Search USGS⌕ Search

SEARCH · Search USGS

Results for “Data”

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 991 records · Page 55Linked to original sources
USGS2015

Archive of Sidescan Sonar and Swath Bathymetry Data Collected During USGS Cruise 13CCT04 Offshore of Petit Bois Island, Gulf Islands National Seashore, Mississippi, August 2013

In August of 2013, the U.S. Geological Survey conducted a geophysical survey offshore of Petit Bois Island, Mississippi. This effort was part of the U.S. Geological Survey Gulf of Mexico Science Coordination partnership with the U.S. Army Corps of Engineers to assist the Mississippi Coastal Improvements Program and the Northern Gulf of Mexico Ecosystem Change and Hazards Susceptibility Project, by mapping the shallow geologic stratigraphic framework of the Mississippi Barrier Island Complex. This geophysical survey will provide additional data necessary for scientists to define, interpret, and provide baseline bathymetry and seafloor habitat for this area, and to aid scientists in predicting future geomorphological changes of the islands with respect to climate change, storm impact, and sea-level rise. Furthermore, these data will provide information for barrier island restoration, particularly in Camille Cut, and protection for the historical Fort Massachusetts on Ship Island, Mississippi. The geophysical data were collected during one cruise ( USGS Field Activity Numbers 13CCT04 ) aboard the Research Vessel Tommy Munro offshore along the gulf side of Petit Bois Island, Gulf Islands National Seashore, Mississippi. Data were acquired with the following equipment: a Systems Engineering and Assessment, Ltd., SWATH plus interferometric sonar (468 kilohertz (kHz)), an EdgeTech 424 (4-24 kHz), an EdgeTech 525i chirp subbottom profiling system, and a Klein 3900 sidescan sonar system. This report serves as an archive of the processed interferometric swath bathymetry and sidescan sonar data. Geographic information system data products include an interpolated digital elevation model, an acoustic backscatter mosaic, trackline maps, and point data files. Additional files include error analysis maps, Field Activity Collection System logs, and formal Federal Geographic Data Committee metadata. NOTE: These data are scientific in nature and are not to be used for navigation. Any use of trade names is for descriptive purposes only and does not imply endorsement by the U.S. Government.

Mississippi↗
USGS1987

Water resources data for Arkansas, water year 1986

Water resources data for the 1986 water year for Arkansas consist of records of gage height, discharge, and water quality of streams; water quality of lakes; and water levels and water quality of wells. This report contains discharge records for 48 gaging stations; water quality for 154 stations, 73 partial-record stations, 5 observation wells, and 1 precipitation station, and water levels for 96 observation wells. Also included are data for 85 crest-stage partial record stations. Additional water data were collected at various sites, not part of the systematic data-collection program, and are published as miscellaneous measurements. These data represent that part of the National Water Data System operated by the U.S. Geological Survey and cooperating State and Federal agencies in Arkansas. Records of discharge or gage height of streams, and contents or elevation of lakes were first published in a series of U.S. Geological Survey water-supply papers entitled "Surface Water Supply of the United States." Through September 30, 1960, these water-supply papers were in an annual series and for 1961-65 and 1966-70 were in a 5 -year series. Records of chemical quality, water temperatures, and suspended sediment were published from 1941 to 1970 in an annual series of water-supply papers entitled "Quality of Surface Waters of the United States." Records of ground-water levels were published from 1935 to 1974 in a series of water-supply papers entitled "Ground Water Levels in the United States." Water-supply papers may be consulted in the libraries of the principal cities in the United States or may be purchased from Branch of Distribution, U.S. Geological Survey, 1200 South Eads Street, Arlington, VA 22202. For water years 1961 through 1974, streamflow data were released by the Geological Survey in annual reports on a State-boundary basis. Water-quality records for water years 1964 through 1974 were similarly released, either in separate reports or in conjuction with streamflow records. Beginning with the 1975 water year, water data for streamflow, water quality, and ground water are published as an official Survey report on a State-boundary basis. These official Survey reports carry an identification number consisting of the two-letter State abbreviation, the last two digits of the water year, and the volume number. For example, this report is identified as "U.S. Geological Survey Water Data Report AR-86-1."

Arkansas↗
USGS1996

Water resources data, Arkansas, water year 1995

The Water Resources Division of the U.S. Geological Survey, in cooperation with State, Federal, and other local governmental agencies, obtains a large amount of data pertaining to the water resources of Arkansas each year. These data, accumulated during many water years, constitute a valuable data base for developing an improved understanding of the water resources of the State. Water resources data reported for the 1995 water year for Arkansas consist of records of discharge and water quality (physical measurements and chemical concentrations) of streams; water quality of lakes; and ground-water levels and ground-water quality. Data from selected sites in Missouri and Oklahoma are also included. This report contains daily discharge records for 49surface-water gaging stations and four daily sediment stations; water quality for 146 surface-water stations, 110 ground-water quality wells and springs, 8 ground-water-level observation wells, and one precipitation-quality station. Also included are data for 102 peak-discharge partial-record stations. Additional water data were collected at various sites, not part of the systematic data-collection program, and are published as miscellaneous measurements. These data represent that part of the National Water Information System operated by the U.S. Geological Survey in cooperation with State and Federal agencies in Arkansas.

Arkansas↗
USGS2002

Water Resources Data--Nebraska, Water Year 2002

The Water Resources Discipline of the U.S. Geological Survey (USGS), in cooperation with State and local agencies, obtains a large amount of data pertaining to the water resources of Nebraska each water year. These data, accumulated during many water years, constitute a valuable data base for developing an improved understanding of the water resources of the State. To make these data readily available to interested parties outside the USGS, the data are published annually in this report series entitled ?Water Resources Data - Nebraska.' The Nebraska water resources data report for water year 2002 includes records of stage, discharge, and water quality of streams; stage and/or contents of lakes and reservoirs; and water levels and quality of ground water in wells. This report contains records of stream stage for 3 stations; stream discharge for 96 continuous and 5 crest-state gaging stations, and 3 miscellaneous and 55 low-flow sites; stream water quality for 23 gaging stations and 5 miscellaneous sites; water elevation and/or contents for 1 lake and 1 reservoir; ground-water levels for 43 observation wells; and ground-water quality for 115 wells. These data represent that part of the National Water Data System collected in and near Nebraska by the U.S. Geological Survey and cooperating local, state and Federal agencies.

Water Data Report↗
USGS1987

Water resources data, Texas, water year 1986. Volume 2: San Jacinto River basin, Brazos River basin, San Bernard River basin, and intervening coastal basins

The Water Resources Division of the U.S. Geological Survey, in cooperation with Federal, State, and City agencies, obtains a large amount of data pertaining to the water resources of Texas each water year. Such data, accumulated during many water years, constitute a valuable data base for developing an improved understanding of the water resources of the State. To make these data readily available to interested parties outside the Geological Survey, the data are published annually in three volumes of this report series entitled "Water Resources Data - Texas." This report series includes records of stage, discharge, and water quality of streams and canals; stage, contents, and water quality of lakes and reservoirs. Volume 2 contains records for water discharge at 112 gaging stations; stage only at 9 gaging station; stage and contents at 24 lakes and reservoirs; and water quality at 72 gaging stations. Also included are data for 39 partial-record stations. The data in this report represent that part of the National Water Data System collected by the U.S. Geological Survey and cooperating State and Federal agencies in Texas.

Texas↗
USGS1997

Water resources data, Washington, water year 1996

The Water Resources Division of the U.S. Geological Survey, in cooperation with State, local, and other Federal agencies, obtains a large amount of data pertaining to the water resources of Washington each water year. These data, accumulated during many water years, constitute a valuable data base for developing an improved understanding of the water resources of the State. To make these data readily available to interested parties outside the Geological Survey, the data are published annually in this report series entitled "Water Resources Data Washington." This report includes records on both surface and ground water in the State. Specifically, it contains: (1) Discharge records for 225 stream-gaging stations, 29 partial-record or miscellaneous streamflow stations, and 3 crest-stage partial-record streamflow station; (2) stage only records for 2 gaging stations; (3) stage and (or) content records for 34 lakes and reservoirs; (4) water-quality records for 68 stream, canal and drain sites; (5) water-level records for 4 observation wells; and (6) water quality record for 1 observation well. Records included for stream stages and for ground-water levels are only a small fraction of those obtained during the water year. This series of annual reports for Washington began with the 1961 water year with a report that contained only data relating to the quantities of surface water. For the 1964 water year, a similar report was introduced that contained only data relating to water quality. Beginning with the 1975 water year, the report format was changed to present, in one volume, data on quantities of surface water, quality of surface and ground water, and ground-water levels.

Washington↗
USGS1973

Water resources data for Wyoming, water year 1972; part 2, water quality records

Water-resources data for the 1972 water year for Wyoming include records of data for the chemical and physical characteristics of surface and ground water. Data on the quality of surface water (chemical, temperature, and sediment) were collected from designated sampling sites at predetermined intervals such as once daily, weekly, monthly or less frequently. At some sites water-temperature data were recorded on charts by recorders. Data are given for 138 sampling stations of which 112 are continuous-record stations and 26 are miscellaneous sites. Chemical analyses also are given for 127 ground-water sites. Locations of surface water-quality stations are shown in figure 1 and locations of ground-water sites are shown in figure 2. These data represent that portion of the National Water Data System collected by the U.S. Geological Suryey and cooperating State and Federal agencies in Wyoming. The records were collected and computed by the Water Resources Division of the U.S. Geological Survey under the direction of Robert L. Cushman, district chief. Wyoming district personnel who contributed significantly to the collection and preparation of data included in this report were: E. R. Cox, S. A. Druse, W. G. Hodson, M. E. Lowry, D. J. O'Connell, G. W. Armentrout, Jr., M.A. Crist, B. H. Ringen, W. B. Borchert, W.R. Glass, G. C. Lines, H.B. Fabricius, P. B. McCollam, C. E. Obert, K. G. Polinoski, T. W. Custis, R. L. Muench, J. O. Ragsdale, M. D. Stevens, R .. C. Baumann, D. W. Brownell, R. V. Foresman, J.M. Johnson, J. L. Lynch, M. L. Smalley, and J. P. Kyner. The Geological Survey has published records of chemical quality, water temperatures, and sediment since 1941 in an annual series of water-supply papers entitled, "Quality of Surface Water of the United States." Beginning with the 1964 water year, water-quality records have been released by the Geological Survey in annual reports on a State-boundary basis. These reports are for limited distribution and are designed primarily for rapid release of data shortly after the end of the water year. These records will be published later in Geological Survey water-supply papers.

Wyoming↗
USGS1974

Water resources data for Wyoming, water year 1973, part 2, water quality records

Water-resources data for the 1973 water year for Wyoming include records of data for the chemical and physical characteristics of surface and ground water. Data on the quality of surface water (chemical, temperature, and sediment) were collected from designated sampling sites at predetermined intervals such as once daily, weekly, monthly or less frequently. At some sites water-temperature data were recorded on charts by recorders. Data are given for 133 sampling stations of which 116 are continuous-record stations and 17 are miscellaneous sites. Chemical analyses are also given for 68 ground-water sites. Locations of surface water-quality stations are shown in figure 1 and locations of ground-water sites are shown in figure 2. These data represent that portion of the National Water Data System collected by the U. S. Geological Survey and cooperating State and Federal agencies in Wyoming. The records were collected and computed by the Water Resources Division of the U. S. Geological Survey under the direction of Robert L. Cushman, district chief. Wyoming district personnel who contributed significantly to the collection and preparation of data included in this report were: E. R. Cox, S. A. Druse, W. G. Hodson, M. E. Lowry, D. J. O'Connell, G. W. Armentrout, Jr., M. A. Crist, B. H. Ringen, W. B. Borchert, W. R. Glass, G. C. Lines, H. B. Fabricius, P. B. McCollam, C. E. Obert, K. G. Polinoski, T. W. Custis, R. L. Muench, J. 0. Ragsdale, M. D. Stevens, R. C. Baumann, D. W. Brownell, J. L. Lynch, S. L. Green, M. L. Smalley, and J. P. Kyner. The Geological Survey has published records of chemical quality, water temperatures, and sediment since 1941 in an annual series of water-supply papers entitled, "Quality of Surface Water of the United States." Beginning with the 1964 water year, water-quality records have been released by the Geological Survey in annual reports on a State-boundary basis. These reports are for limited distribution and are designed primarily for rapid release of data shortly after the end of the water year. These records will be published later in Geological Survey water-supply papers.

Wyoming↗
USGS2022

Eyes on the herd: Quantifying ungulate density from satellite, unmanned aerial systems, and GPScollar data

Novel approaches to quantifying density and distributions could help biologists adaptively manage wildlife populations, particularly if methods are accurate, consistent, cost-effective, rapid, and sensitive to change. Such approaches may also improve research on interactions between density and processes of interest, such as disease transmission across multiple populations. We assess how satellite imagery, unmanned aerial system (UAS) imagery, and Global Positioning System (GPS) collar data vary in characterizing elk density, distribution, and count patterns across times with and without supplemental feeding at the National Elk Refuge (NER) in the US state of Wyoming. We also present the first comparison of satellite imagery data with traditional counts for ungulates in a temperate system. We further evaluate seven different aggregation metrics to identify the most consistent and sensitive metrics for comparing density and distribution across time and populations. All three data sources detected higher densities and aggregation locations of elk during supplemental feeding than non-feeding at the NER. Kernel density estimates (KDEs), KDE polygon areas, and the first quantile of interelk distances detected differences with the highest sensitivity and were most highly correlated across data sources. Both UAS and satellite imagery provide snapshots of density and distribution patterns of most animals in the area at lower cost than GPS collars. While satellite-based counts were lower than traditional counts, aggregation metrics matched those from UAS and GPS data sources when animals appeared in high contrast to the landscape, including brown elk against new snow in open areas. UAS counts of elk were similar to traditional ground-based counts on feed grounds and are the best data source for assessing changes in small spatial extents. Satellite, UAS, or GPS data can provide appropriate data for assessing density and changes in density from adaptive management actions. For the NER, where high elk densities are beneath controlled airspace, GPS collar data will be most useful for evaluating how management actions, including changes in the dates of supplemental feeding, influence elk density and aggregation across large spatial extents. Using consistent and sensitive measures of density may improve research on the drivers and effects of density within and across a wide range of species.

Wyoming↗
USGS2021

Integrating telemetry data at several scales with spatial capture–recapture to improve density estimates

Accurate population estimates are essential for monitoring and managing wildlife populations. Mark–recapture sampling methods have regularly been used to estimate population parameters for rare and cryptic species, including the federally listed Mojave desert tortoise ( Gopherus agassizii ); however, the methods employed are often plagued by violations of statistical assumptions, which have the potential to bias density estimates. By incorporating spatial information into conventional density estimation models, spatial capture–recapture (SCR) models can account for common assumption violations such as spatially heterogeneous detection probabilities and temporary emigration when animals leave plots during a survey. We conducted mark–recapture surveys at 10 1-km 2 plots in and adjacent to the Ivanpah Valley of California and Nevada from 2015 to 2019. Locality data were collected concurrently using radio-telemetry and GPS data loggers. GPS data demonstrated that desert tortoises frequently exhibited temporary emigration outside a plot during the survey periods, thereby complicating standard approaches for closed-model density estimation. We integrated mark–recapture survey data for subadults and adults at each plot with corresponding spatial capture locations and supplementary spatial data using a modified SCR model fitted in a Bayesian framework. We compared density estimates modeled with conventional non-spatial methods, as well as three SCR models based on symmetrical usage areas described by various levels and types of supplementary spatial data. The conventional model consistently resulted in inflated estimates of density while the SCR models allowed us to generate spatially corrected estimates for a species where detectability and densities are low. However, we found that if not properly specified, the temporal scale of supplementary data may result in an unintended source of bias in parameter estimates. Integrating spatial data over a larger temporal scale than mark–recapture surveys were conducted resulted in higher detection probabilities and lower density estimates, due to an overestimation of space use. Our results not only demonstrate the importance of accounting for spatial information but also the value of understanding the potential for bias when integrating multiple data sets at different temporal resolutions. The methods presented can be used to enhance monitoring efforts for the Mojave desert tortoise and other species where mark–recapture methods are used.

California, Nevada↗
USGS2023

The incredible HALK: borrowing data for age assignment

Understanding age and growth are important for fisheries science and management; however, age data are not routinely collected for many populations. We propose and test a method of borrowing age–length data across increasingly broader spatiotemporal levels to create a hierarchical age–length key (HALK). We assessed this method by comparing growth and mortality metrics to those estimated from lake–year age–length keys ages using seven common freshwater fish species across the upper Midwestern United States. Levels used for data borrowing began most specifically by borrowing within lake across time and increased in breadth to include data within the Hydrologic Unit Code (HUC) 10 watershed, HUC8 watershed, Level III Ecoregion, and finally a species-wide data ALK using all available data with our study for a species. Median deviation in mean length of age-3 fish was within 1 cm for the most specific HALK levels, and median deviation in total annual mortality was close to 0 for most species when borrowing occurred within HUC10 and HUC8 watersheds. Percent error in growth curves increased with data borrowing, but plateaued—or even decreased—for some species when data borrowing expanded across spatial levels. We present the HALK as a method for gaining age information about a fishery when age data are unavailable.

Illinois, Indiana, Iowa, Michigan, Minnesota, Sout↗
USGS2018

Looking beyond wildlife: Using remote cameras to evaluate accuracy of gridded snow data

The use of remote cameras is widespread in wildlife ecology, yet few examples exist of their utility for collecting environmental data. We used a novel camera trap method to evaluate the accuracy of gridded snow data in a mountainous region of the northeastern US. We were specifically interested in assessing (1) how snow depth observations from remote cameras compare with gridded climate data, (2) the sources of error associated with the gridded data and (3) the influence of spatial sampling on bias. We compared daily observations recorded by remote cameras with Snow Data Assimilation System (SNODAS ) gridded predictions using data from three winters (2014–2016). Snow depth observations were correlated with SNODAS predictions for sites ( R 2 = 0.20) and regions ( R 2 = 0.16), yet we detected factors associated with SNODAS bias at both scales. Specifically, SNODAS underpredicted depths at high elevations, at sites with higher solar radiation, and within conifer‐dominated forest. Depths were most underpredicted at highest elevations, up to 44 and 26 cm on average at the site and region scales, respectively. Bias was greatest when predictions were lowest, occasionally predicting snow absence when depths were >100 cm at camera sites. We also detected breakdowns in accuracy when certain environmental conditions varied within the 1 km 2 SNODAS grid cells. For example, underprediction was greatest when the solar radiation values of camera stations increased relative to the mean of the SNODAS grid cells. This relationship was most prominent in mountainous regions, suggesting that factors which influence solar radiation (e.g. topographic complexity) contribute to SNODAS inaccuracy. We caution using gridded snow data for ecological studies when bias is unknown. We suggest increased sampling to adjust for errors associated with gridded data products that arise from factors, such as forest cover and topographic variability. Increasing resolution and accuracy of climate data will improve predictions of species’ responses to climate change.

New Hampshire, Vermont↗
USGS2018

Enabling global exchange of groundwater data: GroundWaterML2 (GWML2)

GWML2 is an international standard for the online exchange of groundwater data that addresses the problem of data heterogeneity. This problem makes groundwater data hard to find and use because the data are diversely structured and fragmented into numerous data silos. Overcoming data heterogeneity requires a common data format; however, until the development of GWML2, an appropriate international standard has been lacking. GWML2 represents key hydrogeological entities such as aquifers and water wells, as well as related measurements and groundwater flows. It is developed and tested by an international consortium of groundwater data providers from North America, Europe, and Australasia, and facilitates many forms of data exchange, information representation, and the development of online web portals and tools.

Hydrogeology Journal↗
USGS2019

Using maintenance records from a long-term sensor monitoring network to evaluate the relationship between maintenance schedule and data quality

Sensor-based environmental monitoring networks are beginning to provide the large-scale, long-term data required to address important fundamental and applied questions in ecology. However, the data quality from deployed sensors can be difficult and costly to ensure. In this study, we use maintenance records from the 12-year history of Louisiana’s Coastwide Reference Monitoring System (CRMS) to assess the relationship between various dimensions of data quality and the frequency of field visits to the sensors. We use hierarchical Bayesian models to estimate the probability of missing data, the probability that a corrective offset of the sensor is required, and the magnitude of required offsets for water elevation and salinity data. We compared these estimates to predetermined risk thresholds to the help identify maintenance schedules that balanced the efficient use of labor resources without sacrificing data quality. We found that the relationship between data quality and increasing maintenance interval varied across metrics. Additionally, for most metrics, the maintenance interval when the metric’s credible interval and risk threshold intersected varied throughout the year and with wetland type. These results suggest that complex maintenance schedules, in which field visits vary in frequency throughout the year and with environmental context, are likely to provide the best tradeoff between labor cost and data quality. This analysis demonstrates that quantitative assessment of maintenance records can positively impact the sustainability of long-term data collection projects by helping identify new potential efficiencies in monitoring program management.

Louisiana↗
USGS2021

Implications of aggregating and smoothing daily production data on estimates of the transition time between flow regimes in horizontal hydraulically fractured Bakken oil wells

The level to which data are aggregated or smoothed can impact analytical and predictive modeling results. This paper discusses findings regarding such impacts on estimating change points in production flow regimes of horizontal hydraulically fractured shale oil wells producing from the middle member of the Bakken Formation. Change points that signal transitions in flow regimes are important because they subsequently affect estimates of ultimate recovery from wells producing from shale plays. Extending our earlier work, we employ two different statistical approaches, Bacon–Watts Bayesian regression and nonlinear constrained least squares regression, and a designed computational experiment to estimate the time of transition from the transient to the boundary-dominated flow regime for 14 different wells using daily production data rather than aggregated monthly data, as previously considered. The daily data were also smoothed to reduce noise. Computational experiments suggest that both statistical approaches can lead to plausible estimates of the transition point under different data aggregation or smoothing regimes, but that daily data are likely too granular to produce credible estimates. Although the expected value of transition points using smoothed daily data and monthly disaggregated data are generally comparable, the confidence intervals bounding the estimates based on smoothed daily data are generally wider. Our results not only inform the operational practices of oil producers engaged in economic evaluation of their shale resources and additional play development activities, but also the activities of petroleum research groups, government agencies, and financial organizations seeking to improve the trustworthiness of resource projections.

Mathematical Geosciences↗
USGS1983

Synthesis of geophysical data with space-acquired imagery: a review

Geophysical data obtained from ground and airborne platforms have been used in the development of regional geologic models for many years. Space-acquired data and imagery have a shorter but similar history of applications. All these data may be synthesized either manually or digitally. Manual synthesis methods consist of overlaying and comparing maps, whereas digital synthesis methods consist of computer storage and analysis of registered digital data sets. A data base may include topographic, geologic, soils, aeromagnetic, gravity, radiometric, electromagnetic and geochemical data, and Landsat, Seasat, and Heat Capacity Mapping Mission (HCMM) images, all of which can be evaluated individually or compared in multiple layers (overlays). Stereographic models, useful in the correlation and interpretation of geophysical data, have been created from Landsat images by using aeromagnetic, gravity, geochemical, or topographic values to offset Landsat pixels, thus introducing parallax and permitting stereoscopic viewing. Statistical correlation has been used to determine the applicability of specific data sets to the development of geologic or exploration models. Various arithmetic functions have proven useful in developing models from such data sets.

Advances in Space Research↗
USGS2016

Global biodiversity monitoring: from data sources to essential biodiversity variables

Essential Biodiversity Variables (EBVs) consolidate information from varied biodiversity observation sources. Here we demonstrate the links between data sources, EBVs and indicators and discuss how different sources of biodiversity observations can be harnessed to inform EBVs. We classify sources of primary observations into four types: extensive and intensive monitoring schemes, ecological field studies and satellite remote sensing. We characterize their geographic, taxonomic and temporal coverage. Ecological field studies and intensive monitoring schemes inform a wide range of EBVs, but the former tend to deliver short-term data, while the geographic coverage of the latter is limited. In contrast, extensive monitoring schemes mostly inform the population abundance EBV, but deliver long-term data across an extensive network of sites. Satellite remote sensing is particularly suited to providing information on ecosystem function and structure EBVs. Biases behind data sources may affect the representativeness of global biodiversity datasets. To improve them, researchers must assess data sources and then develop strategies to compensate for identified gaps. We draw on the population abundance dataset informing the Living Planet Index (LPI) to illustrate the effects of data sources on EBV representativeness. We find that long-term monitoring schemes informing the LPI are still scarce outside of Europe and North America and that ecological field studies play a key role in covering that gap. Achieving representative EBV datasets will depend both on the ability to integrate available data, through data harmonization and modeling efforts, and on the establishment of new monitoring programs to address critical data gaps.

Biological Conservation↗
USGS2024

A probabilistic approach to training machine learning models using noisy data

Machine learning (ML) models are increasingly popular in environmental and hydrologic modeling, but they typically contain uncertainties resulting from noisy data (erroneous or outlier data). This paper presents a novel probabilistic approach that combines ML and Markov Chain Monte Carlo simulation to (1) detect and underweight likely noisy data, (2) develop an approach capable of detecting noisy data during model deployment, and (3) interpret the reasons why a data point is deemed noisy to help heuristically distinguish between outliers and erroneous data. The new algorithm recognizes that there is no unique way to split the training data into noisy and clean data, and thus produces an ensemble of plausible splits. The algorithm successfully detected noisy data in synthetic benchmark problems with varying complexity and a real-world public supply water withdrawal dataset. The algorithm is generic and flexible, making it suitable for application across a broad range of hydrologic and environmental disciplines.

Environmental Modelling & Software↗