Elucidating ecological complexity: Unsupervised learning determines global marine eco-provinces

Identification of plankton habitats in the North Sea

The definition of an ecological niche makes it possible to anticipate the responses of a species to changing environmental conditions. Broad tolerance limits and a paucity of readily observable niches in the pelagic zone make it difficult to anticipate responses of the plankton community related to anthropogenic or environmental changes. Plankton distributions are closely linked to climate change and shape the seascape for higher trophic levels, so monitoring plankton distributions and defining ecological niches will help to understand and predict ecosystem responses. Here we apply a machine learning autoencoder and a density-based clustering algorithm to high-frequency datasets sampled with a ROTV Triaxus in the North Sea. The results indicate that in this highly dynamic environment, local hydrography prevents niche-based separation of plankton species at the sub-mesoscale, despite the availability of different habitats. Plankton patches were associated with naturally occurring frontal systems and anthropogenically induced upwelling-downwelling dipoles in the vicinity of offshore wind farms (OWFs).

Ecological associations distribution modelling of marine plankton at a global scale

Marine plankton communities form intricate networks of interacting organisms at the base of the food chain, and play a central role in regulating ocean biogeochemical cycles and climate. However, predicting plankton community shifts in response to climate change remains challenging. While species distribution models are valuable tools for predicting changes in species biogeography under climate change scenarios, they generally overlook the key role of biotic interactions, which can significantly shape ecological processes and ecosystem responses. Here, we introduce a novel statistical framework, association distribution modelling (ADM), designed to model and predict ecological associations distribution in space and time. Applied on a Tara Oceans genome-resolved metagenomics dataset, the present-day biogeography of ADM-inferred marine plankton associations revealed four major biogeographic biomes organized along a latitudinal gradient. We predicted the evolution of these biome-specific communities in response to a climate change scenario, highlighting differential responses to environmental change. Finally, we explored the functional potential of impacted plankton communities, focusing on carbon fixation, outlining the predicted evolution of its geographical distribution and implications for ecosystem function.This article is part of the theme issue 'Connected interactions: enriching food web research by spatial and social interactions'.

Classification of eco-zones from the factors and processes controlling phytoplankton biomass

Phytoplankton is of utmost importance to the marine ecosystem and, subsequently, to the Blue Economy. This study aims to explain the reasons for variability of phytoplankton by estimating the dependency of Chlorophyll-a (Chl-a) on various limiting factors using statistics. The global oceans are classified into coherent units that display similar sensitivity to changing parameters and processes using the k-means algorithm. The resulting six clusters are based on the limiting factors (PAR, iron, or nitrate) that modulate Chl-a yield divisions of the oceans, similar to regions of different trophic statuses. The clusters range from the polar and equatorial regions with high nutrient values limited by light, to open oceanic regions in downwelling gyres limited by nutrients. Some clusters also show a high dependency on marine dissolved iron. Further, oceans are also divided into eight clusters based on the processes (stratification, upwelling, topography, and solar insolation) that impact ocean productivity. The study shows that considering temporal variations is crucial for segregating oceans into ecological zones by utilizing correlation of time-series data into classification. Our results provide valuable insights into the regulation of phytoplankton abundance and its variability, which can help in understanding the implications of climate change and other anthropogenic effects on marine biology.

Patterns in the temporal complexity of global chlorophyll concentration

Decades of research have relied on satellite-based estimates of chlorophyll-a concentration to identify oceanographic processes and plan in situ observational campaigns; however, the patterns of intrinsic temporal variation in chlorophyll-a concentration have not been investigated on a global scale. Here we develop a metric to quantify time series complexity (i.e., a measure of the ups and downs of sequential observations) in chlorophyll-a concentration and show that seemingly disparate regions (e.g., Atlantic vs Indian, equatorial vs subtropical) in the global ocean can be inherently similar. These patterns can be linked to the regularity of chlorophyll-a concentration change and the likelihood of anomalous events within the satellite record. Despite distinct spatial changes in decadal chlorophyll-a concentration, changes in time series complexity have been relatively consistent. This work provides different metrics for monitoring the global ocean and suggests that the complexity of chlorophyll-a time series can be independent of its magnitude.

An archetypes approach to malaria intervention impact mapping: a new framework and example application

BACKGROUND

As both mechanistic and geospatial malaria modeling methods become more integrated into malaria policy decisions, there is increasing demand for strategies that combine these two methods. This paper introduces a novel archetypes-based methodology for generating high-resolution intervention impact maps based on mechanistic model simulations. An example configuration of the framework is described and explored.

METHODS

First, dimensionality reduction and clustering techniques were applied to rasterized geospatial environmental and mosquito covariates to find archetypal malaria transmission patterns. Next, mechanistic models were run on a representative site from each archetype to assess intervention impact. Finally, these mechanistic results were reprojected onto each pixel to generate full maps of intervention impact. The example configuration used ERA5 and Malaria Atlas Project covariates, singular value decomposition, k-means clustering, and the Institute for Disease Modeling's EMOD model to explore a range of three-year malaria interventions primarily focused on vector control and case management.

RESULTS

Rainfall, temperature, and mosquito abundance layers were clustered into ten transmission archetypes with distinct properties. Example intervention impact curves and maps highlighted archetype-specific variation in efficacy of vector control interventions. A sensitivity analysis showed that the procedure for selecting representative sites to simulate worked well in all but one archetype.

CONCLUSION

This paper introduces a novel methodology which combines the richness of spatiotemporal mapping with the rigor of mechanistic modeling to create a multi-purpose infrastructure for answering a broad range of important questions in the malaria policy space. It is flexible and adaptable to a range of input covariates, mechanistic models, and mapping strategies and can be adapted to the modelers' setting of choice.

Understanding opposing predictions of Prochlorococcus in a changing climate

Statistically derived species distribution models (SDMs) are increasingly used to predict ecological changes on a warming planet. For Prochlorococcus, the most abundant phytoplankton, an established statistical prediction conflicts with dynamical models as they predict large, opposite, changes in abundance. We probe the SDM at various spatial-temporal scales, showing that light and temperature fail to explain both temporal fluctuations and sharp spatial transitions. Strong correlations between changes in temperature and population emerge only at very large spatial scales, as transects pass through transitions between regions of high and low abundance. Furthermore, a two-state model based on a temperature threshold matches the original SDM in the surface ocean. We conclude that the original SDM has little power to predict changes when Prochlorococcus is already abundant, which resolves the conflict with dynamical models. Our conclusion suggests that SDMs should prove efficacy across multiple spatial-temporal scales before being trusted in a changing ocean.

An Outlook for Deep Learning in Ecosystem Science

Rapid advances in hardware and software, accompanied by public- and private-sector investment, have led to a new generation of data-driven computational tools. Recently, there has been a particular focus on deep learning—a class of machine learning algorithms that uses deep neural networks to identify patterns in large and heterogeneous datasets. These developments have been accompanied by both hype and scepticism by ecologists and others. This review describes the context in which deep learning methods have emerged, the deep learning methods most relevant to ecosystem ecologists, and some of the problem domains they have been applied to. Deep learning methods have high predictive performance in a range of ecological contexts, leveraging the large data resources now available. Furthermore, deep learning tools offer ecosystem ecologists new ways to learn about ecosystem dynamics. In particular, recent advances in interpretable machine learning and in developing hybrid approaches combining deep learning and mechanistic models provide a bridge between pure prediction and causal explanation. We conclude by looking at the opportunities that deep learning tools offer ecosystem ecologists and assess the challenges in interpretability that deep learning applications pose.

Ocean mover's distance: using optimal transport for analysing oceanographic data

Remote sensing observations from satellites and global biogeochemical models have combined to revolutionize the study of ocean biogeochemical cycling, but comparing the two data streams to each other and across time remains challenging due to the strong spatial-temporal structuring of the ocean. Here, we show that the Wasserstein distance provides a powerful metric for harnessing these structured datasets for better marine ecosystem and climate predictions. The Wasserstein distance complements commonly used point-wise difference methods such as the root-mean-squared error, by quantifying differences in terms of spatial displacement in addition to magnitude. As a test case, we consider chlorophyll (a key indicator of phytoplankton biomass) in the northeast Pacific Ocean, obtained from model simulations, in situ measurements, and satellite observations. We focus on two main applications: (i) comparing model predictions with satellite observations, and (ii) temporal evolution of chlorophyll both seasonally and over longer time frames. The Wasserstein distance successfully isolates temporal and depth variability and quantifies shifts in biogeochemical province boundaries. It also exposes relevant temporal trends in satellite chlorophyll consistent with climate change predictions. Our study shows that optimal transport vectors underlying the Wasserstein distance provide a novel visualization tool for testing models and better understanding temporal dynamics in the ocean.

Trophic interactions with heterotrophic bacteria limit the range of Prochlorococcus

Prochlorococcus is both the smallest and numerically most abundant photosynthesizing organism on the planet. While thriving in the warm oligotrophic gyres, Prochlorococcus concentrations drop rapidly in higher-latitude regions. Transect data from the North Pacific show the collapse occurring at a wide range of temperatures and latitudes (temperature is often hypothesized to cause this shift), suggesting an ecological mechanism may be at play. An often used size-based theory of phytoplankton community structure that has been incorporated into computational models correctly predicts the dominance of Prochlorococcus in the gyres, and the relative dominance of larger cells at high latitudes. However, both theory and computational models fail to explain the poleward collapse. When heterotrophic bacteria and predators that prey nonspecifically on both Prochlorococcus and bacteria are included in the theoretical framework, the collapse of Prochlorococcus occurs with increasing nutrient supplies. The poleward collapse of Prochlorococcus populations then naturally emerges when this mechanism of "shared predation" is implemented in a complex global ecosystem model. Additionally, the theory correctly predicts trends in both the abundance and mean size of the heterotrophic bacteria. These results suggest that ecological controls need to be considered to understand the biogeography of Prochlorococcus and predict its changes under future ocean conditions. Indirect interactions within a microbial network can be essential in setting community structure.

Machine Learning-Based Clustering Analysis: Foundational Concepts, Methods, and Applications

Unsupervised learning, the task of clustering observations in such a way that observations within cluster are more similar than those assigned to other clusters is one the central tasks of data science. Its exploratory and descriptive nature make it one of the most underused and underappreciated methods. In the present chapter we describe its core function with applied examples, explore different approaches, and discuss meaningful applications of the approach for the practicing researcher.

Abrupt shifts in 21st-century plankton communities

Marine microbial communities sustain ocean food webs and mediate global elemental cycles. These communities will change with climate; these changes can be gradual or foreseeable but likely have much more substantial consequences when sudden and unpredictable. In a complex virtual marine microbial ecosystem, we find that climate change–driven shifts over the 21st century are often abrupt, large in amplitude and extent, and unpredictable using standard early warning signals. Phytoplankton with unique resource needs, especially fast-growing species such as diatoms, are more prone to abrupt shifts. Abrupt shifts in biomass, productivity, and community structure are concentrated in Atlantic and Pacific subtropics. Abrupt changes in environmental variables such as temperature and nutrients rarely precede these ecosystem shifts, indicating that rapid community restructuring can occur in response to gradual environmental changes, particularly in nutrient supply rate ratios.

Identifying indicator species in ecological habitats using Deep Optimal Feature Learning

Much of the current research on supervised modelling is focused on maximizing outcome prediction accuracy. However, in engineering disciplines, an arguably more important goal is that of feature extraction, the identification of relevant features associated with the various outcomes. For instance, in microbial communities, the identification of keystone species can often lead to improved prediction of future behavioral shifts. This paper proposes a novel feature extractor based on Deep Learning, which is largely agnostic to underlying assumptions regarding the training data. Starting from a collection of microbial species abundance counts, the Deep Learning model first trains itself to classify the selected distinct habitats. It then identifies indicator species associated with the habitats. The results are then compared and contrasted with those obtained by traditional statistical techniques. The indicator species are similar when compared at top taxonomic levels such as Domain and Phylum, despite visible differences in lower levels such as Class and Order. More importantly, when our estimated indicators are used to predict final habitat labels using simpler models (such as Support Vector Machines and traditional Artificial Neural Networks), the prediction accuracy is improved. Overall, this study serves as a preliminary step that bridges modern, black-box Machine Learning models with traditional, domain expertise-rich techniques.

Future phytoplankton diversity in a changing climate

The future response of marine ecosystem diversity to continued anthropogenic forcing is poorly constrained. Phytoplankton are a diverse set of organisms that form the base of the marine ecosystem. Currently, ocean biogeochemistry and ecosystem models used for climate change projections typically include only 2-3 phytoplankton types and are, therefore, too simple to adequately assess the potential for changes in plankton community structure. Here, we analyse a complex ecosystem model with 35 phytoplankton types to evaluate the changes in phytoplankton community composition, turnover and size structure over the 21st century. We find that the rate of turnover in the phytoplankton community becomes faster during this century, that is, the community structure becomes increasingly unstable in response to climate change. Combined with alterations to phytoplankton diversity, our results imply a loss of ecological resilience with likely knock-on effects on the productivity and functioning of the marine environment.

Moving ecological and biogeochemical transitions across the North Pacific

In the North Pacific Ocean, nutrient rich surface waters flow south from the subpolar gyre through a transitional region and into the subtropics. Along the way, nutrients are used, recycled, and exported, leading to lower biomass and a commensurate change in ecosystem structure moving southward. We focus on the region between the two gyres (the Transition Zone) using a coupled biophysical ocean model, remote sensing, floats, and cruise data to explore the nature of the physical, biogeochemical, and ecological fields in this region. Nonlinear interactions between biological processes and the meridional gradient in nutrient supply lead to sharp shifts across this zone. These transitions between a southern region with more uniform biological and biogeochemical properties and steep meridional gradients to the north are diagnosed from extrema in the first derivative of the properties with latitude. Some transitions like that for chlorophyll a (the transition zone chlorophyll front [TZCF]) experience large seasonal excursions while the location of the transitions in other properties moves very little. The seasonal shifts are not caused by changes in the horizontal flow field, but rather by the interaction of seasonal, depth related, forcing with the mean latitudinal gradients. Focusing on the TZCF as a case study, we express its phase velocity in terms of vertical nutrient flux and internal ecosystem processes, demonstrating their nearly equal influence on its motion. This framework of propagating biogeochemical transitions can be systematically expanded to better understand the processes that structure ecosystems and biogeochemistry in the North Pacific and beyond.

High spatial resolution global ocean metagenomes from Bio-GO-SHIP repeat hydrography transects

Detailed descriptions of microbial communities have lagged far behind physical and chemical measurements in the marine environment. Here, we present 971 globally distributed surface ocean metagenomes collected at high spatio-temporal resolution. Our low-cost metagenomic sequencing protocol produced 3.65 terabases of data, where the median number of base pairs per sample was 3.41 billion. The median distance between sampling stations was 26 km. The metagenomic libraries described here were collected as a part of a biological initiative for the Global Ocean Ship-based Hydrographic Investigations Program, or "Bio-GO-SHIP." One of the primary aims of GO-SHIP is to produce high spatial and vertical resolution measurements of key state variables to directly quantify climate change impacts on ocean environments. By similarly collecting marine metagenomes at high spatiotemporal resolution, we expect that this dataset will help answer questions about the link between microbial communities and biogeochemical fluxes in a changing ocean.

Exploring biogeochemical and ecological redundancy in phytoplankton communities in the global ocean

Climate-change-induced alterations of oceanic conditions will lead to the ecological niches of some marine phytoplankton species disappearing, at least regionally. How will such losses affect the ecosystem and the coupled biogeochemical cycles? Here, we couch this question in terms of ecological redundancy (will other species be able to fill the ecological roles of the extinct species) and biogeochemical redundancy (can other species replace their biogeochemical roles). Prior laboratory and field studies point to a spectrum in the degree of redundancy. We use a global three-dimensional computer model with diverse planktonic communities to explore these questions further. The model includes 35 phytoplankton types that differ in size, biogeochemical function and trophic strategy. We run two series of experiments in which single phytoplankton types are either partially or fully eliminated. The niches of the targeted types were not completely reoccupied, often with a reduction in the transfer of matter from autotrophs to heterotrophs. Primary production was often decreased, but sometimes increased due to reduction in grazing pressure. Complex trophic interactions (such as a decrease in the stocks of a predator's grazer) led to unexpected reshuffling of the community structure. Alterations in resource utilization may cause impacts beyond the regions where the type went extinct. Our results suggest a lack of redundancy, especially in the 'knock on' effects on higher trophic levels. Redundancy appeared lowest for types on the edges of trait space (e.g. smallest) or with unique competitive strategies. Though highly idealized, our modelling findings suggest that the results from laboratory or field studies often do not adequately capture the ramifications of functional redundancy. The modelled, often counterintuitive, responses-via complex food web interactions and bottom-up versus top-down controls-indicate that changes in planktonic community will be key determinants of future ocean global change ecology and biogeochemistry.