Mean dynamic topography (MDT) plays an important role in the dynamics of shelf circulation. Coastal tide gauge observations in combination with the latest generation of geoid models are providing estimates of the alongshore tilt of MDT with unprecedented accuracy. Additionally, high-resolution ocean models are providing better representations of nearshore circulation and the associated tilt of MDT along their coastal boundaries. It has been shown that the newly available geodetic estimates can be used to validate model predictions of coastal MDT variability on global and basin scales. On smaller scales, however, there are significant variations in alongshore MDT that are of the same order of magnitude as the accuracy of the geoid models. In this study, we use a regional ocean model of the Gulf of Maine and Scotian Shelf (GoMSS) to demonstrate that the new observations of geodetically referenced coastal sea level can also provide valuable information for the validation of such high-resolution models. The predicted coastal MDT is in good agreement with coastal tide gauge observations referenced to the Canadian Gravimetric Geoid 2013 – Version A (CGG2013a), including a significant tilt of alongshore MDT along the coast of Nova Scotia. Using the validated GoMSS model and two idealized models, we show that this alongshore tilt of MDT can be interpreted in two complementary, and dynamically consistent, ways: in the coastal view, the tilt of MDT along the coast can provide a direct estimate of the average alongshore current. In the regional view, the tilt provides a measure of area-integrated nearshore circulation. This highlights the value of using geodetic MDT estimates for model validation and ocean monitoring.

Subseasonal-to-seasonal (S2S) prediction is a global effort to forecast the state of the atmosphere and ocean with lead times between two weeks and a season. This study explores the feasibility of S2S prediction of the ocean using a variety of tools including statistical analysis, a statistical-dynamical mixed layer model, and a regional, high-resolution ocean circulation model based on physical principles. Ocean predictability on S2S timescales is analyzed by compositing winter sea surface temperature (SST) anomalies in the North Atlantic with respect to the state of the Madden–Julian Oscillation (MJO). It is found that statistically significant, large-scale SST changes, particularly along the eastern seaboard of North America, can be related to the MJO. This signal is shown to be driven by anomalous air–sea heat fluxes caused by atmospheric perturbations in response to the MJO. The high-resolution model of the Gulf of Maine and Scotian Shelf is used to downscale the mean ocean response to the MJO. The model is able to capture the observed relationship between the MJO and SST in the northwest Atlantic. It is also shown that the anomalous atmospheric circulation in response to the MJO leads to anomalous upwelling on the Scotian Shelf. Overall, this study demonstrates that it is feasible, and of value, to use regional ocean models for S2S prediction.

In shallow, tidally dominated regions, overtides and the mean state of the ocean are coupled through their simultaneous generation by nonlinear processes. We present a new method that uses observed overtides (e.g., M-4) and mean currents to independently assess the accuracy of mean dynamic topography (MDT) predicted by ocean models. This is useful in regions where no sufficiently long, geodetically referenced sea level records are available for validation of the predicted MDT. We apply the new method to a regional model of the Gulf of Maine/Scotian Shelf region (GoMSS) and a barotropic, higher resolution model focused on the upper Bay of Fundy (UBoF). We first show that the tides and mean circulation predicted by UBoF are in good agreement with observations and a significant improvement over GoMSS. Next, we use UBoF to demonstrate that observed overtides are useful in selecting the optimal bathymetry and constraining parameters of an ocean model. An accurate bathymetry is critical for capturing the dominant nonlinear processes that generate overtides and control the form of MDT in shallow, tidally dominated regions. Finally, we use the observed overtides to argue that the MDT predicted by UBoF is more realistic than the prediction by GoMSS. In the vicinity of headlands, both horizontal advection and bottom friction in UBoF generate harmonics of the tidal flow and local setdowns of coastal MDT of O $\mathcal{O}$(10 cm). The prediction of such features, validated by observed overtides, can provide guidance in future deployments of tide gauges in support of geoid and ocean model validation.

Operational flood forecast systems are typically based on barotropic models in order to limit computational cost. In this study, a computationally efficient way of adding spatial and temporal variations of water density (henceforth "adding baroclinicity ") to a global barotropic forecast system is described and evaluated. The approach is illustrated using a global high-resolution (1/12 degrees) model with an optimized vertical grid of nine ocean levels. Temperature (T) and salinity (S) for each layer are weakly nudged to forecast fields provided by a coarser resolution (1/4 degrees), data-assimilative ocean model. This adds realistic baroclinic variability with periods exceeding about 15 d to the 1/12 degrees model while allowing the higher frequency variability to evolve freely. The computational cost of running the new model is only 10% higher than the cost of running the two-dimensional barotropic version of the model that forms the basis of the present total water level (TWL) forecast system. The value of adding baroclinicity is demonstrated across the sea level frequency spectrum using hourly observations of TWL made between October 20, 2019 and February 28, 2021 by a global distribution of 211 tide gauges. Large improvements in the tidal residuals (TWL-tide) are shown for periods exceeding 15 d. This is consistent with the known importance of baroclinic processes to sea level variability in this frequency range. For periods between 10 h and 15 d, baroclinicity also improves the predictions of extreme residuals by up to 15 cm, with the largest improvements occurring on narrow shelves. In terms of extremes, adding baroclinicity improves predictions of the three largest positive tidal residuals for 2020 at 89% of the tide gauges with the largest improvements reaching 42 cm. Similar results were found for the three largest negative residuals for each tide gauge record. The importance of modeling baroclinically-modified coastal trapped waves is illustrated for the west coast of North and South America.

Seasonal forecasting of tropical cyclones is a topic of considerable interest to the public, government and private sectors. To improve understanding of the dynamics controlling the predictability of tropical cyclone (TC) activity, and improve the accuracy of forecasts, multiple studies have related TC activity to empirically‐defined indices including the El Niño‐Southern Oscillation, the Atlantic Multidecadal Oscillation, and the North Atlantic Oscillation. These indices were not developed to forecast TC activity but rather summarize other aspects of atmosphere‐ocean variability. In this study we use a statistical approach, based on redundancy analysis, to define two indices related to overall activity and steering of TCs. We focus on North Atlantic TCs that reached tropical storm strength (≥34 kt) between August and October 1948–2016. TC occurrences are binned using an equal area grid that covers the North Atlantic. The redundancy indices are linear combinations of mean sea level pressure for the same season. Cross validation is used to guard against over fitting in the definition of the indices. This approach provides two physically interpretable redundancy indices related to North Atlantic TC activity. The leading redundancy index is used to successfully reconstruct the total number of TCs and the accumulated cyclone energy, over the extended period 1878–2014 using seasonal mean sea level pressures from the National Centers for Environmental Prediction 20th‐Century Reanalysis version 2c. Extensions of the approach for seasonal forecasting are discussed.

The development of a computationally efficient scheme for predicting the global distribution of total water level (TWL) is discussed. The ocean model is barotropic, has a horizontal grid spacing of 1/12 degrees, and is based on the NEMO modeling framework. It is forced by the gravitational potential and hourly atmospheric fields for 2008. Hourly time spacing was required to resolve the S-2 tide in global air pressure and wind. The predicted tide in water deeper than 400 m was nudged to TPXO8 "observations" of tidal elevation or current using a scheme called tidal nudging (Kodaira a al., 2019). The benefit of nudging horizontal velocity in the momentum equation, compared to sea level in the continuity equation, is discussed. Tidal nudging is shown to improve tidal predictions of sea level at the coast, particularly at the S-2 tidal frequency. The predicted radiational S-2 tide in sea level forced solely by the S-2 tide in global air pressure reaches amplitudes exceeding 80 cm. Decreasing the time spacing of the air pressure forcing from 1 h to 3 h reduces the S-2 amplitude in air pressure by a factor of 0.82, consistent with expectations based on Fourier analysis. This highlights the importance of using hourly atmospheric forcing when predicting the global sea level response to atmospheric forcing. The radiational S-2 tide in sea level is subject to strong nonlinear interaction with the gravitational tide, leading to a pronounced attenuation of the radiational S-2 tide. The attenuation is explained by an increase in effective bottom friction at the S-2 frequency due to the presence of the gravitational tide. Four schemes for predicting TWL are evaluated to quantify the impact of tidal nudging and nonlinear interaction of tide and surge. Using TWLs observed by 304 coastal tide gauges, we show it is necessary to include both tidal nudging and nonlinear interaction. Plans for the further development of an operational flood forecast system for the Canadian coast, based on the above model, are discussed.

Concern over increased flooding and the need for earlier and more reliable risk forecasts motivate the continued development of operational forecasts of coastal water level. We report here on results from a year long ensemble of total water level forecasts calculated using a dynamical ocean model forced with ensemble atmospheric forcing and tidal boundary conditions. We focus on the east coast of Canada. The domain includes the Gulf of St. Lawrence, the Labrador Shelf, the Scotian Shelf, and the Gulf of Maine. The water level ensemble is made of a control and 20 perturbed members. Individual forecasts are produced twice daily for 16 days. The novelty of the present study is in the exploration of perturbations of the ocean contributions. In addition to examining how uncertainty in atmospheric forcing maps into flood risk, we also explore the feasibility, and impact, of perturbing the ocean tides. We use a recent case study to demonstrate our findings.

With the long-term goal of developing an ensemble forecast system for coastal flooding, we are developing a dynamically-based, numerical model of the global ocean. The model is based on the NEMO framework and has been used to predict global tides and surges in previous studies. This study focuses on the optimization of the joint prediction of both tides and surges, the two main components of total water level that cause coastal flooding. To improve the predictions of the tide we use a modified form of "spectral nudging". We show this leads to significant improvements in the prediction of the M-2 tide in the open ocean, and also in the shallow regions closer to shore where the model is not nudged. The median value of the vector difference of the tidal amplitude based on sea level observations and a dataassimilative model, and the predictions of our ocean model, is reduced from 11.2 cm to 2.66 cm by the nudging. The improvement deteriorates significantly however if additional tidal constituents are included in the model (most notably S-2). This is explained in terms of spectral leakage between tidal bands associated with the nudging methodology and a straightforward solution is proposed.

A major challenge for managing impacts and implementing effective mitigation measures and adaptation strategies for coastal zones affected by future sea level (SL) rise is our limited capacity to predict SL change at the coast on relevant spatial and temporal scales. Predicting coastal SL requires the ability to monitor and simulate a multitude of physical processes affecting SL, from local effects of wind waves and river runoff to remote influences of the large-scale ocean circulation on the coast. Here we assess our current understanding of the causes of coastal SL variability on monthly to multidecadal timescales, including geodetic, oceanographic and atmospheric aspects of the problem, and review available observing systems informing on coastal SL. We also review the ability of existing models and data assimilation systems to estimate coastal SL variations and of atmosphere-ocean global coupled models and related regional downscaling efforts to project future SL changes.We discuss (1) observational gaps and uncertainties, and priorities for the development of an optimal and integrated coastal SL observing system, (2) strategies for advancing model capabilities in forecasting shortterm processes and projecting long-term changes affecting coastal SL, and (3) possible future developments of sea level services enabling better connection of scientists and user communities and facilitating assessment and decision making for adaptation to future coastal SL change.

The fact that ocean currents must flow parallel to the coast leads to the dynamics of coastal sea level being quite different from the dynamics in the open ocean. The coastal influence of open-ocean dynamics (dynamics associated with forcing which occurs in deep water, beyond the continental slope) therefore involves a hand-over between the predominantly geostrophic dynamics of the interior ocean and the ageostrophic dynamics which must occur at the coast. An understanding of how this hand-over occurs can be obtained by considering the combined role of coastal trapped waves and bottom friction. We here review understanding of coastal trapped waves, which propagate cyclonically around ocean basins along the continental shelf and slope, at speeds which are fast compared to those of baroclinic planetary waves and currents in the open ocean (excluding the large-scale barotropic mode). We show that this results in coastal sea-level signals on western boundaries which, compared to the nearby open-ocean signals, are spatially smoothed, reduced in amplitude, and displaced along the coast in the direction of propagation of coastal trapped waves. The open-ocean influence on eastern boundaries is limited to signals propagating polewards from the equatorial waveguide (although a large-scale diffusive influence may also play a role). This body of work is based on linearised equations, but we also discuss the nonlinear case. We suggest that a proper consideration of nonlinear terms may be very important on western boundaries, as the competition between advection by western boundary currents and a counter-propagating influence of coastal trapped waves has the potential to lead to sharp gradients in coastal sea level where the two effects come into balance.

Assessments of ocean data assimilation (DA) systems and observing system design experiments typically rely on identical or nonidentical twin experiments. The identical twin approach has been recognized as yielding biased impact assessments in atmospheric predictions, but these shortcomings are not sufficiently appreciated for oceanic DA applications. Here we present the first direct comparison of the nonidentical and identical twin approaches in an ocean DA application. We assess the assimilation impact for both approaches in a DA system for the Gulf of Mexico that uses the ensemble Kalman filter. Our comparisons show that, despite a reasonable error growth rate in both approaches, the identical twin produces a biased skill assessment, overestimating the improvement from assimilating sea surface height and sea surface temperature observations while underestimating the value of assimilating temperature and salinity profiles. Such biases can lead to an undervaluation of some observing assets (in this case profilers) and thus a misguided distribution of observing system investments.

Near-inertial oscillations (NIOs) on the inner Scotian shelf are studied using observations, a simple slab model, and two operational shelf circulation models. High-frequency radar and ADCP observations from December 2015 to February 2016 show that individual NIO events forced by time-varying wind stress typically lasted for three to four inertial periods. NIOs with speeds exceeding 0.25 m s(-1) were observed in the offshore part of the study region, but their amplitudes decreased shoreward within similar to 40 km of the coast. The NIOs had spatial scales of similar to 80 and similar to 40 km in the alongshore and cross-shore directions, respectively. The NIO phases varied moving from west to east, consistent with the typical movement of winter storms across the study region. Evolving rotary spectral analysis reveals that the peak frequency f(p) of the NIOs varied with time by similar to 7% of the local inertial frequency. The variation in f(p) can be explained in part by local wind forcing as demonstrated by the slab model. The remaining variation in f(p) can be explained in part by variations in the background vorticity associated with changes in the strength and position of the Nova Scotia Current, an unstable baroclinic boundary current that runs along the coast to the southwest. Two operational shelf circulation models are used to examine the abovementioned features in the high-frequency-radar and ADCP observations. The models reproduce the spatial structure of the NIOs and, in a qualitative sense, the temporal variations of f(p).

Effective data assimilation methods for incorporating observations into marine biogeochemical models are required to improve hindcasts, nowcasts and forecasts of the ocean's biogeochemical state. Recent assimilation efforts have shown that updating model physics alone can degrade biogeochemical fields while only updating biogeochemical variables may not improve a model's predictive skill when the physical fields are inaccurate. Here we systematically investigate whether multivariate updates of physical and biogeochemical model states are superior to only updating either physical or biogeochemical variables. We conducted a series of twin experiments in an idealized ocean channel that experiences wind-driven upwelling. The forecast model was forced with biased wind stress and perturbed biogeochemical model parameters compared to the model run representing the "truth". Taking advantage of the multivariate nature of the deterministic Ensemble Kalman Filter (DEnKF), we assimilated different combinations of synthetic physical (sea surface height, sea surface temperature and temperature profiles) and biogeochemical (surface chlorophyll and nitrate profiles) observations. We show that when biogeochemical and physical properties are highly correlated (e.g., thermocline and nutricline), multivariate updates of both are essential for improving model skill and can be accomplished by assimilating either physical (e.g., temperature profiles) or biogeochemical (e.g., nutrient profiles) observations. In our idealized domain, the improvement is largely due to a better representation of nutrient upwelling, which results in a more accurate nutrient input into the euphotic zone. In contrast, assimilating surface chlorophyll improves the model state only slightly, because surface chlorophyll contains little information about the vertical density structure. We also show that a degradation of the correlation between observed subsurface temperature and nutrient fields, which has been an issue in several previous assimilation studies, can be reduced by multivariate updates of physical and biogeochemical fields.

Dissolved hydrogen measurements were made at high resolution in surface waters along a tropical north Atlantic transect between Guadeloupe and Cape Verde in 2015 (Meteor 116). Parallel water samples acquired to assess the relative abundance of the nifH gene from several types of diazotrophs, indicated that Trichodesmium and UCYN-A were dominant in this region. We show that a high degree of correlation exists between the hydrogen saturations and UCYN-A nifH abundance, and a weak correlation with Trichodesmium. The findings suggest that nitrogen fixation by UCYN-A is a major contributor to hydrogen supersaturations in this region of the ocean. The ratio of hydrogen released to nitrogen fixed has not been determined for this symbiont, but the indications are that it may be high in comparison with the small number of diazotrophs for which the ratio has been measured in laboratory cultures. We speculate that this would be consistent with the diazotroph being an exosymbiont on its haptophyte host. Our high resolution measurements of hydrogen concentrations are capable of illustrating the time and space scales of inferred activity of diazotrophs in near real-time in a way that cannot be achieved by biological sampling and rate measurements requiring incubations with N-15(2). Direct measurement of high resolution spatial variability would be relatively challenging through collection and analysis of biological samples by qPCR, and extremely challenging by N-15-uptake techniques, neither of which methods yields real-time data. Nonetheless, determination of fixation rates still firmly depends on the established procedure of incubations in the presence of N-15(2).

The overall goal is to downscale ocean conditions predicted by an existing global prediction system and evaluate the results using observations from the Gulf of Maine, Scotian Shelf and adjacent deep ocean. The first step is to develop a one-way nested regional model and evaluate its predictions using observations from multiple sources including satellite-borne sensors of surface temperature and sea level, CTDs, Argo floats and moored current meters. It is shown that the regional model predicts more realistic fields than the global system on the shelf because it has higher resolution and includes tides that are absent from the global system. However, in deep water the regional model misplaces deep ocean eddies and meanders associated with the Gulf Stream. This is not because the regional model’s dynamics are flawed but rather is the result of internally generated variability in deep water that leads to decoupling of the regional model from the global system. To overcome this problem, the next step is to spectrally nudge the regional model to the large scales (length scales > 90 km) of the global system. It is shown this leads to more realistic predictions off the shelf. Wavenumber spectra show that even though spectral nudging constrains the large scales, it does not suppress the variability on small scales; on the contrary, it favours the formation of eddies with length scales below the cutoff wavelength of the spectral nudging.

Global M-2 tidal surface currents are predicted using a global baroclinic ocean model with horizontal grid spacing of 1/12 degrees and 19 z-levels in the vertical. After first showing the predicted tidal elevations are in reasonable agreement with observations made by bottom pressure recorders and altimeters, the predicted tidal surface currents are evaluated by comparing them with independent estimates based on observed drifter trajectories. Both predicted and observed tidal surface currents can exceed 0.1 m s(-1) in the deep ocean. Internal tides are shown to make a significant contribution to the predicted tidal surface currents. Phase locking of the surface and internal tides causes spatial changes in the predicted tidal surface currents that vary with approximately the same wavenumber as that of the lowest mode internal tide. Qualitatively similar, small-scale variations are also detected in the observed estimates but the variations do not line up exactly with the predictions. Possible explanations for the mismatch are given. The seasonal variation of M2 tidal surface currents, and the energy conversion rate from surface to internal tides, is also predicted by initializing, and restoring, the model to an observed seasonal climatology of temperature and salinity. Compared to tidal elevation, the seasonal change of tidal surface current can be large (order 10% for each hemisphere). It is caused by seasonal variations in the vertical structure of the baroclinic modes and the energy conversion rate. In the vicinity of major bathymetric features, the seasonal variation of second and higher order modes can be much larger (up to 50%).

Both the American eel (Anguilla rostrata) and European eel (Anguilla anguilla) undertake long-distance migrations from continental waters to their spawning sites in the SargassoSea. Their migration routes and orientation mechanisms remain a mystery. A biophysical particle tracking model was used in this study to simulate their oceanic migration from two release areas: off the Scotian Shelf (Canada) and off the Irish continental shelf. Two plausible swimming-directed behaviours were considered for simulating two different migratory paths: true navigation to specific spawning sites and innate compass orientation towards the vast spawning area. Several combinations of swimming speeds and depths were tested to assess the effect of ocean circulation on resulting migratory pathways of virtual eels (v-eels), environmental conditions experienced along their oceanic migration, and energy consumption. Simulations show that the spawning area can be reached in time by constantly swimming and following a readjusted heading (true navigation) or a constant heading (compass orientation) even at the lowest swimming speed tested (0.2 m s(-1)) for most v-eels. True navigation might not be necessary to reach the spawning area. The ocean currents affect mainly the migration of American v-eels, particularly for swimming speeds lower than 0.8 m s(-1). The ocean circulation increases the variability in the oceanic migration and generally reduces the efficiency of the veels, although positive effects can be possible for certain individuals. The depth range of diel vertical migration (DVM) significantly affects the total energy expenditure due to the water temperature experienced at the various depths. Model results also suggest that energy would not be a limiting factor as v-eels constantly swimming at 0.8 BL s(-1) spent, <25 and 42% of energy available for migration for American and European v-eels, respectively.

With the long-term goal of developing an operational forecast system for total water level, we conduct a hindcast study of global storm surges for Fall 2014 using a baroclinic ocean model based on the NEMO framework. The model has 19 vertical levels, a horizontal resolution of 1/12°, and is forced by hourly forecasts of atmospheric wind and air pressure. Our first objective is to evaluate the model’s ability to predict hourly sea levels recorded by a global array of 257 tide gauges. It is shown that the model can provide reasonable predictions of surges for the whole test period at tide gauges with relatively large tidal residuals (i.e., gauges where the standard deviation of observed sea level, after removal of the tide, exceeds 5 cm). Our second objective is to quantify the effect of density stratification on the prediction of global surges. It is found that the inclusion of density stratification increases the overall predictive skill at almost all tide gauges. The increase in skill for the instantaneous peak surge is smaller. The location for which the increase in overall skill is largest (east coast of South Africa) is discussed in detail and physical reasons for the improvement are given.

Along the Atlantic coast of Nova Scotia sea urchin disease outbreaks are statistically linked to North Atlantic hurricanes and warm sea temperatures. The amoebic pathogen Paramoeba invadens, which causes these disease outbreaks, is unable to withstand typical minimum sea temperatures along this coast, suggesting that it is reintroduced during periods of peak temperatures. Here, we examine hypotheses for mechanisms of introduction or persistence of P. invadens using data on sea urchin disease outbreaks from a 5-yr field experiment (2010-2014), in combination with high-temporal-resolution oceanographic and meteorologic data. Disease outbreaks were observed in 4 yr, with the onset of mass mortality (>50% morbidity or mortality) ranging from mid August to mid October. Physical data suggest that P. invadens originates in warm offshore surface waters that are transported to the coast during strong storms. Disease outbreaks were associated with passing hurricanes in 2 yr (2010 and 2011). In 2012, a disease outbreak occurred in the absence of a storm and following a strong positive anomaly in winter sea temperature, suggesting survival of the pathogen from 2011. In 2014, a disease outbreak occurred in association with a strong storm (nor'easter) that was not a hurricane. Our findings indicate that changing environmental conditions of increasing minimum sea temperatures and increasing intensity of storms may be altering the dynamics of this disease. These inferences remain equivocal, however, indicating the need for high-resolution dynamical modeling of the physical environment and rapid identification of P. invadens in the environment using genetic tools.

We evaluate the mean circulation patterns, water mass distributions, and tropical dynamics of the North and Equatorial Pacific Ocean based on a suite of global ocean-sea ice simulations driven by the CORE-II atmospheric forcing from 1963-2007. The first three moments (mean, standard deviation and skewness) of sea surface height and surface temperature variability are assessed against observations. Large discrepancies are found in the variance and skewness of sea surface height and in the skewness of sea surface temperature. Comparing with the observation, most models underestimate the Kuroshio transport in the Asian Marginal seas due to the missing influence of the unresolved western boundary current and mesoscale eddies. In terms of the Mixed Layer Depths (MLDs) in the North Pacific, the two observed maxima associated with Subtropical Mode Water and Central Mode Water formation coalesce into a large pool of deep MLDs in all participating models, but another local maximum associated with the formation of Eastern Subtropical Mode Water can be found in all models with different magnitudes. The main model bias of deep MLDs results from excessive Subtropical Mode Water formation due to inaccurate representation of the Kuroshio separation and of the associated excessively warm and salty Kuroshio water. Further water mass analysis shows that the North Pacific Intermediate Water can penetrate southward in most models, but its distribution greatly varies among models depending not only on grid resolution and vertical coordinate but also on the model dynamics. All simulations show overall similar large scale tropical current system, but with differences in the structures of the Equatorial Undercurrent. We also confirm the key role of the meridional gradient of the wind stress curl in driving the equatorial transport, leading to a generally weak North Equatorial Counter Current in all models due to inaccurate CORE-II equatorial wind fields. Most models show a larger interior transport of Pacific subtropical cells than the observation due to the overestimated transport in the Northern Hemisphere likely resulting from the deep pycnocline. (C) 2016 Elsevier Ltd. All rights reserved.