The simulation of wildland fire behavior, the prediction of wildfire hazards, and the decision on fuel treatment strategies all require accurate, high-precision, and up-to-date fuel maps. Existing maps of vegetation or fuel generally show coarse categories such as floristic composition and limited fuel types, but not vertical fuel structures and amounts. Although airborne Lidar data can provide reliable estimates of canopy traits, they are far away from the ground, so that laser light will be attenuated or unable to penetrate dense leaf layers, making it hard to assess the fuel beneath the forest canopy. Drone-based Lidar, can fly at low altitudes through the forest and deliver dense point cloud data, is ideal for collecting forest fuel structures and the amount of fuel in vertical layers. To quantify the small spatial scale changes in surface fuels, we used the Zenmuse L1 Lidar onboard the DJI Matrice 300 to observe the fuel structure and loading at the Blodgett Forest Research Station in the foothills of the California Sierra Nevada mountains. Two 5m x 5m square sample areas with different fuel compositions and loads were selected and scanned using drone-based Lidar at a height of 30�40m before and after a prescribed fire. Site 1 has lower fuel loads and compositions than site 2. Then, we filtered surface fuels from point cloud data using height measurement as a proxy. Fuel loads were calculated using random forest models and adjusted by field survey records. Comparing the fuel data before and after the prescribed burn, the results indicate that at site 1, the majority of the surface fuel was converted to debris and ash, while at site 2, the near-surface shrubs decreased and the vegetation structure became simpler. This method accurately captures the changes in fuel structure and fuel loads, and it demonstrates how drone-based Lidar can be used to fill gaps in existing fuel maps, improve their accuracy, and build comprehensive forest fuel libraries. These data products will facilitate future improvements to the performance of fire behavior or predictive models and support fuel treatment decisions.
In cold regions, the longevity of lake and reservoir ice cover is reduced by rising global temperatures, with significant impacts on aquatic ecosystems. For example, the early break-up of the ice cover leads to an accelerated development of summer stratification and higher surface temperatures, which can result in lower dissolved oxygen concentrations in the water and more sustained evaporative losses. In order to diagnose these present and future changes, it is useful to employ modeling. While physically based lake models can be a good alternative for this purpose, they are generally too demanding in terms of input data to be deployed on a large scale in cold regions, which are therefore often difficult to access and with limited in situ monitoring. While lake ice has been studied for a long time, reservoir ice is rather poorly documented and differs from lake ice by significant water level fluctuations and outflows required for turbining. The objective of this study is to explore a simple and large-scale approach to model the ice cover duration of reservoirs in cold regions. The ice cover duration is estimated from satellite images, using an automated Sentinel-2A image processing algorithm that differentiates ice from water on the Google Earth Engine platform. The algorithm is applied to 479 northern hemisphere hydroelectric reservoirs from the Global Reservoirs and Dams (GRanD v1.3) database that are located north of the January 0�C isotherm. This allows for the creation of an entirely new database of phenology dates (ice-on and ice-off) associated with hydroelectric reservoirs from 2018 to 2021. These phenology dates, in combination with air temperature data from weather stations and/or reanalysis (ERA5-Land), are then used to model the ice cover duration of these reservoirs using the Stefan equation, which is based on daily air temperatures. Various parameters influencing ice formation are explored such as accumulated freezing degree days, climatic zones and reservoir morphometric parameters. A multivariate statistical study correlating physical and morphometric reservoir parameters with phenology dates is also performed. Finally, an ice thickness measurement campaign conducted during the winter of 2021-2022 at the Romaine-2 reservoir in Quebec provided a smaller-scale and more detailed application of the methodology.
Archie's equation is widely used to estimate porosity from electrical log data. On the other hand, Archie's equation is not valid for rocks containing conductive clay minerals, and the estimation of porosity using the equation may significantly overestimate porosity for such rocks. In this study, we estimated porosity along a scientific drilling located in a volcanic region by combining electrical log and core data considering the effect of excess conductivity. The target borehole was drilled through the Futagawa fault, the source fault of the 2016 Kumamoto earthquake mainshock. Electrical log was conducted in this borehole from a depth range of 302-660 m, excluding a section of 383-399 m where log data is not available. The rocks in this borehole consist of sedimentary and altered volcanic rocks, which are thought to contain conductive clay minerals, and are classified into eight units based on their lithology. To account excess conductivity effects, the apparent formation factor, Fa was converted to the corrected formation factor, Fc. The conversion using the Waxman and Smits (1968) model requires a depth profile of BQv, which represents the excess conductivity term. Since this BQv cannot be obtained from log data, we conducted resistivity measurements of core samples saturated by solutions with different salinities and estimated their BQv based on the relationship between rock electrical conductivity and pore fluid conductivity. One representative core sample was selected in each lithological unit for the multiple conductivity-salinity tests for calculating the BQv, and we assumed that the BQv of the representative sample is applicable throughout each lithological unit. Consequently, the BQv depth profile enabled conversion from Fa to Fc at each depth. Then utilizing Archie's equation showing Fc-porosity relation, porosities were estimated at all depths where electrical log was conducted. The depth profile of estimated porosity generally agreed with the core porosity measured in previous study in this borehole. Estimated porosity from Archie's equation using Fa was not very consistent with the core porosity, our results demonstrate the importance to consider excess conductivity for properly evaluating the formation's porosity. .
Electrical resistivity has been a reliable, and relatively inexpensive way to visualize the near surface earth. After field acquisition, it is often difficult to estimate the volume of anomalies in the post-processing stages away from the field with no direct measurements of the subsurface such as local borehole data. The software ResIPy (v3.3.3) allows for onsite processing with immediate images of electrical resistivity distribution of the near surface structures. For a groundwater exploration survey in Sentinel, OK, electrical resistivity was used. Processing the data using ResIPy showed possible groundwater and clay lenses locations in the study area. The groundwater zone was distinguished from clay lenses with its higher resistivity compared to the low resistivity pockets close to the surface. The water sample from a nearby well has a 22.5 ?.m resistivity value and the resistive groundwater anomaly zone lies at depth of 15m to 20m, consistent with data from a water well close to the survey site. Creek water near the survey area has a 6.22 ?.m resistivity. Nearby well data also shows clay lenses atop of a sandy layer at depths of 0m to 7m with an average thickness of 4m. Here we show how any anomaly�s volume can be calculated using the attributes of ResIPy�s meshing algorithm. Meshing usually discretizes the subsurface into small tetrahedrons with known volumes known as cells. The volumes of each cell can be calculated with simple volume equations for tetrahedra or any other shape specified by the user post inversion and used to estimate the sum of the anomaly volume that lies within a group of cells. We developed a simple GUI module and added that to ResIPy, so that the users can simply define a resistivity (or conductivity) threshold after the inversion and see the volume of the anomaly that appears within the defined threshold. In this study, we used this approach to estimate the volume of clay lenses that lie above the aquifer in the target area. Our findings show that an overall 858 m3 of clay with an average thickness of 7m (consistent with the well data) is available at the site.
Yellowstone National Park (YNP) contains one of the world�s most highly concentrated areas of active geysers and other hydrothermal features (e.g. mud pots, hydrothermal pools). Despite numerous studies focusing on imaging these hydrothermal and meteoric waters, geophysical mapping of the near-surface hydrologic flow via Self-Potential (SP) measurements is rarely attempted. In this study, we present recent SP measurements collected from 2015 through 2021 from Old Faithful (i.e. cone geyser), Spouter Geyser (i.e. fountain geyser) and phase separation pool systems within Norris Geyser Basin (i.e. Two Pools) and Sentinel Meadows. These data are processed, analyzed, and modeled to answer the basic question of how either hydrothermal or meteoric recharge subsurface waters move into and around these systems. SP measurements were acquired using 56 non-polarizing, Ag-AgCl2 electrodes equally spaced along lines on the surface to map electric currents and both positive and negative voltage anomalies. Multiple and/or Time-Lapse SP data were collected in incremental, parallel lines that ran over the source to allow spatial mapping, pseudo 3D gridding and uncertainty analyses. Data were diurnally corrected utilizing two electrode positions within the line. An AGI SuperSting R8� recorded SP data in an �automatic� setting where all positions were sampled within 10 minutes. Both the absolute SP and SP difference curves derived from the raw SP data were modeled using COMSOL Multiphysics� Software to produce a 3D image conceptualizing the groundwater flow system. Forward modeling results observed in the simulated 2D curves match both positive and negative voltage anomalies in and around the geyser and phase separation systems. The Self-Potential method helps constrain groundwater flow and compare the dynamics among similarly labelled and structured systems. Understanding and modeling these hydraulic flow paths provides insight into the subsurface dynamics of these features, compliments near-surface geophysical imaging and provides an opportunity to meaningfully predict potential areas where the hazard of geothermal explosions may occur.
Electrical resistivity tomography (ERT) is a well-known geophysical tool for mapping salinity problems, mainly in coastal aquifers. High-resolution ERT images provide essential information for the characterization of these salinity-affected groundwater-bearing zones. Present study centres around salinity affected area of the Manadarmani-Contai region of West Bengal, India, where ERT data was collected using multiple arrays. After data collection, one significant processing step includes the selection of a suitable inversion scheme to generate a more reliable subsurface model. As for different arrays, the same inversion scheme leads to different results depending on the data sensitivity, signal strength, background noise etc. Thus, to overcome this limitation, we have jointly inverted Dipole-Dipole and Wenner arrays. As dipole-dipole has more no. of data points than the Wenner array, Jointly inverted results are slightly biased towards the dipole-dipole array. Although these arrays have different sensitivity patterns, the combination of this joint inversion provides good lateral resolution and shows a saline clay layer about 24 m below in the depth section. Further, Inversion results can be more refined by adopting different norms. Keywords: ERT, Joint inversion, Salinity, Aquifers.
Low frequency electrical relaxation phenomenon can be observed in porous materials and is related to the ability of grain surfaces to conduct locally bound electrical charge. Bound charges associated with the electrical double layer and ions in solution polarize in the direction of the electric field. Measurements of this relaxation phenomenon can be a useful tool to quantify the specific surface area of a given sample remotely. To date this type of frequency dependent measurement has been limited to galvanic observations at the laboratory and field scales. Transitioning from galvanic to standoff observations through electromagnetic induction sensors. Previously, we have presented investigations where measurements of the secondary magnetic field response from porous media in the frequency range of 100 kHz through 20 MHz were examined for evidence of relaxation signatures using a high-frequency electromagnetic induction sensor (HFEMI). Here, we evaluate a newly constructed low-frequency electromagnetic induction sensor (LFEMI), capable of measurements in the 3 Hz to 1 kHz range. These systems have demonstrated performance for metal detection in the form of ohmic relaxation. In this work, we present results of experiments designed to observe surface relaxation responses in porous materials as a result of bound charge transfer, i.e. surface conductivity rather than Ohmic conductivity. The ability to measure this phenomenon remotely has the potential for broad impact within the subsurface interrogation communities including proximal soil sensing, environmental monitoring, geotechnical assessment, and hydrogeologic investigations.
The frequency and intensity of tropical cyclones over the Arabian Sea increased over the past three decades, possibly due to the warming of the Arabian Sea. Fourteen cyclones (wind speed ? 63 km/h) were reported in the 1990s, 16 in the 2000s, and 18 in the 2010s. During this period, 12 cyclones hit the southern coast of the Arabian Peninsula, three of which were categorized as extremely severe cyclonic storms (ESCS) (wind speed ?167 km/h) and one as a super cyclonic storm (wind speed ? 222 km/h). Those storms brought vast amounts of rainfall inland and could have produced sizable groundwater recharge. We estimated the change in terrestrial water storage from Gravity Recovery and Climate Experiment (GRACE) and GRACE-Follow-On (GRACE-FO) and made correlations with Global Precipitation Measurement mission (GPM) data. We accomplished those goals for three cyclones that hit the Arabian Peninsula in 2011 (Keila), 2015 (Chapala), and 2018 (Luban) in three main steps. (1) track/select cyclones with a high wind speed (? 63 km/h) during the GRACE and GRACE-FO operational period (April 2002 to Dec 2021); (2) delineate areas that received precipitation exceeding 10 mm during cyclone activity, and estimating precipitation over each of those areas; and (3) compute the increase in GRACETWS before and after each of the three cyclones. Our preliminary findings reveal the following: (1) cyclones induce extreme precipitation, 2-4 times the largest recorded seasonal precipitation events. (2) The precipitation was extensive (Keila: 6 km3; Chapala: 6.6 km3, Luban: 2.25 km3), it extended over vast areas (Keila: 75,731 km2; Chapala: 94,273 km2; and Luban: 58,944 km2), and produced a noticeable increase in GRACETWS (Keila: 1.3 km3, Chapala: 3.4 km3, and Luban: 0.25 km3). (3) The increase in GRACETWS over both cyclones, Keila and Luban, was depleted over a short period (<2 months) but lasted for over three months for the Chapala cyclone, possibly due to the presence of thick alluvial aquifers and outcrops of the deep aquifer (Umm-Er-Radhuma) in areas affected by cyclone Chapala. We are currently investigating the utility of higher temporal resolution GRACETWS (Sakumura et al., 2016) in capturing the rapid changes in TWS directly after cyclones make landfall.
The use of geoelectrical methods for investigating variations in wetlands soil properties could dramatically expand our understanding of wetlands� hydrological and biogeochemical functions. However, existing petrophysical relationship may be limited in wetland soils due to their hydric nature. Hence, this study focuses on advancing our mechanistic understanding of wetland soils� electrical response in other to adapt existing petrophysical models for predicting wetlands soil properties including moisture, and organic matter from measured electrical signals. In this study, we use an EM-38-MK2 operated at a 14.5 kHz frequency to obtain the spatial distribution of soil electrical conductivity and identify soil boundaries. Core samples were collected at 0-30 cm depth at each soil boundary using an AMS soil bulk density sampler. We performed spectral induced polarization (SIP) measurement on 18 soil cores in the laboratory using a frequency range of 0.01 Hz to 10 kHz. All samples were partially saturated using 78 ?S/cm of brine (NaCl) solution prior to the SIP measurements. We recorded the impedance (z) and phase shift (?) response using four electrodes and computed the real and imaginary complex conductivity. Using the same setup for the SIP, we estimated the formation factor (F) for each sample and obtained F values between 4.35 and 11.67 through repeated measurement of the bulk and fluid resistivity at different salinity concentration. We also measured the physical-chemical properties of the soil including the soil texture, moisture content, organic matter, bulk density, and porosity. Our result shows a strong correlation (71% & 85%) with ?' and measured soil properties, but no correlation with ?'' at 1.14 Hz. We then evaluate ? variations with measure soil properties where we observe a strong correlation (63% & 65%). This study indicated that wetland soil electrical response has the strongest effects on ?' and ? variations which is foremost arising from the variation in moisture content. And other apparent strong relations such as SOM and soil textures (clay content) were due to a high correlation with moisture content.
Groundwater is being extracted for a managed aquifer recharge (MAR) pilot project utilizing riverbank filtration and groundwater transfer and injection located near the Tallahatchie River at Shellmound, MS. The extraction well for the project is located about 40 m away from the river and screened into the Mississippi River Valley Alluvial (MRVA) aquifer. The MAR project aims is to pump groundwater and induce infiltration from the Tallahatchie River by means of bank filtration. The MRVA aquifer is highly heterogeneous and typically varies from confined to unconfined within short distances. It is important to characterize the MRVA aquifer, but existing geophysical information is incomplete. Previously collected Airborne Electromagnetic (AEM) data did not provide continuous high-resolution data of the subsurface near the extraction well and were muted close to the river due to power lines. Bore logs provide discrete data and cannot be used for continuous spatial characterization of the aquifer. Electrical Resistivity Tomography (ERT) provides high-resolution subsurface images that can be used for aquifer characterization, such as determining aquifer extent, identifying preferential flow paths, and detecting subsurface voids or variations in lithology. ERT fills the gaps in bore logs and AEM data at the study site. In this study, data for the three ERT profiles were collected with 4 m electrode spacing. Data for the 444 m long ERT profile-1 were collected across and on the west side of the river, starting from the riverbank with the Wenner array types. For the 668 m long ERT profile-2 and profile-3, data were collected with a roll along of 50% overlapping along the river using the Dipole-Dipole array types. ERT data were inverted with the finite element method and Cholesky Decomposition solver with Dirichlet boundary conditions. The inverted ERT profile-1 and profile-2 were calibrated with the extraction well, and nearby monitoring well bore log to delineate the aquifer. Available AEM data for the extraction site was also used to calibrate the inverted ERT profiles for aquifer delineation. Profile-1 shows the MRVA aquifer transitioned from confined to unconfined in the agricultural field about 380 m west of the river. Profile-2 shows that the aquifer is confined along the profile and thicker on the northern side compared to the south side of the extraction well, and this profile also shows four higher resistivity anomaly zones whose depths range from 20 to 32 m. These higher resistivity zones might be attributed to preferential groundwater flow paths or loosely packed sand and gravel zones. Profile-3 shows the aquifer varies from confined to unconfined in multiple places along this north-south profile west of the river. This study characterizes and delineates the vertical extent of the MRVA aquifer near the extraction well and will provide an improved understanding of groundwater flow to support the assessment of the MAR project.
The paucity of information on the role subsurface hydrology plays in nutrient (Nitrogen and Phosphorous) sequestration within wetlands is a subject that requires attention. The objective of this research is to develop a coupled surface � subsurface water and contaminant transport flow models using HydroGeosphere (HGS), to determine the fate and transport of leached nutrients and how it affects downstream hydrology. The first phase consists of characterizing the geology using electrical resistivity tomography (ERT), electromagnetic induction (EMI) and the ground penetrating radar (GPR). A SuperSting R8 resistivity meter, an EM38-MK2 and the Sensors and Software Pulse EKKO system were used for the ERT, EMI and GPR respectively. The geophysical result showed a top layer of 1.5 m thick lying on top of a sandy layer of about 3 m thickness. The sand is underlain by a glacial lacustrine � glacial till that is about 2.5m thick. The top layer is composed mainly of clayey loam with lenses of silty loam material. GPR radargram show an approximate depth to water table of 0.65m which was useful in understanding the flow regime at the study site. The EMI displayed a spatial distribution of electrical conductivity values across the study area. Areas of low conductivity corresponded with silty loam material whereas high conductive areas were identified as clayey loam. The EMI data was used to delineate regions of silty loam topsoil which are potential zones of direct infiltration. Constant head double ring infiltrometer test was conducted on different sections of the site to estimate the infiltration ratef the soil. The equilibrium saturated hydraulic conductivity from the infiltrometer test was estimated as 6.67 10-6 ms-1 for the silty loam and 5.79 10-8 ms-1 for the clayey loam. Further study will estimate other hydraulic parameters of the sand such as the average linear velocity, hydraulic retardation and dispersion coefficients to calibrate the flow and transport model of the site. Knowledge of the hydraulic properties of both the surface and subsurface media will help to understand how the interaction between the surface and subsurface hydrology enhance the nutrient retention capacity of the soil which is vital for wetland restoration globally.
Soil parameters assessment is essential for sustainable civil engineering constructions. Collecting soil samples from the borehole and obtaining geotechnical parameters in a laboratory is time-consuming, expensive, and laborious. In the present work, Electrical resistivity tomography (ERT) data provided an alternative way to get geotechnical parameters (Plastic index, Moisture content, and Friction angle) to characterize the subsurface properties in less time. ABEM Terrameter LS 2 instrument used for ERT data acquisition in the coal seam area; The Profile length was 400 m with an electrode spacing of 10 m. Inversion of ERT data performed through the Res2dinv program based on the smoothness-constrained least-square method. Approximate true resistivity values were obtained from the inversion technique with reasonable root mean square (RMS) error. For high resolution, the results were interpolated with every 2 m interval. Polynomial regression fitting approximation was used for cohesion variation with resistivity values. All resistivity data was analyzed and demonstrated its relationship with geotechnical properties. Electrical resistivity correlated well with geotechnical parameters. Finally, the low-strength soil and the main factors that affected resistivity contrast in the coal seam zone were interpreted. This study could assist civil and mining engineers in future constructions.
Site response study is a technique that is helpful for evaluating soil sediments of the surficial crust on strong ground motion. The Kanto earthquake that happened in 1923 showed us that surficial geology plays a major role in controlling damages caused by destructive earthquakes. But it took decades long to accept the fact over the wide earthquake engineering community. In this study, site characterization has been done by using Nakamura�s Horizontal to Vertical Spectral Ratio Technique (HVSR). A small 1��1� region is chosen over the Kanazawa which belongs to Chubu region, Japan. Ten years of strong ground motion data is obtained from KiK NET maintained by NIED (National Research Institute for Earth Science and Disaster Resilience). Waveforms of duration 120 seconds are chosen following the onset of P arrivals. A ten seconds window is marked following the onset of S-waves. Fast fourier Transformation (FFT) of the ten second window converted the time series waveforms in the frequency domain. The Horizontal to Vertical Spectral Ratio (HVSR) is obtained by averaging each Horizontal and Vertical component recorded by the stations present over the region. Out of 18 stations present over the region, each event has been recorded by a minimum of 6 stations. Plotting of H/V spectral ratio with a frequency range from 1 to 10 Hz gives us the Predominant frequency (fpeak) over the regions which was found between 2.1 and 9.92 Hz. The predominant frequency values show peaks at a lower frequency at sites corresponding to class C to D whereas shows fpeak values at higher frequencies for class C to B which shows that with an increase in compaction of sediments fpeak increases here the sediment ranges from stiff soil to soft rock. The VS30 values are needed to correlate the surficial lithology. Site characterization study using this method will help to control loss and damages caused by destructive earthquakes.
Groundwater is a precious resource to irrigate the crops in developing countries. This research was conducted in Chiniot District of Pakistan for assessing the groundwater strata using an integrated GIS and resistivity survey method. For this purpose, vertical electrical sounding (VES) resistivity data was obtained using an ABEM Terrameter SAS 4000 with a Schlumberger electrode configuration in which half current electrode spacing (AB/2) ranges from 2 to 180 m and half potential electrode spacing (MN/2) ranges from 0.5 to 20 m were used. A computer model �IX1D v2� was used to determine soil layer thicknesses and resistivities for each VES point and then a Rockworks model (Rockware, Inc.) was also used to develop a conceptual 3-D lithological logs model of research area based on determined soil layers thicknesses values. Geo hydraulic parameters (hydraulic conductivity and transmissivity) and Dar Zarrouk parameters (transverse unit resistance and longitudinal unit conductance) were determined based on the modeling results. ArcGIS Pro was used for mapping the soil aquifer layers� thicknesses and resistivities to specify the soil aquifer layers lithology and groundwater quality zones as well as for mapping of geo hydraulic parameters to estimate well yield and water bearing capacities of aquifer layers and Dar Zarrouk parameters to estimate groundwater potential zones with respect to VES point regions. Based on aquifer layers resistivities, hydraulic conductivities, transmissivities, transverse unit resistance and longitudinal unit conductance, eight VES points were considered marginal quality of groundwater zone having intermediate well yield of �30 to 60 liter/sec�, medium water bearing capacity and medium potential aquifer through fine sand and clay formation while fourteen VES points were considered fresh quality of groundwater zone having high well yield �>60 liter/sec�, high water bearing capacity and high potential aquifer through coarse sand and gravel formation. Four VES points were considered saline quality of groundwater zone having low well yield �6 to 30 liter/sec�, low water bearing capacity and low potential aquifer through clay formation.
The global landmine and unexploded ordnance (UXO) crisis is a $200 billion dollar problem that causes 15,000-20,000 casualties annually. A terrain-adaptable, geophysics-based approach to detection is needed. Single-sensor unmanned aerial vehicle (UAV) methods often produce dangerous false flags, while current multi-sensor systems are land-based, expensive to manufacture, and cumbersome to transport to remote sites. Understanding how environmental conditions, host overburden lithology, and vegetation affect geophysical signatures of buried objects is crucial for making progress and reducing false flags. We are developing a multi-sensor package that can be integrated on a quad-copter unmanned aerial vehicle (UAV) that evaluates terrain and environmental conditions using machine learning, and weights on-board geophysical instrument(s) data to obtain the most accurate landmine detection performance. One of the techniques we are incorporating is magnetometery, which is often utilized in both ground and aerial landmine detection surveys. Slung-loaded total field magnetometers are commonly used in UAV-based surveys, but the lack of component information limits the ability to infer the orientation, distance, and magnetic moment of detected anomalies. Fluxgate magnetometers provide component information but suffer from calibration errors and noise when used in isolation. To overcome some of the practical limitations of a single fluxgate magnetometer, we have developed a magnetic gradiometer system consisting of four 3-component fluxgate magnetometers arranged in a tetrahedral configuration. By examining the gradients between each component of each fluxgate, we are able to streamline calibration efforts, and mitigate spatial variations that result from space weather and cultural noise sources. Here we present initial results from our magnetic gradiometer system for detecting inert landmines and proxy UXO in a controlled sand bed testing site and compare them to synthetic data. Finally, we discuss the practicality and utility of machine learning as a tool for anomaly detection and classification. We gratefully acknowledge the support of NSF Grant No. IIP2044611 and DoD NDSEG Fellowship.