SEARCH · Geology Reports

Results for “Algorithms”

Search indexed USGS publications on groundwater, aquifers, geologic maps, mineral resources and earthquakes. Explore source records by subject and place.

SourceResearch connectionFrom yearThrough year

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 541 records · Page 30Linked to original sources

USGS2025

Advancements in satellite observations of inland and coastal waters: Building towards a global validation network

The use of satellite-based remote sensing imagery for water quality monitoring of inland and coastal waters has become widespread over the last few decades, with the expansion of, and investment in, operational Earth-observing missions. Satellite-based sensors are uniquely suited to provide synoptic, system-wide water quality parameter estimates that supplement traditional field-based sampling methods. The remote sensing of water quality parameter estimates is particularly valuable in systems with high temporal and spatial variability, as well as in areas that are difficult to access, or where agencies lack funding for routine monitoring. However, optically complex inland and coastal waters pose additional challenges for developing robust remote sensing retrieval models for optical properties and water quality parameters. One of the biggest challenges is collecting high quality field measurements that are used to calibrate and validate the retrieval algorithms. Here, we present the current status of satellite missions, field methods that include instruments used and commonly measured parameters, and repositories of historical field data that are relevant to inland and coastal water studies. We then present data requirements for model validation and highlight gaps in validation coverage. Finally, we provide suggestions for future field campaigns to improve coordination with remote sensing data collection and to ensure that field data is well suited for use in model or algorithm development.

Remote Sensing↗

USGS2015

Categorisation of northern California rainfall for periods with and without a radar brightband using stable isotopes and a novel automated precipitation collector

During landfall of extratropical cyclones between 2005 and 2011, nearly 1400 precipitation samples were collected at intervals of 30-min time resolution with novel automated collectors at four NOAA sites in northern California [Alta (ATA), Bodega Bay (BBY), Cazadero (CZD) and Shasta Dam (STD)] during 43 events. Substantial decreases were commonly followed hours later by substantial increases in hydrogen isotopic composition ( δ 2 H VSMOW where VSMOW is Vienna Standard Mean Ocean Water) and oxygen isotopic composition ( δ 18 O VSMOW ) of precipitation. These variations likely occur as pre-cold frontal precipitation generation transitions from marine vapour masses having low rainout to cold cloud layers having much higher rainout (with concomitant brightband signatures measured by an S-band profiling radar and lower δ 2 H VSMOW values of precipitation), and finally to shallower, warmer precipitating clouds having lower rainout (with non-brightband signatures and higher δ 2 H VSMOW values of precipitation), in accord with ‘seeder–feeder’ precipitation. Of 82 intervals identified, a remarkable 100.5 &permil; decrease in δ 2 H VSMOW value was observed for a 21 January 2010 event at BBY. Of the 61 intervals identified with increases in δ 2 H VSMOW values as precipitation transitioned to shallower, warmer clouds having substantially less rainout (the feeder part of the seeder–feeder mechanism), a remarkable increase in δ 2 H VSMOW value of precipitation of 82.3 &permil; was observed for a 10 February 2007 event at CZD. All CZD and ATA events having δ 2 H VSMOW values of precipitation below −105 &permil; were atmospheric rivers (ARs), and of the 13 events having δ 2 H VSMOW values of precipitation below −80 &permil;, 77 % were ARs. Cloud echo-top heights (a proxy for atmospheric temperature) were available for 23 events. The mean echo-top height is greater for higher rainout periods than that for lower rainout periods in 22 of the 23 events. The lowest δ 2 H VSMOW of precipitation of 28 CZD events was −137.9 &permil; on 16 February 2009 during an AR with cold precipitating clouds and very high rainout with tops >6.5 km altitude. An altitude effect of −2.5 &permil; per 100 m was measured from BBY and CZD δ 2 H VSMOW data and of −1.8 &permil; per 100 m for CZD and ATA δ 2 H VSMOW data. We present a new approach to categorise rainfall intervals using δ 2 H VSMOW values of precipitation and rainfall rates. We term this approach the algorithmic-isotopic categorisation of rainfall, and we were able to identify higher rainout and/or lower rainout periods during all events in this study. We conclude that algorithmic-isotopic categorisation of rainfall can enable users to distinguish between tropospheric vapour masses having relatively high rainout (typically with brightband rain and that commonly are ARs) and vapour masses having lower rainout (commonly with non-brightband rain).

California↗

USGS2015

A comparative study of surface waves inversion techniques at strong motion recording sites in Greece

Surface wave method was used for the estimation of Vs vs depth profile at 10 strong motion stations in Greece. The dispersion data were obtained by SASW method, utilizing a pair of electromechanical harmonic-wave source (shakers) or a random source (drop weight). In this study, three inversion techniques were used a) a recently proposed Simplified Inversion Method (SIM), b) an inversion technique based on a neighborhood algorithm (NA) which allows the incorporation of a priori information regarding the subsurface structure parameters, and c) Occam's inversion algorithm. For each site constant value of Poisson's ratio was assumed (ν=0.4) since the objective of the current study is the comparison of the three inversion schemes regardless the uncertainties resulting due to the lack of geotechnical data. A penalty function was introduced to quantify the deviations of the derived Vs profiles. The Vs models are compared as of Vs(z), Vs30 and EC8 soil category, in order to show the insignificance of the existing variations. The comparison results showed that the average variation of SIM profiles is 9% and 4.9% comparing with NA and Occam's profiles respectively whilst the average difference of Vs30 values obtained from SIM is 7.4% and 5.0% compared with NA and Occam's.

Conference Paper↗

USGS2009

Detection and characterization of benthic filamentous algal stands ( Cladophora sp.) on rocky substrata using a high-frequency echosounder

A high-frequency echosounder was used to detect and characterize percent cover and stand height of the benthic filamentous green alga Cladophora sp. on rocky substratum of the Laurentian Great Lakes. Comparisons between in situ observations and estimates of the algal stand characteristics (percent cover, stand height) derived from the acoustic data show good agreement for algal stands that exceeded the height threshold for detection by acoustics (~7.5 cm). Backscatter intensity and volume scattering strength were unable to provide any predictive power for estimating algal biomass. A comparative analysis between the only current commercial software (EcoSAV™) and an alternate method using a graphical user interface (GUI) written in MATLAB® confirmed previous findings that EcoSAV functions poorly in conditions where the substrate is uneven and bottom depth changes rapidly. The GUI method uses a signal processing algorithm similar to that of EcoSAV but bases bottom depth classification and algal stand height classification on adjustable thresholds that can be visualized by a trained analyst. This study documents the successful characterization of nuisance quantities of filamentous algae on hard substrate using an acoustic system and demonstrates the potential to significantly increase the efficiency of collecting information on the distribution of nuisance macroalgae. This study also highlights the need for further development of more flexible classification algorithms that can be used in a variety of aquatic ecosystems.

Great Lakes↗

USGS2026

Origins, evolutions, and future directions of Landsat science products for advancing global inland water and coastal ocean observations

In April 2020, the U.S. Geological Survey (USGS) Earth Resources Observation and Science (EROS) Center introduced a Level 2 provisional Aquatic Reflectance (AR) product for the Landsat 8 Operational Land Imager (OLI), marking the initial phase in developing a standardized global product for Landsat-derived surface water measurements. The goal of USGS EROS aquatic product research and development is to prepare for an operational processing architecture for Landsat Collection 3 in the late 2020s that will enable use of quality-controlled data for emerging Landsat aquatic science applications. To achieve this, we released a subset of the Landsat 8/9 provisional AR products (Crawford et al., 2025, https://doi.org/10.5066/P14MBBRM) and examined its general performance through the Science Algorithms to Operations (SATO) framework alongside quantitative assessment using community made inland water data records (GLObal Reflectance community dataset for Imaging and optical sensing of Aquatic environments, GLORIA) and radiometric coastal validation platforms (NASA’s Ocean Color component of the Aerosol Robotic Network, AERONET-OC). Variability within the validation datasets indicate that the performance of the Landsat 8/9 provisional AR retrieval is highly context-dependent; errors are minimal in optically simple waters (e.g., clear to moderately turbid coastal waters) but increase considerably in optically complex waters where factors such as elevated levels of turbidity, chlorophyll (Chl a ) concentrations, or colored dissolved organic matter (CDOM) dominate the water column. Additionally, this paper examines key algorithmic considerations for atmospheric correction, highlighting factors that influence accuracy, scalability, and computational efficiency necessary for collection processing in the operational Landsat Product Generation System (LPGS). This paper is intended to communicate with aquatic scientists, satellite oceanographers, and the broader Earth observation community on the origins, requirements, challenges, successes, and future objectives for operationalizing global AR data products for Landsat satellite missions.

Earth System Science Data↗

USGS2025

Review and synthesis of the applications of machine learning to coalbed methane recovery

Over the last 30 years, a substantial literature has evolved on the use of machine learning (ML) to assess, predict, and improve the efficiency of coalbed methane (CBM) recovery. In the United States, the production of CBM declined as shale gas production matured, but CBM continues to be an important energy resource in other parts of the world. ML applications that have the potential to improve CBM reservoir management and production forecasts, and to increase exploration and operational efficiency, are still of significant interest. The integration of geostatistical techniques into the CBM ML applications has been largely absent but represents an opportunity for improvement. The literature demonstrates the widespread interest in, and applicability of, ML algorithms applied to CBM problems, and that they continue to result in improvements in predictive performance. However, (1) much of the research is more academic than operational, (2) many results are based on simulations, or small or proprietary datasets, (3) ML performance information can be inconsistent and sometimes entirely omitted, (4) most methodologies are unique to the specific CBM situation and likely not generalizable, (5) no standard data repositories are available to directly compare the performance of competing algorithms, and (6) the spatial component is often omitted. Finally, relatively new ML protocols involving causality analysis and reinforced learning, as well as hybrid workflows combining both supervised and unsupervised learning, are anticipated to dominate the future investigations. Integration of geostatistical and geospatial analysis with ML should enhance performance.

Book chapter↗

USGS2006

Planning for robust reserve networks using uncertainty analysis

Planning land-use for biodiversity conservation frequently involves computer-assisted reserve selection algorithms. Typically such algorithms operate on matrices of species presence?absence in sites, or on species-specific distributions of model predicted probabilities of occurrence in grid cells. There are practically always errors in input data?erroneous species presence?absence data, structural and parametric uncertainty in predictive habitat models, and lack of correspondence between temporal presence and long-run persistence. Despite these uncertainties, typical reserve selection methods proceed as if there is no uncertainty in the data or models. Having two conservation options of apparently equal biological value, one would prefer the option whose value is relatively insensitive to errors in planning inputs. In this work we show how uncertainty analysis for reserve planning can be implemented within a framework of information-gap decision theory, generating reserve designs that are robust to uncertainty. Consideration of uncertainty involves modifications to the typical objective functions used in reserve selection. Search for robust-optimal reserve structures can still be implemented via typical reserve selection optimization techniques, including stepwise heuristics, integer-programming and stochastic global search.

Ecological Modelling↗

USGS1994

Earthquake classification, location, and error analysis in a volcanic environment: implications for the magmatic system of the 1989-1990 eruptions at redoubt volcano, Alaska

Determination of the precise locations of seismic events associated with the 1989-1990 eruptions of Redoubt Volcano posed a number of problems, including poorly known crustal velocities, a sparse station distribution, and an abundance of events with emergent phase onsets. In addition, the high relief of the volcano could not be incorporated into the hypoellipse earthquake location algorithm. This algorithm was modified to allow hypocenters to be located above the elevation of the seismic stations. The velocity model was calibrated on the basis of a posteruptive seismic survey, in which four chemical explosions were recorded by eight stations of the permanent network supplemented with 20 temporary seismographs deployed on and around the volcanic edifice. The model consists of a stack of homogeneous horizontal layers; setting the top of the model at the summit allows events to be located anywhere within the volcanic edifice. Detailed analysis of hypocentral errors shows that the long-period (LP) events constituting the vigorous 23-hour swarm that preceded the initial eruption on December 14 could have originated from a point 1.4 km below the crater floor. A similar analysis of LP events in the swarm preceding the major eruption on January 2 shows they also could have originated from a point, the location of which is shifted 0.8 km northwest and 0.7 km deeper than the source of the initial swarm. We suggest this shift in LP activity reflects a northward jump in the pathway for magmatic gases caused by the sealing of the initial pathway by magma extrusion during the last half of December. Volcano-tectonic (VT) earthquakes did not occur until after the initial 23-hour-long swarm. They began slowly just below the LP source and their rate of occurrence increased after the eruption of 01:52 AST on December 15, when they shifted to depths of 6 to 10 km. After January 2 the VT activity migrated gradually northward; this migration suggests northward propagating withdrawal of magma from a plexus of dikes and/or sills located in the 6 to 10 km depth range. Precise relocations of selected events prior to January 2 clearly resolve a narrow, steeply dipping, pencil-shaped concentration of activity in the depth range of 1-7 km, which illuminates the conduit along which magma was transported to the surface. A third event type, named hybrid, which blends the characteristics of both VT and LP events, originates just below the LP source, and may reflect brittle failure along a zone intersecting a fluid-filled crack. The distribution of hybrid events is elongated 0.2-0.4 km in an east-west direction. This distribution may offer constraints on the orientation and size of the fluid-filled crack inferred to be the source of the LP events. ?? 1994.

Journal of Volcanology and Geothermal Research↗

USGS2000

Wildlife tradeoffs based on landscape models of habitat

It is becoming increasingly clear that the spatial structure of landscapes affects the habitat choices and abundance of wildlife. In contrast to wildlife management based on preservation of critical habitat features such as nest sites on a beach or mast trees, it has not been obvious how to incorporate spatial structure into management plans. We present techniques to accomplish this goal. We used multiscale logistic regression models developed previously for neotropical migrant bird species habitat use in South Carolina (USA) as a basis for these techniques. Based on these models we used a spatial optimization technique to generate optimal maps (probability of occurrence, P = 1.0) for each of seven species. To emulate management of a forest for maximum species diversity, we defined the objective function of the algorithm as the sum of probabilities over the seven species, resulting in a complex map that allowed all seven species to coexist. The map that allowed for coexistence is not obvious, must be computed algorithmically, and would be difficult to realize using rules of thumb for habitat management. To assess how management of a forest for a single species of interest might affect other species, we analyzed tradeoffs by gradually increasing the weighting on a single species in the objective function over a series of simulations. We found that as habitat was increasingly modified to favor that species, the probability of presence for two of the other species was driven to zero. This shows that whereas it is not possible to simultaneously maximize the likelihood of presence for multiple species with divergent habitat preferences, compromise solutions are possible at less than maximal likelihood in many cases. Our approach suggests that efficiency of habitat management for species diversity can by maximized for even small landscapes by incorporating spatial context. The methods we present are suitable for wildlife management, endangered species conservation, and nature reserve design.

Conference Paper↗

USGS2001

Recognition of fiducial surfaces in lidar surveys of coastal topography

A new method for the recognition and mapping of surfaces in coastal landscapes that provide accurate and low variability topographic measurements with respect to airborne lidar surveys is described and demonstrated in this paper. Such surfaces are herein termed "fiducial" because they can represent reference baseline morphology in Studies of coastal change due to natural or anthropogenic causes. Non-fiducial surfaces may also be identified in each separate lidar survey to be used in a given geomorphic change analysis. Sites that are non-fiducial in either or both lidar surveys that bracket the time period under investigation may be excluded from consideration in subsequent calculations of survey-to-survey elevation differences to eliminate spurious indications of landscape change. This new analysis method, or lidar fiducial surface recognition (LFSR) algorithm, is intended to more fully enable the non-ambiguous Use of topographic lidar in a range of coastal investigations. The LFSR algorithm may be widely applied, because it is based solely on the information inherent in the USGS/NASA/NOAA airborne topographic lidar coverage that exists for most of the contiguous U.S. coastline.

Photogrammetric Engineering and Remote Sensing↗

USGS2002

Hydroacoustic Current Meters for the Measurement of Discharge in Shallow Rivers and Streams

The U.S. Geological Survey (USGS) is evaluating the use of hydroacoustic current meters for making discharge measurements in shallow rivers and streams. The USGS historically has made discharge measurements in shallow rivers using mechanical, impellor-type current meters attached to a wading rod. The evaluation project has focused on three categories of hydroacoustic meters: an acoustic Doppler velocimeter (ADV) called a Flowtracker3, an acoustic Doppler velocity profiler (BoogieDopp), and bottom-tracking acoustic Doppler current profilers (ADCPs). The USGS role in this project includes providing USGS discharge-computation methods and algorithms to instrument manufacturers and evaluating instruments in the laboratory and field. An ADV (Flowtracker) designed for making discharge measurements in shallow rivers, has been tested in a USGS tow tank and was found to meet USGS calibration standards for mechanical, impellor-type current meters. The Flowtracker was field tested by USGS offices in five states; the tests were conducted by comparing discharge measurements made with the ADV to discharge measurements made with mechanical, impellor-type current meters. In general, the comparisons of Flowtracker performance to mechanical-meter results were favorable. An acoustic Doppler velocity profiler (BoogieDopp) is being evaluated for making discharge measurements in shallow rivers. The Boogiedopp will measure vertical velocity profiles at stationary positions across a channel, and the velocity profiles will be used to compute discharge. Discharge-computation software based on USGS methods and algorithms is under development for the acoustic Doppler velocity profiler. The USGS will evaluate bottom-tracking ADCPs from two manufacturers for making discharge measurements in shallow water. The bottom-tracking feature allows ADCPs to compute discharge from a moving platform as the platform moves across the channel.

Conference Paper↗

USGS2004

Extracting transient Rayleigh wave and its application in detecting quality of highway roadbed

This paper first explains the tau-p mapping method of extracting Rayleigh waves (LR waves) from field shot gathers. It also explains a mathematical model of physical character parameters of quality of high-grade roads. This paper then discusses an algorithm of computing dispersion curves using adjacent channels. Shear velocity and physical character parameters are obtained by inversion of dispersion curves. The algorithm using adjacent channels to calculating dispersion curves eliminates average effects that exist by using multi-channels to obtain dispersion curves so that it improves longitudinal and transverse resolution of LR waves and precision of non-invasive detection, and also broadens its application fields. By analysis of modeling results of detached computation of the ground roll and real examples of detecting density and pressure strength of a high-grade roadbed, and by comparison of shallow seismic image method with borehole cores, we concluded that: 1 the abnormal scale and configuration obtained by LR waves are mostly the same as the result of shallow seismic image method; 2 an average relative error of density obtained from LR waves inversion is 1.6% comparing with borehole coring; 3 transient LR waves in detecting density and pressure strength of a high-grade roadbed is feasible and effective.

Conference Paper↗

USGS1992

A semi-implicit finite difference model for three-dimensional tidal circulation,

A semi-implicit finite difference formulation for the numerical solution of three-dimensional tidal circulation is presented. The governing equations are the three-dimensional Reynolds equations in which the pressure is assumed to be hydrostatic. A minimal degree of implicitness has been introduced in the finite difference formula so that in the absence of horizontal viscosity the resulting algorithm is unconditionally stable at a minimal computational cost. When only one vertical layer is specified this method reduces, as a particular case, to a semi-implicit scheme for the solutions of the corresponding two-dimensional shallow water equations. The resulting two- and three-dimensional algorithm is fast, accurate and mass conservative. This formulation includes the simulation of flooding and drying of tidal flats, and is fully vectorizable for an efficient implementation on modern vector computers.

Conference Paper↗

USGS2016

Consistent and efficient processing of ADCP streamflow measurements

The use of Acoustic Doppler Current Profilers (ADCPs) from a moving boat is a commonly used method for measuring streamflow. Currently, the algorithms used to compute the average depth, compute edge discharge, identify invalid data, and estimate velocity and discharge for invalid data vary among manufacturers. These differences could result in different discharges being computed from identical data. Consistent computational algorithm, automated filtering, and quality assessment of ADCP streamflow measurements that are independent of the ADCP manufacturer are being developed in a software program that can process ADCP moving-boat discharge measurements independent of the ADCP used to collect the data.

Conference Paper↗

USGS2004

Preparing for THEMIS controlled global Mars mosaics

We have begun work to prepare for producing controlled 2001 Mars Odyssey THEMIS infrared (IR) and visible (VIS) global mosaics of Mars. This effort is being coordinated with colleagues from Arizona State University and on the THEMIS team who plan to address radiometric issues in making such mosaics. We are concentrating on geometric issues. Several areas of investigation are now in progress, including: a) characterizing the absolute pointing accuracy of THEMIS images; b) investigating whether automatic tie point matching algorithms could be used to provide connections between overlapping THEMIS images; c) developing algorithms to allow for the photogrammetric (bundle) adjustment of the THEMIS IR (line scanner) camera images. Our primary goal in this pilot study effort will be to make several test control THEMIS mosaics and better determine which methods could be used, which require development, and what level of effort is required, in order to make large regional or global controlled THEMIS mosaics.

Conference Paper↗

USGS2021

Model structural uncertainty quantification and hydrogeophysical data integration using airborne electromagnetic data

A irborne electromagnetic (AEM) data are used to estimate large - scale model structural geometry, i.e. the spatial distribution of different lit hological units based on assumed or estimated resistivity - lithology relationships, and the uncertainty in those structures given imperfect measurements. Geophysically derived estimates of model structural uncertainty are then combined with hydrologic obse rvations to assess the impact of model structural error on hydrologic calibration and prediction errors. Using a synthetic numerical model, we describe a sequential hydrogeophysical approach that: (1) uses Bayesian Markov chain Monte Carlo (McMC) methods to produce a robust estimate of uncertainty in electrical resistivity parameter s , (2) combines geophysical parameter uncertainty estimates with borehole observations of lithology to produce probabilistic estimates of model structural uncertainty over the e ntire AEM survey area using geostatistical sequential indicator simulation algorithms, and (3) uses model structural estimates along with hydrologic observations to quantify both hydrologic parameter and prediction uncertainty using a second McMC sampling algorithm. Results of simulations will be presented that illustrate the complete workflow from geophysical parameter uncertainty analysis to the impact of model structural uncertainty on hydrologic parameter estimates.

Conference Paper↗

USGS2020

New operational national satellite burned area product

Introduction Lack of consistent spatial and temporal fire information with relevant spatial resolution hinders land management and broad-scale assessments of fire activity, especially in the eastern United States and the Great Plains where fi re is important ecologically and culturally. Remote sensing can be used to monitor fi re activity, augment existing fi re data, and fill information gaps. In particular, Landsat offers one of the most complete time series of remote sensing data sets as the Landsat satellites with spectral bands useful for mapping fi res and burn severity have been operational since 1984. Furthermore, Landsat satellite imagery collect data at a resolution useful for on-the-ground comparisons and management decisions. Methods A gradient-boosting regression model algorithm was used to predict burn probabilities (BP), indicating the likelihood that a pixel had burned in a fire. Then the algorithm translated the burn probability images to burn classification (BC) images using thresholding and region growing. Burned areas smaller than 5 acres were removed to reduce noise. The BP and BC products were generated for Landsat scenes collected from 1984 to present with <= 80% cloud cover and are available for download from USGSs EarthExplorer website (www.earthexplorer. usgs.gov). Annual summaries were also produced from the scene-level BC and BP products. These include: 1) the maximum burn probability value from all scenes in a year, (2) the burn classification count, or the number of times a pixel was classified as burned in a year, (3) the burn date, or the day of year of the first Landsat scene a pixel was classified as burned, and (4) a burn classification which removes burned areas still visible from previous years with a filter. Results The Landsat Burned Area Products identified 183% more burned area than the Monitoring Trends in Burn Severity and 56% more burned area than the MODIS MCD64A1.006 products2. From 1984 through 2018, annual burned area mapped by the Landsat Burned Area Products averaged 30,000 km2, ranged between 14,000 km2 in 1991 and 46,500 km2 in 2012, and increased over time at a rate of 356 km2/year. The Landsat Burned Area products consistently map burned areas 5 acres or larger over time while capturing unburned islands within fi res. Such characteristics are critical for understanding long-term impacts of fi res on human and natural systems. The products are being routinely produced as new Landsat data are collected and demonstrate the utility remote sensing offers for monitoring landscape change.

Report↗

USGS2023

Camera trap distance sampling survey design, Andersen Airforce Base, Guam

Reliable population estimates of animal density is one of the most elementary needs for the control and management of wildlife, particularly for introduced ungulates on oceanic islands. On Guam, Philippine deer (Rusa marianna) and wild pigs (Sus scrofa; wild boar and descendants of domestic pigs) cause agricultural and ecological damage and are hunted for recreational, nutritional, and cultural uses. Most common population estimation methods are based on capture-recapture and related methods that require marking or uniquely identifying individuals. Capturing, marking, and either recapturing or resighting individuals repeatedly is labor intensive and expensive. In many situations marking or individually distinguishing animals is not feasible, necessitating estimating densities and abundance from unmarked animal populations. Motion-triggered camera traps are a relatively low-cost approach that can be used to generate presence/pseudo-absence and indices of relative abundance on multiple species simultaneously. We used distance sampling with camera traps to estimate deer and pig densities from non-independent observations of unmarked animals while accounting for imperfect detection where some present individuals are not detected. We present methods to (1) process the digital imagery data automatically for species detection and species categorization using a machine learning algorithm, (2) automatically estimate distance to detected species using a separate machine learning algorithm, and (3) estimate densities using distance sampling with camera trap methods. We compare accuracy statistics and results of ungulate densities estimated from automated methods to those estimated from manual assessment. We collected 7,695 videos: 381 videos contained deer and 377 contained pigs. The object detection and identification model performed well with overall accuracy above 80% and F1 scores above 0.9. The hazard-rate key detection function was chosen for deer and pigs based on Akaike’s information criterion accounting for overdispersion. Deer density estimates were 0.53 ± 0.20 deer/ha with higher density in the Plateau area than the Tarague area of Guam. Pig density estimates were 0.53 ± 0.32 pigs/ha, also with higher densities in the Plateau area than the Tarague area. Coefficients of variation ranged from 0.38 to 1.15, and greater numbers of camera traps would be required for pigs than deer to achieve desired coefficients of variation. On average, 101.9 ± 82.3 deer and 131.6 ± 118.8 pigs were detected per day. Microsite heterogeneity affected densities where orientation-specific estimates were less precise than estimates made with the full dataset. We developed a camera trap survey design based on standard camera trapping sampling protocols using motion-activated, digital cameras and determined that distance sampling methods using camera traps produce reliable densities of unmarked deer and pigs on Guam. Our camera trap survey design is based on a regularly sized trapping grid that is generalizable and can be expanded to survey other areas of Guam.

Guam↗