USGS Science⌕ Search

SEARCH · USGS Science

Results for “Algorithms”

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

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

At least 163 records · Page 9Linked to original sources
USGS2012

Combining satellite-based fire observations and ground-based lightning detections to identify lightning fires across the conterminous USA

Lightning fires are a common natural disturbance in North America, and account for the largest proportion of the area burned by wildfires each year. Yet, the spatiotemporal patterns of lightning fires in the conterminous US are not well understood due to limitations of existing fire databases. Our goal here was to develop and test an algorithm that combined MODIS fire detections with lightning detections from the National Lightning Detection Network to identify lightning fires across the conterminous US from 2000 to 2008. The algorithm searches for spatiotemporal conjunctions of MODIS fire clusters and NLDN detected lightning strikes, given a spatiotemporal lag between lightning strike and fire ignition. The algorithm revealed distinctive spatial patterns of lightning fires in the conterminous US While a sensitivity analysis revealed that the algorithm is highly sensitive to the two thresholds that are used to determine conjunction, the density of fires it detected was moderately correlated with ground based fire records. When only fires larger than 0.4 km 2 were considered, correlations were higher and the root-mean-square error between datasets was less than five fires per 625 km 2 for the entire study period. Our algorithm is thus suitable for detecting broad scale spatial patterns of lightning fire occurrence, and especially lightning fire hotspots, but has limited detection capability of smaller fires because these cannot be consistently detected by MODIS. These results may enhance our understanding of large scale patterns of lightning fire activity, and can be used to identify the broad scale factors controlling fire occurrence.

IEEE Journal of Selected Topics in Applied Earth O↗
USGS2014

Comparing species distribution models constructed with different subsets of environmental predictors

Aim To assess the usefulness of combining climate predictors with additional types of environmental predictors in species distribution models for range-restricted species, using common correlative species distribution modelling approaches. Location Florida, USA Methods We used five different algorithms to create distribution models for 14 vertebrate species, using seven different predictor sets: two with bioclimate predictors only, and five ‘combination’ models using bioclimate predictors plus ‘additional’ predictors from groups representing: human influence, land cover, extreme weather or noise (spatially random data).We use a linear mixed-model approach to analyse the effects of predictor set and algorithm on model accuracy, variable importance scores and spatial predictions. Results Regardless of modelling algorithm, no one predictor set produced significantly more accurate models than all others, though models including human influence predictors were the only ones with significantly higher accuracy than climate-only models. Climate predictors had consistently higher variable importance scores than additional predictors in combination models, though there was variation related to predictor type and algorithm. While spatial predictions varied moderately between predictor sets, discrepancies were significantly greater between modelling algorithms than between predictor sets. Furthermore, there were no differences in the level of agreement between binary ‘presence–absence’ maps and independent species range maps related to the predictor set used. Main conclusions Our results indicate that additional predictors have relatively minor effects on the accuracy of climate-based species distribution models and minor to moderate effects on spatial predictions. We suggest that implementing species distribution models with only climate predictors may provide an effective and efficient approach for initial assessments of environmental suitability.

Florida↗
USGS2026

The potential impact of three-dimensional distributed slip models derived from real-time GNSS data on the performance of the ShakeAlert earthquake early warning system for slab interface earthquakes

The ShakeAlert® earthquake early warning (EEW) system is designed to warn users of imminent strong ground motion with sufficient time to take protective actions. ShakeAlert currently uses three algorithms to characterize the earthquake source. One estimates the location and magnitude using the first few seconds of the P wave and, while fast, tends to underestimate magnitude for M w 7.0+ earthquakes. A second estimates the location, orientation, length, and corresponding magnitude of a line source using observed peak ground acceleration and contributes primarily to M w 5.5+ earthquakes. The third infers earthquake magnitude from peak ground displacement measured using Global Navigation Satellite System (GNSS) data and offers nonsaturating magnitudes for M w 7.0+ earthquakes. Other EEW algorithms exist that infer temporally evolving spatially variable slip on a 3D fault surface from real‐time GNSS data, information that might enable more accurate and timely alerts in the event of large‐magnitude subduction interface earthquakes. Here, we evaluate the potential contribution of one such algorithm, BEFORES ( Minson et al. , 2014 ), to improve ShakeAlert performance through a simulated real‐time implementation of Bayesian evidence‐based fault orientation and real‐time earthquake slip (BEFORES) and other ShakeAlert algorithms using data for eight M w 7.6+ earthquakes. The test results demonstrate that BEFORES can produce well‐constrained and accurate magnitude estimates as soon as or sooner than other EEW algorithms, in turn enabling it to increase the amount of warning time users receive in many cases. However, with a modified Mercalli intensity (MMI) threshold of 3.5, which is commonly used for issuing alerts, BEFORES would tend to alert large geographic regions that did not feel strong shaking (MMI 6+). This effect can be mitigated using a higher alert threshold of MMI 4.5 without negative impact on the amount of warning time obtainable with BEFORES.

Bulletin of the Seismological Society of America↗
USGS2022

Testing the ShakeAlert earthquake early warning system using synthesized earthquake sequences

We test the behavior of the United States (US) West Coast ShakeAlert earthquake early warning (EEW) system during temporally close earthquake pairs to understand current performance and limitations. We consider performance metrics based on source parameter and ground‐motion forecast accuracy, as well as on alerting timeliness. We generate ground‐motion times series for synthesized earthquake sequences from real data by combining the signals from pairs of well‐recorded earthquakes (⁠4.4≤M≤7.1⁠) using time shifts ranging from −60 to +180 s. We examine fore‐ and aftershock sequences, near‐simultaneous events in different source regions, and simulated out‐of‐network and offshore earthquakes. We find that the operational ShakeAlert algorithms Earthquake Point‐source Integrated Code (EPIC) and Finite‐Fault Rupture Detector (FinDer) and the Propagation of Local Undamped Motion (PLUM) method perform largely as expected: EPIC provides the best source location estimates and is often fastest but can underestimate magnitudes or, in extreme cases, miss large earthquakes; FinDer provides real‐time line‐source models and unsaturated magnitude estimates for large earthquakes but currently cannot process concurrent events and may mislocate offshore earthquakes; PLUM identifies pockets of strong ground motion, but can overestimate alert areas. Implications for system performance are: (1) spatially and temporally close events are difficult to identify separately; (2) challenging scenarios with foreshocks that are close in space and time can lead to missed alerts for large earthquakes; and (3) in these situations the algorithms can often estimate ground motion better than source parameters. To improve EEW, our work suggests revisiting the current algorithm weighting in ShakeAlert, to continue developments that focus on using ground‐motion data to aggregate alerts from multiple algorithms, and to investigate methods to optimally leverage algorithm ground‐motion estimates. For testing and certification of EEW performance in ShakeAlert and other EEW systems where applicable, we also suggest that 25 of our 73 scenarios become part of the baseline data set.

California↗
USGS2011

Derived crop management data for the LandCarbon Project

The LandCarbon project is assessing potential carbon pools and greenhouse gas fluxes under various scenarios and land management regimes to provide information to support the formulation of policies governing climate change mitigation, adaptation and land management strategies. The project is unique in that spatially explicit maps of annual land cover and land-use change are created at the 250-meter pixel resolution. The project uses vast amounts of data as input to the models, including satellite, climate, land cover, soil, and land management data. Management data have been obtained from the U.S. Department of Agriculture (USDA) National Agricultural Statistics Service (NASS) and USDA Economic Research Service (ERS) that provides information regarding crop type, crop harvesting, manure, fertilizer, tillage, and cover crop (U.S. Department of Agriculture, 2011a, b, c). The LandCarbon team queried the USDA databases to pull historic crop-related management data relative to the needs of the project. The data obtained was in table form with the County or State Federal Information Processing Standard (FIPS) and the year as the primary and secondary keys. Future projections were generated for the A1B, A2, B1, and B2 Intergovernmental Panel on Climate Change (IPCC) Special Report on Emissions Scenarios (SRES) scenarios using the historic data values along with coefficients generated by the project. The PBL Netherlands Environmental Assessment Agency (PBL) Integrated Model to Assess the Global Environment (IMAGE) modeling framework (Integrated Model to Assess the Global Environment, 2006) was used to develop coefficients for each IPCC SRES scenario, which were applied to the historic management data to produce future land management practice projections. The LandCarbon project developed algorithms for deriving gridded data, using these tabular management data products as input. The derived gridded crop type, crop harvesting, manure, fertilizer, tillage, and cover crop products are used as input to the LandCarbon models to represent the historic and the future scenario management data. The overall algorithm to generate each of the gridded management products is based on the land cover and the derived crop type. For each year in the land cover dataset, the algorithm loops through each 250-meter pixel in the ecoregion. If the current pixel in the land cover dataset is an agriculture pixel, then the crop type is determined. Once the crop type is derived, then the crop harvest, manure, fertilizer, tillage, and cover crop values are derived independently for that crop type. The following is the overall algorithm used for the set of derived grids. The specific algorithm to generate each management dataset is discussed in the respective section for that dataset, along with special data handling and a description of the output product.

Open-File Report↗
USGS2016

QRev—Software for computation and quality assurance of acoustic doppler current profiler moving-boat streamflow measurements—Technical manual for version 2.8

The software program, QRev applies common and consistent computational algorithms combined with automated filtering and quality assessment of the data to improve the quality and efficiency of streamflow measurements and helps ensure that U.S. Geological Survey streamflow measurements are consistent, accurate, and independent of the manufacturer of the instrument used to make the measurement. Software from different manufacturers uses different algorithms for various aspects of the data processing and discharge computation. The algorithms used by QRev to filter data, interpolate data, and compute discharge are documented and compared to the algorithms used in the manufacturers’ software. QRev applies consistent algorithms and creates a data structure that is independent of the data source. QRev saves an extensible markup language (XML) file that can be imported into databases or electronic field notes software. This report is the technical manual for version 2.8 of QRev.

Open-File Report↗
USGS2000

A three-dimensional finite-volume Eulerian-Lagrangian Localized Adjoint Method (ELLAM) for solute-transport modeling

This report documents the U.S. Geological Survey Eulerian-Lagrangian Localized Adjoint Method (ELLAM) algorithm that solves an integral form of the solute-transport equation, incorporating an implicit-in-time difference approximation for the dispersive and sink terms. Like the algorithm in the original version of the U.S. Geological Survey MOC3D transport model, ELLAM uses a method of characteristics approach to solve the transport equation on the basis of the velocity field. The ELLAM algorithm, however, is based on an integral formulation of conservation of mass and uses appropriate numerical techniques to obtain global conservation of mass. The implicit procedure eliminates several stability criteria required for an explicit formulation. Consequently, ELLAM allows large transport time increments to be used. ELLAM can produce qualitatively good results using a small number of transport time steps. A description of the ELLAM numerical method, the data-input requirements and output options, and the results of simulator testing and evaluation are presented. The ELLAM algorithm was evaluated for the same set of problems used to test and evaluate Version 1 and Version 2 of MOC3D. These test results indicate that ELLAM offers a viable alternative to the explicit and implicit solvers in MOC3D. Its use is desirable when mass balance is imperative or a fast, qualitative model result is needed. Although accurate solutions can be generated using ELLAM, its efficiency relative to the two previously documented solution algorithms is problem dependent.

Water-Resources Investigations Report↗
USGS2015

Efficient wetland surface water detection and monitoring via Landsat: Comparison with in situ data from the Everglades Depth Estimation Network

The U.S. Geological Survey is developing new Landsat science products. One, named Dynamic Surface Water Extent (DSWE), is focused on the representation of ground surface inundation as detected in cloud-/shadow-/snow-free pixels for scenes collected over the U.S. and its territories. Characterization of DSWE uncertainty to facilitate its appropriate use in science and resource management is a primary objective. A unique evaluation dataset developed from data made publicly available through the Everglades Depth Estimation Network (EDEN) was used to evaluate one candidate DSWE algorithm that is relatively simple, requires no scene-based calibration data, and is intended to detect inundation in the presence of marshland vegetation. A conceptual model of expected algorithm performance in vegetated wetland environments was postulated, tested and revised. Agreement scores were calculated at the level of scenes and vegetation communities, vegetation index classes, water depths, and individual EDEN gage sites for a variety of temporal aggregations. Landsat Archive cloud cover attribution errors were documented. Cloud cover had some effect on model performance. Error rates increased with vegetation cover. Relatively low error rates for locations of little/no vegetation were unexpectedly dominated by omission errors due to variable substrates and mixed pixel effects. Examined discrepancies between satellite and in situ modeled inundation demonstrated the utility of such comparisons for EDEN database improvement. Importantly, there seems no trend or bias in candidate algorithm performance as a function of time or general hydrologic conditions, an important finding for long-term monitoring. The developed database and knowledge gained from this analysis will be used for improved evaluation of candidate DSWE algorithms as well as other measurements made on Everglades surface inundation, surface water heights and vegetation using radar, lidar and hyperspectral instruments. Although no other sites have such an extensive in situ network or long-term records, the broader applicability of this and other candidate DSWE algorithms is being evaluated in other wetlands using this work as a guide. Continued interaction among DSWE producers and potential users will help determine whether the measured accuracies are adequate for practical utility in resource management.

Florida↗
USGS2021

Learning augmented methods for matching: Improving invasive species management and urban mobility

With the success of machine learning, integrating learned models into real-world systems has become a critical chal- lenge. Naively applying predictions to combinatorial opti- mization problems can incur high costs, which has motivated researchers to consider learning augmented algorithms that can make use of faulty or incomplete predictions. Inspired by two matching problems in computational sustainability where data are abundant, we consider the learning augmented min weight matching problem where some nodes are revealed online while others are known a priori, e.g., by being pre- dicted by machine learning. We develop an algorithm that is able to make use of this extra information and provably im- proves upon pessimistic online algorithms. We evaluate our algorithm on two settings from computational sustainability – the coordination of opportunistic citizen scientists for inva- sive species management and the matching between taxis and riders under uncertain trip duration predictions. In both cases, we perform extensive experiments on real-world datasets and find that our method outperforms baselines, showing how learning augmented algorithms can reliably improve solu- tions for problems in computational sustainability

New York↗
USGS2022

Predicting geothermal favorability in the western United States by using machine learning: Addressing challenges and developing solutions

Previous moderate- and high-temperature geothermal resource assessments of the western United States utilized weight-of-evidence and logistic regression methods to estimate resource favorability, but these analyses relied upon some expert decisions. While expert decisions can add confidence to aspects of the modeling process by ensuring only reasonable models are employed, expert decisions also introduce human bias into assessments. This bias presents a source of error that may affect the performance of the models and resulting resource estimates. Our study aims to reduce expert input through robust data-driven analyses and better-suited data science techniques, with the goals of saving time, reducing bias, and improving predictive ability. We present six favorability maps for geothermal resources in the western United States created using two strategies applied to three modern machine learning algorithms (logistic regression, support-vector machines, and XGBoost). To provide a direct comparison to previous assessments, we use the same input data as the 2008 U.S. Geological Survey (USGS) conventional moderate- to high-temperature geothermal resource assessment. The six new favorability maps required far less expert decision-making, but broadly agree with the previous assessment. Despite the fact that the 2008 assessment results employed linear methods, the non-linear machine learning algorithms (i.e., support-vector machines and XGBoost) produced greater agreement with the previous assessment than the linear machine learning algorithm (i.e., logistic regression). It is not surprising that geothermal systems depend on non-linear combinations of features, and we postulate that the expert decisions during the 2008 assessment accounted for system non-linearities. Substantial challenges to applying machine learning algorithms to predict geothermal resource favorability include severe class imbalance (i.e., there are very few known geothermal systems compared to the large area considered), and while there are known geothermal systems (i.e., positive labels), all other sites have an unknown status (i.e., they are unlabeled), instead of receiving a negative label (i.e., the known/proven absence of a geothermal resource). We address both challenges through a custom undersampling strategy that can be used with any algorithm and then evaluated using F1 scores.

western United States↗
USGS2022

Optimizing surveys of fall-staging geese using aerial imagery and automated counting

Ocular aerial surveys allow efficient coverage of large areas and can be used to monitor abundance and distribution of wild populations. However, uncertainty around resulting population estimates can be large due to difficulty in visually identifying and counting animals from aircraft, as well as logistical challenges in estimating detection probabilities. Photographic aerial surveys can mitigate these challenges and can allow flight at higher altitudes to minimize disturbance of birds and improve safety for surveyors. We evaluated a photographic aerial survey that incorporated a systematic sampling design with automated photo capture and processing for fall-staging geese at Izembek Lagoon, Alaska, in 2017–2019. Ocular aerial surveys have been completed at Izembek Lagoon for >40 years. For the new photo survey, we used a commercial system to automatically trigger cameras at preset points. We then applied a machine-learning algorithm trained to automatically identify and count geese in our photos, manually corrected those counts, and quantified the algorithm's accuracy. We translated corrected counts into density and extrapolated mean density across the entire lagoon to estimate total population size for Pacific brant ( Branta bernicla ) and cackling geese ( B. hutchinsii ). The automated algorithm undercounted geese, but successfully identified the small subset of photos containing geese. Manual correction was therefore needed only for photos automatically identified as containing geese, allowing substantial reduction of workload. Manually-corrected, photo-based estimates of Pacific brant and cackling goose population sizes were larger and more precise than ocular estimates in all 3 years. To reduce costs with little penalty for variance around population estimates, the photographic survey design could be optimized by reducing the number of transects to ~67% of the current number while still manually correcting all photos in which the automated algorithm detected geese. Further years of both ocular and photo surveys would be needed to calibrate the photo estimates against the >40-year timeseries of the ocular survey, after which the photo series could successfully guide management of Pacific brant. As technologies continue to advance, we expect photographic surveys with automated counting to be easily implemented and advantageous to many monitoring programs.

Alaska↗
USGS2021

Random forest

This entry defines and discusses the random forest machine learning algorithm. The algorithm is used to predict class or quantities for target variables using values of a set of predictor variables. It uses decision trees that are generated from bootstrap sampling of the training data set to create a "forest". The entry discusses the algorithm steps, the interpretative tools of the resulting model, current areas of research, and its limitations. Applications to the quantitative geosciences are reviewed as well as availability of software to implement the algorithm.

Book chapter↗
USGS2007

Generating an image of dispersive energy by frequency decomposition and slant stacking

We present a new algorithm for calculating an image of dispersive energy in the frequency-velocity (f-v) domain. The frequency decomposition is first applied to a shot gather in the offset-time domain to stretch impulsive data into pseudo-vibroseis data or frequency-swept data. Because there is a deterministic relationship between frequency and time in a sweep used in the frequency decomposition, the first step theoretically completes the transform from time to frequency. The slant stacking is then performed on the frequency-swept data to complete the transform from offset to velocity. This simple two-step algorithm generates an image of dispersive energy in the f-v domain. The straightforward transform only uses offset information of data so that this algorithm can be applied to data acquired with arbitrary geophone-acquisition geometry. Examples of synthetic and real-world data demonstrate that this algorithm generates accurate images of dispersive energy of the fundamental as well as higher modes. ?? Birkha??user Verlag, Basel, 2007.

Pure and Applied Geophysics↗
USGS2024

Monarch butterfly (Danaus plexippus) roost site-selection criteria and locations east of the Appalachian Mountains, U.S.A.

The monarch butterfly is a flagship species and pollinator whose populations have declined by 85% in the recent two decades. Their largest population overwinters in Mexico, then disperses across eastern North America during March to August. During September-December, they return south using two flyways, one that spans the central United States and another that follows the Atlantic coast. Migrating monarchs fly diurnally and roost in groups nocturnally. We sought to determine the criteria this species uses to select roost sites, and the landscape context where those sites are found. We developed species distribution models of the landscape context of Atlantic flyway roost sites via citizen scientist observations and environmental variables that affect monarchs in the adult stage prior to migration, using two algorithms (Maximum Entropy and Genetic Algorithm for Ruleset Prediction). We developed two model validation methods: a citizen scientist smartphone application and peer-informed comparisons with aerial imagery. Proximity to surface water, elevation, and vegetative cover were the most important criteria for monarch roost site selection. Our model predicted 2.6 million ha (2.9% of the study area) of suitable roosting habitat in the Atlantic flyway, with the greatest availability along the Atlantic coastal plain and Appalachian Mountain ridges. Conservation of this species is difficult, as monarchs range over both large areas and various habitat types, and most current monarch research and conservation efforts are focused on the breeding and overwintering periods. These models can serve to help prioritize surveys of roosting sites and conservation efforts during the monarchs’ fall migration.

Journal of Insect Behavior↗
USGS2008

Development of a three-dimensional, regional, coupled wave, current, and sediment-transport model

We are developing a three-dimensional numerical model that implements algorithms for sediment transport and evolution of bottom morphology in the coastal-circulation model Regional Ocean Modeling System (ROMS v3.0), and provides a two-way link between ROMS and the wave model Simulating Waves in the Nearshore (SWAN) via the Model-Coupling Toolkit. The coupled model is applicable for fluvial, estuarine, shelf, and nearshore (surfzone) environments. Three-dimensional radiation-stress terms have been included in the momentum equations, along with effects of a surface wave roller model. The sediment-transport algorithms are implemented for an unlimited number of user-defined non-cohesive sediment classes. Each class has attributes of grain diameter, density, settling velocity, critical stress threshold for erosion, and erodibility constant. Suspended-sediment transport in the water column is computed with the same advection-diffusion algorithm used for all passive tracers and an additional algorithm for vertical settling that is not limited by the CFL criterion. Erosion and deposition are based on flux formulations. A multi-level bed framework tracks the distribution of every size class in each layer and stores bulk properties including layer thickness, porosity, and mass, allowing computation of bed morphology and stratigraphy. Also tracked are bed-surface properties including active-layer thickness, ripple geometry, and bed roughness. Bedload transport is calculated for mobile sediment classes in the top layer. Bottom-boundary layer submodels parameterize wave-current interactions that enhance bottom stresses and thereby facilitate sediment transport and increase bottom drag, creating a feedback to the circulation. The model is demonstrated in a series of simple test cases and a realistic application in Massachusetts Bay.

Computers & Geosciences↗
USGS2016

Challenges for mapping cyanotoxin patterns from remote sensing of cyanobacteria

Using satellite imagery to quantify the spatial patterns of cyanobacterial toxins has several challenges. These challenges include the need for surrogate pigments – since cyanotoxins cannot be directly detected by remote sensing, the variability in the relationship between the pigments and cyanotoxins – especially microcystins (MC), and the lack of standardization of the various measurement methods. A dual-model strategy can provide an approach to address these challenges. One model uses either chlorophyll -a (Chl -a ) or phycocyanin (PC) collected in situ as a surrogate to estimate the MC concentration. The other uses a remote sensing algorithm to estimate the concentration of the surrogate pigment. Where blooms are mixtures of cyanobacteria and eukaryotic algae, PC should be the preferred surrogate to Chl -a . Where cyanobacteria dominate, Chl -a is a better surrogate than PC for remote sensing. Phycocyanin is less sensitive to detection by optical remote sensing, it is less frequently measured, PC laboratory methods are still not standardized, and PC has greater intracellular variability. Either pigment should not be presumed to have a fixed relationship with MC for any water body. The MC-pigment relationship can be valid over weeks, but have considerable intra- and inter-annual variability due to changes in the amount of MC produced relative to cyanobacterial biomass. To detect pigments by satellite, three classes of algorithms (analytic, semi-analytic, and derivative) have been used. Analytical and semi-analytical algorithms are more sensitive but less robust than derivatives because they depend on accurate atmospheric correction; as a result derivatives are more commonly used. Derivatives can estimate Chl -a concentration, and research suggests they can detect and possibly quantify PC. Derivative algorithms, however, need to be standardized in order to evaluate the reproducibility of parameterizations between lakes. A strategy for producing useful estimates of microcystins from cyanobacterial biomass is described, provided cyanotoxin variability is addressed.

Harmful Algae↗
USGS2022

Spectral mixture analysis for surveillance of harmful algal blooms (SMASH): A field-, laboratory-, and satellite-based approach to identifying cyanobacteria genera from remotely sensed data

Algal blooms around the world are increasing in frequency and severity, often with the possibility of adverse effects on human and ecosystem health. The health and economic impacts associated with harmful algal blooms, or HABs, provide compelling rationale for developing new methods for monitoring these events via remote sensing. Although concentrations of chlorophyll- a and key pigments like phycocyanin are routinely estimated from satellite images and used to infer algal or cyanobacterial cell counts, current methods are unable to provide information on the taxonomic composition of a bloom. This study introduced a new approach capable of differentiating among genera based on their reflectance characteristics: Spectral Mixture Analysis for Surveillance of HABs, or SMASH. The foundation of SMASH is a multiple endmember spectral mixture analysis (MESMA) algorithm that takes a library of cyanobacteria endmembers and a hyperspectral image as input and estimates the fractional abundance of each genus, plus water, on a per-pixel basis. Importantly, we assume that the water column consists of only pure water and cyanobacteria, implying that our linear spectral unmixing models do not account for other optically active constituents such as suspended sediment and colored dissolved organic matter (CDOM). We used reflectance spectra for 12 genera measured under a microscope to populate an algal spectral library and applied the SMASH workflow to satellite images from four waterbodies across the United States. Normalized spectral separability scores indicated that the 12 genera were distinct from one another and the MESMA algorithm reproduced known input fractions for simulated mixtures that included all pairwise combinations of genera and water. We used Upper Klamath Lake as an example to illustrate data products generated via SMASH: maps of the normalized difference chlorophyll index and cyanobacterial index, a MESMA-based classification of algal genera, fraction images for each endmember, and a root mean square error (RMSE) image that summarizes uncertainty. For Upper Klamath Lake, these outputs highlighted a complex algal bloom featuring several genera, primarily Aphanizomenon , and intricate spatial patterns associated with gyres. The maximum RMSE constraint imposed on the MESMA algorithm provided a means of avoiding false positive detection of genera not present in a waterbody but must not be set so low as to leave much of an image unclassified in cases where genera included in the library are present. Comparison of endmember fractions with relative biovolumes calculated from field samples indicated that taxonomic information from SMASH was consistent with field observations. For example, the algorithm successfully identified Microcystis in Owasco Lake but avoided misclassifying Asterionella , a genus not yet included in our library, in Detroit Lake. This proof-of-concept investigation demonstrates the potential of SMASH to enhance our understanding of algal blooms, particularly with respect to their spatial and temporal dynamics.

New York, Oregon, Texas↗
USGS2007

Field test comparison of an autocorrelation technique for determining grain size using a digital 'beachball' camera versus traditional methods

This extensive field test of an autocorrelation technique for determining grain size from digital images was conducted using a digital bed-sediment camera, or 'beachball' camera. Using 205 sediment samples and >1200 images from a variety of beaches on the west coast of the US, grain size ranging from sand to granules was measured from field samples using both the autocorrelation technique developed by Rubin [Rubin, D.M., 2004. A simple autocorrelation algorithm for determining grain size from digital images of sediment. Journal of Sedimentary Research, 74(1): 160-165.] and traditional methods (i.e. settling tube analysis, sieving, and point counts). To test the accuracy of the digital-image grain size algorithm, we compared results with manual point counts of an extensive image data set in the Santa Barbara littoral cell. Grain sizes calculated using the autocorrelation algorithm were highly correlated with the point counts of the same images (r2 = 0.93; n = 79) and had an error of only 1%. Comparisons of calculated grain sizes and grain sizes measured from grab samples demonstrated that the autocorrelation technique works well on high-energy dissipative beaches with well-sorted sediment such as in the Pacific Northwest (r2 ??? 0.92; n = 115). On less dissipative, more poorly sorted beaches such as Ocean Beach in San Francisco, results were not as good (r2 ??? 0.70; n = 67; within 3% accuracy). Because the algorithm works well compared with point counts of the same image, the poorer correlation with grab samples must be a result of actual spatial and vertical variability of sediment in the field; closer agreement between grain size in the images and grain size of grab samples can be achieved by increasing the sampling volume of the images (taking more images, distributed over a volume comparable to that of a grab sample). In all field tests the autocorrelation method was able to predict the mean and median grain size with ???96% accuracy, which is more than adequate for the majority of sedimentological applications, especially considering that the autocorrelation technique is estimated to be at least 100 times faster than traditional methods.

Sedimentary Geology↗