experiments, the CVs for an artificial “wild” zooplankton sample decreased, indicating that the pure-sample CVs are conservative estimates of subsampling error. For the wild sample, again the Folsom splitter was the most accurate and precise, and the HML beaker technique was the fastest. Both the pure- and wild-sample experiments indicated an occasional, inherent unpredictability in subsampling zooplankton in terms of bias, precision, and time. The literature clearly showed an inverse relationship between subsample N and subsampling error.
Using the HML beaker technique and a pure sample of fish eggs it was determined that error increased with the number of splits; the results were inconclusive as to whether or not error is reduced by a device which gives multiple subsamples. The difference in speed between repeated halving and obtaining multiple subsamples was negligible.
From the literature it was determined that subsampling error is not always small compared with variation among samples, as is generally assumed. It is suggested that in any zooplankton study which employs subsampling, the contribution of subsampling error to total variance should be determined for specific target taxa and considered when sample abundance estimates are made.
Discussed are the idea, examples, problems, and applications of convexity. Topics include historical examples, definitions, the John-Loewner ellipsoid, convex functions, polytopes, the algebraic operation of duality and addition, and topology of convex bodies. (KR)
Ocean acidification and warming will be most pronounced in the Arctic Ocean. Aragonite shell-bearing pteropods in the Arctic are expected to be among the first species to suffer from ocean acidification. Carbonate undersaturation in the Arctic will first occur in winter and because this period is also characterized by low food availability, the overwintering stages of polar pteropods may develop into a bottleneck in their life cycle. The impacts of ocean acidification and warming on growth, shell degradation (dissolution), and mortality of two thecosome pteropods, the polar _Limacina helicina_ and the boreal _L. retroversa_, were studied for the first time during the Arctic winter in the Kongsfjord (Svalbard). The abundance of _L. helicina_ and _L. retroversa_ varied from 23.5 to 120 ind m−2 and 12 to 38 ind m−2, and the mean shell size ranged from 920 to 981 μm and 810 to 823 μm, respectively. Seawater was aragonite-undersaturated at the overwintering depths of pteropods on two out of ten days of our observations. A 7-day experiment [temperature levels: 2 and 7 °C, pCO2 levels: 350, 650 (only for _L. helicina_) and 880 μatm] revealed a significant pCO2 effect on shell degradation in both species, and synergistic effects between temperature and pCO2 for _L. helicina_. A comparison of live and dead specimens kept under the same experimental conditions indicated that both species were capable of actively reducing the impacts of acidification on shell dissolution. A higher vulnerability to increasing pCO2 and temperature during the winter season is indicated compared with a similar study from fall 2009. Considering the species winter phenology and the seasonal changes in carbonate chemistry in Arctic waters, negative climate change effects on Arctic thecosomes are likely to show up first during winter, possibly well before ocean acidification effects become detectable during the summer season.