A full waveform inversion of the kinematics of a 3-D source inside an artificial rock plate is presented. The source is provided by a piezoelectric disc embedded inside a gypsum plate, excited by an arbitrary electrical signal. Elastic waves emitted from the source propagated through the plate and were recorded by an array of wide-band high-fidelity microseismic displacement sensors. These microseismic data are then inverted by deconvolving the recorded signals with the theoretical Green's functions to obtain the kinematics of the source modeled by force couples. The force time functions thus inverted are justified qualitatively by the known mechanical behavior and compared quantitatively to simplified electro-acoustic equations. For the first time, the combined use of a wide-band high-fidelity sensor, Green's functions incorporating Q damping, effective point source theory, and linear deconvolution yield physically justified time functions of the source kinematics. (c) 2004 Elsevier Ltd. All rights reserved.

Water shortages, particularly evident in the state of California, emphasize the need for a better hydrologic understanding, and improved water management techniques. The majority of the state's water originates in the Sierra Nevada as snow, melting throughout the year to meet the needs of various stakeholders. Current measurement techniques are unable to resolve variability of the snowpack at the basin scale, and snowmelt processes are not well captured by existing hydrologic models. A system-level solution is introduced to facilitate scientific understanding and water management decisions in basins of the Sierra Nevada. The core of this thesis focuses on expanding current sensing methods at the hydrologic catchment (1-2km2) scale to develop an improved understanding of the mountain water balance. It is shown that Wireless Sensor Networks (WSN) offer an ultra low-power, cost-effective solution to instrument catchment-scale regions. An explicit WSN deployment strategy is derived, leveraging in-situ network statistics of the Packet Delivery Ratio, and Received Signal Strength Indicator to optimize network performance. A hydrologic variability analysis is carried out on the dataset collected by the network, showing temporal stability in the variability of snowdepth at the catchment scale, and validating the use of a stratified sampling approach based on the even instrumentation of physiographic variables. To derive estimates of Snow Water Equivalent (SWE) at the larger basin scale, an estimation framework based on Gaussian Processes, and an optimal sampling strategy, which maximizes Mutual Information, is described. The sensor placement method outperforms a number of other sampling designs, reducing estimation error by up to 100mm. A computationally tractable, Hybrid System model of snow dynamics is introduced and shown to accurately reflect the various stages of snowmelt when compared to observations in the Sierra Nevada. The need for improved sensing and estimation procedures is highlighted by an analysis of the effects of improved SWE estimates on basin-scale streamflow forecasting methods. The use of historically reconstructed SWE data shows that streamflow forecasting error could be reduced by nearly 10% through improved SWE estimates.

Hydrologic modeling is heavily used in water resources for advancing scientific process understanding and supporting operational decision making, but the capabilities of these models are challenged by climate change, land-use changes, and population growth. Novel ground-based and remotely sensed data sources can provide high-quality and reliable information that can enhance hydrologic models to meet current demands. Here, I examine how these data sources can be leveraged to facilitate and improve use of three major model types: data-driven, physically based, and conceptual, which each fall in a different range of complexity (the extent to which a model represents or simplifies physical processes) and scale (the number of simulation or data points). I focus on the state of California, a region that faces significant water management challenges due to its arid Mediterranean climate, highly seasonal precipitation, and high variability in annual precipitation totals. Chapter II investigates snowpack response to rain-on-snow events by leveraging expanded ground-based wireless-sensor networks to drive a machine-learning algorithm. Supported by the high temporal and spatial resolution of these data, I identified a nonlinear and temporally variable relationship between Leaf Area Index and snow-depth change during rain-on-snow events. Chapter III demonstrates how remotely sensed spatial maps can be used to identify the optimal spatial distribution of input data in a physically based hydrologic model through use of a Gaussian-Mixture Model. This novel method addresses issues of overparameterization and improves the efficiency of model design for physically based models while still being grounded in physical principles. Finally, Chapter IV applies new spatially distributed data to a conceptual hydrologic model to quantify the impact of predictable versus unpredictable shifts in the precipitation-runoff relationship during droughts. Collectively, this work demonstrates that 1) data that are gathered strategically by location and water-balance component is most helpful in improving or developing new models; 2) remotely sensed data are most effective when calibrated or coupled with ground-based sensors; 3) scale continues to be a challenge both on the modeling and data sides, whether scaling up physically based models or scaling down conceptual models; and 4) model improvement relies not only on data collection but on making the data usable and easily accessible, particularly for operational applications.

The dynamic transition from slow to rapid sliding along a frictional interface is of interest to geophysicists, engineers and scientists alike. In our direct shear experiment, we simulated a pre-existing frictional fault similar to those occurring naturally in the Earth. Laboratory experiments have been used in the past to successfully link the fields of rock mechanics and seismology and have been able to produce better estimates of seismic hazard. These laboratory studies found that prior to large earthquakes a nucleation region grows outwardly. Within this region shear stress is overcome through the accumulation of slip across the fault. This phenomena is referred to as slip-weakening. Once the region grows large enough an earthquake ensues. The growth phase of this nucleation zone is called the premonitory phase. Newer seismological instruments are showing that this phase actually carries seismic signatures (originally thought to be an aseismic process). This precursory seismicity is observed as foreshocks and migrating swarm-like foreshocks occur at the fringe of this expanding nucleation front. The previous laboratory studies, used to develop the major phenomenological earthquake models, did not observe this precursory seismicity, possibly due to the lack of sensors capable of measuring this high-frequency phenomenon. The laboratory study reported here has incorporated appropriate sensors that can detect foreshock events on the fringe of a nucleation zone prior to a gross fault rupture (main shock).

\par \ident During loading we observed foreshocks sequences as slip transitioned from slow to rapid sliding. These laboratory-induced foreshocks showed similar acoustic characteristics and spatio-temporal evolution as those detected in nature. Through direct observation (video camera), foreshocks were found to be the rapid, localized (millimeter length scale) failure of highly stresses asperities formed along the interface. The interface was created by the meshing of two rough polymethyl methacrylate (PMMA) bodies in a direct shear configuration. A carefully calibrated pressure sensitive film was used to map the contact junctions (asperities) throughout the interface at a range of applied normal loads $F_{n}$. Foreshocks were found to coalesce in a region of the fault that exhibited a more dense distribution of asperities (referred to as the \emph{seismogenic} region).

\par \indent Microscopy of the interface in the \emph{seismogenic} region displayed a variety of surface roughness at various length scales. This may have been introduced from the surface preparation techniques use to create a mature interface. The mature interface consisted of `flat-topped' asperity regions with separating sharp valleys. The `flat-topped' sections spanned millimetric length scales and were considerably flatter (nanometric roughness) that the roughness exhibited at longer length scales (tens of millimeters). We believe that the smoother, `flat-topped' sections were responsible for the individual asperity formation (determining their size and strength), whereas the longer length scale roughness influenced the asperity-asperity interaction during the nucleation phase. Asperities in the \emph{seismogenic} region where shown to exist close enough to each other so that elastic communication (through the off-fault material) could not be neglected.

\par \indent Prior to gross fault rupture (i.e. mainshock), we measured the propagation of a slow nucleating rupture into the relatively `locked', \emph{seimsogenic} region of the fault. Slow slip dynamics were captured using slip sensors placed along the fault that measured a non-uniform slip profile leading up to failure. We found that the propagation of the slow rupture into the locked region was dependent on the normal force $F_{n}$. Higher $F_{n}$ was found to slow the propagation of shear rupture into the locked region. Within the relatively `locked' region, a noticeable increase in size and a more compact spatial-temporal distribution of foreshocks were measured when $F_{n}$ was increased.

\par \indent In order to develop an understanding of the relationship between $F_{n}$ and the resistance of the fault to slow rupture, a quasi-static finite element (FE) model was developed. The model used distributions of asperities measured directly from the pressure sensitive film in a small section of the interface where foreshocks coalesced; specifically, the region where the slowly propagating slip front encountered the more dense distribution of asperities. A single asperity was modeled and followed the Cattaneo partial slip asperity solution. As the shear force increased along the fault, the asperities in this model were able to accommodate tangential slip by entering a partial sliding regime; the central contact of the asperities remained adhered while sliding accumulated along its periphery. Partial slip on the asperity propagated inwards as the shear force was incrementally increased. A further increase in the shear force caused the asperity to enter a full sliding condition. Increasing confining loads caused increased stiffness and increased capacity to store potential shear strain energy -- a possible measure of the `degree of coupling' between the fault surfaces. Physics from the numerical model followed the qualitative observations made using photometry of asperities along the interface, which visualized asperities in the `locked' region -- larger asperities remained stuck throughout the loading cycle and the light transmitted through individual asperities decreased from the periphery as shear loads increased.

\par \indent The numerical partial slip, quantified by the potential energy stored by the asperity, increased relative to the normal pressure $p$. Asperity-asperity interactions were modeled along the interface using a quasi-static analysis. Progression of slip into the asperity field was increasingly inhibited as the normal confining force $F_{n}$ was increased. The computational model provided an explanation as to why an increased confining force $F_{n}$ could result in an increased resistance to slow rupture as well as an increased potential for larger foreshocks within the resistive, relatively `locked' section of a fault. The experiments and modeling presented in this study lay the foundation for more innovative laboratory work that could potentially improve the phenomenological models currently used to estimate earthquake hazard.

The use of wireless sensor networks(WSN) to study the mountain hydrologic cycle has been proposed as a supplement to existing systems used for ground-based hydrologic and ecological monitoring. Historically, the study of mountain hydrology and the water cycle has been largely observational, with meteorological forcing and hydrological variables extrapolated from a few infrequent manual measurements. Recent developments in Internet of Things (IoT) technology are revolutionizing the field of mountain hydrology. Low-power WSNs can now generate denser data in real-time and for a fraction of the cost of labor-intensive manual measurement campaigns. This research details the requirements and different technical options, describes the technology deployed in the American River basin, and discusses the methods associated with modeling large-scale environmental monitoring in extreme conditions. The American

River Hydrologic Observatory (ARHO) project has deployed fourteen low-power wireless IoT networks throughout the American River basin to monitor California’s snowpack. A network of sensors for spatially representative water-balance measurements was developed and deployed across the 2154 km2 snow-dominated portion of the upper American River basin, primarily to measure changes in snowdepth and soil-water storage, air temperature, and humidity. The WSNs, each has 10 measurement nodes that were strategically placed within a 1-km2 area, across different elevations, aspects, slopes and canopy covers.

The research evaluates the accuracy of a machine-learning based path loss model to estimate the expected radio transmission distances. The model is trained on 42,157,324 RSSI samples collected over seven months from the ARHO WSN. The 2218 links in the network span across the upper portion of the American River basin and are deployed in a complex environment, with large variations of terrain attributes and vegetation coverage. The model is based on an ensemble regression tree machine learning algorithm (Random Forest). The independent variables used in the model include path distance, canopy coverage, terrain variability, and path angle. The accuracy of this model is compared to several well-known canonical and empirical propagation models. This model showed a 37% reduction in the average prediction error compared to the canonical/empirical model with the best performance. The research presents an in-depth discussion on the strengths and limitations of the proposed approach as well as opportunities for further research.

Compared to existing operational sensor installations, the ARHO WSN reduces hydrologic uncertainty in at least three ways. First, redundant measurements improved estimation of

lapse rates for air and dew-point temperature. Second, distributed measurements captured local variability and constrained uncertainty in the air and dew-point temperature, snow accumulation and derived hydrologic attributes important for modeling and prediction. Third, the distributed relative-humidity measurements offer a unique capability to monitor upper- basin patterns in dew-point temperature and better characterize elevation gradient of water vapor-pressure deficit. Network statistics during the first year of operation demonstrated that the ARHO WSN was robust for cold, wet and windy conditions in the basin.

Using daily dew-point temperature and the amount of snow accumulation at each node to estimate the fraction of rain versus snow resulted in an underestimate of total precipitation be- low the 0 o C dew-point elevation, which averaged 1730 m across 10 precipitation events. Blending lower-elevation rain-gauge data with higher-elevation sensor-node data for each event pro- vided precipitation estimates that were on average 15-30% higher than using either set of measurements alone. Using data from the current operational snow-pillow sites give even lower estimates of basin-wide precipitation. Given the increasing importance of liquid precipitation in a

warming climate, a strategy that blends distributed measurements of both liquid and solid precipitation will provide more-accurate basin-wide precipitation estimates. The distributed, representative sensor-network measurements also improve upon operational estimates of snow- pack water storage, snowmelt amount and snowmelt timing across the basin.

Rapidly varying forces, such as those associated with impact, rapid crack propagation, and fault rupture, are sources of stress waves which propagate through a solid body. This dissertation investigates how properties of a stress wave source can be identified or constrained using measurements recorded at an array of sensor sites located far from the source. This methodology is often called the method of acoustic emission and is useful for structural health monitoring and the noninvasive study of material behavior such as friction and fracture. In this dissertation, laboratory measurements of 1-300 mm wavelength stress waves are obtained by means of piezoelectric sensors which detect high frequency (10 kHz- 3MHz) motions of a specimen's surface, picometers to nanometers in amplitude. Then, stress wave source characterization techniques are used to study ball impact, drying shrinkage cracking in concrete, and the micromechanics of stick-slip friction of Poly(methyl methacrylate) (PMMA) and rock/rock interfaces.

In order to quantitatively relate recorded signals obtained with an array of sensors to a particular stress wave source, wave propagation effects and sensor distortions must be accounted for. This is achieved by modeling the physics of wave propagation and transduction as linear transfer functions. Wave propagation effects are precisely modeled by an elastodynamic Green's function, sensor distortion is characterized by an instrument response function, and the stress wave source is represented with a force moment tensor. These transfer function models are verified though calibration experiments which employ two different mechanical calibration sources: ball impact and glass capillary fracture. The suitability of the ball impact source model, based on Hertzian contact theory, is experimentally validated for small (~1 mm) balls impacting massive plates composed of four different materials: aluminum, steel, glass, and PMMA.

Using this transfer function approach and the two mechanical calibration sources, four types of piezoelectric sensors were calibrated: three commercially available sensors and the Glaser-type conical piezoelectric sensor, which was developed in the Glaser laboratory. The distorting effects of each sensor are modeled using autoregressive-moving average (ARMA) models, and because vital phase information is robustly incorporated into these models, they are useful for simulating or removing sensor-induced distortions, so that a displacement time history can be retrieved from recorded signals. The Glaser-type sensor was found to be very well modeled as a unidirectional displacement sensor which detects stress wave disturbances down to about 1 picometer in amplitude.

Finally, the merits of a fully calibrated experimental system are demonstrated in a study of stress wave sources arising from sliding friction, and the relationship between those sources and earthquakes. A laboratory friction apparatus was built for this work which allows the micro-mechanisms of friction to be studied with stress wave analysis. Using an array of 14 Glaser-type sensors, and precise models of wave propagation effects and the sensor distortions, the physical origins of the stress wave sources are explored. Force-time functions and focal mechanisms are determined for discrete events found amid the "noise" of friction. These localized events are interpreted to be the rupture of micrometer-sized contacts, known as asperities. By comparing stress wave sources from stick-slip experiments on plastic/plastic and rock/rock interfaces, systematic differences were found. The rock interface produces very rapid (<1 microsecond) implosive forces indicative of brittle asperity failure and fault gouge formation, while rupture on the plastic interface releases only shear force and produces a source more similar to earthquakes commonly recorded in the field. The difference between the mechanisms is attributed to the vast differences in the hardness and melting temperatures of the two materials, which affect the distribution of asperities as well as their failure behavior. With proper scaling, the strong link between material properties and laboratory earthquakes will aid in our understanding of fault mechanics and the generation of earthquakes and seismic tremor.

[1] The self-potential (SP) response during hydraulic fracturing of intact Sierra granite was investigated in the laboratory. Excellent correlation of pressure drop and SP suggests that the SP response is created primarily by electrokinetic coupling. For low pressures, the variation of SP with pressure drop is linear, indicating a constant coupling coefficient (Cc) of - 200 mV/MPa. However, for pressure drops > 2 MPa, the magnitude of the Cc increases by 80% in an exponential trend. This increasing Cc is related to increasing permeability at high pore pressures caused by dilatancy of microcracks and is explained by a decrease in the hydraulic tortuosity. Resistivity measurements reveal a decrease of 2% prior to hydraulic fracturing and a decrease of similar to 35% after fracturing. An asymmetric spatial SP response created by injectate diffusion into dilatant zones is observed prior to hydraulic fracturing, and in most cases this SP variation revealed the impending crack geometry seconds before failure. At rupture, injectate rushes into the new fracture area where the zeta potential is different than in the rock porosity, and an anomalous SP spike is observed. After fracturing, the spatial SP distribution reveals the direction of fracture propagation. Finally, during tensile cracking in a point load device with no water flow, a SP spike is observed that is caused by contact electrification. However, the time constant of this event is much less than that for transients observed during hydraulic fracturing, suggesting that SP created solely from material fracture does not contribute to the SP response during hydraulic fracturing.