Visioning channel evolution for river management: Toward a functional decision support framework

Improvements in simulating and communicating the evolutionary trajectory of river morphology in response to environmental forcing over multi‐decadal timeframes would foreshadow the development of “foresight competency” in river management, whereby resource managers could strategically plan toward the most preferred of several plausible futures. Of the six steps in foresight competency, visioning, which involves translating scientific forecasts into a format useable by resource managers via a user‐friendly and interactive decision support tool that supports transparent decision‐making, is the least well developed. The approach requires converting forecasting model outputs into metrics of channel evolution that highlight transitions either within or between channel morphology states. Here, seven process‐based state transition metrics are proposed covering channel planform, morphological stability, corridor belt width, floodplain connectivity, bank erosion rate, bedform habitat diversity, and ecohydraulic diversity. To aid decision support, the metrics are converted into graphical indicators that are intuitive for management use and assembled into several prototype dashboard‐style graphical user interfaces designed to facilitate interactivity. A proof‐of‐concept illustration is provided and priorities in development toward a fully operational decision support tool are discussed. Such developments are critical in ensuring the practical relevance of geomorphology.

ability to simulate and communicate anticipated channel evolutionary responses to changes in environmental forcing factors such as climate, land uses, and water resource management activities over management timeframes of decades to a century or more.Such improvements would foreshadow the development of foresight competency (e.g., Hines et al., 2017;Hines & Bishop, 2006) into the field of river management, allowing managers to assemble several contingent predictions (i.e., predictions subject to certain assumed changes in condition, Wilcock & Iverson, 2003), that allows them to envisage several plausible futures (Voros, 2003) in channel evolution.Consequently, management actions could be taken to avoid certain unlikely or undesirable possible futures (e.g., severe channel instability), instead instigating strategic decisionmaking directed toward a preferred future condition (e.g., dynamically sustained channel diversity) which may not be the most probable future based on recent trends in the boundary conditions (Figure 1).
Best practice in achieving foresight competency is argued to consist of six steps involving framing, scanning, forecasting, visioning, planning, and acting competencies (Hines et al., 2017;Hines & Bishop, 2006;Hines & Zindato, 2016).This model can be adapted for use in strategic river management (Figure 2).The two initial steps-framing the project's objectives and baseline conditions and scanning for information on past changes and likely future changes in environmental drivers and emerging issues-are reasonably well developed: there are frequently four fundamental questions in modern river management (after Montgomery et al., 1995) involving how the landscape worked in the past, what its history has been, what its current condition is, and what are the desired future states.In addition, numerous studies have now retrospectively reconstructed channel evolutionary trajectories over the centennial past using multiple historical data sources (review in Downs & Piégay, 2019).The final two steps of the foresight process are likewise well developed.Systematic guidance on process-based river management planning (or designing, Hines et al., 2017) is provided in several textbooks (e.g., Brierley & Fryirs, 2005, 2008;Downs & Gregory, 2004;Roni & Beechie, 2012) and integrated, catchment-based policies exist at national and supra-national scales (e.g., the EU Water Framework Directive, EC, 2000).Acting (or adapting) to implement management plans is generally subject to well-developed regulatory overview by national environment agencies.Far less well developed are aspects that synthesize baseline information and various possible futures via scenario-based forecasting (or futuring) of river channel evolution, and visioning the modeled outcomes to maximize the chance of aligning channel evolution with more favorable future states in terms of factors such as channel stability, flood risk, and habitat provision, and away from other plausible but undesirable futures.
Modeling river channel evolution in response to changes in environmental drivers is subject to multiple challenges (Lotsari et al., 2015, p. 26).Scale is a paramount concern-high complexity models focused on short-term, high-resolution simulation of hydraulics and channel morphology are generally too computationally complex to work over management time and space scales, but "reduced complexity models" (RCMs) designed to simulate landscape evolution over centuries to millennia may be ill-suited for managers who require forecasts with locational and temporal specificity.There is some promise in combining RCMs with hydraulic drivers to produce greater reach-scale specificity (e.g., Coulthard et al., 2013;Poeppl et al., 2019;Ramirez et al., 2020;Ziliani & Surian, 2012), but there remains a "desperate" need (Lane et al., 2007) for tools that predict the medium-term response of channel morphology to changing sediment delivery dynamics.One prospect is "intermediate complexity models" that simulate decadal scale channel evolution using either simplified equations to represent processes such as channel

Research Impact Statement
We outline a decision support framework for integrating decadal scale modeling of channel morphology evolution to benefit strategic planning of river management and restoration.
Beyond such advances in channel evolutionary forecasting, foresight competency requires visioning the outcome via decision support systems (DSS) that allow managers to make choices informed by the predicted evolution of channel morphology.Whereas predictive models produce applicable information, only with the addition of DSS is the contribution truly applied.DSS are vital in realizing the transformative potential of predictive models.Defined as tools that help decision makers solve unstructured or semistructured decision problems using an easily understood, user-friendly, and interactive computer system capable of incorporating the decision makers own insights (Gorry & Scott Morton, 1971;Matthies et al., 2007), DSS add transparency to decision-making processes (McIntosh et al., 2011).DSS require components to manage data, one or more predictive models to provide forecasts, and an interface that facilitates interactivity (Alter, 1980).This is particularly important in strategic planning where scenario analysis and simulations are critical and form the basis of interactions between the systems analyst and the environmental manager (Matthies et al., 2007).In this regard, while the challenge of knowledge transfer in river restoration is recognized (e.g., Biron et al., 2018), examples of DSS in river management are very much in their infancy.Examples include using structured decision trees in place of a predictive model to communicate aspects of the current biophysical condition in support of river management (Boitsidis et al., 2006;Gurnell et al., 2020;Marttunen et al., 2019;Shuker et al., 2012) and approaches to related riverine concerns such as biological integrity (Beck et al., 2019), non-point pollution (Dorner et al., 2007), fish population modeling (Freeman et al., 2013), and engineering priorities (Schultze et al., 2019).But almost no DSS support channel evolution.Nascent developments toward this end include an expert opinion-driven approach for advising managers of the likelihood of overbank inundation, channel avulsion, scour hole development and bank and bench stability (Glendining & Pollino, 2012), and the use of hydraulic geometry relationships to provide management-relevant indicators of morphology change according to different water management scenarios (Van de Waal & Rowntree, 2010).
With forecasting abilities in river management likely to improve as intermediate complexity numerical models better simulate river channel evolution, this research instead focuses on the almost entirely neglected visioning component by developing a decision support tool that accommodates channel evolution over planning timeframes.The tool uses inputs from both "ends" of the foresight process (Figure 2), forecasting state transition changes in river morphology using a hydrodynamic numerical model applied over decadal timeframes and utilizing knowledge about the strategic planning requirements of environmental managers involved in river management.The nexus between these starting points is developed into a framework for decision support containing, first, a set of channel evolution metrics that parameterize channel evolution F I G U R E 2 A best practice foresight competency model for strategic foresight in river basin management (developed from Hines & Zindato, 2016;Hines et al., 2017).Critically, the underdeveloped "visioning" component includes the process of decision support, the focus herein.
from a numerical model.The metrics are translated into decision support indicators suitable for strategic application of channel evolution in river management.The indicators are communicated via a dashboard-style graphical user interface to portray threshold-based risks related to common concerns in riverine planning, engineering, restoration, and conservation.A case study provides illustration of the overall proofof-concept framework.Finally, we consider the prospects for this framework to add a suitable visioning component of foresight competency regarding channel evolution into river management.

| Forecasting channel evolution for strategic visioning
Channel evolution is contingent on the channel's history and position within the watershed, with evolution occurring within a set of imposed boundary conditions (Fryirs & Brierley, 2013) that prescribe a fluvial processes "domain" (Montgomery, 1999) of "normal" or expected behaviors.Over decades-to-century timeframes, boundary conditions are most likely to change progressively from changes in climate, land use, and resource management, although step changes may result, for instance, following dam building or the connection of urban drainage.In general, progressive change in boundary conditions implies progressive morphological evolution but the existence of extrinsic and intrinsic thresholds (Schumm, 1973) in geomorphic "states" opens the prospect of abrupt morphological response across thresholds and into a new state.As such, foresight in channel evolution needs to be capable of describing within-state and between-state transitions and, ideally, to explain such transitions.This implies the use of a "state transition model" basis for characterizing evolution (review in Phillips & Van Dyke, 2017).
Channel evolution forecasting to date has focused on one of two approaches to state transition, based either on pattern or process.Patternbased approaches link repeatedly observed and generalizable sequences of channel morphology, typified by changes in the channel's widthdepth ratio and bed elevation, to changes in causal mechanisms influencing upstream to downstream sediment transport processes and the likelihood for mass failure of the channel banks (Schumm et al., 1984;Simon & Hupp, 1986;Simon, 1989; see also Brierley et al., 2002;Fryirs & Brierley, 2000 for a related approach).Such "channel evolution models" (CEMs) have been widely applied but their empirical generality versus the "importance of place-based contingencies in shaping adjustment trajectories" (Van Dyke, 2013, p. 763) means that numerous exceptions, shortcuts, and resetting of the evolutionary sequence have been noted (e.g., Cluer & Thorne, 2014;Doyle et al., 2002;Hawley et al., 2012;Major et al., 2019;Thompson et al., 2016;Toone et al., 2012).
Process-based approaches to state transitions have generally used one of two methods.The first is regime theory, wherein analytical predictions for channel morphology state are derived from governing conditions of sediment flux, streamflow, and channel boundary conditions (Eaton & Millar, 2017).The approach is well suited for use in channel design (e.g., Copeland, 1994;Soar & Thorne, 2001, 2011;Stroth et al., 2017), and recent developments have considered the characterization of river evolution (e.g., Davidson & Eaton, 2018;Eaton & Millar, 2017;Lammers & Bledsoe, 2018), but it ignores the underlying hydraulic "driver" for process transitions.Conversely, physically based models capable of simulating meander migration (e.g., Parker et al., 2011) or hydraulically based models linked to "near-mechanical" processbased equations for change (e.g., Langendoen et al., 2016) are more precise but are computationally demanding and have highly exacting data requirements.Both situations are thus problematic when providing "foresight" (i.e., horizon-scanning) perspectives for river management over planning timeframes and lengthy reaches.
An alternative, intermediate complexity, hybrid approach is utilized here based on a simplified, process-based hydrodynamic and sediment transport CEM linked to seven metrics representing channel state.The one-dimensional forecasting tool, FRAME (Future River Analysis and Management Evaluation, Soar et al., 2023) is derived from the SIAM (Sediment Impact Analysis Methods) approach for characterizing channel morphology change according to imbalances in reach-scale sediment transport (e.g., Biedenharn et al., 2006), and the Potamod model (Cox et al., 2015) that integrates SIAM with the HEC-RAS computer program (Gibson & Little, 2006) to predict vertical changes in channel morphology.FRAME uses simplified cross sections, each representative of an equal channel length, and is driven by annualized flow duration curves to enable simulations over multiple decades.
Currently, FRAME accounts for bed material load only, with wash load assumed to pass entirely through the modeled reach with no influence on morphological response.Bed and bank material gradations are initially input by the user at locations along the main channel, where data are available, as a series of stepped bed and bank material change points.Then, following each model time step, FRAME employs a simple two-layer scheme for mixing the bed material in response to bed erosion and deposition to allow for active layer fining or coarsening over time.
Additional flow and sediment inputs/outputs are simulated by geomorphic rules associated with tributaries, diversion/inflow points, and dike fields.FRAME achieves sediment mass balance on a grain size class basis and conserves mass by carrying excess material or transport capacity downstream in cases where cross-sectional change is inhibited from full adjustment.Numerical stability is achieved using a sub-annual time step that allows morphological changes to evolve gradually via convergence on sediment continuity for prevailing conditions.Calculation of bed material transport includes an optional hiding-exposure factor and FRAME includes the ability to switch to (and from) a cohesive bed scour submodule based on excess specific stream power where cohesive material or bedrock become exposed (or buried) with degradation of (or renewed deposition over) the bed.FRAME currently simulates vertical bed adjustments only, with channel lateral adjustment functionality planned to feature in a future revision.
While development and initial testing of FRAME has focused on the Lower Mississippi River (LMR), the model is not restricted to large rivers (see Cox et al., 2023), although a program of further testing and applications will reveal if there are limitations associated with river scale and related hydrological behavior.Computations and assumptions of FRAME associated with hydrological inputs, hydraulics, sediment dynamics, and morphological adjustments are discussed in detail by Soar et al. (2023).
In scenario modeling of river futures, FRAME outputs the annual evolution of the bed profile and bed material composition together with a suite of hydrologic, hydraulic, and sedimentological data that facilitate the estimation of the channel state metrics (see next section).

| Visioning for decision support: Metrics, indicators, dashboard
While decision support tools for the geomorphology of channel evolution might eventually be nested within a far broader "environmental DSS" that includes the social processes of choice, such as for river rehabilitation (e.g., Reichert et al., 2007), the focus here is solely on the intelligence and design aspects of DSS (McIntosh et al., 2011).Underpinning such technical elements requires a clear conceptual understanding of the constraints and capabilities resulting from the predictive "engine" driving the decision support tool, as follows.
"Visioning" of channel evolution requires more than predicting state transition channel adjustments.It requires that results are translated from their initial scientific formulation into a series of metrics that directly and expediently address the needs of environmental managers involved in strategic planning for river futures.The requirement is to address the three fundamental uses of fluvial geomorphology in river management, namely sustainable land-use planning, hazard avoidance/asset maximization, and conservation management (e.g., river restoration) (Downs & Booth, 2011).In addition, whereas river evolution can be considered as "five-dimensional," related to changes in vertical, lateral, length, and "fractal" (i.e., textural) changes over time, numerical models are limited to predicting change only in a finite number of the channel's "degrees of freedom" or "modes of self-adjustment" (e.g., Hey, 1997;Lane, 1955Lane, , 1957;;Maddock, 1970) according to the model's formulation.For instance, FRAME is currently a one-dimensional hydrodynamic model capable of providing direct and derived insights related to evolutionary changes in flow properties, bed elevations, and the grain size texture of the channel bed.Lateral and length changes cannot be simulated.Thus, channel evolutionary metrics must address the needs of environmental managers while acknowledging the constraints arising from the capabilities of the available forecasting model.Seven prospective metrics that inform such dimensions of evolution are outlined in the next section.
While the metrics provide a fundamental outcome of a model's foresight capabilities, decision support requires that the metrics are subsequently transformed into indicators that visually address concerns typical for those involved in strategic river management.Indicators should ideally be communicated as probabilities of crossing a threshold between current morphological state and projected future state (i.e., state transitions) in the form of positive or negative risk statements related to the river's evolutionary trajectory.Assuming that the decision support tool is run by an analyst (e.g., geomorphologist or engineer) before passing to an environmental manager or decision maker (Rizzoli & Young, 1997), it makes sense to display the indicators via an interactive graphical user interface (GUI) of clear "at a glance" changes that encourage interaction and widespread adoption of the tool.We thus conceive of the GUI as a "heart-rate monitor"-style dashboard that indicates to managers the expectations of changes in time and space in proximity to thresholds in parameter states.Interactivity is also important in decision support, requiring the capacity to simulate future scenarios by changing boundary conditions (e.g., flow regimes, land use, and resources management), and by providing the ability to simulate instream engineering measures through manipulating the channel morphology.
Providing a GUI as an interactive dashboard thus provides an element of "gamification" in foresight competency.
The visioning step thus combines metrics, indicators, and a dashboard to communicate forecast outcomes in a manner meaningful to planners, managers, and engineers.The overall framework for decision support thus forms a meta-model of rules, constraints, models, and theories by which information is gathered, analyzed, and output to the river manager.The current proof-of-concept meta-model for decision support (RUBRIC, RUles-Based morphological Response In river Channels) is outlined in Figure 3.This meta-model is, in effect, a diagnostic procedure for channel assessment to support river management (earlier versions include Sear et al., 1995, their figure 6;and Montgomery & MacDonald, 2002, their figure 2) extended to include communication of risks and user interactivity.

| ME TRI C S OF CHANNEL E VOLUTI ON
Below, we identify seven metrics designed to encompass a comprehensive range of channel state adjustments.In general, the translation of model forecasts into foresight metrics should be automated and analytically based, to eliminate preparatory data processing, and delay the point at which interpretative decisions are required, respectively.By preference, metrics drawn from physically based studies should provide the strongest theoretical justification and responsiveness to simulations and, ideally, metrics should be communicated as likelihoods of state transition changes between current and projected future conditions.Not all metrics will be required in all applications, and not all metrics will show decadal-scale responsiveness in every river type.Metrics identified for this task are outlined in Table 1 and discussed briefly below.

| Metrics related to land-use planning
In land-use planning, the lateral aspects of river evolution are critical because they relate to the land area required to ensure healthy natural functioning of the river and/or to ensure that (for instance) proposed river corridor land uses will not create avoidable river-related risks.Such lateral extents are variously described as the streamway or river corridor or, more vividly, a freedom space that allows "room for the river" (Cals et al., 1998).The management challenge lies in fulfilling the "erodible corridor concept" (ECC), trying "…to create a balance between the environmental benefits derived from allowing the river to migrate freely (within the corridor) and the economic benefits derived from protecting property and infrastructure (outside the corridor)" (Piegay et al., 2005, p. 786).Essentially, the ECC defines the land "recycled" during the foresight period.Various ECC extents can be distinguished (Piegay et al., 2005) ranging from the full width of the active floodplain to a minimum negotiated mobility space that does not discount the prospect of needing erosion control.For strategic foresight, the "functional" corridor reflecting unconstrained channel activity in the next 5-10 decades may be most useful.
Three prospective erodible corridor metrics are identified.First, it is highly informative to know whether the river evolutionary trajectory is likely to cross a threshold in planform type.There are significant implications for river corridor width if a single-thread channel is likely to convert to a multithreaded configuration (or vice versa).Commonly, the risk of changing predominant channel planform type is based on "discriminant" function equations that, for instance, separate meandering and braided planforms using parameters such as discharge, slope, width, and grain size (e.g., Beechie & Imaki, 2014;Bledsoe & Watson, 2001;Ferguson, 1987;Knighton & Nanson, 1993;Leopold & Wolman, 1957; Where: Step-pool(n = 171): Parker , 1976;Richardson & Thorne, 2001).Here, we adopt the equations of Eaton et al. (2010), which determine thresholds between fundamentally stable single-thread channels, unstable multithread channels, and transitional anabranching channels.We focus on the initial shift from single to multithreaded channels because the number of anabranches does not significantly affect the risk associated with such changes.
While the chance of planform shift is generally low, it may become more likely under conditions of increased climate extremes or management interventions.Conversely, a river predicted to be of an alternate type to that observed may indicate a significant problem for managers of channel planform disequilibrium (example in Downs et al., 2013).
Second, especially where channels are less mobile (likely leading to floodplain development close to the channel edge), it would be useful to foresee potential trends in channel morphology, for instance in relation to channel incision or aggradation and prospective riverbank failure.
One such metric is Watson et al.'s (2002) dimensionless stability state diagram, comprising a two-threshold state transition index based on changes in "hydraulic stability" and "bank stability" (Figure 4).As FRAME provides incremental values of changing bed height, hydraulic stability (N h ) is estimated directly by the model, indicating evolutionary trends toward aggradation or degradation.Geotechnical bank stability (N g ) is given as the ratio between existing bank height and the critical bank height (i.e., h/H c ) at the same angle.Simon's (1995) simplified approach to bank (in)stability risk is used here, with critical bank height (H c ) based on Carson and Kirkby (1972) and the bank height (h) derived as a byproduct of the bed elevation change.
Third, for single thread rivers, a "within-state" transition metric providing foresight on meandering activity as it defines the erodible corridor could provide potentially important knowledge regarding land requirements and prospective erosion hazard during management timeframes.
Projected rates of lateral adjustment and implied meander belt width changes can help planners specify the "active erodible corridor" width for environmental and economic benefit.Metric options include classifying relative planform stability (e.g., Church, 2006), empirical characterization of meander migration data based on extensive field data (e.g., Lagasse et al., 2004), a gridded time-weighted locational probability approach (Graf, 2000;Tiegs et al., 2005;Tiegs & Pohl, 2005), or physically based meander development models either simplified (e.g., Lammers & Bledsoe, 2018) or based on "excess velocity" (e.g., Ikeda et al., 1981;Parker et al., 2011).Choice of metric is restricted by the functionality of the supporting forecasting model.Tracking the movement of individual meander bends is beyond the current capability inherent to FRAME, but there is considerable potential to simulate reach-scale sinuosity evolution (essentially, channel lengthening) as a proxy for forecasting the magnitude of lateral migration and floodplain occupancy.A widely reported challenge in developing any model-based meandering metric, though, is the need for calibration to parameterize the riverbank erodibility coefficients (e.g., Micheli & Kirchner, 2002;Micheli et al., 2004; review in Castro-Bolinga & Fox, 2018).

| Metrics related to minimizing hazards or maximizing river-based environmental assets
In hazard assessment and asset maximization, channel evolution metrics are likely to relate to changes in flood risk and bank stability.In flow duration curves rather than to sequences of event hydrographs, a suitable metric would be the annual percentage time that overtopping flows occur, based on the conveyance capacity of the channel at overtopping stage which will progressively alter according to projected bed-level changes.As results are reported on an annual basis, they are best interpreted in terms of changing inundation trends over decadal time spans, and in the knowledge that certain cross sections are more prone to overbank flow than others.Whether forecast changes are viewed positively or negatively will depend on the river management aims-in areas with populated floodplains, increased flood risk is a hazard whereas in restoring channel-floodplain connectivity, increasing inundation is a critical asset for habitat enhancement.
Similarly, substantive rates of bank erosion may require engineering intervention if they occur near critical infrastructure but should be encouraged to foster healthy river and riparian functioning in the absence of infrastructure (e.g., Florsheim et al., 2008) or where bank protection may cause more undesirable changes elsewhere.Multiple approaches exist including reconnaissance methods (e.g., Thorne, 1998), extrapolations from historical analysis of planform change (e.g., Lagasse et al., 2004), statistical methods using locally weighted logistic regressions (Varouchakis et al., 2016), indices based on the physical characteristics of river banks (e.g., the bank erosion hazard index, Rosgen, 2001), and process-based modeling based on the mechanical stability of the bank (e.g., Osman & Thorne, 1988;Simon et al., 2000Simon et al., , 2011;;Simon & Collison, 2002).The latter approach would provide erosion rate predictions useful in foresight of risk to near-channel infrastructure but representing the detailed data inputs associated with bank erosion modeling sits awkwardly in the context of requirements for a long-term forecasting model such as FRAME.Therefore, a simplified, annually integrated bank erosion function for FRAME is planned.

| Metrics related to river conservation
Evolutionary concerns in river conservation often relate to the "fractal" dimension of river environments defined by the riverbed's grain size texture and flow hydraulics.Such factors are intimately linked to aquatic habitat potential and thus the likelihood of sustaining or restoring (or losing) valued aquatic biodiversity.
At the reach scale, there is thus a potentially valuable metric relating to changes in aquatic habitat resulting from changing bedform habitat state, based on the assumption that changes in channel morphology will alter the alluvial structure of the channel bed that provides Shields stress required to entrain the median bed particle size).While the Buffington (2012) channel type dataset consists of nearly 1200 data points extracted from over 90 publications, class types in nature are far from exclusively defined in q* and q * b space (see Figure 5).The metric is thus determined probabilistically using binary logistic regression with the likelihood of the channel type derived from multiple independent predictor variables (significant variables frequently included channel slope, S; relative submergence, h*; and channel width, w, see Table 1).
F I G U R E 5 State diagram separating reach types from the Montgomery-Buffington classification using energetic principles (Buffington, 2012).
Relatively limited predictive success of the equations suggests that there may be infrequently collected channel morphology variables that are important in determining channel type.Changes over time for this metric may indicate increasingly suboptimal conditions for the original mix of aquatic flora and fauna characterizing that bedform habitat type, or possibilities for a new instream aquatic ecology.Changes would highlight the need for further, more detailed examination to judge whether the changes are welcomed or not.
A second conservation metric relates to the ecohydraulic diversity status of the river, aiming to determine whether channel evolution forecasts will improve or degrade conditions for one or more valued instream species (most likely native fish).Where threatened and endangered species exist, instream habitat suitability can be a critical element in determining whether management activities are allowed.Instream hydraulic habitat assessments are usually based on variations in flow velocity (v), flow depth (D), and bed surface grain sizes (d)-for instance, in models like PHABSIM (Bovee, 1982;Milhous et al., 1989) and CASIMIR (Jorde et al., 2001;Mounton et al., 2007;Noack et al., 2013), compared against empirically derived habitat suitability curves for species and life stages of interest.However, such curves need to be regionally derived and are rarely available.As a simplified alternative, FRAME provides cross-sectionally averaged values for v, D, and d on annual time steps that allows metrics to be specified in proportion to the yearly flow regime.Thus, metrics were developed for relative changes in hydraulic suitability according to changes in the annualized median and interquartile ranges of both v and D, and absolute changes in grain size texture d, with measures for both relatively coarse particles (using 50-84th percentiles) and for the finer fraction (using the 16th percentile).For many fish species, such percentile ranges would likely represent metrics of suitability and unsuitability, respectively, but they could be altered easily where, for example, valued sand-dwelling fishes exist.Again, changes indicated by this "horizon scanning" conservation metric may alert the river manager of the need for a detailed assessment of fish habitat suitability.

| DA S HBOARD DE VELOPMENT
As indicated in Figure 3, visioning demands that the metrics of channel evolution are transformed into visually accessible indicators and communicated via a graphical dashboard to provide a comprehensive, high-level overview of forecasts.Indicators were developed to depict, wherever possible, the likelihood of state transition changes as a ratio.Display dashboards became popular as scorecards of key performance indicators in the late 1990s but have recently become ubiquitous in smartphone apps, personal tracking devices, etc. Dashboards typically have an operational or analytical focus, with the latter used, as here, to help the user interpret data, analyze trends, and drive decision-making.
The success of an analytical dashboard depends both on choosing the right form of data representation (graphs, bar charts, etc.) but also various aspects of design.This includes the consistency and grouping of related content, the logic and simplicity of the page layout, using color and white space appropriately, using rates, percentages, and ratios rather than raw data, providing a suitable balance between data and interactivity, and amplifying the signal at the expense of noise (Bunting & Siegal, 2017).
Indicators developed from the five functional metrics are described in Table 2.The indicators and thus the initial proof-of-concept dashboard were designed to function in an Excel spreadsheet, facilitating easy transference of output data from FRAME (also Excel-based) and without requiring design skills related to specialist software.Excel supports an increasing range of dashboard functionality.Dashboards were designed with interactivity in mind, wherein the user would simulate scenarios of climate or land-use change by altering input variables in FRAME (i.e., step 8 of the meta-model framework, Figure 3).Because the output datasets comprise multiple indicators evolving in both space and time, they are difficult to condense to a single display.As such, several proof-of-concept dashboards were designed highlighting (1) a userchosen year displaying all indicators as they vary upstream to downstream and (2) a user chosen cross section displaying all indicators as they change through the timeframe of the simulation.Related to a testbed illustration using the LMR, Figure 6 exemplifies two dashboard designs for (a) an arbitrary cross section through time (cross section 17, River Mile (RM) 466) and (b) along the reach for a hypothetical year of interest (2075).Both the aerial imagery and menu systems are placeholders for intended future interactivity (see Section 6).

| APPLI C ATI ON
A proof-of-concept application of the RUBRIC process was conducted on a 305-km reach of the LMR between the Arkansas confluence and Vicksburg, MS (River Mile 576 to 389).The model used 40 baseline cross sections derived from a 2004 hydrographic survey, the daily mean discharge record from the Arkansas City gaging station and a representative bed material particle size distribution developed from multiple point samples collected from 1966 to 1974.It was broadly calibrated against comparison of 2004 and 2013 hydrographic surveys and specific gage data.Continuous bank protection severely curtails the prospect of both riverbank erosion and meander belt evolution, matching potential morphological evolution in the reach to the five metrics that can be currently derived from the FRAME model.Four scenarios were run for a 60-year period (notionally 2020-2080).The first used repeat cycles of recent flow conditions.The second scenario simulated generally higher flow conditions after 2040 (based on regional climate change projections, Krysanova et al., 2017;Huang et al., 2020) using selected upper percentile annual flow duration curves from the recent past.The third used recent flows in combination with a hypothetical flow diversion structure halfway along the reach, and the fourth combined the flow diversion structure with higher flow conditions.Details are provided in Table 3.
Overall, these scenarios indicate relatively limited morphological evolution of the LMR which is unsurprising given the reach's continuous bank protection.With a continuation of recent flow conditions, a general bed-level equilibrium is predicted in terms of balanced mild degradation and aggradation and this outcome changes little under the wetter conditions of scenario 2 (Table 3).Adding a flow diversion structure (scenario 3) shifts the reach-wide morphological response toward aggradation under current flow conditions, and with higher flows post-2040 (scenario 4), there is a minor reversion toward degradation (Table 3).Perhaps unsurprisingly, the impact of a hypothetical flow structure diverting one-quarter of high flows from the reach is of greater bed level significance than somewhat higher flow magnitudes.
Illustrating the results via indicator plots in the prototype dashboards, the morphological stability indicator shows the bed alternating annually between small amounts of degradation and aggradation for cross section 17 and along the reach in 2075 (x-axis of left-hand plots in Figure 6) and, assisted by the bank revetments, overwhelmingly stable banks (y-axis factor of safety >1) for both.Flow overtopping durations through time are indicated as intermittent at cross section 17, according to flow year (central plot in Figure 6a).In Figure 6b, the year 2075 under hypothetically wetter conditions causes high durations of overtopping flow in 2075, with reductions downstream of the imposed midreach flow diversion structure.The metric of ecohydraulic diversity is displayed using multiple indicators (right-hand panels in Figure 6) with little trend in the coarser (d 50 -d 84 , red bars) and finer (d 16 , red trace) grain sizes, a result that appears consistent with the limited morphological change.There is a mild cyclicity of velocities and depths for cross section 17 over the 2020-2080 period (green and purple plots in Figure 6a, respectively) which presumably correspond to the relatively wetter or drier flow conditions of individual years.Figure 6b indicates a relative suppression of velocity and depth ranges downstream of the diversion (RM 480) by 2075 which is logically consistent with the diversion of higher discharges.
Neither figure suggests a change in planform type over time (for cross section 17, Figure 6a, blue bars) or space (in 2075, Figure 6b, blue bars), but the channel is consistently predicted to be multithreaded by a large ratio.This opens the intriguing prospect that bank revetments prevent multiple channel threads developing in the LMR or, more mundanely, that the LMR sits outside of the range of test environments incorporated into Eaton et al.'s (2010) metric.A similar interpretation can be advanced for the channel bedform types shown with a 100% likelihood of being dune-ripple (green bars in Figure 6) based on Buffington's (2012) aquatic habitat metric.First, this is a reasonable expectation for the LMR but, second, low gradient channels in the Buffington dataset are always predicted to be dune-ripple types.It is likely, of course, that for the indicators developed using primarily gravel-bed river datasets, there may be only limited sensitivity in sand-bed applications.
Overall, the prototype dashboards indicate the potential for displaying channel evolution metrics, even within a spreadsheet format, but also point toward some of the current limitations and extensions required to develop a user-friendly system, as explored below.
TA B L E 2 Proof-of-concept indicators for five functional metrics of channel evolution, see Figure 6 for arrangement as dashboard.

Channel planform
Predictions of single or multithreaded channel planform displayed on a ratio scale where 1 is the threshold value.Values above the threshold (multithread prediction) are indicated in a different color than those below, with the ratio value indicating the proximity to the threshold "Clustered column/line combination" graph

Channel morphology
Two-dimensional plots (year or cross section focus), indicating ratios for changing bed elevation (x-axis) and changing factor of safety in bank stability (y-axis).The four quadrants of the display are illustrated in different color points or backgrounds to highlight differences.The points indicate proximity to the threshold condition of one or both ratios "Scatter" and "scatter/stacked area combination" graphs

Floodplain connectivity
Plots indicating percentage time that channel flows are predicted to be outof-bank on an annual basis.Plots are coded to different colors according to the percentage overtopping, providing a graphically distinctive separation of relative overtopping extents between year or cross sections

| PROS PEC TS
In attempting to use a foresight competency approach (Hines & Bishop, 2006) to accommodate channel evolution dynamics and so benefit a wide range of river management issues, including those related to river corridor space requirements, bed and bank erosion, flood risk, riverbed habitat potential and ecohydraulic diversity, we have tackled the highly under-developed visioning component (see Figure 2), advancing a proof-of-concept decision support tool designed to integrate strategic forecasts of channel evolution into the management realm.The visioning process (RUBRIC, Figure 3) is based on a comprehensive set of channel evolution metrics derived, where possible, from established is encouraged by translating the channel evolution metrics into a rational system for indicating evolutionary changes via threshold-based graphical indicators displayed using a dashboard-style GUI (Figure 6).Scenario building is built-in to the forecasting-visioning system, but user-led interactivity requires operationalizing (see below).
The current proof-of-concept research has involved parallel developments in both the RUBRIC visioning system and the FRAME forecasting tool to provide functionality within the foresight process (Figure 7a).At present, the forecasting component exists separate to the visioning component with Excel-based processes for data export from FRAME, data import and metrics calculation, indicators calculation, and dashboard display (Figure 7b).Scenario setting within FRAME allows alteration of external controls by making changes to the annual flow duration curve and/or sediment regime to simulate wide-ranging future scenarios such as climate change, watershed build-out, dam building, or removal.Flow and sediment can be added or removed to simulate tributary inputs and flow abstraction or import.Morphological adjustments can be made to simulate management actions.They include the ability to alter cross sections, to add or remove instream structures, and a simple system for altering input bank parameters to simulate changes in land use in the riparian corridor or the addition or removal or revetments.
Near-term advances in visioning capability require improvements in model-based forecasting capabilities and efforts to improve dashboard design and functionality, including those resulting from discussions with river managers.Priority developments for the forecasting tool include adding elements of lateral and length adjustment to permit computation of the erodible corridor and riverbank erosion rate metrics.This will require modification of the meta-model underlying the visioning process (i.e., Figure 3) to facilitate a robust "transient response" sequence for multidimensional channel adjustments, for which there is little literature guidance.Dashboard design improvements will consider the development of a third dashboard type highlighting comparisons between multiple scenarios for single indicators and displaying indicators within a mapping system of satellite imagery (e.g., using Google Earth kml/kmz data layers) to provide visually more intuitive indications of change.
Critically, functionality improvements must include integration of the metric and indicator calculations into the FRAME forecasting tool, creating an integrated forecasting and visioning tool that allows managers to set prospective scenarios and/or view subsets of the indicators via a drop-down menu system (Figure 7c).Facilitating such "back-directed" user interactivity is essential for encouraging user uptake of the decision support tool.
Pursuing foresight competency for channel evolution in river management has involved not only developments in forecasting-visioning capabilities but reflection on the adequacy of the framing-scanning components, that is, on the robustness of science underpinning the core metrics of channel evolution.It became apparent just how little fluvial geomorphology research is ever directed at testing concepts such as the physically based relationships used here.For instance, the bedform habitat type metric used the foremost dataset on channel types available (Buffington, 2012) but overlapping conditions between types (Figure 5) results in generally limited success in statistically separating bedform types (equations in Table 1).This may reflect the high levels of stochasticity in bedform typing, but also suggests that geomorphologists should trial new variables to predict channel morphology and planform.For instance, consistent with Lane's (1955)  7 | CON CLUS IONS Hines and Zindato (2016) note that the first three steps in the six-step foresight competency model (Figure 2) involve "Mapping," that is, constructing alternative futures, while the second three are about "Influencing" actions taken to shape the future.As such, developing a closely coupled forecasting-visioning process (as here) is critical in linking the two phases and underpins the translation of plausible future predictions into the actions necessary to achieve them.The research here thus initiates the prospect of foresight competency in river management by enabling managers to trial various plausible or possible future scenarios in river morphology evolution (Voros, 2003: see Figure 1), conceivably allowing them to influence actions that steer evolution toward a preferred future of resilient functional and sustainable ecosystem attributes instead of the most likely future under current conditions.Such capability has potentially massive value.
A foresight competency approach can provide a platform for translating scientific forecasts into a format usable by resource managers.
The presented metrics, graphical indicators, and dashboard design ideas represent an initial attempt at indicating how such visioning capability might be achieved in practice.However, the approach is highly demanding of the underlying database of empirical and theoretical knowledge in fluvial geomorphology.More broadly, foresight competency has great promise as a conceptual framework for applications of fluvial geomorphology, focusing attention on the "full" six-step ensemble of knowledge requirements necessary for strategic applications.The framework thus addresses the call by Gregory and Lewin (2015) for a re-examination of the conceptual basis of geomorphology, and for the provision of concepts as models that can be tested operationally and that contribute more to forecasting than introspective interpretation.Whereas the modern science of geomorphology coalesced around systems concepts in the 1960s, in the early 21st Century, the relevance of geomorphology could be underpinned by translating applicable knowledge through systems of decision support in order to deliver foresight competency.
relation to flood risk, the frequency of floodplain inundation (i.e., the hydrological connectivity of the channel to its floodplain) will change both as a function of non-stationary climate and changes in channel capacity.Using FRAME, where channel evolution responds to imposed annual F I G U R E 4 Dimensionless morphology adjustment state diagram with quadrants (numbered) indicating stability states (adapted from Watson et al., 2002).CEM x annotations represent corresponding state in the channel evolution model (CEM; e.g., Simon, 1989) and gray shading is hypothesized as the ideal zone for channel design.
the framework for aquatic habitat type.The proposed metric originates in a state diagram of channel types derived from a regime equationbased interpretation(Buffington, 2012, see alsoBuffington & Montgomery, 2022) of the widely used "Montgomery-Buffington" channel classification system(Montgomery & Buffington, 1997).Channel types are based on changes in dimensionless unit rates for discharge (q*) and sediment supply (q * b ) (Figure5), derived as functions of channel slope (S), "relative submergence" (h* = h/d 50 , where h = channel depth and d 50 = median bed material particle size), and excess Shields stress (τ*/ * c50 , where τ* = Shields stress generated at bankfull [τ*] and * c50 = critical analyses about state transition changes in channel morphology.Constraints arise both from the outputs possible from a computationally efficient hydrodynamic computer model that forecasts channel (bed) evolution over decadal time scales (i.e., FRAME), and from the assumed requirements of environmental managers involved in planning and enacting river management and restoration activities.End user adoption F I G U R E 6 Prototype dashboards for indicator display relating to the Lower Mississippi River test reach (see next), illustrating (a) channel evolution through time for an individual cross section (17) under a continuation of flows from the recent past (resulting from scenario 1 in Table 3) and (b) evolution along the reach for an individual year (2075, downstream is toward the left) with a flow diversion in place and wetter flow conditions after 2040 (i.e., scenario 4).The central panel includes indicators for the planform (blue bars), floodplain connectivity duration (multicolor bars), and bedform type (green bars); the left-hand panel provides the channel morphology indicator scatterplot and the right-hand panel provides the ecohydraulic diversity indicators (top to bottom: grain size distribution, annualized interquartile velocity range, and annualized interquartile flow depths).

"
notion of channel equilibrium,Beechie and Imaki (2014) achieved better predictive success in distinguishing channel planform types by including several variables to reflect potential sediment supply.Echoing recent debates about the predictability of earth systems (NASEM, 2020), we can infer that greater thought is required in optimizing river morphology observations, considering what is measured, how it is parameterized, what instruments are needed, and whether prior statistical and modeling tools are adequate.Advances in fluvial geomorphology should eventually facilitate metrics of greater accuracy and precision for predicting state transition morphology transformations.At present, the current metrics would likely create significant signal to noise ratios in the uncertainty predictions that would be desirable additions to the graphical indicators.Furthermore, the paucity of channel evolution data reduces prospects for running retrospective simulations on well-constrained case studies as a validation check on the accuracy and precision of FRAME as a predictive tool, although several tests are planned.TA B L E 3 Drivers for channel evolution scenario testing for the Lower Mississippi River, 2020-2080.change": recent-past flows with higher flows after 2040One 20-year cycle of annual flow duration curves 1994-2013, followed by three 13-year sequences of high flow years (87-98th percentile annual flow duration curves) Recent past" flow conditions + opening of a diversion at XS 20 (RM 480), diverting 25% of channel flows with 25% sediment removal efficiency in flows greater than 11

F
I G U R E 7 (a) Current conceptualization of the FRAME-RUles-Based morphological Response In river Channels (RUBRIC) interface as it contributes to achieving the two central steps for foresight competency in river management; (b) summary of the current computational process interface between FRAME and RUBRIC; (c) conceptualization of next phase developments with metric and indicator calculations integrated within FRAME allowing "back-directed" interactivity.
Management metrics proposed as practicable decision support tools.