Search

Summary Infrastructure must become more resilient as the global climate changes and also more affordable in the economic and political context of a post-COVID world. We can solve this dual challenge and drive global infrastructure investment into a more sustainable direction by taking our cues from Nature.

Globally, rising seas, coastal erosion, extended dry periods, and flooding contribute to decreased water security and increased disaster incidence. Nature-Based Solutions (NBS) are increasingly advanced as innovative responses to promote adaptation and build resilience, and they are arguably more sustainable than traditional gray infrastructure. There is a growing body of information regarding the material, social, and technological advances that constitute NBS and the ways in which nature can complement traditional built infrastructure. However, critical gaps remain. Promoting a coupled systems approach, we explore fundamental challenges, including issues of participation and equity, economic valuation, scalar mismatches, the integration of natural and built infrastructure, and governance. NBS do not entail quick solutions, and to reach their full potential NBS require a fundamental rethinking of society’s relationship with nature.

River flows exert dominant controls on in-stream biota. Quantifying linkages between hydrology and biology is important for assessing the effects of flow alteration on ecological functions. Hydrologic models are often used to quantify these flow-ecology relationships and guide management actions. Traditional model calibration techniques typically focus on a best overall fit criterion that may not be suitable for environmental flow applications where certain elements of the flow regime exert a dominant influence on biotic composition. We present an approach for hydrologic model calibration that improves the accuracy of calculated flow metrics known to be significant drivers of ecosystem response. First, we developed regional flow-ecology relationships based on streamflow gage and benthic macroinvertebrate data from southern California to determine which streamflow metrics best explain variability in taxonomic and trait-based biotic indices. Next, we developed and calibrated a series of hydrologic models to minimize error in these important flow metrics. For our study sites, flow flashiness and low flow frequency (indicative of drying) were found to best explain biotic condition. Hydrologic models calibrated specifically to minimize errors in these flow metrics predicted macroinvertebrate indices better than models calibrated to maximize fit to the overall flow regime. This ecological-calibration approach requires some a priori knowledge of flow-ecology relationships, but it produces results that can improve assessment of the impacts of changing flow regimes on biota and guide the development of strategies to mitigate ecological degradation.

Excessive river erosion and sedimentation threatens critical infrastructure, degrades aquatic habitat, and impairs water quality. Tools for predicting the magnitude of erosion, sedimentation, and channel evolution processes are needed for effective mitigation and management. We present a new numerical model that simulates coupled river bed and bank erosion at the watershed scale. The model uses modified versions of Bagnold’s sediment transport equation to simulate bed erosion and aggradation, as well as a simplified Bank Stability and Toe Erosion Model (BSTEM) to simulate bank erosion processes. The model is mechanistic and intermediate complexity, accounting for the dominant channel evolution processes while limiting data requirements. We apply the model to a generic test case of channel network response following a disturbance and the results match physical understanding of channel evolution. The model was also tested on two field data sets: below Parker Dam on the lower Colorado River and the North Fork Toutle River (NFTR) which responded dramatically to the 1980 eruption of Mount St. Helens. It accurately predicts observed channel incision and bed material coarsening on the Colorado River, as well as observations for the upstream 18 km of the NFTR watershed. The model does not include algorithms for extensive lateral migration and avulsions and therefore did not perform well in the lower NFTR where the channel migrated across a wide valley bottom. REM is parsimonious and useful for simulating network scale channel change in single thread systems responding to disturbance.

An ecohydrological stream type classification was developed to improve decision making for ephemeral and intermittent streams at four military reservations in the southwestern U.S.: Fort Irwin, Yuma Proving Ground (YPG), Fort Huachuca, and Fort Bliss. Agglomerative hierarchical cluster analysis was used to classify stream reaches by ecohydrologic properties (vegetation, hydrologic, and geomorphic attributes derived using geographic information system analyses), and Classification and Regression Trees (CART) were used to determine thresholds for each variable for a predictive model. Final stream types were determined from statistical analyses, cluster validity tests, examination of mapped clusters, and site knowledge. Climate regime and geomorphology were most important for YPG and Fort Irwin where annual precipitation is low. Vegetation variables were important at Fort Bliss and hydrologic variables were important at Fort Huachuca where higher annual precipitation and a bimodal rainfall pattern occur. The classification results and input variables are spatially linked to specific stream reaches, allowing managers to identify locations with similar attributes to support management actions. These methods enable the development of a stream type classification in gauged or ungauged watersheds and for areas where intensive field data collection is not feasible.

The broad scope and definition of sustainability has perplexed assessment of water infrastructure systems, especially for the purpose of directing engineering practices when quantified criteria are desired. An input-oriented data envelopment analysis (DEA) was improved to benchmark the relative sustainability of the water infrastructure of 157 cities in China. The DEA calculates a single sustainability score using seven inputs and five outputs that represent the economic, resource and environmental dimensions of sustainability. Overall, 69 out of the 157 sampled systems obtained high sustainability scores. Eight specific efficiency indicators based on individual DEA input to output ratio were evaluated to identify the causes of performance differences. Compared to water supply systems, the performance of wastewater treatment plants has greater influence on the sustainability score of the overall system. For all systems, the sustainability scores are more sensitive to sludge production and electricity consumption than capital investment and removal efficiency of treatment processes. The DEA provides guidelines to cities for setting priorities in order to meet specific sustainability criteria. Statistical analysis indicates that the overall sustainability score primarily depends on the system scale, meteorological conditions such as air temperature and rainfall, and source water quality.

The Colorado Front Range flood of September 2013 induced a diverse range of geomorphic changes along numerous stream corridors, providing an opportunity to assess responses to a large flood in a semiarid landscape. We defined six classes of geomorphic change related to peak unit stream power and valley confinement for 531 stream reaches over 226 km, spanning a gradient of channel scales and slope. Geomorphic change was generally driven by erosion of channel margins in confined reaches and by a combination of deposition and erosion in unconfined reaches. The magnitude of geomorphic change typically increased with unit stream power ( ω ), with greater responses observed in unconfined channels. Cumulative logit modeling indicated that total stream power or unit stream power, unit stream power gradient, and valley confinement are significant predictors of geomorphic response for this flood event. Based on this dataset, thresholds for geomorphic adjustment were defined. For channel slopes ω > 230 W/m 2 (16 lb/ft-s; at least 10% of the investigated sites experienced substantial channel widening) and a credible potential for avulsions, braiding, and loss of adjacent road embankments associated with ω > 480 W/m 2 (33 lb/ft-s; at least 10% of the investigated sites experienced such geomorphic change). Infrequent to numerous eroded banks were very likely with ω > 700 W/m 2 (48 lb/ft-s), with substantial channel widening or major geomorphic change shifting from credible to likely. Importantly, in reaches where there were large reductions in ω as the valley form shifted from confined to relatively unconfined, large amounts of deposition-induced, reach-scale geomorphic change occurred in some locations at relatively low ω . Additionally, alluvial channels with slopes > 3% had greater resistance to geomorphic change, likely caused by armoring by larger bed material and increased flow resistance from enhanced bedforms. Finally, we describe how these results can potentially be used by practitioners for assessing the risk of geomorphic change when evaluating current or planned conditions.

Measuring precipitation in semi-arid landscapes is important for understanding the processes related to rainfall and run-off; however, measuring precipitation accurately can often be challenging especially within remote regions where precipitation instruments are scarce. Typically, rain-gauges are sparsely distributed and research comparing rain-gauge and RADAR precipitation estimates reveal that RADAR data are often misleading, especially for monsoon season convective storms. This study investigates an alternative way to map the spatial and temporal variation of precipitation inputs along ephemeral stream channels using Normalized Difference Vegetation Index (NDVI) derived from Landsat Thematic Mapper imagery. NDVI values from 26 years of pre- and post-monsoon season Landsat imagery were derived across Yuma Proving Ground (YPG), a region covering 3,367 km2 of semiarid landscapes in southwestern Arizona, USA. The change in NDVI from a pre-to post-monsoon season image along ephemeral stream channels explained 73% of the variance in annual monsoonal precipitation totals from a nearby rain-gauge. In addition, large seasonal changes in NDVI along channels were useful in determining when and where flow events have occurred.

Summary The decision to use either daily-averaged or sub-daily streamflow records has the potential to impact the calculation of sediment transport metrics and stream channel design. Using bedload and suspended load sediment transport measurements collected at 138 sites across the United States, we calculated the effective discharge, sediment yield, and half-load discharge using sediment rating curves over long time periods (median record length = 24 years) with both daily-averaged and sub-daily streamflow records. A comparison of sediment transport metrics calculated with both daily-average and sub-daily stream flow data at each site showed that daily-averaged flow data do not adequately represent the magnitude of high stream flows at hydrologically flashy sites. Daily-average stream flow data cause an underestimation of sediment transport and sediment yield (including the half-load discharge) at flashy sites. The degree of underestimation was correlated with the level of flashiness and the exponent of the sediment rating curve. No consistent relationship between the use of either daily-average or sub-daily streamflow data and the resultant effective discharge was found. When used in channel design, computed sediment transport metrics may have errors due to flow data resolution, which can propagate into design slope calculations which, if implemented, could lead to unwanted aggradation or degradation in the design channel. This analysis illustrates the importance of using sub-daily flow data in the calculation of sediment yield in urbanizing or otherwise flashy watersheds. Furthermore, this analysis provides practical charts for estimating and correcting these types of underestimation errors commonly incurred in sediment yield calculations.

The characteristics of hydroclimatic risk in the 21st Century are rapidly changing. Increases in extreme weather events and population densities alter exposure to floods and droughts. Water infrastructure is unable to keep pace and deterministic models can mislead. Yet, predominant strategies for managing risk continue to follow historical precedent, striving to tame nature's outbursts and mitigate disasters through conventional engineering structures. In this Viewpoint article, scholars from three disciplines address underlying limitations in contemporary risk management, each of which is rooted in the concept of hubris. They argue that effective risk management must extend beyond hubris-thinking to an approach based on humility. Extending from the three types of limitations—described as Disconnect from Nature, Engineers of a Fixed Nature. and Modelers of Nature— a collective and synthetic consideration provides insights into potential pathways forward for both science and practice. More satisfactory strategies for managing hydroclimatic risk will require a shift from a hubris-based paradigm to a more humility-based approach.

Rising sea levels and growing coastal populations are intensifying interactions at the land-sea interface. To stabilize upland and protect human developments and infrastructure from coastal hazards, property owners commonly emplace hard armoring structures, such as bulkheads and revetments, along the estuarine shoreline. While research regarding the ecological and economic impacts that accompany hard armoring implementation has garnered significant attention, there have been limited efforts to document the extent of hard armoring or identify drivers of hard armoring patterns. This study addresses a knowledge gap in the documentation and understanding of hard armoring patterns and their drivers at the scale of individual landowners, by assessing the distribution of hard armoring along the Georgia coastline. Using a set of environmental and socioeconomic predictor variables in coordination with a novel fine-scale census of hard armoring, we present a parsimonious statistical model that predicts the probability of hard armoring emplacement at the parcel level. Our model has a high correct classification accuracy and we find that a variety of interacting environmental and socioeconomic influences contribute to patterns of shoreline armoring; in particular, the presence or absence of armoring on a neighboring parcel has a strong association with the probability of hard armoring. Our model also suggests that continued sea level rise and coastal population growth could trigger increases in armoring implementation into the future. Thus, we underscore the importance of considering both the current and potential future distributions of hard armoring when evaluating associated impacts of sea level rise on ecological systems and in socioeconomic evaluations. For example, evolving distributions of hard armoring should be considered in predictions of salt marsh sustainability through inland migration. In addition to providing the first parcel-level census analysis of hard armoring at a scope relevant to local and regional government decision-making, this study also presents a methodology adaptable to assessing patterns of hard armoring in other study areas. With improved understanding of hard armoring distributions, SLR response plans can be fully informed to design more efficient scenarios for both urban development and coastal ecosystems.

Evaluating and designing channel spanning structures for successful fish passage requires description of hydraulic conditions at scales meaningful to fish. We describe novel approaches combining fish movement data and hydraulic descriptions from a three-dimensional computational fluid dynamics model to examine the physical processes that limit upstream movement of trout across 3 unique in-stream structures at a whitewater park (WWP) in Lyons, Colorado. These methods provide a continuous and spatially explicit description of velocity, depth, vorticity, and turbulent kinetic energy along potential fish swimming paths in the flow field. Logistic regression analyses indicate a significant influence of velocity and depth on limiting passage success and accurately predict greater than 87% of observed fish movements. However, vorticity, turbulent kinetic energy, and a cost function do not significantly affect passage success. Unique combinations of depth and velocity at each WWP structure reflect variation in passage success. The methods described in this study provide a powerful approach to quantify hydraulic conditions at a scale meaningful to a fish and to mechanistically evaluate the effects of hydraulic structures on fish passage. The results of these analyses can be used for management and design guidance, and have implications for fishes with lesser swimming abilities.

Traditional approaches to water resource engineering have sought to maintain a static, optimized state of system performance in providing reliable water supplies, energy, and flood protection. However, delivery of these services has been associated with the disruption of freshwater ecosystem functioning, driving global-scale declines of biodiversity and the loss of ecosystem services. Climate change is presenting new challenges for water and ecosystem managers alike. Yet, climate change is also creating new opportunities to consider ecological resilience in the design and management of water systems. Here, we describe a set of climate-informed ecological resilience principles and associated indicators, which can support integration of ecosystem needs within water resource engineering decision-making. These have the potential to guide climate-adaptive water resource management while also provisioning broad benefits to both people and ecosystems in a shifting operating environment.

Summary Semiarid channels exhibit an extreme sensitivity to upstream urban development, particularly in unconfined valleys with unprotected grades. For example, one of our study streams in southern California has increased its cross-sectional area by nearly 14-fold relative to its pre-developed channel form in a watershed that has been only lightly developed (10.4% imperviousness). Multivariate regression models of cross-sectional channel enlargement at 61 sites were highly dependent on the ratio of post- to pre-urban sediment-transport capacity over cumulative duration simulations of 25 yrs (Lr), which explained nearly 60% of the variance. The proximity of a channel hard point such as bedrock or artificial grade control was also significant, indicating that channel enlargement increased moving upstream from grade control. The enlargement models point to the importance of balancing the post-developed sediment transport to the pre-developed setting over an entire range of flows rather than a single flow in order to reduce the risk of adverse channel responses to hydromodification. The need for controlling a wide range of flows was underscored by logistic-regression analyses that indicated a high risk of instability in systems with Lr > 1, especially for fine-grained systems (i.e., d50

summary In 15 mountain stream reaches containing instream wood, we characterized velocity and flow resistance at bankfull through low flows. These data were: (1) used to assess the accuracy of previously published velocity prediction techniques for high-gradient channels; and (2) were combined with field data from other studies to develop general methodologies for predicting velocity and flow resistance in alluvial and mixed alluvial-bedrock channels both with and without step-forming instream wood. Velocity and flow resistance were poorly predicted by variables characterizing grain size and relative grain submergence. Conversely, methods based on detrended standard deviation of bed elevations (rz) and relative bedform submergence (h/rz) explained up to 84% of the variance in the measured flow resistance coefficients and 97% of the variance in dimensionless velocity. With an average velocity of 0.44 m/s for the collected measurements, velocity was predicted with RMS (root mean square) error as low as 0.071 m/s (16% of average) when discharge and bedform geometry is known and 0.10 m/s (23%) when only bedform geometry is known. Additionally, an empirical relationship indicates V=u � ¼ h=r z, supporting previouslypublished laboratory findings using a field-based dataset in complex high-gradient channels. Interactions between instream wood and clasts result in substantially enhanced step heights and flow resistance. This compound effect defies description by grain size and relative grain submergence. However, rz and h/rz quantify variability due to both clasts in combination with wood and clasts alone, providing relatively accurate predictions for the tested dataset and indicating substantial predictive capabilities in channels where bedforms are the primary source of flow resistance. Published by Elsevier B.V.

s u m m a r y Forty-three US Geological Survey gauges with records greater than �15 yrs located in watersheds less than �250 km 2 were used to model the effects of suburbanization on streams in semi-arid southern California. The watersheds spanned a gradient of urban development, ranging 0–23% total impervious area in 2001. With little flow control at the subdivision scale, most impervious area in the region is relatively well-connected to surface-drainage networks and hydrologically effective. Consequently, total impervious area was an effective hydrologic surrogate for urbanization, emerging from an expansive array of geospatially-derived hydrologic variables as a statistically-significant ( p < 0.05) predictor of instantaneous peak-flow rates at the 1.5- and 2-yr recurrence intervals and the durations of all geomorphically-important flows. To represent the effects of urbanization on flow durations, we developed duration density functions by using power functions (typical R 2

Mountain snowpack is the main source of water in the semi-arid Colorado River Basin (CRB), and while the demands for water are increasing, competing and often conflicting, the supply is limited and has become increasingly variable over the 20th Century. Greater variability is believed to contribute to lower accuracy in water supply forecasts, plus greater variability violates the assumption of stationarity, a fundamental assumption of many methods used in water resources engineering planning, design and management. Thus, it is essential to understand the underpinnings of hydroclimatic variability in order to accurately predict effects of climate changes and effectively meet future water supply challenges. A new methodology was applied to characterized time series of temperature, precipitation, and streamflow (i.e., historic and reconstructed undepleted flows) according to the three climate regimes that occurred in CRB during the 20th Century. Results for two tributaries in the Upper CRB show that hydroclimatic variability is more deterministic than previously thought because it entails complementary temperature and precipitation patterns associated with wetter or drier conditions on climate regime and annual scales. Complementary temperature and precipitation patterns characterize climate regime type (e.g., cool/wet and warm/dry), and the patterns entail increasing or decreasing temperatures and changes in magnitude and timing of precipitation according to the climate regime type. Accompanying each climate regime on annual scales are complementary temperature (T) and precipitation (P) patterns that are associated with upcoming precipitation and annual basin yield (i.e., total annual flow volume at a streamflow gauge). Annual complementary T and P patterns establish by fall, are detectable as early as September, persist to early spring, are related to the relative magnitude of upcoming precipitation and annual basin yield, are unique to climate regime type, and are specific to each river basin. Thus, while most of the water supply in the Upper CRB originates from winter snowpack, statistically significant indictors of relative magnitude of upcoming precipitation and runoff are evident in the fall, well before appreciable snow accumulation. Results of this study suggest strategies that may integrated into existing forecast methods to potentially improve forecast accuracy and advance lead time by as much as six months (i.e., from April 1 to October 1 of the previous year). These techniques also have applications in downscaling climate models and in river restoration and management.

article i nfo The combined influence of tree-clearing, road construction, snowmaking, and machine-grading can cause increased flow and sediment loads along streams in or adjacent to commercial ski resorts. These changes to stream channels can increase bank failures, bed material size, pool scour, and, in extreme cases, channel incision. We used field data from the White River National Forest in Colorado, which includes several major ski resorts, to test the hypothesis that ski slope development causes a significant difference in bank stability, undercut banks, fine sediment, wood load, pool residual depth, and particle size (D84) between the ski area project streams and reference streams. We further hypothesize that the changes in a stream are mitigated by the density and type of vegetation growing along the banks. A significant difference is defined as a project stream that is outside the range of variability of the reference streams. To test these hypotheses, we surveyed channel conditions, channel dimensions, and vegetation along 47 stream reaches (200-300 m in length). Twenty-four of these streams are within ski areas (project streams), either adjacent to or downstream from ski slopes. Twenty-three reference streams with very little to no development in their basins are used to define reference conditions of bank stability, bank undercutting, bank height, wood load, pool residual depth, sediment size, and vegetation structure. A combination of statistical techniques, including Principal Components Analysis and Classification and Regression Tree Analysis, was used to assess the controls on stream channel morphology and to analyze the differences between project and reference streams. Project streams that are significantly different than reference streams have a combination of a higher percentage of fine sediment, smaller pool residual depth, and higher percentage of unstable banks. The impacted project streams have bed material derived from granitic rocks and a lower density of understory vegetation. These results show the importance of considering vegetative and geologic influences on channel form and processes when assessing impacts of land use change. Roads and machine-grading have the most significant impact on the streams, causing bed fining and pool filling. These data and results will help in revising a forest management plan to provide guidelines for planning and development of ski areas on public lands.

Using daily discharge data from the USGS, we analyzed how hydrologic regimes vary with land use in four large hydrologic regions that span a gradient of natural land cover and precipitation across the continental United States. In each region we identified small streams (contributing areab282 km 2 ) that have continuous daily streamflow data. Using a national database, we characterized the composition of land cover of the watersheds in terms of aggregate measures of agriculture, urbanization, and least disturbed (“natural”). We calculated hydrologic alteration using 10 ecologically-relevant hydrologic metrics that describe magnitude, frequency, and duration of flow for 158 watersheds within the Southeast (SE), Central (CE), Pacific Northwest (NW), and Southwest (SW) hydrologic regions of the United States. Within each watershed, we calculated percent cover for agriculture, urbanized land, and least disturbed land to elucidate how components of the natural flow regime inherent to a hydrologic region is modified by different types and proportions of land cover. We also evaluated how dams in these regions altered the hydrologic regimes of the 43 streams that have pre- and post-dam daily streamflow data. In an analysis of flow alteration along gradients of increasing proportion of the three land cover types, we found many regional differences in hydrologic responses. In response to increasing urban land cover, peak flows increased (SE and CE), minimum flows increased (CE) or decreased (NW), duration of near-bankfull flows declined (SE, NW) and flow variability increased (SE, CE, and NW). Responses to increasing agricultural land cover were less pronounced, as minimum flows decreased (CE), near-bankfull flow durations increased (SE and SW), and flow variability declined (CE). In a second analysis, for three of the regions, we compared the difference between least disturbed watersheds and those having either N15% urban and N25% agricultural land cover. Relative to natural land cover in each region, urbanization either increased (SE and NW) or decreased (SW) peak flows, decreased minimum flows (SE, NW, and SW), decreased durations of near-bankfull flows (SE, NW, and SW), and increased flow variability (SE, NW, and SW). Agriculture had similar effects except in the SE, where near-bankfull flow durations increased. Overall, urbanization appeared to induce greater hydrologic responses than similar proportions of agricultural land cover in watersheds. Finally, the effects of dams on hydrologic variation were largely consistent across regions, with a decrease in peak flows, an increase in minimum flows, an increase in nearbankfull flow durations, and a decrease in flow variability. We use this analysis to evaluate the relative degree to which land use has altered flow regimes across regions in the US with naturally varying climate and natural land cover, and we discuss the geomorphic and ecological implications of such flow modification. We end with a consideration of what elements will ultimately be required to conduct a more comprehensive national assessment of the hydrologic responses of streams to land cover types and dams. These include improved tools for modeling hydrologic metrics in ungauged watersheds, incorporation of high-resolution geospatial data

Streamflow prediction in ungauged basins provides essential information for water resources planning and management and ecohydrological studies yet remains a fundamental challenge to the hydrological sciences. A methodology is presented for stratifying streamflow regimes of gauged locations, classifying the regimes of ungauged streams, and developing models for predicting a suite of ecologically pertinent streamflow metrics for these streams. Eighty-four streamflow metrics characterizing various flow regime attributes were computed along with physical and climatic drainage basin characteristics for 150 streams with little or no streamflow modification in Colorado, Washington, and Oregon. The diverse hydroclimatology of the study area necessitates flow regime stratification and geographically independent clusters were identified and used to develop separate predictive models for each flow regime type. Multiple regression models for flow magnitude, timing, and rate of change metrics were quite accurate with many adjusted R2 values exceeding 0.80, while models describing streamflow variability did not perform as well. Separate stratification schemes for high, low, and average flows did not considerably improve models for metrics describing those particular aspects of the regime over a scheme based on the entire flow regime. Models for streams identified as ‘snowmelt’ type were improved if sites in Colorado and the Pacific Northwest were separated to better stratify the processes driving streamflow in these regions thus revealing limitations of geographically independent streamflow clusters. This study demonstrates that a broad suite of ecologically relevant streamflow characteristics can be accurately modeled across large heterogeneous regions using this framework. Applications of the resulting models include stratifying biomonitoring sites and quantifying linkages between specific aspects of flow regimes and aquatic community structure. In particular, the results bode well for modeling ecological processes related to high-flow magnitude, timing, and rate of change such as the recruitment of fish and riparian vegetation across large regions.

The field of ecoinformatics is concerned with gaining a greater understanding of complex ecological systems. Many ecoinformatic tools, including artificial neural networks (ANNs), can shed important insights into the complexities of ecological data through pattern recognition and prediction; however, we argue that ecological knowledge has been used in a very limited fashion to shape the manner in which these approaches are applied. The present study provides a simple example of using ecological theory to better direct the use of neural networks to address a fundamental question in aquatic ecology—how are local stream macroinvertebrate communities structured by a hierarchy of environmental factors operating at multiple spatial scales? Using data for 195 sites in the western United States, we developed single-scale, multi-scale and hierarchical multi-scale neural networks relating EPT (Orders: E phermeroptera, P lecoptera, T richoptera) richness to environmental variables quantified at 3 spatial scales: entire watershed, valley bottom (100s–1000s m), and local stream reach (10s–100s m). Results showed that models based on multiple spatial scales greatly outperformed single-scale analyses ( R = 0.74 vs. R¯ = 0.51) and that a hierarchical ANN, which accounts for the fact that valley- and watershed-scale drivers influence local characteristics of the stream reach, provided greater insight into how environmental factors interact across nested spatial scales than did the non-hierarchical multi-scale model. Our analysis suggests that watershed drivers play a greater role in structuring local macroinvertebrate assemblages via their direct effects on local-scale habitats, whereas they play a much smaller indirect role through their influence on valley-scale characteristics. For the hierarchical model, the strongest predictors of EPT richness included descriptors of climate, land-use and hydrology at the watershed scale, land-use at the valley scale, and substrate characteristics and riparian cover at the reach scale. In summary, our results highlight the importance of incorporating environmental hierarchies to better understand and predict local patterns of macroinvertebrate assemblage structure in stream ecosystems. More generally, our case study serves to emphasize how incorporating prior ecological knowledge into ANN model structure can strengthen the relevance of ecoinformatic techniques for the broader scientific community.

A large and geographically diverse data set consisting of meandering, braiding, incising, and post-incision equilibrium streams was used in conjunction with logistic regression analysis to develop a probabilistic approach to predicting thresholds of channel pattern and instability. An energy-based index was developed for estimating the risk of channel instability associated with specific stream power relative to sedimentary characteristics. The strong significance of the 74 statistical models examined suggests that logistic regression analysis is an appropriate and effective technique for associating basic hydraulic data with various channel forms. The probabilistic diagrams resulting from these analyses depict a more realistic assessment of the uncertainty associated with previously identified thresholds of channel form and instability and provide a means of gauging channel sensitivity to changes in controlling variables.