Search

The growing push for open data resulted in an abundance of data for coastal researchers, which can lead to problems for individual researchers related to data discoverability. One solution is to explicitly develop services for coastal researchers to help curate data for discovery, hosting discussions around reuse, community building, and finding collaborators. To develop the idea of a coastal data curation service, we investigate aspects of the UNESCO International Coastal Atlas Network member sites that could be used to build a curation service. We develop a minimal example of a coastal data curation service, deploy this as a website, and describe the next steps to move beyond the prototype phase. We envision a coastal data curation service as a way to cultivate a community focused on coastal data discovery and reuse.

Classifying images using supervised machine learning (ML) relies on labeled training data—classes or text descriptions, for example, associated with each image. Data‐driven models are only as good as the data used for training, and this points to the importance of high‐quality labeled data for developing a ML model that has predictive skill. Labeling data is typically a time‐consuming, manual process. Here, we investigate the process of labeling data, with a specific focus on coastal aerial imagery captured in the wake of hurricanes that affected the Atlantic and Gulf Coasts of the United States. The imagery data set is a rich observational record of storm impacts and coastal change, but the imagery requires labeling to render that information accessible. We created an online interface that served labelers a stream of images and a fixed set of questions. A total of 1,600 images were labeled by at least two or as many as seven coastal scientists. We used the resulting data set to investigate interrater agreement: the extent to which labelers labeled each image similarly. Interrater agreement scores, assessed with percent agreement and Krippendorff's alpha, are higher when the questions posed to labelers are relatively simple, when the labelers are provided with a user manual, and when images are smaller. Experiments in interrater agreement point toward the benefit of multiple labelers for understanding the uncertainty in labeling data for machine learning research.

After decades of study and significant data collection of time-varying swash on sandy beaches, there is no single deterministic prediction scheme for wave runup that eliminates prediction error – even bespoke, locally tuned predictors present scatter when compared to observations. Scatter in runup prediction is meaningful and can be used to create probabilistic predictions of runup for a given wave climate and beach slope. This contribution demonstrates this using a data-driven Gaussian process predictor; a probabilistic machine-learning technique. The runup predictor is developed using 1 year of hourly wave runup data (8328 observations) collected by a fixed lidar at Narrabeen Beach, Sydney, Australia. The Gaussian process predictor accurately predicts hourly wave runup elevation when tested on unseen data with a root-mean-squared error of 0.18 m and bias of 0.02 m. The uncertainty estimates output from the probabilistic GP predictor are then used practically in a deterministic numerical model of coastal dune erosion, which relies on a parameterization of wave runup, to generate ensemble predictions. When applied to a dataset of dune erosion caused by a storm event that impacted Narrabeen Beach in 2011, the ensemble approach reproduced ∼85 % of the observed variability in dune erosion along the 3.5 km beach and provided clear uncertainty estimates around these predictions. This work demonstrates how data-driven methods can be used with traditional deterministic models to develop ensemble predictions that provide more information and greater forecasting skill when compared to a single model using a deterministic parameterization – an idea that could be applied more generally to other numerical models of geomorphic systems.

We performed a series of laboratory experiments to study the growth and development of wave-generated ripples under constant and varying conditions. There are two aspects of our study. First, under time varying wave conditions, we measure the rate of ripple adjustment and the equilibrium mean ripple wavelength before and after the change in forcing. Our experiments with increasing wave forcing (i.e., increasing wave amplitude or wave period) lead to faster ripple growth and larger final wavelengths. Decreasing wave forcing leads to shorter ripple wavelengths and hysteresis. Existing predictors provide a reasonable description of both the final equilibrium spacing and of the spacing evolution after wave conditions are changed. The presence of defects (crestline irregularities) impacts the final equilibrium ripple wavelength. Second, we investigate the role of initial perturbation on ripple growth and geometry. Larger initial perturbations in the bed lead to faster growth of ripples and larger initial ripple wavelength. Our results suggest that irregularities at the pattern scale, such as defects and initial bed perturbations, are likely to impact ripple evolution and can be a control factor on hysteresis.

Despite the importance of coastal dunes to many low-lying coastal communities and ecosystems, our understanding of how both climatic and anthropogenic pressures affect foredune evolution on time scales of years to decades is relatively poor. However, recently developed coupled numerical modeling tools have allowed for the exploration of the erosion and growth of coastal foredunes on time scales of hours to years. For example, Windsurf is a new process-based numerical modeling system (Cohn et al. 2019a) that simulates the evolution of dune-backed sandy coastal systems in response to wave, wind, and water level forcings. CReST, developed as a front-end interface to Windsurf, aims to add the ability to incorporate beach nourishment and dune construction, beach and dune grading, dune grass planting scenarios, dune grass removal, and the presence of hard engineering structures into the model framework to better account for the complex dynamics of managed coastlines. Initial model sensitivity tests suggest that the model provides a flexible framework to investigate the complex interactions between beaches and dunes for a variety of exploratory and applied applications.

Coastal foredunes provide the first line of defense against rising sea levels and storm surge and for this reason there is increasing interest in understanding and modeling foredune formation and post-storm recovery. However, there is limited observational data available to provide empirical guidance for the development of model parameterizations. To provide guidance for improved representation of dune grass growth in models, we conducted a two-year multi-species transplant experiment on Hog Island, VA, U.S.A. and measured the dependence of plant growth on elevation and distance from the shoreline, as well as the relationship between plant growth and sand accumulation. We tracked total leaf growth (length) and aboveground leaf length and found that Ammophila breviligulata (American beachgrass) and Uniola paniculata (sea oats) grew more than Spartina patens (saltmeadow cordgrass) by a factor of 15% (though not statistically significant) and 45%, respectively. Our results also suggest a range of basal/frontal area ratios (an important model parameter) from 0.5-1 and a strong correlation between transplant growth and total sand deposition for all species at the scale of two years, but not over shorter temporal scales. Distance from the shoreline and elevation had no effect on transplant growth rate but did have an effect on survival. Based on transplant survival, the seaward limit of vegetation at the end of the experiment was approximately 30 m from the MHWL and at an elevation of 1.43 m, corresponding to inundation less than 7.5% of the time according to total water level calculations. Results from this experiment provide evidence for the dune-building capacity of all three species, suggesting S. patens is not a maintainer species, as previously thought, but rather a moderate dune builder even though its growth is less stimulated by sand deposition than A. breviligulata and U. paniculata.

Deliberate, real‐time human interventions into geomorphic processes are a phenomenon that no off‐the‐shelf numerical model of morphodynamics is built to capture. We suggest that active, responsive human processes that affect sediment transport during major storm events be included in evolving efforts to model geomorphic change.

Streaming social media provides a real-time glimpse of extreme weather impacts. However, the volume of streaming data makes mining information a challenge for emergency managers, policy makers, and disciplinary scientists. Here we explore the effectiveness of data learned approaches to mine and filter information from streaming social media data from Hurricane Irma's landfall in Florida, USA. We use 54,383 Twitter messages (out of 784K geolocated messages) from 16,598 users from Sept. 10 - 12, 2017 to develop 4 independent models to filter data for relevance: 1) a geospatial model based on forcing conditions at the place and time of each tweet, 2) an image classification model for tweets that include images, 3) a user model to predict the reliability of the tweeter, and 4) a text model to determine if the text is related to Hurricane Irma. All four models are independently tested, and can be combined to quickly filter and visualize tweets based on user-defined thresholds for each submodel. We envision that this type of filtering and visualization routine can be useful as a base model for data capture from noisy sources such as Twitter. The data can then be subsequently used by policy makers, environmental managers, emergency managers, and domain scientists interested in finding tweets with specific attributes to use during different stages of the disaster (e.g., preparedness, response, and recovery), or for detailed research., 46 pages, 11 Figures

Barrier islands and spits tend to migrate landward in response to sea-level rise through the storm-driven process of overwash, but overwash flux depends on the height of the frontal dunes. Here, we explore this fundamental linkage between dune dynamics and barrier migration using the new model Barrier3D. Our experiments demonstrate that discontinuous barrier retreat is a prevalent behavior that can arise directly from the bistability of foredune height, occurring most likely when the storm return period and characteristic time scale of dune growth are of similar magnitudes. Under conditions of greater storm intensity, discontinuous retreat becomes the dominant behavior of barriers that were previously stable. Alternatively, higher rates of sea-level rise decrease the overall likelihood of discontinuous retreat in favor of continuous transgression. We find that internal dune dynamics, while previously neglected in exploratory barrier modeling, are an essential component of barrier evolution on time scales relevant to coastal management.

Overwash is a key mechanism controlling the flux of sediment from the front of a barrier island to the top and back of an island during a storm event. The process is essential for barrier environments to maintain their height and width relative to sea level. Barrier topography and vegetation – and also road networks and buildings – can direct overwash flow, and thus the shape and size of sedimentary deposits that overwash leaves behind. Controls on overwash deposition have been examined more closely in natural settings than in developed zones. But overwash poses a major hazard to coastal infrastructure, and accurate prediction of storm impacts requires quantitative insight into the dynamics of overwash morphology in built settings. Here, we compare barrier floodplain controls across a range of spatial "fabrics", both natural and built (e.g., sparse to dense vegetation coverage; sparse to dense configurations of roads and buildings), to explore how these fabrics affect scaling relationships for overwash morphology. Integrating empirical measurements from post-storm imagery, trials of an analogue model in a small experimental basin, and results from a numerical toy model, we identify thresholds at which floodplain fabrics cause scaling relationships to change, or "break". Our findings illustrate a continuum in overwash pattern formation between endogenous self-organisation and exogenous forcing templates, and set up further inquiry into the dynamics of flood deposition in built environments.

We use Genetic Programming (GP), a type of Machine Learning (ML) approach, to predict the total and infragravity swash excursion using previously published datasets that have been used extensively in swash prediction studies. Three previously published works with a range of new conditions are added to this dataset to extend the range of measured swash conditions. Using this newly compiled dataset we demonstrate that a ML approach can reduce the prediction errors compared to well-established parameterizations and therefore it contributes to the error in coastal hazards assessment (e.g. coastal inundation). Predictors obtained using GP can also be physically sound and replicate the functionality and dependencies of previous published formulas. Overall, we show that ML techniques are capable of both improving predictability (compared to classical regression approaches) and providing physical insight into coastal processes.

The Geological Society of America Bulletin was an early home for quantitative geomorphology research. Although geomorphology papers are not uniformly the highest cited papers in the Bulletin, many show ‘delayed recognition’—they garner only few citations directly after publication, before suddenly being widely and numerously cited (sometimes decades after publication). I focus here on (1) algorithmically detecting cases of delayed recognition in geomorphology literature from the Bulletin and (2) providing insight into why delayed recognition occurred for these papers.

International audience; Beaches around the world continuously adjust to daily and seasonal changes in wave and tide conditions, which are themselves changing over longer timescales. Different approaches to predict multi-year shoreline evolution have been implemented; however, robust and reliable predictions of shoreline evolution are still problematic even in short-term scenarios (shorter than decadal). Here we show results of a modelling competition, where 19 numerical models (a mix of established shoreline models and machine learning techniques) were tested using data collected for tairua beach, new Zealand with 18 years of daily averaged alongshore shoreline position and beach rotation (orientation) data obtained from a camera system. in general, traditional shoreline models and machine learning techniques were able to reproduce shoreline changes during the calibration period (1999-2014) for normal conditions but some of the model struggled to predict extreme and fast oscillations. During the forecast period (unseen data, 2014-2017), both approaches showed a decrease in models' capability to predict the shoreline position. this was more evident for some of the machine learning algorithms. A model ensemble performed better than individual models and enables assessment of uncertainties in model architecture. Research-coordinated approaches (e.g., modelling competitions) can fuel advances in predictive capabilities and provide a forum for the discussion about the advantages/disadvantages of available models. Quantitative prediction of beach erosion and recovery is essential to planning resilient coastal communities with robust strategies to adapt to erosion hazards. Over the last decades, research efforts to understand and predict shoreline evolution have intensified as coastal erosion is likely to be exacerbated by climatic changes 1-5. The social and economic burden of changes in shoreline position are vast, which has inspired development of a growing variety of models based on different approaches and techniques; yet current models can fail (e.g. predicting erosion in accreting conditions). The challenge for shoreline models is, therefore, to provide reliable, robust and realistic predictions of change, with a reasonable computational cost, applicability to a broad variety of systems, and some quantifiable assessment of the uncertainties.

Wave-generated ripples are macroscopic roughness elements that influence fluid flow and sediment transport. For a major group of ripples (orbital ripples), morphology (height and wavelength) is set by the wave conditions. In natural conditions, where wave forcing is highly variable, ripple morphology is frequently changing. We investigate the rate of morphological change after changes in wave conditions (i.e., wave height and period), using laboratory experiments with a step-change in wave forcing. The adjustment time of ripple morphology to new conditions –– the hysteresis time scale (hereafter referred to simply as hysteresis) –– is proportional to changes in wave orbital diameter, and the coefficient of proportionality differs for decreasing and increasing orbital diameter. When the Shields parameter is lower than a threshold (0.043), there is no change in ripple morphology (or changes are extremely slow). In addition, we find the presence of defects (irregularities from straight parallel crestlines) reduces the hysteresis time scale. The dependence of hysteresis on defect density implies that a larger density of defects results in faster ripple adjustment, confirming previous theoretical and numerical model results.

Despite decades of regulatory efforts in the United States to decrease vulnerability in developed coastal zones, exposure of residential assets to hurricane damage is increasing — even in places where hurricanes have struck before. Comparing plan-view footprints of individual residential buildings before and long after major hurricane strikes, we find a systematic pattern of ‘building back bigger’ among renovated and new properties. US coastlines that are exposed to hurricanes are subject to a myriad of regulations regarding building and rebuilding of structures, yet satellite imagery shows that the footprint of residential buildings increases after hurricane events for both new and renovated structures. Such an effect poses a challenge for vulnerable coastlines to build resilience in the face of growing hazards and houses.

Encroachment of woody plants into grasslands has occurred worldwide and includes coastal ecosystems. This conversion process is mediated by seed dispersal patterns, environmental filtering, and biotic interactions. As spatiotemporally heterogeneous, harsh environments, barrier islands present a unique set of challenges for dispersal and establishment. Environmental conditions act as filters on dispersed seeds, thereby influencing encroachment and distribution patterns. Seldom have patterns of propagule dispersal been considered in the context of woody encroachment. We quantified dispersal and post‐dispersal processes of an encroaching woody population of Morella cerifera relative to directional rate of encroachment and observed distribution patterns on an Atlantic coastal barrier island with strong environmental filtering. We analyzed historic foredune elevation as a proxy for reduced interior environmental stress. The dispersal kernel was leptokurtic, a common characteristic of expanding populations, but rate of encroachment has slowed since 2005. Expansion pattern was related to foredune elevation, which limits encroachment below a threshold elevation. This difference between dispersal kernel behavior and encroachment rate is due to limited availability of suitable habitat for Morella and temporal variability in chlorides during the time of germination. Our results demonstrate that processes mediating seeds and seedling success must be accounted for to better understand establishment patterns of encroaching woody plants.

International audience; Predictions of shoreline change are of great societal importance, but models tend to be tested and tuned for the specific site of interest. To overcome this issue and test the ability of numerical models to simulate shoreline change over the medium scale (order of years) we have organized a non-competitive competition where participants were given data to train their model (1999-2014) and data to predict seasonal to inter-annual future changes (2014-2017). Participants were shown the observed shoreline changes only after submission of their modelling results. Overall, 19 numerical models were tested, the vast majority falling in the broad categories of "hybrid models" or "machine learning". Models were able to reproduce the mean characteristics of shoreline change but often failed to reproduce the observed rapid changes induced by storms.

Coastal dunes arise from feedbacks between vegetation and sediment supply. Species-specific differences in plant functional morphology affect sand capture and dune shape. In this study, we build on research showing a relationship between dune grass species and dune geomorphology on the US central Atlantic Coast. This study seeks to determine the ways in which four co-occurring dune grass species (Ammophila breviligulata, Panicum amarum, Spartina patens, Uniola paniculata) differ in their functional morphology and sand accretion. We surveyed the biogeography, functional morphology, and associated change in sand elevation of the four dune grass species along a 320-kilometer distance across the Outer Banks. We found that A. breviligulata had dense and clumped shoots, which correlated with the greatest sand accretion. Coupled with fast lateral spread, it tends to build tall and wide foredunes. Uniola paniculata had fewer but taller shoots and was associated with ~42% lower sand accretion. Coupled with slow lateral spread, it tends to build steeper and narrower dunes. Panicum amarum had similar shoot densities and associated sand accretion to U. paniculata despite its shorter shoots, suggesting that shoot density is more important than morphology. Finally, we hypothesize, given the distributions of the grass species, that foredunes may be taller and wider and have better coastal protection properties in the north where A. breviligulata is dominant. If under a warming climate A. breviligulata experiences a range shift to the north, as appears to be occurring with U. paniculata, changes in grass dominance and foredune morphology could make for more vulnerable coastlines.

For years, publication in a peer-reviewed scientific journal has served as the official announcement that a piece of research has entered the accepted body of scientific knowledge. However, chances are good that the data, method innovations, and preliminary conclusions contained in a scientific paper have been circulating among the researchers’ colleagues for a year or more before the paper is officially published. By the time a journal issue goes to the printing press, the research may have made significant progress beyond what the published papers document. Preprints—published, non-peer-reviewed scholarly papers that often precede publication in a peer-reviewed journal—have been a part of science since at least the 1960s [Cobb, 2017]. ArXiv, arguably the most prominent preprint system, has served the physics, mathematics, and computer science communities for more than 25 years.

Beach nourishment, a method for mitigating coastal storm damage or chronic erosion by deliberately replacing sand on an eroded beach, has been the leading form of coastal protection in the United States for four decades. However, investment in hazard protection can have the unintended consequence of encouraging development in places especially vulnerable to damage. In a comprehensive, parcel‐scale analysis of all shorefront single‐family homes in the state of Florida, we find that houses in nourishing zones are significantly larger and more numerous than in non‐nourishing zones. The predominance of larger homes in nourishing zones suggests a positive feedback between nourishment and development that is compounding coastal risk in zones already characterized by high vulnerability.

On sandy coastlines, foredunes provide protection from coastal storms, potentially sheltering low areas—including human habitat—from elevated water level and wave erosion. In this contribution we develop and explore a one-dimensional model for coastal dune height based on an impulsive differential equation. In the model, coastal foredunes continuously grow in a logistic manner as the result of a biophysical feedback and they are destroyed by recurrent storm events that are discrete in time. Modeled dunes can be in one of two states: a high “resistant-dune” state or a low “overwash-flat” state. The number of stable states (equilibrium dune heights) depends on the value of two parameters, the nondimensional storm frequency (the ratio of storm frequency to the intrinsic growth rate of dunes) and nondimensional storm magnitude (the ratio of total water level during storms to the maximum theoretical dune height). Three regions of phase space exist (1) when nondimensional storm frequency is small, a single high resistant-dune attracting state exists; (2) when both the nondimensional storm frequency and magnitude are large, there is a single overwash-flat attracting state; (3) within a defined region of phase space model dunes exhibit bistable behavior—both the resistant-dune and the low overwash-flat states are stable. Comparisons to observational studies suggest that there is evidence for each state to exist independently, the coexistence of both states (i.e., segments of barrier islands consisting of overwash-flats and segments of islands having large dunes that resist erosion by storms), as well as transitions between states.

Numerical models rely on the parameterization of processes that often lack a deterministic description. In this contribution we demonstrate the applicability of using machine learning, a class of optimization tools from the discipline of computer science, to develop parameterizations when extensive data sets exist. We develop a new predictor for near-bed suspended sediment reference concentration under unbroken waves using genetic programming, a machine learning technique. We demonstrate that this newly developed parameterization performs as well or better than existing empirical predictors, depending on the chosen error metric. We add this new predictor into an established model for inner-shelf sorted bedforms. Additionally we incorporate a previously reported machine-learning-derived predictor for oscillatory flow ripples into the sorted bedform model. This new "hybrid" sorted bedform model, whereby machine learning components are integrated into a numerical model, demonstrates a method of incorporating observational data (filtered through a machine learning algorithm) directly into a numerical model. Results suggest that the new hybrid model is able to capture dynamics previously absent from the model – specifically, two observed pattern modes of sorted bedforms. Lastly we discuss the challenge of integrating data-driven components into morphodynamic models and the future of hybrid modeling.

A range of computer science methods termed machine learning (ML) enables the extraction of insight and quantitative relationships from multidimensional datasets. Here, we review the use of ML on supervised regression tasks in studies of coastal morphodynamics and sediment transport. We examine aspects of ‘what’ and ‘why’, such as ‘what’ science problems ML tools have been used to address, ‘what’ was learned when using ML, and ‘why’ authors used ML methods. We find a variety of research questions have been addressed, ranging from small-scale predictions of sediment transport to larger-scale sand bar morphodynamics and coastal overwash on a developed island. We find various reasons justify the use of ML, including maximize predictability, emulation of model components, the need for smooth and continuous nonlinear regression, and explicit inclusion of uncertainty. The expanding use of ML has allowed for an expanding set of questions to be addressed. After reviewing the studies we outline a set of best practices for coastal researchers using machine learning methods. Finally we suggest possible areas for future research, including the use of novel machine learning techniques and exploring open data that is becoming increasingly available.

Accelerated sea level rise and the potential for an increase in frequency of the most intense hurricanes due to climate change threaten the vitality and habitability of barrier islands by lowering their relative elevation and altering frequency of overwash. High-density development may further increase island vulnerability by restricting delivery of overwash to the subaerial island. We analyzed pre-Hurricane Sandy and post-Hurricane Sandy (2012) lidar surveys of the New Jersey coast to assess human influence on barrier overwash, comparing natural environments to two developed environments (commercial and residential) using shore-perpendicular topographic profiles. The volumes of overwash delivered to residential and commercial environments are reduced by 40% and 90%, respectively, of that delivered to natural environments. We use this analysis and an exploratory barrier island evolution model to assess long-term impacts of anthropogenic structures. Simulations suggest that natural barrier islands may persist under a range of likely future sea level rise scenarios (7–13 mm/yr), whereas developed barrier islands will have a long-term tendency toward drowning.

Because barriers are low-lying and dynamic landforms, they are especially sensitive to changing environmental conditions. The continued existence of barriers will depend on the degree to which these landforms can maintain elevation above sea level while also migrating landward. We are increasingly learning that ecomorphodynamic interactions (i.e., interactions between morphology, fluid dynamics, and/or sediment transport with biological processes) as well as couplings between barrier and back-barrier environments play a critical role in determining how barrier systems will evolve as sea level rises, storm intensity increases, and the species composition of coastal vegetation changes in the future. For example, the effectiveness of storms in building elevation and moving a barrier landward is determined, in large part, by the morphology of the coastal foredune (i.e., the seaward-most dune), which is itself a product of couplings between vegetation and sediment transport processes. The cross-shore and alongshore shape of coastal foredunes, in the presence of shoreline erosion or shoreline accretion, is influenced by the distance from the shoreline that vegetation can grow, the rates of lateral and vertical vegetation growth of dune-building vegetation, as well as the dependence of vegetation growth on dune slope. In addition, as storm frequency increases relative to the rate at which dunes can grow, dunes, and therefore local barrier elevation, may become bistable, tending to be in either a high dune/barrier or low dune/barrier state. When dunes are low, storms can effectively increase barrier elevation and move a barrier landward over time leading also to the potential for increased connectivity to back-barrier marshes, which are vulnerable to drowning as sea level rises. In this case, sand delivered to back-barrier marshes can, for a time, allow back-barrier marshes to persist under conditions in which they would otherwise disappear, thereby benefitting the entire barrier-marsh system. Here we provide a synthesis of model results—tested against observations—that demonstrate these findings, illustrating the importance of feedbacks between vegetative and sediment transport processes, and couplings between landscape units, in influencing the future evolution of barrier-marsh systems.

We use a machine learning approach to seek an accurate, physically sound predictor, to estimate the mean velocity for open-channel flow when submerged arrays of rigid cylinders (model vegetation) are present. A genetic programming routine is used to find a robust relationship between relevant properties of the model vegetation and flow parameters. We use published data from laboratory experiments covering a broad range of conditions to obtain an equation that matches the performance of other predictors from recent literature in terms of accuracy, while showing a less complex structure. We also investigate how different criteria for data selection, as well as the size of the data set used to train the algorithm, influences the accuracy of the resulting predictors. Our results show that a proper use of Machine-Learning techniques does not only provide empirical correlations, but can yield physically sound models as representative of the physical processes involved. We provide a clear, thorough example of the application of GP, its advantages and shortcomings, to encourage the use of data-driven techniques as part of the data analysis process, and to address common misconceptions of machine learning as simple correlation techniques or physically senseless statistical analysis.

We use a machine learning approach based on genetic programming to predict noncohesive particle settling velocity. The genetic programming routine is coupled to a novel selection algorithm that determines training data from a collected database of published experiments (985 measurements). While varying the training data set size and retaining an invariant validation set we perform multiple iterations of genetic programming to determine the least data needed to train the algorithm. This method retains a maximum quantity of data for testing against published predictors. The machine learning predictor for settling velocity performs better than two common predictors in the literature and indicates that particle settling velocity is a nonlinear function of all the provided independent variables: nominal diameter of the settling particle, kinematic viscosity of the fluid, and submerged specific gravity of the particle.

We have developed an exploratory model of plan view, millennial-scale headland and bay evolution on rocky coastlines. Cross-shore coastline relief, or amplitude, arises from alongshore differences in sea cliff lithology, where durable, erosion-resistant rocks protrude seaward as headlands and weaker rocks retreat landward as bays. The model is built around two concurrent negative feedbacks that control headland amplitude: (1) wave energy convergence and divergence at headlands and bays, respectively, that increases in intensity as cross-shore amplitude grows and (2) the combined processes of beach sediment production by sea cliff erosion, distribution of sediment to bays by waves, and beach accumulation that buffers sea cliffs from wave attack and limits further sea cliff retreat. Paired with the coastline relief model is a numerical wave transformation model that explores how wave energy is distributed along an embayed coastline. The two models are linked through genetic programming, a machine learning technique that parses wave model results into a tractable input for the coastline model. Using a pool of 4800 wave model simulations, genetic programming yields a function that relates breaking wave power density to cross-shore headland amplitude, offshore wave height, approach angle, and period. The goal of the coastline model is to make simple, but fundamental, scaling arguments on how different variables (such as sea cliff height and composition) affect the equilibrium cross-shore relief of headland and bays. The model's generality highlights the key feedbacks involved in coastline evolution and allows its equations (and model behaviors) to be easily modified by future users.

We integrate published data sets of field and laboratory experiments of wave ripples and use genetic programming, a machine learning paradigm, in an attempt to develop a universal equilibrium predictor for ripple wavelength, height, and steepness. We train our genetic programming algorithm with data selected using a maximum dissimilarity selection routine. Thanks to this selection algorithm; we use less data to train the genetic programming software, allowing more data to be used as testing (i.e., to compare our predictor vs. common prediction schemes). Our resulting predictor is smooth and physically meaningful, different from other machine learning derived results. Furthermore our predictor incorporates wave orbital ripples that were previously excluded from empirical prediction schemes, notably ripples in coarse sediment and long wavelength, low height ripples (‘hummocks’). This new predictor shows ripple length to be a weakly nonlinear function of both bottom orbital excursion and grain size. Ripple height and steepness are both nonlinear functions of grain size and predicted ripple length (i.e., bottom orbital excursion and grain size). We test this new prediction scheme against common (and recent) predictors and the new predictors yield a lower normalized root mean squared error using the testing data. This study further demonstrates the applicability of machine learning techniques to successfully develop well performing predictors if data sets are large in size, extensive in scope, multidimensional, and nonlinear.

Many studies focus on the emergence and development of rhythmic landscape patterns. In this contribution we explore the different behaviors found as patterns evolve; the trajectories that patterns exhibit as they transit from infinitesimal-amplitude perturbation to a statistically steady state (or in some cases to continued statistical evolution). The variety of behaviors observed, either through field and laboratory experiments or numerical modeling, can be reduced to four classes: (a) simple stabilization where predictions based on the initial growth of small perturbations corresponds with the characteristics of patterns observed in nature; (b) significant pattern coarsening en route to saturated wavelength, where non-linear interactions between finite-amplitude pattern elements change the geometric properties of a pattern as it approaches steady-state; (c) perpetual coarsening where the wavelength associated with the emerging pattern continues to grow over time and is only limited by physical boundaries or external constrains; (d) slow evolution toward a different attractor, a novel behavior observed in numerical modeling that involves profound temporal changes in pattern characteristics. Within these classes we also observe generalizable non-linear behaviors: dependence on initial conditions, the emergence of pattern-scale variables such as pattern defects, and the presence of multiple stable states. Predicting the shape of patterns to come remains a challenge – one that we suggest requires a range of modeling approaches to address both initial instabilities and the emergent properties of evolving patterns, which involve disparate forms of non-linear interactions. Consideration of generic system behaviors at the pattern scale could enhance future pattern formation studies, facilitating appropriate pairings of analysis approaches and pattern-evolution modes. Copyright © 2013 John Wiley & Sons, Ltd.

Are you looking to communicate science more broadly? Start by editing articles on the Internet's most popular general reference work.