The poorly understood relationship between specific land uses, consequent altered sediment loads and runoff, and the downstream river channel response to these altered patterns lies at the heart of the problem of management decisions regarding cumulative watershed effects. Most rivers in California have experienced altered sediment supply due to land use. The pristine state of these channels is unknown, and because channels do not respond Immediately to inputs of sediment, it can not be assumed that current channel conditions reflect current sediment loads. This makes stream rehabilitation projects, which are becoming increasingly commonplace. subject to considerable uncertainty as to their outcome. There are many aspects to this relationship between land use, sediment loads, runoff. and channel response that are challenging, including: 1) the prediction of the rate and grain size of altered sediment yield caused by a particular disturbance; 2) the prediction of the changes in runoff amounts, including how these effects influence hillslope erosion rates; and 3) the prediction of stream channel changes that include effects on surface grain size distributions, sediment storage, channel geometry. bank instability, sediment loads, and sediment concentrations. The pressure now to make decisions in advance of complete understanding of this relationship suggests that the problem be turned around: is it possible to examine a channel reach of interest and with some relatively simple observations anticipate how it will respond to altered sediment yields?

The work accomplished here was oriented toward devising methodologies to predict probable channel response to altered runoff and sediment supply due to land use by "reading" rivers. Our approach was based on our recent experimental results that demonstrated a quantitative relationship between sediment supply and the degree and spatial extent of surface armoring in gravel bedded rivers (Dietrich, et al., 1989). The theory we proposed (see Appendix A for complete paper) argues that under an Imposed boundary shear stress, such as that which occurs at bankfull discharge, grain size adjustments modify critical boundary shear stress and, consequently. the excess stress that controls bedload transport rates. When sediment supply is high for the imposed shear stress. the grain size of the bed will be close to that of the load. In contrast, if the supply is reduced, the surface will coarsen, thereby causing reduction in transport rate in concert with the reduced load. Hence the degree of stream surface armoring may be used to define a stream's state of adjustment relative to sediment supply.

Three specific research objectives were proposed in Project W-734. The first was to test in a natural river the validity of recent theories for initial motion of heterogeneous mixtures of gravels and sand size sediment. This objective was accomplished through a combination of tracer studies and direct measurement of friction angles of individual particles. The second objective was to test our theory regarding grain size adjustments (Dietrich, et al., 1989) by measuring vertical and lateral sorting variation in relation to sediment supply. (The theory assumes that the median grain size of the bed is the scale for critical shear stress of the bed and that selective entrainment can occur, causing grain size adjustments to develop from supply variations.) This work was done in conjunction with another project which examined several rivers. The work funded here focussed primarily on one study site and examined it in greater detail. The third objective was to examine whether alternating zones of high and low rates of sediment movement along a particular creek could be identified. This was done through a combination of mapping and tracer studies.

As proposed, this work was accomplished in close collaboration with Laurel Collins, geologist for the East Bay Regional Park District.

Abstract

What Controls Shallow Landslide Size Across Landscapes?

by Dino Bellugi

Doctor of Philosophy in Earth and Planetary Science

and the Designated Emphasis in Computational Science and Engineering

University of California, Berkeley

Professor William E. Dietrich, Chair

Shallow landslides that usually involve only the colluvial soil mantle, are a widespread phenomenon in the United States and the world. Often triggered by extreme precipitation events, they can be the primary sources of debris flows, and are generally a threatening source of hazards, causing loss of life, destruction of property, and affecting communities all across the nation. Shallow landslides also play an important role in landscape evolution, dominating erosion in steeper landscape, unleashing debris flows that carve valley networks, and delivering sediment to rivers. The two primary aspects affecting the impact of shallow landslides, both in terms of downstream hazard and their geomorphic significance, are their location and size.

Theoretical and observational research has provided some insight on the controls on the size of shallow landslides. It has been observed that landslide exhibit a smaller size in grasslands than in forested areas and that landslides were smaller in areas where root strength decreased as a result land use change. The parameters that are most relevant for the occurrence of rainfall-triggered shallow landslides are slope, pore pressure, root and soil strength, and soil depth. Theoretical analyses have suggested that a decline in root strength results in failures having lower minimum lengths and widths, while low gradients, low pore water pressures, or high soil friction result in failures having higher lengths and widths. However, few if any studies examine the controls on both landslide location and size across a landscape. I hypothesize that the co-organization of landscape properties, such as slope, soil depth, pore pressure, and root reinforcement, controls the size and location of shallow landslides.

We currently lack mechanistic models for specifically predicting shallow landslide size across landscapes, thus reducing the effectiveness of landslide hazard delineation, and inhibiting our ability to formulate and apply mechanistic models for landslide flux and surface erosion. One reason for this is the one-dimensional representation of slope stability, generally applied in existing regional scale applications. Such a representation cannot produce discrete landslides and thus cannot make predictions on landslide size. Furthermore, one-dimensional approaches cannot include lateral effects which are known to be important in defining instability. These limitations can be addressed by a three-dimensional slope stability model, but its application to a landscape is challenging. Whereas the one-dimensional slope stability at a location can be determined independently of its dimensions and surroundings, multi-dimensional analyses require the treatment of discrete shapes. As these shapes are not known a priori, a search algorithm is required. This is a non-trivial problem, whose naïve solution (i.e. an exhaustive search) is of exponential complexity, rendering the problem effectively intractable at any relevant scale. Any new procedure must be sufficiently general to evolve with current understanding, but with a parsimonious parameterization in order to be compatible with available data. The procedure must be computationally efficient to be applicable at scales large enough to be relevant for geomorphological and hazard related questions, yet at sufficiently fine resolution to capture the fundamental mechanics of slope failure.

In this dissertation I develop a procedure which couples a novel slope stability model that captures the basic physics of shallow landsliding, with a new and efficient search algorithm based on spectral graph theory that can predict discrete shallow landslides. In order to apply this procedure at the regional scale, I define sub-models to produce the required data, when they are not available at the necessary resolution. These sub-models extract topographic attributes, compute the spatial distributions of soils, and estimate the root reinforcement and pore pressure fields. I define formal framework to evaluate the performance of the procedure, based on information retrieval theory. This procedure should advance our understanding and prediction capability, enabling me test the hypothesis that the co-organization of landscape properties, such as slope, soil depth, pore pressure, and root reinforcement, controls the size and location of shallow landslides. As these properties are mostly dictated by topography, I hypothesize that topography exerts a first order control on both location and size.

In chapter two, I present a multi-dimensional stability model framework that can be applied to landscapes at the regional scale. This slope stability model is mechanistic but not so mechanistic that its application becomes impracticable. It is fully three-dimensional in the treatment of the forces acting on a discretized slope element and it is statically determinate. The model considers the effects of root cohesion and pore pressures, and includes the effects of earth pressure in a manner that is compatible with natural slopes. Finally, this model is easily applicable to spatially gridded data, and requires only a modest parameterization facilitated by procedures defined to obtain spatially explicit parameter fields at the required resolution. The slope stability model allows for the characterization of the forces acting on all the boundaries resulting from the discretization of a landscape into slope element blocks (and thus the role of each block in the stability of the landscape). However, it requires a deterministic search procedure that is able to select discrete least-stable combinations of slope elements across a landscape to obtain meaningful shallow landslide predictions.

In chapter three, I define a procedure that can for the first time predict discrete landslides. Its foundation is a search algorithm based on spectral graph theory that can efficiently provide a good approximation for an otherwise intractable problem. This procedure relies on a slope stability model, as well as sub-models and data for, among others, topography, soil depth, and root strength, discussed in the previous chapter. However, the procedure is general, and is not confined by their choice: as better models and data emerge, the procedure can be easily modified to take advantage of such improvements. A formal framework is defined to evaluate the performance of the procedure, based on information retrieval theory. Applying the procedure to a synthetic landscape illustrates how landslide size and location are affected by the hetereogeneity of parameters such as root strength and pore pressure.

In chapter four, the procedure is applied to an instrumented catchment in the Oregon Coast Range using field-measured physical parameters, successfully predicting the size and location of the shallow landslide which destroyed the site during a storm in November, 1996. The procedure was then applied to a larger study area using modeled physical parameters, under a suite of diverse hydrological scenarios. The application of the procedure results in, and is able to reproduce the distribution of sizes and locations observed during the ten years of research at the site. Performance is quantified using a set of information retrieval measures, performing significantly better than a random classifier, demonstrating the applicability of the procedure.

In chapter five, a sensitivity analysis is performed to explore the controls on shallow landslide size and location. Rainfall, vegetation, soil, and topographic characteristics are systematically varied, resulting in probability density functions of predicted landslide size and location. I find that increasing precipitation or soil depth results in an increased number of predicted landslides. In contrast, increasing soil strength through root reinforcement or friction angle results in a decrease in the number of predicted landslides. Increasing soil depth results in predicted landslides being preferentially located in locations with steep slopes, while increasing soil strength results in predicted landslides being preferentially located in locations with high drainage area. Precipitation affects characteristic landslide location differently: if lateral re-distribution of water is dominant, landslides are predominantly found in locations with high drainage area; in contrast, when vertical infiltration dominates they are predominantly found in areas with steep slopes. Predicted characteristic size increases with increased precipitation and with increased root strength. However, it decreases when the increased strength results from an increase of the soil friction angle. Under uniform soil thickness, characteristic size decreases with increasing soil depth. When soil thickness distributions are instead controlled by topography, increasing soil depth causes the predicted characteristic landslide size to first increase and then to decrease, after a critical value, reflecting the stabilization effect of very thick soils.

In chapter six the effects of the fine scale variability of root strength on slope stability are examined, using a method which could be extended to represent the impact of spatial variability of landslide-relevant parameters on landslide size, location, and abundance. A simple dynamic hydrological model is used in combination with a ten-minute rainfall intensity record for a landslide-triggering storm. When comparing with a map of debris flows which occurred during the storm, the procedure predicts landslides in the observed debris flow source areas. Although over-prediction is greatly reduced, there remain a considerable number of predicted landslides in areas which did not fail during the storm event. Regardless, this is a promising result, as it suggests that this procedure is capable of capturing the timing of landslides (as well as their size and location), given a sufficiently resolved characterization of the hydrology.

I find that the spatial structure of soil depth, pore pressure, and root strength determines the areas favorable to landsliding that can be exploited by rain storms, resulting in the characteristic size and location distributions of rainfall-triggered landslides. Varying these controlling properties, even uniformly, changes the spatial distribution of these areas in the landscape. This results in new characteristic distributions of landslide size and location, as landslides sample different parts of the landscape. This reveals the first-order control exerted by topography on shallow landslides. Furthermore, the general spatial pattern of landsliding did not fundamentally change with the introduction of stochastic variability in root strength or with variations in the mechanism of pore pressure generation. This highlights the fundamental role played by the topographically-controlled distribution of soil thickness in defining landslide location.

Understanding hazards posed by rainfall-triggered shallow landslides requires predicting where landslides will occur, when they will occur, how big will they be, how fast they will mobilize, and how far will they go. This research constitutes a significant step in this direction by providing some of the first coupled predictions of where and how big landslides are, and demonstrating that capturing their timing is well within reach. By coupling this procedure with climate and vegetation models we can now explore the impact of climate and land use change on the landsliding regime. By integrating the procedure into a landscape evolution model we can then explore how, over longer time scales, landslides shape a landscape.

Millions of dollars are spent annually in California on erosion control measures after wildland fires despite the lack of quantitative evidence that such measures are needed or beneficial. Application of grass seeding and other measures after a fire are often justified on the basis of a perceived concern about the risk of debris and mud flows endangering property and lives. While the effectiveness of such measures can be seriously questioned, the risk may nonetheless be significant. We have conducted a two year study to test and expand a recently developed physically-based model for predicting the spatial pattern of landslide potential. This model can be applied at any time and serve as part of planning document to prepare for the effects of fire. The model uses digital elevation data and is based on simple assumptions about runoff and slope stability mechanisms; in it's simplest form the model is parameter free and easily calibrated.

We first tested the model in the Oakland firestorm area. Here high resolution digital elevation data were available and 78 landslide scars had been mapped in a 12.12 km2 area. A simple ratio of effective precipitation rate to the ability of the soil to transmit the water is calculated from the model for the condition of instability. The smaller the ratio, the higher the potential instability. As we have found elsewhere, there is threshold value which delineates all observed scars in the Oakland hill. During the study, the Highway 41 fire occurred in the San Luis Obispo area, consuming 196 km2 of landslide prone topography. Aerial photographs were taken, and a detailed survey using global positioning techniques was accomplished to guide the construction of a detailed digital elevation model of the site. This past winter after the fire was one of the wettest of record and a total of 467 shallow landslides were mapped, with 82% occurring in the unburned areas and most of the landslides in the burned area occurring where the 11 grass seeding was most effective. Once we obtain the digital data we will use the landslide model to test various hypotheses for this unexpected, but important result.

Field studies suggest that bedrock incision by granular flows may be the primary process cutting valleys in steep, unglaciated landscapes. The mechanisms of granular flow incision, however, are not well quantified. Here I present a suite of laboratory experiments describing processes and rates of bedrock erosion by granular flows. In the first study, experiments in a 56 cm diameter, 15 cm wide rotating drum test the hypothesis that bedrock erosion is related to grain inertial stresses which scale with shear rate and particle size. In 67 experimental runs, the eroded depth of the bedrock sample varied with inertial stresses in the granular flow to a power less than 1.0, and inversely with the bedrock strength. The second study used a new debris flow flume facility comprised of a 4-meter diameter, 80-cm wide vertically rotating drum to measure mean and fluctuating normal forces at the base of granular flows. The mean bulk force scaled with the static weight of the flow, while the variance of the force was a function of grain diameter, flow velocity, and matrix fluid properties. I show that the square of the impulse, related to kinetic energy transferred to the bed from granular collisions, can be quantified as the variance of the force signal. My results provide the first quantitative relationships between a metric for the collisional energy at the boundary and measurable properties of field-scaled flows. In the third study, I measured erosion of synthetic bedrock samples in the 4 meter diameter drum to test three models for the relationship between bedrock erosion rate and measured basal forces: (1) erosion by impact wear resulting from forces due to bulk inertial solid stress (2) erosion by sliding wear scaled by bulk normal force (3) erosion by impact wear scaled by the square of the impulse exerted on the bed. Based on my experimental observations, I propose a debris flow erosion rule that includes components of both sliding and impact wear, whose relative importance is scaled by experimentally-tested variables. Finally, as part of my investigations of controls on boundary forces and bedrock wear, I observed grain segregation processes and fluid-sediment interactions that were previously undescribed in the literature. These included lateral oscillations of the flow front and the formation of asymmetric coarse-particle gyres. I documented the first order controls on segregation by particle size, boundary conditions, fluid content, and fluid viscosity.

The amount of water runoff need to evolve landscapes is rarely assessed. Empirical studies correlate erosion rate to runoff or mean annual precipitation, but rarely is the full history of a landscape known such that it is possible to assess how much water was required to produce it. While this may not seem to be of primary importance on Earth where water is commonly plentiful, on Mars the amount of water to drive landscape evolution is a key question. Here we tackle this question through a series of five chapters, one devoted to field work at Meteor Crater, another to laboratory experiments about controlling processes, and then two chapters on analysis of landforms and implications of water runoff on Mars (associated with the Mars Science Laboratory mission to Gale Crater), and then we complete this effort with a consideration of how we can reliably assign relative timing between events resulting in small depositional features. What follows below is a summary of what is found in each chapter.

Meteor Crater, a 4.5 km2 impact crater that formed ~50,000 years ago in northern Arizona, has prominent gully features on its steep walls that appear similar to some gullies found on Mars. At the crater bottom, there are over 30 meters of lake sediments from a lake that disappeared ~10,000 to 11,000 years ago, indicating the transition from the Pleistocene to the current, drier climate. A combination of fieldwork, cosmogenic dating, and topographic analysis of LiDAR data show that debris flows, not seepage erosion and fluvial processes as previously suggested in the literature, drove gully incision during their formation period of ~40,000 years before the onset of the Holocene. Runoff from bare bedrock source areas high on the crater wall cut into lower debris mantled slopes, where the runoff bulked up and transformed into debris flows that carried boulders down to ~5 to 8 degree slopes, leaving distinct boulder lined levees and lobate tongues of terminal debris deposits that crisscrossed on the lower slopes. We hypothesize that the fine material, likely generated in the impact, and deposited with the coarse debris on the lower portion of the crater wall, is key to this bulking up process as flows cut across the deposits. Fluvial processes following the debris flow gullies extended alluvial deposits to the crater floor and contributed to lake infilling. Cosmogenic dating confirms that most of the modification of the crater walls occurred before the early Holocene. To account for the 75 distinct deposits currently lying on the crater floor, debris flow frequency would be about 1 event every 17 years, assuming debris flow activity terminated ~10,000 years ago. Assuming a water-to-rock ratio of 0.2 at the time of transport, it would have taken ~100,000 m3 of water to transport the ~500,000 m3 of debris flow deposits on the crater floor. Given the 4.5 km2 size of the crater, this extensive erosion would require less than 0.02 m of total runoff, or the equivalent of just 0.001 mm/year over a 40,000 year period. This insignificant amount of water was likely packaged into short-lived storm or snow-melt events when debris flows were generated. Much more runoff did occur, as evidenced by the lake and fluvial deposits, as well as the likely cool, wet conditions of the late Pleistocene. This suggests only a small fraction of the total runoff is needed to do considerable geomorphic evolution, producing strongly gully-scared crater walls. Currently, only minor fluvial modification of the gully networks occurs.

To test the hypothesis that fine-grained material is a necessary component behind sustained mobilization in granular debris flows, we performed a series of experiments in a 4 m diameter vertically rotating drum. Specifically, we assessed the role of fines and coarse grain size distribution on the rate of pore fluid pressure generation and its dissipation upon cessation of movement of the granular mass. We varied the amount of fines, from no fines to amounts found in debris flow deposits, and varied the coarse grain size distribution, from a single grain size (10 mm) to a range found in a natural flow deposit. The mixtures with fines contents close to those found in actual debris flows had elevated pore pressures, indicating they were almost fully liquefied, though the highest excess pore pressures were generated with the combination of fine material and a wide coarse grain size distribution. A boulder was also placed upon the fines-rich mixtures just after cessation of motion and it was observed that the pore fluid pressure rose instantly, bearing most of the grain's weight, and then the pressure slowly declined. When the same boulder was placed on the water-gravel flows, there was no change in fluid pressure, as the mass was supported entirely by grain-grain contacts. Our observations, combined with observations of others in the literature, inform the conceptual model presented herein where we hypothesize that the dilational separation of particles during debris flow shear leads to coarse particles (in this case particles >12 mm) slowly settling through the highly viscous non-Newtonian fluid. As the fines-rich mixture has a yield strength of ~5 Pa, we calculate that particles ~12 mm and smaller remain suspended. This combined effect requires the fluid to sustain the weight of the particles, leading to pore pressures equal to the entire weight of the solid and fluid mass.

As part of the MSL science team, we focused on understanding the geomorphology and its implications for the hydrologic history of the landing site for the Curiosity rover, which was located at the distal end of the Peace Vallis fan in Gale Crater. The Peace Vallis fan covers 80 km2 and is fed by a 730 km2 catchment. Valley incision into accumulated debris delivered sediment through a relatively low-density valley network to a mainstem channel to the fan. An estimated total fan volume of 0.9 km3 matches the calculated volume of removal due to valley incision (0.8 km3). The fan profile is weakly concave up with a mean slope of 1.5% for the lower portion. Runoff (discharge/watershed area) to produce the fan is estimated to be more than 600 m, perhaps as much as 6000 m, indicating a hydrologic cycle that likely lasted at least 1000's of years. Atmospheric precipitation, possibly snow, not groundwater seepage produced the runoff. Based on topographic data, Peace Vallis fan likely onlapped the rise onto which Curiosity landed, Bradbury Rise, and spilled into a topographic low to the east of the rise. This argues that the light-toned fractured terrain within this topographic low corresponds to the distal deposits of Peace Vallis fan and in such a setting lacustrine deposits were predicted (and later confirmed by observations from the rover).

Previous studies had suggested that there might have been large lakes in Gale. This history is important to understanding Mars' climate history and the volumes of water that may have passed through or covered the sediments encountered by Curiosity. Here we use improved imagery, topography, and ground-based observations from the Curiosity rover to map deltas, fans and gullies, which leads to a proposed history of lakes in Gale. We report evidence for at least three distinct large lakes within Gale, all which occurred after the crater's central mound, Aeolis Mons, reached close to its current topographic form. From corresponding deltaic deposits off the southern rim of Gale crater and the southern flank of Aeolis Mons, we identified the highest lake level at -3280 meters, which would have had a mean depth of 700 meters. The larger context provided by craters near Gale suggests that this lake was derived from regionally sourced water from the south. The next lake level, which was established after a period of drying with subsequent lake level decline before rewetting and lake level rise, is defined by four deltaic features as well as the termination of gullies around the northern rim of Gale. This second lake corresponds to an elevation of -3980 m, has a mean depth of 300 meters, and was likely sourced from water more local to Gale. Lake levels then rose to the third lake level, at -3780 m, with a mean depth of 400 meters. Two deltaic deposits derived from sediment from the rim of Gale define this lake. This last lake declined sufficiently quickly to prevent incision into any of the deltaic forms around the crater, though hydrologic activity likely continued as evidenced by a time of fan building, including the Peace Vallis fan, around the crater. Quantification of the fan and delta volumes and their gully sources on the crater wall and rim, and on Aeolis Mons, suggest that these discrete deposits contain essentially all of the sediment mobilized from the upslope source areas. Importantly for the MSL mission, this suggests that most of the sediments the rover has and will encounter experienced at least a few cycles of drying and wetting.

The depositional features (i.e. alluvial fans and deltas) within Gale Crater are small but they likely record a time history of lakes and other depositional events within the crater. An obvious question to ask is: can crater counting, which has been used to date much of Mars' surface, be used on such small surfaces to provide accurate age dating? Here we introduce a simple simulation model to quantify the effects of sample area size and crater obliteration effects on age estimates derived from crater size-frequency distributions. Our results show, quantitatively, that crater counting to estimate age of small surfaces (<1000 km2) has unavoidable large uncertainties. This arises because the small craters, which are most numerous, are eliminated due to erosion and infilling and there is a low probability of counting large craters on small sample areas. These effects on small and large craters leads to a narrow range of crater sizes in which the correct age may be reflected in the crater density functions that are used to date surfaces. This range decreases with decreasing surface area and increasing erosion and infilling. Steps in the isochron data, in which crater density data are shifted downward for some crater sizes to a different density size relationship, are often cited as indicators of resurfacing events. We find, however, that such steps occurred randomly in ~5% of the crater size-frequency distributions we generated. This can lead to large errors when determining which tangential isochron is used to assign a surface age. Our modeling, which accounts for obliteration of craters and a reduced chance of encountering large craters on small and young surfaces, suggests that in general the least reliable ages occur between 1 and 3 Ga years. Younger surfaces preserve the smaller craters and older surfaces ones collect larger ones, both improving the probability of obtaining the correct age. For areas less than 1000 km2 and true ages of 1 to 3 Ga, there is only a 20 to 40% probability of the calculated age lying within 0.25 Ga of the true value, where as for surfaces smaller than 500 km2 (most mapped Mars fans are smaller) the corresponding probabilities drop to about 10 to 20%. To highlight the constraints on dating small deposits, we applied this model to cratering data from the Peace Vallis fan (~80 km2) and one of the larger deltaic deposits within Gale crater (117 km2), which deposited at different elevations and likely record different hydrologic events. Even though portions of the observed crater count data in each case follow the proposed crater density isochrons, the uncertainties for such small areas are so large as to make the age assignment unreliable.

Gravel-bed rivers commonly display distinct sorting patterns on their beds. Visually, this heterogeneity often appears to form an organization of textural patches or facies. Patches can be distinguished into &ldquofree patches,&rdquo which are zones of sorted material that move freely, such as bedload sheets; &ldquoforced patches,&rdquo which are areas of sorting forced by topographic controls; and &ldquofixed patches&rdquo of bed material rendered immobile through localized coarsening that remain fairly persistent through time. The local bed surface grain size, and therefore bed surface patchiness, exerts considerable influence on local bed mobility, bedload transport rates, hydrodynamic roughness, and benthic microhabitats. Despite the ecological and morphodynamic importance of bed surface patchiness, we lack accurate and objective methods to delineate patches, and have little understanding of how patches form, evolve, and interact.

This dissertation investigates the mechanisms responsible for the formation of bed surface patches. First, two sets of flume experiments are used to explore how fixed and free patches respond to reductions in sediment supply, and to show that sediment supply is a primary control on the distribution of fixed patches and the dynamics of bedload sheets. A near-field scale flume experiment is then used to show how bed topography, the flow field, and the sediment transport field interact to form forced patches. Size-selective, cross-stream sediment transport is shown to be a critical mechanism responsible for the development of forced bed surface patches. A high-resolution dataset of spatial grain size distributions collected during this experiment is then analyzed with several clustering methods to develop an objective method of delineating patches. These analyses suggest that rivers have a tendency to form a finite number of patch types that become distributed throughout the reach. Lastly, a two-dimensional morphodynamic model simulating flow, mixed-grain size sediment transport, and bed evolution is presented and used to show that forced patches interact with the evolving bed and, through their effects on the flow field, have a considerable impact on morphologic evolution.

On large lowland rivers, overbank flows can deposit as much as 40% of the river's sediment load across adjacent floodplain environments on an annual basis. The spatial pattern of this deposition and the corresponding grain size distributions are not well known, despite the importance of understanding floodplain evolution, the generation of alluvial architecture, and the spread of contaminants. The lack of field data of floodplain deposition also limits our ability to evaluate models of floodplain sedimentation. Here I report floodplain overbank depositional patterns and grain sizes on the Strickland River in Papua New Guinea, compare these data with similar data from the Fly River, to which the Strickland drains, and explore distinct differences between the two rivers using a simple 1-D diffusion-advection model to predict the field observations.

The Strickland and Fly Rivers join at just 6 m above sea level, and have experienced the same Holocene sea-level rise. Historically, the Strickland has carried about seven times as much sediment load and twice as much water discharge as the Fly. As a result of this higher sediment flux, the shorter lowland Strickland River floodplain should be more developed in response to Holocene sea-level rise and, consequently, should presently be storing proportionately less sediment than the floodplain of the middle Fly River. A field campaign to collect samples to document floodplain sedimentation rates was conducted in 2003. Mine-derived elevated Pb and Ag concentrations in 111 shallow (<1 m) floodplain cores collected in 2003 were used to determine sediment deposition rates across the lower Strickland floodplain. Observed sediment deposition rates decrease across the floodplain with distance from the channel bank, and the average rate of deposition was 1.4 cm/yr over the first 1 km.

The magnitude of overbank deposition along the lowland sand-bedded Strickland resulted in a ~13% loss of the total sediment load, corresponding to overbank deposition of about ~0.05% of the total load per km of channel length of the mainstem. Deposition rates over the first 1 km from the channel bank on the Strickland were about ten times higher than those estimated for natural sediment loads on the Fly. However, the proportion of the sediment load deposited per channel length on the Strickland was less than that of the Fly (0.09%/km of mainstem channel length) due to an extensive network of tributary and tie channels that convey sediment to the floodplain on the Fly. Furthermore, the lateral migration of the Strickland channel was ~5 times that on the Fly, such that most overbank deposits on the Strickland were returned to the channel, causing the net loss of sediment to the floodplain to be small.

In addition to the differences in the magnitude of deposition, the pattern of declining floodplain deposition with increasing distance from the channel differed between the two river systems: deposition decreased exponentially with distance from the channel on the Fly River floodplain and non-exponentially on the Strickland River floodplain. In order to assess the processes controlling this difference in the pattern of overbank deposition, sediment grain size distribution needed to be assessed. Particle size distributions for floodplain samples from the same cores used to determine deposition rates were measured using a Coulter LS laser particle sizer. Sediments from the Fly River floodplain did not vary much in size with distance from the channel and sand comprises about 5% of samples at all distances from the channel. This contrasted significantly with sediment size distributions across the Strickland floodplain, where sediments rapidly fined with distance from the channel. There was little sand or coarse silt present beyond 400 m from the channel and clay and fine silt comprised the majority of sediment beyond this distance.

Extensive studies of the Fly and Strickland River floodplain deposits motivated further study of the mechanisms of deposition. The differences in both the rate of deposition and the variation of grain size as a function of distance from the channel were hypothesized to be due to a difference in the relative importance of diffusive and advective transport of sediment from the channel to the floodplain. A 1-D numerical model was developed to investigate this hypothesis. The model indicated that lateral advection was the primary transport mechanism of sediment across both floodplains and that the key difference in depositional pattern may be explained by flocculation of all sediment particle sizes transported on to the swampy, highly-organic waters of Fly River floodplain and flocculation of primarily fine sediment transported across the less-swampy Strickland River floodplain. Lateral variability in accretion from overbank deposition can be explained by a mechanistically-derived exponential function that is solved by balancing a simple advection-deposition transport formula, but the parameters needed to evaluate the relationship are difficult to measure directly. Model parameters were estimated from the limited data available. A depositional length-scale was used to calculate the possible duration of active deposition, the effective settling velocity of flocculated particles, and floodplain flow velocity. The non-exponential deposition pattern observed on the Strickland River floodplain was likely the result of deposition of a distribution of particles sizes and a distribution of effective settling velocities. The observed non-exponential accretion on the Strickland River floodplain arose from the sum of the exponential functions for each particle size. The flocculated particles on the Fly River floodplain likely had a single effective settling velocity resulting in the observed exponential deposition pattern and lack of fining in overbank deposits.

The Fly and Strickland Rivers serve as a unique natural laboratory where relatively undeveloped, large lowland river systems with similar climate, geology, ecology and history can be studied to examine the fundamental processes controlling the evolution of alluvial valleys. The extensive analysis of floodplain sediment cores demonstrated that the rate, pattern and proportional loss of sediment overbank was quite different on the two systems as a result of differing sediment loads, subsequent adjustment to Holocene sea level rise, and, probably, the relative importance of flocculation of sediments transported overbank. Despite these differences, the overbank deposition pattern on both river systems can be describe by a simple advection-settling model. The pattern of deposition can be used to evaluate the model parameters and quantify key variables such as particle and flow characteristics which are logistically difficult to obtain on large floodplains. This understanding enhances our ability to evaluate floodplain sediment transport dynamics based upon depositional records and identifies key measurements that should be incorporated into future studies.

Earth's critical zone includes plants as well as the heterogeneous near-surface layers into which they are rooted: the soil, saprolite, and weathered bedrock. It is within the critical zone (CZ) that water is stored and then released to streams and withdrawn by vegetation, highlighting its importance to humans, ecosystems, and the functioning of the Earth system in general. The research presented here is motivated by the global-scale challenges of mapping and predicting subsurface critical zone structure and understanding how that structure impacts water routing and storage for terrestrial ecosystems. It addresses these challenges by shedding light on inter-related ecohydrologic puzzles presented at the regional-scale across natural landscapes in the Eel River watershed: What explains the persistence and dominance of a particular species of oak, Quercus garryana, across savannas typified in the dry season by rolling golden hills mantled by senesced annual grasses? What explains the sharp ecotone that extends for hundreds of kilometers between this oak savanna community and a dense evergreen forest in a region of similar climate? Why did plant communities in this region fare better than others in the face of recent extreme drought?

By studying Quercus garryana's ecophysiology, I show that the oak is extremely water-limitation tolerant, which explains its ability to persist where a thin subsurface critical zone provides limited water storage capacity. Through sapflow monitoring on mature trees inhabiting an upslope position in the Central belt melange of the Franciscan, I reveal that the oaks maintain high rates of transpiration throughout the summer dry season, even as pre-dawn water potentials dropped to very low levels (below -3 MPa). At this site, Douglas fir has not encroached upon the oak groves like it has elsewhere throughout their mutual range. This is presumably due to Douglas fir's lower water limitation tolerance, and anticipates Quercus garryana's likely persistence in a warmer, drier future along western North America.

Through geospatial analysis coupled with intensive hillslope- and catchment-scale hydrological observation, I show that under a similar climate, adjacent landscapes across an extensive region of the Northern California Coast Ranges are either evergreen forest or deciduous oak savannah depending on their subsurface lithology. This is due to lithologically controlled contrasts in the extent of bedrock weathering and water storage capacity and thereby plant‐-available moisture in the summer dry season: a thick subsurface critical zone stores ample moisture and supports evergreen forest, whereas an adjacent thin subsurface critical zone sheds wet season rains and sustains savannah. The thick subsurface critical zone occurs in the modestly deformed shales and sandstones of the Coastal belt of the Franciscan; the thin subsurface critical zone occurs in the intensely deformed mud-matrix melange of the Central belt of the Franciscan. An important difference between the rock types is the thickness of the weathered bedrock layer; in contrast, the soil layer is of similar thickness. The thicker weathered bedrock at the Coastal belt site results in a larger rock moisture reservoir, i.e., unsaturated water residing within the weathered bedrock, which is primarily responsible for sustaining summer transpiration after shallow soils have dried. Extensive drilling was used to characterize the subsurface weathering extent and water storage dynamics, and was complemented by point--scale precipitation monitoring, catchment-wide remotely sensed evapotranspiration, and stream gauging to determine annual water budgets. Finally, late--summer dry season pre-dawn water potential observations in the dominant trees at both sites were used to compare subsurface water availability and show that where the weathered bedrock was thicker and hosted more rock moisture, the forest was less water stressed.

The observation of plant-available water in the dry season being mediated by the water storage capacity of the critical zone motivated a new hypothesis for how the subsurface can regulate plant response to drought: where annual rainfall reliably exceeds subsurface storage capacity, plant productivity and water use in summer should be insensitive to winter rainfall variability, because the subsurface water storage is replenished in wet and dry years alike. A simple ecohydrologic model for water storage in seasonally dry, Mediterranean climates predicts that storage will be replenished most reliably where the average precipitation is high relative to the subsurface storage capacity and winter evapotranspiration, and where year-to-year variability in precipitation is low. In order to test whether this storage-capacity limited behavior indeed decouples plants from inter-annual rainfall variability, running winter water balances were calculated for all unimpaired and generally undisturbed rain-dominated Mediterranean USGS catchments, using runoff, gridded precipitation and remotely sensed evapotranspiration as input data. These water balances revealed that sites in the Northern California Coast Ranges exhibited storage-capacity limitation, unlike most of the rest of the state. At all of the sites in the Northern California Coast Ranges, the consistent summer plant water supply resulted in plant insensitivity to rainfall variability, as revealed by remotely sensed summer plant greenness. Such sites are inherently resilient to meteorological drought, helping to explain the lack of mortality in the region in the extreme 2011-2016 drought.

The critical zone is defined as the thin outer veneer of Earth’s terrestrial surface, extending from the top of the vegetation canopy to the base of weathered bedrock. Very little is known about how the critical zone is structured and how its structure controls the storage, transport, and chemical evolution of the biosphere’s most important resource- water. In hilly or mountainous landscapes, the critical zone often includes tens of meters of weathered rock beneath the surface and this weathered rock hosts a dynamic hydrologic system that is virtually unexplored. Below weathered bedrock, lies an unmapped three dimensional fresh bedrock surface, Zb, that defines the bottom boundary of the critical zone. This dissertation develops novel theory to predict how this fresh bedrock surface is structured across ridge and valley topography and illustrates, through a field study, how that structure influences the routing of water within the landscape. I report, for the first time, how the structure and hydrologic dynamics of the critical zone vary across an entire hillslope, from channel to topographic divide.

Current models for development of the critical zone emphasize top-down processes associated with infiltrating waters and gases, as well as fracturing due to the differential stresses generated by topography. I propose a distinctly different theory, which enables a prediction of the thickness of weathered bedrock across a landscape. I hypothesize that as fresh bedrock, saturated with nearly stagnant fluid, is advected upward into the near-surface through uplift and erosion, channel incision produces a lateral head gradient within the fresh bedrock inducing drainage towards the channel. Drainage of the fresh bedrock causes weathering through drying (i.e. repeated cycles of wetting and drying) and permits the introduction of atmospheric and biotically controlled acids and oxidants such that the boundary between weathered and unweathered bedrock is set by the uppermost elevation of undrained fresh bedrock, Zb.

At steady-state the rate at which fresh bedrock crosses the Zb boundary is equal to the channel incision rate (which commonly is less than 1 mm/yr). Hence, this slow drainage of fresh bedrock, progressively allowing weathering to proceed, exerts a “bottom up” control on the advance of the weathering front. The thickness of the weathered zone is calculated as the difference between the predicted topographic surface profile (driven by erosion) and the predicted groundwater profile (driven by drainage of fresh bedrock). For the steady state, soil-mantled case, a coupled analytical solution arises in which both profiles are driven by channel incision. Lithology of the fresh bedrock influences the thickness of the weathered zone through the saturated hydraulic conductivity of the bedrock. Measurements of rate processes and topography, as well as depth to fresh bedrock at the divide can be used to estimate the saturated hydraulic conductivity and porosity of the fresh bedrock. Two non-dimensional numbers corresponding to the mean hillslope gradient and mean groundwater table gradient emerge and their ratio defines the proportion of the hillslope relief that is unweathered. The model predicts a thickening of the weathered zone upslope and consequently, a progressive upslope increase in the residence time of bedrock in the weathered zone. Despite its simplicity, the model makes testable predictions and is consistent with field data from three sites.

To investigate how the critical zone is structured across a hillslope and how water is routed throughout the critical zone, I conducted an intensive field investigation on a steep (average 30 degree), actively eroding (0.2-0.4 mm/yr), 135 m long soil-mantled hillslope within the Northern California Coast Ranges (referred to as Rivendell). The 4000 m2 hillslope is located within the 17 km2 Elder Creek watershed, in the Angelo Coast Range Reserve. The hillslope is forested with up to 60 m tall Douglas fir (Pseudotsuga menziesii) and mixed evergreen hardwoods including live oak (Quercus wislizeni), madrone (Arbutus menziesii), and California bay (Umbellularia californica), and is underlain by vertically dipping argillite with sandstone interbeds. The climate is seasonally dry, and characterized by warm, dry summers (May- Sept) and cool, wet winters within which all of the precipitation (1800 mm mean annual precipitation) falls.

A network of 12 wells, as deep as 30 m, were drilled across the hillslope into fresh bedrock and an extensive sensor network of over 750 sensors records soil moisture and rock moisture, and meteorological and groundwater conditions across the site. Streamflow at the base of the hillslope is recorded at a United States Geologic Survey station a short distance upstream. To document the spatial and temporal dynamics of rock moisture, I performed periodic neutron probe surveys within deep wells.

Drilling revealed a 4-25 m thick zone of variably weathered, fractured bedrock underlying, thin (<50 cm) soils. Intensely fractured argillite forms a saprolite in the upper 4 m, below which fracture density, porosity, and mechanical strength decreases with depth. Fresh bedrock at the base of the profile (revealed through large increases in standard penetration resistance and an absence of signs of oxidative weathering) bounds the weathered zone from below. The boundary between unweathered and weathered rock, Zb, is progressively deeper upslope, forming an upslope thickening wedge of fractured, weathered bedrock that is increasingly weathered upslope.

The seasonal addition of rainfall to this structured weathering profile, leads to the development of three distinct hydrologic zones: a near surface 4-18 m thick zone that remains unsaturated year round, a 4-15 m thick seasonally saturated zone that fluctuates largely within the same elevations year after year, and a zone which remains chronically saturated below an annually repeatable minimum water table position.

A significant consequence of the development of the weathering front into bedrock is that infiltrating rainfall travels through and is stored within weathered rock as rock moisture. Rock moisture is the exchangeable water within unsaturated weathered and fractured bedrock. It has been identified as an important source of moisture to vegetation, but is poorly documented due to its inaccessibility and therefore remains an unaccounted for, but important, component of the hydrologic cycle. Here, for the first time, I directly document the spatial and temporal dynamics of rock moisture throughout the critical zone.

Periodic surveys in deep wells reveal a seasonal cycle of rock moisture addition and depletion across the hillslope. This cycle begins with the first rains that mark the end of the dry season, which advance moisture into the soil and often up to 1 m into the weathered bedrock. Subsequent rains advance a wetting front through the upper 5-12 m of the profile, where increases in rock moisture storage are proportional to the addition of rainfall. In some instances, these early wet season storms generate a small, rapid but short-lived response of the water table.

Once cumulative rainfall has caused the local rock moisture storage to reach a capacity beyond which rock moisture no longer increases, groundwater responds to rainfall. Further incoming water is passed rapidly, via fracture flow, to the groundwater table. This rock moisture storage capacity, which is observed to be approximately the same each year, increases upslope from 85 to 615 mm, corresponding to the upslope increase in weathering of the bedrock. The average rock moisture storage in the chronically unsaturated zone across the hillslope is about 280 mm. The upslope increase in rock moisture storage needed to initiate the seasonal groundwater response leads to the condition where, early in the wet season, runoff is generated from the lower part of the hillslope while the upper part of the hillslope is still gaining moisture.

Once rock moisture is seasonally elevated, all infiltrating precipitation travels vertically through soil, saprolite, and weathered rock (we observe no overland flow or saturated flow within the soil). The timing of the rapid response of the groundwater system (~ hours) is highly variable for a given depth and does not appear to depend on travel distance to the water table. Additional storms throughout the wet season do not alter the structure or magnitude of rock moisture storage within the hillslope. Rock moisture storage is most significant in the upper 5-12 m and diminishes with depth to a zone where no detectable changes in rock moisture are observed despite the rise and fall of the water table within this zone. Rock moisture may occur as water along fracture surfaces or as water that penetrates the matrix blocks bounding fractures. At depth, the constant saturation of matrix blocks leads to the dominance of fracture flow, which drives the rapid (10-5 to 10-3 m/s) and significant (up to 11 m in a single storm) rise of the water table. On average, rock moisture changes of only 5% are needed to achieve saturation in the seasonally saturated zone. The dynamic and responsive, fracture dominated groundwater system leads to 97-99% of runoff in Elder Creek occurring during the wet season.The final storm of the wet season marks the initiation of the slow decline of rock moisture and groundwater within the hillslope throughout the long (>120 days) dry season.

The 30-130 mm of seasonal soil moisture storage is rapidly depleted within the first sev- eral weeks following the final storm. Approximately 12 weeks into a typically 18-week dry season, less than 15 mm of soil moisture remains, while up to 120 mm (53 mm average across the hillslope) is stored in the upper 10 m as rock moisture. An annual cycle is repeated each year: drying to a characteristic value and then wetting to a similar rock moisture storage capacity. Because the precipitation exceeds the rock moisture storage capacity even in dry years, the spatial pattern and magnitude of dry season rock moisture are insensitive to the precipitation magnitude and temporal pattern of the immediately preceding wet season, even in a significant drought year (approximately half mean an- nual precipitation). The drop in groundwater level over the last 2 months of the dry season (equivalent to approximately 5-50 mm) is also similar in different years. Runoff of just 5 mm per month is sufficient to sustain flow in Elder Creek and support salmonid populations.

Rock moisture, a direct consequence of the alteration of bedrock in the near surface, is a virtually unknown and unmapped component of the hydrologic balance. Here, I show that year after year, 30-60% of the incoming precipitation is stored seasonally as rock moisture. Nearly all of that water must be used in transpiration.Hence, rock moisture is a major source of water for vegetation. Because incoming rainfall first restores this moisture content before generating runoff, even in strong drought years, the rock moisture is available, and provides drought resilience. The ecohydrologic function of the critical zone at this site must therefore be divided between 1) near surface rock moisture storage that controls the exchange of gases and solutes and supports dry season transpiration and 2) the fracture dominated seasonally perched groundwater system that routes most precipitation as runoff and controls both peak and low streamflow and aquatic habitat. The dual function of seasonal storage and rapid transmission of water is the defining feature of the rock moisture system.

The rock moisture dynamics documented here explain several previously observed processes at Rivendell. The seasonal build up of rock moisture leads to mixing of waters and damping of the stable isotope signature of storm events. Within the dynamic rock moisture zone, seasonal stimulation of subsurface microbial communities (as expressed in gas composition) occurs and cation exchange processes likely drive the solute chemistry of water recharged to the groundwater and drained as runoff. Rock moisture is likely important in a wide range of settings. I propose here the possibility that there may be a co-evolution of vegetation and critical zone structure wherein the water extraction by trees accelerates weathering of the bedrock, which in- creases rock moisture retention. Further field characterization is needed in a range of climates lithologies, and tectonic settings to document the critical zone structure, its rock moisture characteristics, and ecosystem dynamics. Modeling has begun to account for rock moisture dynamics and runoff through fractured bedrock, however, further field investigation will help guide models that predict critical zone development over sufficiently large areas to inform regional hydrologic, climate, and ecologic models.

It is surprising that gravel-bed rivers meander. In laboratory settings using cohesionless sediment transported as bedload, braiding inevitably emerges. The easily entrained outer bank sediments lead to relatively rapid bank erosion, the inner bank bar reaches it shallowest depth towards the center of the channel, the flow diverges, and braiding arises. These observations have led to the inference that meandering requires some bank strength to slow outer bank erosion and allow the inner bank to keep pace with outer bank retreat. Commonly it has been suggested that vegetation can provide that outer bank strength. While this view has led recently to numerical modeling incorporating vegetation strength, no studies had successfully created meandering rivers in the laboratory such that controlling mechanisms could be explored. Nor has there been an effort to delineate under what general conditions gravel bedded meanders are found in the field. These knowledge gaps are particularly important to stream restoration work because the creation of morphologically-stable gravel-bedded meanders is a common goal.

This dissertation uses flume experiments and a compilation of field data to explore the conditions that support meandering in gravel-bed rivers. Using alfalfa sprouts as model vegetation, sand as scaled-down gravel, and a lightweight plastic as scaled-down sand, I created for the first time a self-maintaining, laterally migrating meandering river with cutoff loops in a 6.1-m wide, 17-m long laboratory flume. The channel, 0.4 m wide, had a sinuosity of 1.1-1.2 and transported sediment with a median size of 0.78 mm. The sinuosity of the channel increased as bends grew, decreased as the channel cut off, and was regenerated while the channel maintained a steady width. In this experiment, we found that a steady bankfull flow was sufficient to sustain a meandering planform, and that application of higher peak flows caused the channel to widen because bank erosion outpaced bar growth. Coarse sediment was exchanged between eroding banks and the next bar downstream with little net downstream flux, consequently to prevent aggradation and avulsion at the upstream end of the experimental reach required turning off the coarse sediment feed. The input of fine sediment was crucial for blocking chute channels (a locus for braiding), filling point bars downstream of the bend apex, and plugging abandoned channels following cutoff. Hence, sustained growth of bends and development of meandering requires more than just sufficient bank strength to slow the outer bank erosion rate.

A compilation of 166 gravel-bedded meanders from the literature shows that gravel meanders primarily occur in lower gradient reaches extending from mountain ranges, high-elevation valleys, and in humid areas influenced by glaciation or glacio-fluvial outwash. Unexpectedly, analysis of Google Earth imagery showed that about 1/3 of the dataset lacked cutoff scars and other evidence of active migration. Gravel bedded meanders with median surface grain size> 10 mm had low Shields stress, particularly for meanders with cutoffs and without islands. Gravel bedded meanders with median surface grain sizes < 10 mm had higher Shields stresses that were transitional between coarser gravel bedded meanders and finer sand bedded meanders. Calculation of potential sediment transport rates show that for 16 of the 18 gravel bedded meanders with migration rate data, the sediment flux from bank erosion exceeds this transport capacity, suggesting that bank erosion is a primary source of sediment. Collectively this shows that gravel bedded meanders can be classed into different types that reflect the relative migration rates and associated Shield stress and that generally gravel-bedded meanders are associated with low Shield stresses and relatively low sediment supply.

Because of the importance of sediment supply in meandering rivers, I also investigated the morphological effects of doubling supply to a self-formed sinuous channel with 9 bends that was fixed in place. Five of these bends developed bar-pool topography. The doubling of supply caused the channel to steepen by 33%, and the topographic response of bars was limited to shoaling of two pools as they filled with coarse sediment. One of the bars also extended upstream. Three other bends had limited changes to an increase in supply, likely due to the fixed walls that supported wide bars and narrow, deep pools. The maximum aggradation was greater than a channel depth, indicating that for a similar supply increase in a freely migrating channel would lead to avulsion. This experiment supports further the interpretation that gravel-bedded meanders are associated with relatively low sediment supply.

The final experiment replicated our first alfalfa experiment but used a constant discharge and sediment supply with much denser alfalfa, in order to explore the effect of increased bank strength on channel morphodynamics. The denser alfalfa led to a narrower channel, but the migration rate was similar to our first experiment. As the channel migrated, associated decreases in the water surface slope caused the style of migration to change from focused on the entire bend to focus at the bend apex, creating skewed bends. As the bend skewness increased, the sediment flux at the flume outlet decreased, and most of the sediment flux out of the experimental reach was the lightweight plastic (model sand) rather than the coarse sediment supply. This created a channel with a sinuosity of 1.6, but then the channel aggraded and cut off, decreasing the sinuosity (to 1.2) and a corresponding increased slope. Following the cutoffs, the sediment transport rates and channel migration rates increased. The steady flow discharge and dense alfalfa limited cutoffs until aggradation altered floodplain hydrology to suppress alfalfa growth and create relatively weak paths through the floodplain.

Taken together these studies show that for gravel bedded rivers to meander three conditions are needed: 1) additional bank strength (here from vegetation) to slow erosion rates' 2) low sediment supply and correspondingly low Shields stresses to prevent aggradation and avulsion as sinuosity increases; and 3) fine sediment to fill chutes and the downstream end of bars. Variable peak discharges are not necessary but may contribute to cutoffs that maintain a steeper slope able to transport the sediment supplied to the reach. These observations suggest that creating a self-maintaining laterally migrating gravel-bedded river as a restoration outcome will require an assessment of likely coarse and fine sediment supply as well as the ability of vegetation to provide sufficient bank strength.