Integrating machine learning with otolith isoscapes: Reconstructing connectivity of a marine fish over four decades.

Stable isotopes are an important tool to uncover animal migration. Geographic natal assignments often require categorizing the spatial domain through a nominal approach, which can introduce bias given the continuous nature of these tracers. Stable isotopes predicted over a spatial gradient (i.e., isoscapes) allow a probabilistic and continuous assignment of origin across space, although applications to marine organisms remain limited. We present a new framework that integrates nominal and continuous assignment approaches by (1) developing a machine-learning multi-model ensemble classifier using Bayesian model averaging (nominal); and (2) integrating nominal predictions with continuous isoscapes to estimate the probability of origin across the spatial domain (continuous). We applied this integrated framework to predict the geographic origin of the Northwest Atlantic mackerel (Scomber scombrus), a migratory pelagic fish comprised of northern and southern components that have distinct spawning sites off Canada (northern contingent) and the US (southern contingent), and seasonally overlap in the US fished regions. The nominal approach based on otolith carbon and oxygen stable isotopes (δ13C/δ18O) yielded high contingent classification accuracy (84.9%). Contingent assignment of unknown-origin samples revealed prevalent, yet highly varied contingent mixing levels (12.5-83.7%) within the US waters over four decades (1975-2019). Nominal predictions were integrated into mackerel-specific otolith oxygen isoscapes developed independently for Canadian and US waters. The combined approach identified geographic nursery hotspots in known spawning sites, but also detected geographic shifts over multi-decadal time scales. This framework can be applied to other marine species to understand migration and connectivity at a high spatial resolution, relevant to management of unit stocks in fisheries and other conservation assessments.

Topological clustering of multilayer networks.

Multilayer networks continue to gain significant attention in many areas of study, particularly due to their high utility in modeling interdependent systems such as critical infrastructures, human brain connectome, and socioenvironmental ecosystems. However, clustering of multilayer networks, especially using the information on higher-order interactions of the system entities, still remains in its infancy. In turn, higher-order connectivity is often the key in such multilayer network applications as developing optimal partitioning of critical infrastructures in order to isolate unhealthy system components under cyber-physical threats and simultaneous identification of multiple brain regions affected by trauma or mental illness. In this paper, we introduce the concepts of topological data analysis to studies of complex multilayer networks and propose a topological approach for network clustering. The key rationale is to group nodes based not on pairwise connectivity patterns or relationships between observations recorded at two individual nodes but based on how similar in shape their local neighborhoods are at various resolution scales. Since shapes of local node neighborhoods are quantified using a topological summary in terms of persistence diagrams, we refer to the approach as clustering using persistence diagrams (CPD). CPD systematically accounts for the important heterogeneous higher-order properties of node interactions within and in-between network layers and integrates information from the node neighbors. We illustrate the utility of CPD by applying it to an emerging problem of societal importance: vulnerability zoning of residential properties to weather- and climate-induced risks in the context of house insurance claim dynamics.

The recurring impact of storm disturbance on black sea bass (Centropristis striata) movement behaviors in the Mid-Atlantic Bight.

Storm events are a significant source of disturbance in the Middle Atlantic Bight, in the Northwest Atlantic, that cause rapid destratification of the water column during the late summer and early fall. Storm-driven mixing can be considered as a seasonal disturbance regime to demersal communities, characterized by the recurrence of large changes in bottom water temperatures. Black sea bass are a model ubiquitous demersal species in the Middle Atlantic Bight, as their predominantly sedentary behavior makes them ideal for tagging studies while also regularly exposing them to summer storm disturbances and the physiological stresses associated with thermal destratification. To better understand the responsiveness of black sea bass to storm impacts, we coupled biotelemetry with a high-resolution Finite Volume Community Ocean Model (FVCOM). During the summers of 2016-2018, 8-15 black sea bass were released each year with acoustic transponders at three reef sites, which were surrounded by data-logging receivers. Data were analyzed for activity levels and reef departures of black sea bass, and fluctuations in temperature, current velocity, and turbulent kinetic energy. Movement rates were depressed with each consecutive passing storm, and late-season storms were associated with permanent evacuations by a subset of tagged fish. Serial increases in bottom temperature associated with repeated storm events were identified as the primary depressor of local movement. Storm-driven increases in turbulent kinetic energy and current velocity had comparatively smaller, albeit significant, effects. Black sea bass represents both an important fishery resource and an indicator species for the impact of offshore wind development in the United States. Their availability to fisheries surveys and sensitivity to wind turbine impacts will be biased during periods of high storm activity, which is likely to increase with regional climate change.

Complementing the power of deep learning with statistical model fusion: Probabilistic forecasting of influenza in Dallas County, Texas, USA.

Influenza is one of the main causes of death, not only in the USA but worldwide. Its significant economic and public health impacts necessitate development of accurate and efficient algorithms for forecasting of any upcoming influenza outbreaks. Most currently available methods for influenza prediction are based on parametric time series and regression models that impose restrictive and often unverifiable assumptions on the data. In turn, more flexible machine learning models and, particularly, deep learning tools whose utility is proven in a wide range of disciplines, remain largely under-explored in epidemiological forecasting. We study the seasonal influenza in Dallas County by evaluating the forecasting ability of deep learning with feedforward neural networks as well as performance of more conventional statistical models, such as beta regression, autoregressive integrated moving average (ARIMA), least absolute shrinkage and selection operators (LASSO), and non-parametric multivariate adaptive regression splines (MARS) models for one week and two weeks ahead forecasting. Furthermore, we assess forecasting utility of Google search queries and meteorological data as exogenous predictors of influenza activity. Finally, we develop a probabilistic forecasting of influenza in Dallas County by fusing all the considered models using Bayesian model averaging.

Detection of the effects of stormwater control measure in streams using a Bayesian BACI power analysis.

The unpredictable timing and magnitude of precipitation events and the spatiotemporal variability of constituent concentrations are major complications to effective monitoring of watershed nutrient and sediment loads. Furthermore, detecting small changes in constituent loads in response to implementation of Stormwater control measures (SCMs) against natural variability is a challenge. Nevertheless, regulatory frameworks that direct reductions of pollutants to streams frequently depend on the ability to quantify changes in loads after management interventions. The before-after-control impact (BACI) sampling design is often used to assess the effects of an environmental change made at a known point in time. However, this approach may be complicated to apply to nutrient and sediment loads in streams as the relative impact of SCMs on nutrient concentration conditional on the long term variability of discharges has not been evaluated. Multi-scale monitoring studies that provide estimates of the natural temporal and spatial variability of discharge and concentrations could provide useful information in designing a BACI study. Here we use data from the Baltimore Long Term Ecological Research (LTER) sites and urban restoration sites to develop multiple statistical measures of the effectiveness of a given monitoring scheme in revealing the hypothesized restoration effects in terms of hydrology and nutrient loads. Stratified sampling over baseflow and stormflow and the use of multiple control streams were useful tools to detect long term cumulative reductions in concentrations due to SCMs. Moderate reductions in concentration (20%), however, were not detectable with the design options considered. We emphasize that appropriate pre-planning of monitoring schemes and sampling frequency is essential to determine if the effects on constituent loads resulting from a given watershed restoration activity are measurable.

Time series models of decadal trends in the harmful algal species Karlodinium veneficum in Chesapeake Bay.

The harmful dinoflagellate, Karlodnium veneficum, has been implicated in fish-kill and other toxic, harmful algal bloom (HAB) events in waters worldwide. Blooms of K. veneficum are known to be related to coastal nutrient enrichment but the relationship is complex because this HAB taxon relies not only on dissolved nutrients but also particulate prey, both of which have also changed over time. Here, applying cross-correlations of climate-related physical factors, nutrients and prey, with abundance of K. veneficum over a 10-year (2002-2011) period, a synthesis of the interactive effects of multiple factors on this species was developed for Chesapeake Bay, where blooms of the HAB have been increasing. Significant upward trends in the time series of K. veneficum were observed in the mesohaline stations of the Bay, but not in oligohaline tributary stations. For the mesohaline regions, riverine sources of nutrients with seasonal lags, together with particulate prey with zero lag, explained 15%-46% of the variation in the K. veneficum time series. For the oligohaline regions, nutrients and particulate prey generally showed significant decreasing trends with time, likely a reflection of nutrient reduction efforts. A conceptual model of mid-Bay blooms is presented, in which K. veneficum, derived from the oceanic end member of the Bay, may experience enhanced growth if it encounters prey originating from the tributaries with different patterns of nutrient loading and which are enriched in nitrogen. For all correlation models developed herein, prey abundance was a primary factor in predicting K. veneficum abundance.

Bootstrap quantification of estimation uncertainties in network degree distributions.

We propose a new method of nonparametric bootstrap to quantify estimation uncertainties in functions of network degree distribution in large ultra sparse networks. Both network degree distribution and network order are assumed to be unknown. The key idea is based on adaptation of the "blocking" argument, developed for bootstrapping of time series and re-tiling of spatial data, to random networks. We first sample a set of multiple ego networks of varying orders that form a patch, or a network block analogue, and then resample the data within patches. To select an optimal patch size, we develop a new computationally efficient and data-driven cross-validation algorithm. The proposed fast patchwork bootstrap (FPB) methodology further extends the ideas for a case of network mean degree, to inference on a degree distribution. In addition, the FPB is substantially less computationally expensive, requires less information on a graph, and is free from nuisance parameters. In our simulation study, we show that the new bootstrap method outperforms competing approaches by providing sharper and better-calibrated confidence intervals for functions of a network degree distribution than other available approaches, including the cases of networks in an ultra sparse regime. We illustrate the FPB in application to collaboration networks in statistics and computer science and to Wikipedia networks.

Year-round spatiotemporal distribution of harbour porpoises within and around the Maryland wind energy area.

Offshore windfarms provide renewable energy, but activities during the construction phase can affect marine mammals. To understand how the construction of an offshore windfarm in the Maryland Wind Energy Area (WEA) off Maryland, USA, might impact harbour porpoises (Phocoena phocoena), it is essential to determine their poorly understood year-round distribution. Although habitat-based models can help predict the occurrence of species in areas with limited or no sampling, they require validation to determine the accuracy of the predictions. Incorporating more than 18 months of harbour porpoise detection data from passive acoustic monitoring, generalized auto-regressive moving average and generalized additive models were used to investigate harbour porpoise occurrence within and around the Maryland WEA in relation to temporal and environmental variables. Acoustic detection metrics were compared to habitat-based density estimates derived from aerial and boat-based sightings to validate the model predictions. Harbour porpoises occurred significantly more frequently during January to May, and foraged significantly more often in the evenings to early mornings at sites within and outside the Maryland WEA. Harbour porpoise occurrence peaked at sea surface temperatures of 5°C and chlorophyll a concentrations of 4.5 to 7.4 mg m-3. The acoustic detections were significantly correlated with the predicted densities, except at the most inshore site. This study provides insight into previously unknown fine-scale spatial and temporal patterns in distribution of harbour porpoises offshore of Maryland. The results can be used to help inform future monitoring and mitigate the impacts of windfarm construction and other human activities.