Search USGS⌕ Search

SEARCH · Search USGS

Results for “Precision Agriculture”

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.

123 records · Page 7Linked to original sources
USGS2018

A 30-m landsat-derived cropland extent product of Australia and China using random forest machine learning algorithm on Google Earth Engine cloud computing platform

Mapping high resolution (30-m or better) cropland extent over very large areas such as continents or large countries or regions accurately, precisely, repeatedly, and rapidly is of great importance for addressing the global food and water security challenges. Such cropland extent products capture individual farm fields, small or large, and are crucial for developing accurate higher-level cropland products such as cropping intensities, crop types, crop watering methods (irrigated or rainfed), crop productivity, and crop water productivity. It also brings many challenges that include handling massively large data volumes, computing power, and collecting resource intensive reference training and validation data over complex geographic and political boundaries. Thereby, this study developed a precise and accurate Landsat 30-m derived cropland extent product for two very important, distinct, diverse, and large countries: Australia and China. The study used of eight bands (blue, green, red, NIR, SWIR1, SWIR2, TIR1, and NDVI) of Landsat-8 every 16-day Operational Land Imager (OLI) data for the years 2013–2015. The classification was performed by using a pixel-based supervised random forest (RF) machine learning algorithm (MLA) executed on the Google Earth Engine (GEE) cloud computing platform. Each band was time-composited over 4–6 time-periods over a year using median value for various agro-ecological zones (AEZs) of Australia and China. This resulted in a 32–48-layer mega-file data-cube (MFDC) for each of the AEZs. Reference training and validation data were gathered from: (a) field visits, (b) sub-meter to 5-m very high spatial resolution imagery (VHRI) data, and (c) ancillary sources such as from the National agriculture bureaus. Croplands versus non-croplands knowledge base for training the RF algorithm were derived from MFDC using 958 reference-training samples for Australia and 2130 reference-training samples for China. The resulting 30-m cropland extent product was assessed for accuracies using independent validation samples: 900 for Australia and 1972 for China. The 30-m cropland extent product of Australia showed an overall accuracy of 97.6% with a producer’s accuracy of 98.8% (errors of omissions = 1.2%), and user’s accuracy of 79% (errors of commissions = 21%) for the cropland class. For China, overall accuracies were 94% with a producer’s accuracy of 80% (errors of omissions = 20%), and user’s accuracy of 84.2% (errors of commissions = 15.8%) for cropland class. Total cropland areas of Australia were estimated as 35.1 million hectares and 165.2 million hectares for China. These estimates were higher by 8.6% for Australia and 3.9% for China when compared with the traditionally derived national statistics. The cropland extent product further demonstrated the ability to estimate sub-national cropland areas accurately by providing an R 2 value of 0.85 when compared with province-wise cropland areas of China. The study provides a paradigm-shift on how cropland maps are produced using multi-date remote sensing. These products can be browsed at www.croplands.org and made available for download at NASA’s Land Processes Distributed Active Archive Center (LP DAAC) https://www.lpdaac.usgs.gov/node/1282 .

ISPRS Journal of Photogrammetry and Remote Sensing↗
USGS2011

Rapid assessment of rice seed availability for wildlife in harvested fields

Rice seed remaining in commercial fields after harvest (waste rice) is a critical food resource for wintering waterfowl in rice-growing regions of North America. Accurate and precise estimates of the seed mass density of waste rice are essential for planning waterfowl wintering habitat extents and management. In the Sacramento Valley of California, USA, the existing method for obtaining estimates of availability of waste rice in harvested fields produces relatively precise estimates, but the labor-, time-, and machineryintensive process is not practical for routine assessments needed to examine long-term trends in waste rice availability. We tested several experimental methods designed to rapidly derive estimates that would not be burdened with disadvantages of the existing method. We first conducted a simulation study of the efficiency of each method and then conducted field tests. For each approach, methods did not vary in root mean squared error, although some methods did exhibit bias for both simulations and field tests. Methods also varied substantially in the time to conduct each sample and in the number of samples required to detect a standard trend. Overall, modified line-intercept methods performed well for estimating the density of rice seeds. Waste rice in the straw, although not measured directly, can be accounted for by a positive relationship with density of rice on the ground. Rapid assessment of food availability is a useful tool to help waterfowl managers establish and implement wetland restoration and agricultural habitat-enhancement goals for wintering waterfowl. ?? 2011 The Wildlife Society.

Wildlife Society Bulletin↗
USGS2008

Reducing sedimentation of depressional wetlands in agricultural landscapes

Depressional wetlands in agricultural landscapes are easily degraded by sediments and contaminants accumulated from their watersheds. Several best management practices can reduce transport of sediments into wetlands, including the establishment of vegetative buffers. We summarize the sources, transport dynamics, and effect of sediments, nutrients, and contaminants that threaten wetlands and the current knowledge of design and usefulness of grass buffers for protecting isolated wetlands. Buffer effectiveness is dependent on several factors, including vegetation structure, buffer width, attributes of the surrounding watershed (i.e., area, vegetative cover, slope and topography, soil type and structure, soil moisture, amount of herbicides and pesticides applied), and intensity and duration of rain events. To reduce dissolved contaminants from runoff, the water must infiltrate the soil where microbes or other processes can break down or sequester contaminants. But increasing infiltration also diminishes total water volume entering a wetland, which presents threats to wetland hydrology in semi-arid regions. Buffer effectiveness may be enhanced significantly by implementing other best management practices (e.g., conservation tillage, balancing input with nutrient requirements for livestock and crops, precision application of chemicals) in the surrounding watershed to diminish soil erosion and associated contaminant runoff. Buffers require regular maintenance to remove sediment build-up and replace damaged or over-mature vegetation. Further research is needed to establish guidelines for effective buffer width and structure, and such efforts should entail a coordinated, regional, multi-scale, multidisciplinary approach to evaluate buffer effectiveness and impacts. Direct measures in "real-world" systems and field validations of buffer-effectiveness models are crucial next steps in evaluating how grass buffers will impact the abiotic and biotic variables attributes that characterize small, isolated wetlands. ?? 2008 The Society of Wetland Scientists.

Wetlands↗
USGS2011

Use of airborne hyperspectral imagery to map soil parameters in tilled agricultural fields

Soil hyperspectral reflectance imagery was obtained for six tilled (soil) agricultural fields using an airborne imaging spectrometer (400–2450 nm, ~10 nm resolution, 2.5 m spatial resolution). Surface soil samples ( n = 315) were analyzed for carbon content, particle size distribution, and 15 agronomically important elements (Mehlich-III extraction). When partial least squares (PLS) regression of imagery-derived reflectance spectra was used to predict analyte concentrations, 13 of the 19 analytes were predicted with R 2 > 0.50, including carbon (0.65), aluminum (0.76), iron (0.75), and silt content (0.79). Comparison of 15 spectral math preprocessing treatments showed that a simple first derivative worked well for nearly all analytes. The resulting PLS factors were exported as a vector of coefficients and used to calculate predicted maps of soil properties for each field. Image smoothing with a 3 × 3 low-pass filter prior to spectral data extraction improved prediction accuracy. The resulting raster maps showed variation associated with topographic factors, indicating the effect of soil redistribution and moisture regime on in-field spatial variability. High-resolution maps of soil analyte concentrations can be used to improve precision environmental management of farmlands.

Applied and Environmental Soil Science↗
USGS2012

Evaluating prediction uncertainty of areas contributing recharge to well fields of multiple water suppliers in the Hunt-Annaquatucket-Pettaquamscutt River Basins, Rhode Island

Three river basins in central Rhode Island-the Hunt River, the Annaquatucket River, and the Pettaquamscutt River-contain 15 production wells clustered in 4 pumping centers from which drinking water is withdrawn. These high-capacity production wells, operated by three water suppliers, are screened in coarse-grained deposits of glacial origin. The risk of contaminating water withdrawn by these well centers may be reduced if the areas contributing recharge to the well centers are delineated and these areas protected from land uses that may affect the water quality. The U.S. Geological Survey, in cooperation with the Rhode Island Department of Health, began an investigation in 2009 to improve the understanding of groundwater flow and delineate areas contributing recharge to the well centers as part of an effort to protect the source of water to these well centers. A groundwater-flow model was calibrated by inverse modeling using nonlinear regression to obtain the optimal set of parameter values, which provide a single, best representation of the area contributing recharge to a well center. Summary statistics from the calibrated model were used to evaluate the uncertainty associated with the predicted areas contributing recharge to the well centers. This uncertainty analysis was done so that the contributing areas to the well centers would not be underestimated, thereby leaving the well centers inadequately protected. The analysis led to contributing areas expressed as a probability distribution (probabilistic contributing areas) that differ from a single or deterministic contributing area. Groundwater flow was simulated in the surficial deposits and the underlying bedrock in the 47-square-mile study area. Observations (165 groundwater levels and 7 base flows) provided sufficient information to estimate parameters representing recharge and horizontal hydraulic conductivity of the glacial deposits and hydraulic conductance of streambeds. The calibrated value for recharge to valley-fill deposits was 27.3 inches per year (in/yr) and to upland till deposits was 18.7 in/yr. Calibrated values for horizontal hydraulic conductivity of the valley-fill deposits ranged from 20 to 480 feet per day (ft/d) and of the upland till deposits was 16.2 ft/d. Calibrated values of streambed hydraulic conductance ranged from 10,000 to 52,000 feet squared per day. Values of recharge and horizontal hydraulic conductivity of the valley-fill deposits were the most precisely estimated, whereas the horizontal hydraulic conductivity of till deposits was the least precisely estimated. Simulated areas contributing recharge to the well centers on the basis of the calibrated model ranged from 0.19 to 1.12 square miles (mi 2 ) and covered a total area of 2.79 mi 2 for average well center withdrawal rates during 2004-08 (235 to 1,858 gallons per minute (gal/min)). Simulated areas contributing recharge for the maximum well center pumping capacities (800 to 8,500 gal/min) ranged from 0.37 to 3.53 mi2 and covered a total area of 7.99 mi2 in the modeled area. Simulated areas contributing recharge extend upgradient of the well centers to upland till and to groundwater divides. Some areas contributing recharge include small, isolated areas remote from the well centers. Relatively short groundwater traveltimes from recharging locations to discharging wells indicated the wells are vulnerable to contamination from land-surface activities: median traveltimes ranged from 2.9 to 5.0 years for the well centers, and 78 to 93 percent of the traveltimes were 10 years or less for the well centers. Land cover in the areas contributing recharge includes a substantial amount of urban land use for the two well centers in the Hunt River Basin, agriculture and sand and gravel mining uses for the well center in the Annaquatucket River Basin, and, for the well center in the Pettaquamscutt River Basin, land use is primarily undeveloped. Model-prediction uncertainty was evaluated using a Monte Carlo analysis. The parameter variance-covariance matrix from nonlinear regression was used to create parameter sets that reflect the uncertainty of the parameter estimates and the correlation among parameters. The remaining parameters representing the glacial deposits (vertical anisotropy of valley-fill deposits and of till deposits, maximum groundwater evapotranspiration, and hydraulic conductance for headdependent cells representing a groundwater divide) that could not be estimated with nonlinear regression were incorporated into the variance-covariance matrix using prior information on parameters. Thus the uncertainty analysis was an outcome of calibrating the parameters to available observations and to information that the modeler provided. A water budget and model-fit statistical criteria were used to assess parameter sets so that prediction uncertainty was not overestimated. Because of the effects of parameter uncertainty, the size of the probabilistic contributing areas for each well center for both average and maximum pumping rates was larger than the size of the deterministic contributing areas for the well center. Thus, some areas not in the deterministic contributing area may actually be in the contributing area, including additional areas of urban and agricultural land use. Generally, areas closest to the well centers with short groundwater traveltimes are associated with higher probabilities, whereas areas distant from the well centers with long groundwater traveltimes are associated with lower probabilities. The deterministic contributing areas generally corresponded to areas associated with high probabilities (greater than 50 percent). Areas associated with low probabilities extended long distances along groundwater divides in the uplands remote from the well centers.

Rhode Island↗
USGS2026

Estimating submersed aquatic vegetation biomass using a hierarchical Bayesian model of ordinal measurements and structural zeros

This study explores the prediction of a continuous target variable that is primarily measured as an ordinal outcome. Interest in such prediction may occur given data from double- or two-phase sampling designs, where an auxiliary variable is measured at all sampling units and a target and predictor variable, X , at a subset of those units. This study focuses on the case where X is nonnegative, is conditional on a zero process, and is sampled using a cluster sampling design. We model the ordinal outcome using cumulative logistic regression and the conditional regressor X , following log transformation, in a Bayesian setting. Estimation of model parameters and prediction of missing X is generally accurate and precise, particularly with sufficiently large Pr(X > 0) and associated variances. This study is motivated by a need to accurately estimate the biomass of submersed aquatic vegetation on the Upper Mississippi River, where two types of measurements for biomass are available: inexpensive, ordinal biomass scores and expensive, continuous diver-harvested biomass data. Here, X represents plant biomass and Pr(X > 0) is the usual species site occupancy measure. When fit using biomass data from the submerged aquatic vegetation species Vallisneria americana Michx, our model estimates parameters within the parameter space region where the model performed acceptably using synthetic data. We expect this model to appeal to investigators with ordered outcomes, cluster designs, and one or more continuous, positive predictors.

Journal of Agricultural, Biological and Environmen↗
USGS2024

Investigation of land cover within wetland complexes at Dixie Meadows, Churchill County, Nevada, from October 2015 to January 2022

The U.S. Geological Survey investigated land cover at subannual time steps within six wetland areas in Dixie Valley, Churchill County, Nevada, from October 2015 to January 2022. As requested by the U.S. Fish and Wildlife Service, we used aerial photography and satellite remote sensing data to map surface water and other land cover types within the wetland complexes. We identified five land cover classes using the green normalized difference vegetation index (gNDVI) and its inverse relationship to the normalized difference water index (NDWI) within three U.S. Department of Agriculture National Agriculture Imagery Program aerial images (acquired in 2015, 2017, and 2019) and 110 European Space Agency Sentinel-2 satellite images (acquired 2015–2022). The relative wetness of soil conditions within each land cover class is estimated by comparison to previously published observations of relative conductivity measured by 79 field-based sensors within the wetlands from 2019 to 2021. We mapped the areal coverage of the five land cover classes for approximately 385 acres (1,559,000 square meters [m²]) comprising six individual wetland complexes as well as a larger 1,298- acre (5,254,000-m 2 ) area of interest inclusive of the wetland complexes and adjacent landscape. Land cover of open water (Class 5) primarily within ponds at one of the wetland complexes comprised 8,333 m 2 , on average, of the wetland complexes. Land cover of mixed shallow surface water, saturated soil, and vegetation (Class 4) comprised 111,723 m 2 on average of the wetland complexes. Land cover of dense green vegetation canopy cover (Class 3) that often (46 percent of observations) had underlying surface water or saturated soil conditions comprised 592,522 m 2 on average of the wetland complexes. The remaining areas of the wetland complexes not mapped as these three land cover types (Classes 2 and 1) had sparse vegetation or bare soil cover and commonly (greater than or equal to 67 percent of observations) had dry soil conditions. The investigation of land cover detailed in this report could inform future efforts to map land cover more precisely via higher resolution remote sensing or ground-based surveying or could be incorporated with other environmental monitoring data to characterize habitat and hydrology of the wetland complexes at Dixie Meadows.

Nevada↗
USGS2018

Estimating metal concentrations with regression analysis and water-quality surrogates at nine sites on the Animas and San Juan Rivers, Colorado, New Mexico, and Utah

The purpose of this report is to evaluate the use of site-specific regression models to estimate metal concentrations at nine U.S. Geological Survey streamflow-gaging stations on the Animas and San Juan Rivers in Colorado, New Mexico, and Utah. Downstream users could use these regression models to determine if metal concentrations are elevated and pose a risk to water supplies, agriculture, and recreation. Multiple linear-regression models were developed by relating metal concentrations in discrete water-quality samples to continuously monitored streamflow and surrogate parameters (specific conductance, pH, turbidity, and water temperature) collected at the U.S. Geological Survey stations. Models were developed for dissolved and total concentrations of aluminum, arsenic, cadmium, copper, iron, lead, manganese, and zinc using water-quality samples collected from 2005 to 2017 by several Federal, State, Tribal, and local agencies using different collection methods and analytical laboratories. Model performance varied but, in general, models for dissolved metals did not perform as well as those for total metals. Dissolved metals generally were correlated to specific conductance or streamflow and total metals generally were better correlated with turbidity. Explanatory variables in the models reflected hydrologic and geochemical processes within the basin. A larger number of regression models were statistically significant for the most upstream sites, where metal concentrations were elevated by drainage from abandoned mines and mineralized bedrock. Models generally did not perform as well at downstream sites, especially for dissolved metals, which occurred at lower concentrations than at the upstream sites. In the lower reaches of the rivers, the input of more alkaline water from tributaries and groundwater reduced metal solubility and diluted metal concentrations. The number and distribution of samples in the calibration datasets also may have been a factor in model development. At some sites on the San Juan River, calibration datasets were more limited and did not represent the full range of observed hydrologic and water-quality conditions, especially during storm events in summer and fall. Recommendations for model use are given based on estimates of model precision, biases, and adequacy of the calibration datasets in terms of the number of samples and representativeness of the observed range of streamflow and water-quality conditions.

Colorado, New Mexico, Utah↗
USGS2001

Estimation of population size using open capture-recapture models

One of the most important needs for wildlife managers is an accurate estimate of population size. Yet, for many species, including most marine species and large mammals, accurate and precise estimation of numbers is one of the most difficult of all research challenges. Open-population capture-recapture models have proven useful in many situations to estimate survival probabilities but typically have not been used to estimate population size. We show that open-population models can be used to estimate population size by developing a Horvitz-Thompson-type estimate of population size and an estimator of its variance. Our population size estimate keys on the probability of capture at each trap occasion and therefore is quite general and can be made a function of external covariates measured during the study. Here we define the estimator and investigate its bias, variance, and variance estimator via computer simulation. Computer simulations make extensive use of real data taken from a study of polar bears (Ursus maritimus) in the Beaufort Sea. The population size estimator is shown to be useful because it was negligibly biased in all situations studied. The variance estimator is shown to be useful in all situations, but caution is warranted in cases of extreme capture heterogeneity.

Journal of Agricultural, Biological, and Environme↗
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↗
USGS2009

Simulation of Groundwater-Level and Salinity Changes in the Eastern Shore, Virginia

Groundwater-level and salinity changes have been simulated with a groundwater model developed and calibrated for the Eastern Shore of Virginia. The Eastern Shore is the southern part of the Delmarva Peninsula that is occupied by Accomack and Northampton Counties in Virginia. Groundwater is the sole source of freshwater to the Eastern Shore, and demands for water have been increasing from domestic, industrial, agricultural, and public-supply sectors of the economy. Thus, it is important that the groundwater supply be protected from overextraction and seawater intrusion. The best way for water managers to use all of the information available is usually to compile this information into a numerical model that can simulate the response of the system to current and future stresses. A detailed description of the geology, hydrogeology, and historical groundwater extractions was compiled and entered into the numerical model. The hydrogeologic framework is composed of a surficial aquifer under unconfined conditions, a set of three aquifers and associated overlying confining units under confined conditions (the upper, middle, and lower Yorktown-Eastover Formation), and an underlying confining unit (the St. Marys Formation). An estimate of the location and depths of two major paleochannels was also included in the framework of the model. Total withdrawals from industrial, commercial, public-supply, and some agricultural wells were compiled from the period 1900 through 2003. Reported pumpage from these sources increased dramatically during the 1960s and 70s, up to currently about 4 million gallons per day. Domestic withdrawals were estimated on the basis of population census districts and were assigned spatially to the model on the assumption that domestic users are located close to roads. A numerical model was created using the U.S. Geological Survey (USGS) code SEAWAT to simulate both water levels and concentrations of chloride (representing salinity). The model was calibrated using 605 predevelopment and transient water-level observations that are associated predominantly with 20 observation nests of wells sited across the study area. Sampling for groundwater chemistry at these sites revealed that chloride has not increased significantly in the last 20 years. Environmental tracers in the samples also indicated that the water in the surficial aquifer is typically years to decades old, whereas water in the confined aquifers is typically centuries to millennia old. The calibration procedure yielded distributions of hydraulic conductivity and storage coefficients of the aquifers and confining units that are based on 21 pilot points, but vary smoothly across the study area. The estimated values are consistent with other measurements of these properties measured previously on cores and during hydraulic tests at various well fields. Simulations performed with the model demonstrated that the calibrated model can reproduce the observed historical water levels fairly well (R2 = 0.93). The chloride concentrations were also simulated, but a match with chloride concentrations was more difficult to achieve (R2 = 0.16) because of the lack of sufficient data and the unknown exact behavior of the entire transition zone in the millennia leading up to the present day. Future pumping scenarios were simulated through 2050, with pumping set to either 2003 rates or total permitted withdrawal rates. Water levels in 2050 are predicted to be lower than current levels by a few feet where stresses are currently heaviest but potentially by tens of feet if total permitted withdrawals are extracted at current low-stressed sites. Simulations of chloride concentrations through 2050 revealed some potential for seawater intrusion in the areas of Cape Charles, Chincoteague, east of the town of Exmore, and east of the town of Accomac, but precise estimates of concentration increases are highly uncertain. Simulation results were also used to estimate that the down

Scientific Investigations Report↗
USGS2012

Uncertainty quantification for environmental models

Environmental models are used to evaluate the fate of fertilizers in agricultural settings (including soil denitrification), the degradation of hydrocarbons at spill sites, and water supply for people and ecosystems in small to large basins and cities—to mention but a few applications of these models. They also play a role in understanding and diagnosing potential environmental impacts of global climate change. The models are typically mildly to extremely nonlinear. The persistent demand for enhanced dynamics and resolution to improve model realism [17] means that lengthy individual model execution times will remain common, notwithstanding continued enhancements in computer power. In addition, high-dimensional parameter spaces are often defined, which increases the number of model runs required to quantify uncertainty [2]. Some environmental modeling projects have access to extensive funding and computational resources; many do not. The many recent studies of uncertainty quantification in environmental model predictions have focused on uncertainties related to data error and sparsity of data, expert judgment expressed mathematically through prior information, poorly known parameter values, and model structure (see, for example, [1,7,9,10,13,18]). Approaches for quantifying uncertainty include frequentist (potentially with prior information [7,9]), Bayesian [13,18,19], and likelihood-based. A few of the numerous methods, including some sensitivity and inverse methods with consequences for understanding and quantifying uncertainty, are as follows: Bayesian hierarchical modeling and Bayesian model averaging; single-objective optimization with error-based weighting [7] and multi-objective optimization [3]; methods based on local derivatives [2,7,10]; screening methods like OAT (one at a time) and the method of Morris [14]; FAST (Fourier amplitude sensitivity testing) [14]; the Sobol' method [14]; randomized maximum likelihood [10]; Markov chain Monte Carlo (MCMC) [10]. There are also bootstrapping and cross-validation approaches.Sometimes analyses are conducted using surrogate models [12]. The availability of so many options can be confusing. Categorizing methods based on fundamental questions assists in communicating the essential results of uncertainty analyses to stakeholders. Such questions can focus on model adequacy (e.g., How well does the model reproduce observed system characteristics and dynamics?) and sensitivity analysis (e.g., What parameters can be estimated with available data? What observations are important to parameters and predictions? What parameters are important to predictions?), as well as on the uncertainty quantification (e.g., How accurate and precise are the predictions?). The methods can also be classified by the number of model runs required: few (10s to 1000s) or many (10,000s to 1,000,000s). Of the methods listed above, the most computationally frugal are generally those based on local derivatives; MCMC methods tend to be among the most computationally demanding. Surrogate models (emulators)do not necessarily produce computational frugality because many runs of the full model are generally needed to create a meaningful surrogate model. With this categorization, we can, in general, address all the fundamental questions mentioned above using either computationally frugal or demanding methods. Model development and analysis can thus be conducted consistently using either computation-ally frugal or demanding methods; alternatively, different fundamental questions can be addressed using methods that require different levels of effort. Based on this perspective, we pose the question: Can computationally frugal methods be useful companions to computationally demanding meth-ods? The reliability of computationally frugal methods generally depends on the model being reasonably linear, which usually means smooth nonlin-earities and the assumption of Gaussian errors; both tend to be more valid with more linear

SIAM News↗
USGS2021

Global cropland-extent product at 30-m resolution (GCEP30) derived from Landsat satellite time-series data for the year 2015 using multiple machine-learning algorithms on Google Earth Engine cloud

Executive Summary Global food and water security analysis and management require precise and accurate global cropland-extent maps. Existing maps have limitations, in that they are (1) mapped using coarse-resolution remote-sensing data, resulting in the lack of precise mapping location of croplands and their accuracies; (2) derived by collecting and collating national statistical data that are often subjective, leading to substantial uncertainties in cropland-area estimates, as well as their locations; and (3) extracted from one or more classes of a land use–land cover product in which cropland classes are not the focus of mapping, leading to their mixing with other classes and creating significant errors of omission and commission. These limitations can be overcome by producing high-resolution cropland-extent maps using satellite-sensor data, such as Landsat 30-m resolution or higher. The most fundamental cropland product is the high-resolution cropland-extent map because all higher level cropland products, such as crop-watering method (that is, whether crops are irrigated or rainfed), crop types, cropping intensities, cropland fallows, crop productivity, and crop-water productivity, are dependent on a precise and accurate cropland-extent product. Given these realities, the overarching goal of this study was to produce a Landsat satellite-derived global cropland-extent product at 30-m resolution. The work, which involved a paradigm shift in how global cropland-extent maps are produced, involved the following five key steps: (1) petabyte-scale computing that involved multiyear, 8- to 16-day, time-series Landsat 30-m resolution data for the global land surface; (2) composition of analysis-ready data (ARD) cubes; (3) creation of a large global-reference data hub for machine learning; (4) use of multiple machine-learning algorithms (MLAs) by writing software and computing in the cloud; and (5) Google Earth Engine (GEE) cloud computing. The five key steps involved nine distinct phases. First, the world was segmented into 74 agroecological zones (AEZs). Second, Landsat 8- to 16-day data were used to time-composite 10-band (blue, green, red, near-infrared, short-wave infrared band 1, short-wave infrared band 2, thermal infrared, enhanced vegetation index, normalized difference water index, and normalized difference vegetation index) Landsat 30-m resolution data cubes for every 2- to 4-month time period during 3- to 4-year periods (stated as nominal-year 2015 or, simply, 2015), along with two additional 30-m resolution bands (Shuttle Radar Topography Mission elevation, and slope) in each of the 74 AEZs. Third, more than 100,000 reference-training data samples were collected using ground data (some of which were collected using a mobile application), as well as submeter- to 5-m-resolution, very high-resolution imagery sourced from other reliable sources. Fourth, reference-training data were used to create a knowledge base for separating cropland from noncropland. Fifth, MLAs such as the pixel-based supervised random forest and support-vector machines were written on the GEE using Python and JavaScript. Sixth, object-based recursive hierarchical segmentation algorithm was used, in addition to MLAs, to overcome uncertainties. Seventh, MLAs used the knowledge base to classify and separate cropland from noncropland. Eighth, accuracy assessment was conducted by generating error matrices for each of the 74 AEZs using 19,171 independent validation-data samples. Ninth, cropland areas were computed for all countries of the world and compared with United Nation’s (UN’s) Food and Agricultural Organization (FAO) and other national statistics. The outcome was a Landsat-derived global cropland-extent product at 30-m resolution (GCEP30), which has an overall accuracy of 91.7 percent. For the cropland class, producer’s accuracy was 83.4 percent, and user’s accuracy was 78.3 percent. GCEP30 calculated (using direct pixel count) the global net-cropland area (GNCA) for the year 2015 as 1.873 billion hectares (~12.6 percent of the Earth’s terrestrial area). The continental cropland distribution as a percentage of GNCA was Asia, 33 percent; Europe, 25.5 percent; Africa, 16.7 percent; North America, 14.4 percent; South America, 8.1 percent; and Australia and Oceania, 2.4 percent. The worldwide cropland areas in GCEP30 for 2015 were higher by 236 to 299 million hectares (Mha) compared to national statistics reported elsewhere for the same year (for example, in Food and Agriculture Organization’s corporate statistical database [FAOSTAT] and in the monthly irrigated and rainfed crop areas [MIRCA] database). The global cropland area reported for 2015 increased by 344 Mha (22.5 percent), compared to the year 2000. During the same period (2000–2015), the world’s population increased by 20 percent. Whereas some of these areal increases are real increases in cropland areas, others are due to the types of data, methods, and approaches used. Using the highest known resolution (compared to previous coarse-resolution global products) enabled this study to capture fragmented croplands. Coarse-resolution data compute areas on the basis of subpixels, which, for a large proportion of certain land use–land cover classes, will show only a certain percentage of the total pixel area as actual area. Subpixel areas can lead to substantial uncertainties in area computation, as determining the exact fraction of cropland areas within a coarse-resolution pixel is resource intensive and subject to errors. Other innovations in GCEP30 include reference-data hubs, machine learning, and cloud computing. Cropland areas in 214 countries, territories, departments, and regions were calculated for the year 2015 using GCEP30, on the basis of UN’s global administrative unit layers (GAUL) boundaries. The 10 leading countries in terms of cropland area (as a percentage of the GNCA) were India (9.6 percent), United States (8.95 percent), China (8.82 percent), Russia (8.32 percent), Brazil (3.42 percent), Ukraine (2.32 percent), Canada (2.29 percent), Argentina (2.05 percent), Indonesia (2 percent), and Nigeria (1.91 percent). Together, these 10 countries occupy 50 percent of the global cropland, and they have 52 percent of the global population. Their combined cropland area increased by 2 percent between 2000 and 2015, compared to the substantial increase in population of 517 million (15.5 percent). Together, India, United States, China, and Russia encompass 36 percent of the total area. In the United States and Canada, from 2000 to 2015, cropland decreased by about 2 percent, whereas their populations increased by 14 and 13 percent, respectively. The additional food requirements in these 10 countries, which are caused by increased populations, as well as increasing nutritional demands, are met by production increases in existing cropland or through virtual food trade, or both. More than 18 countries, territories, departments, or regions had 60 percent or more of their geographic area as cropland: Republic of Moldova, San Marino, and Hungary had more than 80 percent of the country’s area as cropland; Denmark, Ukraine, Ireland, and Bangladesh, 70 to 80 percent; and Uruguay, Netherlands, United Kingdom, Spain, Lithuania, Poland, Gaza Strip, Czechia, Italy, India, and Azerbaijan, 60 to 70 percent. Europe and South Asia can be considered agricultural capitals of the world, on the basis of their percentages of geographic area as cropland. United States, China, and Russia, which all have high cropland areas, are ranked second, third, and fourth in the world; India is ranked first. However, the amount of cropland as a percentage of the country’s geographic area is relatively very low for United States (18.3 percent), China (17.7 percent), and Russia (9.5 percent), whereas it is 60.5 percent for India. Most African and South American countries, territories, departments, or regions have less than 15 percent of their geographic area as cropland. China and India together house 36 percent of the world’s population; however, between 2000 and 2015, the amount of China’s cropland area fell by 18.9 percent, owing to urban expansion and the abandonment of farmlands caused by demographic changes (that is, the movement of population from villages to cities). In contrast, China’s population grew by 10 percent. The amount of India’s cropland increased by 8.5 percent, whereas its population grew by 20 percent. This study showed that, out of the 10 leading cropland countries, Ukraine, Nigeria, Russia, and Indonesia showed an 18 to 31 percent increase in cropland areas, on the basis of GCEP30 by the year 2015, compared to 2000. Nigeria’s cropland area increased by 25 percent, and its population increased by 31 percent in the same period. In these countries, food security is maintained by cropland expansion, productivity increases, and virtual food trade. Nevertheless, this trend of increasing net-cropland area and productivity will likely become difficult to maintain, owing to diminishing arable lands and plateauing of 50 years of continual yield increases, requiring policymakers to explore novel and data-supported approaches to solving future food security issues. The GCEP30 product, which can be browsed at full resolution at www.croplands.org , has been released for public download and use through U.S. Geological Survey (USGS)–National Aeronautics and Space Administration (NASA) Land Processes Distributed Active Archive Center (see https://lpdaac.usgs.gov/news/release-of-gfsad-30-meter-cropland-extent-products/ ).

Professional Paper↗
USGS2024

A comparative analysis of OpenET for evaluating evapotranspiration in California almond orchards

The almond industry in California faces water management challenges that are being exacerbated by droughts, climate change, and groundwater sustainability legislation. The Tree-crop Remote sensing of Evapotranspiration eXperiment (T-REX) aims to explore opportunities to improve precision irrigation management for woody perennial cropping systems. Almond orchards in the California Central Valley were equipped with eddy covariance flux measurements to evaluate satellite remote sensing-based evapotranspiration (RSET) models. OpenET provides high-resolution (30-m spatial and daily temporal) RSET data, synthesizing decades of research for practical water management. This study provides an evaluation of OpenET performance at six almond sites covering a large range in soils, age, and variety. It also compares OpenET ensemble evapotranspiration (ET) data with applied irrigation and precipitation records over an additional 148 almond orchards located in the Central Valley of California. Results show OpenET models, including the ensemble ET value, produced reasonable and actionable ET values, with overall coefficient of determination (R 2 ) and mean absolute error values of 0.73- and 0.95-mm d −1 at the daily time step, respectively. However, given the temporal sampling of Landsat (8-day revisit) and the interpolation methods used, the assessed ET models had difficulty in capturing short-term variability in almond ET; for example, the rapid decline in measured ET observed as a response to lack of irrigation preceding and during almond harvest. The study also drew attention to the spatial complexity in scenarios where irrigated orchards are surrounded by hot/dry areas, causing discrepancies between measured and modeled ET values. In comparison with irrigation records, OpenET ensemble ET was capable of quantifying water input (applied irrigation + precipitation) in almond orchards to within 13 % when evaluating monthly data. Initial results presented here reinforce the idea that RSET models, such as in OpenET, are powerful tools, yet their application requires nuanced understanding and careful consideration of local conditions.

California↗
USGS2025

Fluvial sediment dynamics in the Shoshone River and tributaries around Willwood Dam, Park County, Wyoming

Sedimentation affects many of the aging reservoirs in the United States. Dams and water diversions from rivers have been central elements of infrastructure supporting agricultural irrigation in the arid and semiarid regions of the Western United States for more than a century. The Willwood Irrigation District diversion dam (hereafter referred to as “Willwood Dam”) in Park County, Wyoming, is approximately 12 miles northeast of Cody, Wyo.; has a structural height of 70 feet; and impounds the Shoshone River for diversion into the Willwood Canal. Willwood Dam is part of a larger irrigation scheme supported by water storage in the much larger Buffalo Bill Dam, which is approximately 20 miles upstream. In October 2016, renovation construction activities at Willwood Dam and the Willwood Canal caused an unplanned evacuation of nearly 96,000 cubic yards of fine sediment. The fine sediment release in 2016 raised concerns that ongoing sediment management at Willwood Dam could impose limits on the long-term health of the aquatic ecosystem and fish populations. The U.S. Geological Survey, in cooperation with Wyoming Department of Environmental Quality and Willwood Work Groups 2 and 3, initiated an investigation of the dynamics of sediment transport in the Shoshone River and selected tributaries between Buffalo Bill Dam and Willwood Dam. The goal of the study was to quantify sediment transport into and out of Willwood Dam on an annual, seasonal, and event basis to better understand the relative quantities of sediment coming from natural sources and human activities on the landscape. The study ran from March 2019 through October 2021 and used observations of streamflow, turbidity, and acoustic backscatter collected at streamgages upstream and downstream from Willwood Dam to quantify suspended-sediment loads into and out of the dam during irrigation and fallow seasons, precipitation-runoff events, and deliberate sediment releases. Each tributary’s relative contribution to the sediment load upstream from Willwood Dam was examined using discrete measurements of suspended-sediment concentration and bedload during irrigation and fallow seasons, precipitation events, and stable conditions. Analysis of daily precipitation and temperature data indicated that conditions in the study area during the 2019 agricultural year were wetter and colder than period of record normal, and drier and near normal temperatures for the 2020 and 2021 agricultural years. Not all sediment load records between 2019 and 2021 are complete because of rejected observations (outliers), instrument failures or fouling, and instrument removal for calibrations. Statistical modeling of suspended-sediment concentration using paired values of turbidity and acoustic backscatter produced four models that, after refinement, had coefficients of determination indicating that more than 84 percent of the variance was explained by either turbidity or acoustic backscatter. A system of rules was developed to select the model predictions based on the seasonal operations of Willwood Dam, assumptions about the grain sizes mobilized during these operations, and assumed accuracy of the models at the downstream streamgage (Shoshone River below Willwood Dam, near Ralston, Wyo. [streamgage 06284010]) under different operational conditions. The sediment budget between upstream and downstream estimates of loads was interpreted using the mean predicted values bound by their respective model prediction intervals. When mean predicted loads of one streamgage were contained in the prediction intervals of the other streamgage, and vice-versa, difference in the sediment budget were interpreted as “indeterminate.” Modeled sediment load balances demonstrated the depositional and erosional behaviors expected from the conceptual model of dam operations whereby sediment tends to accumulate during irrigation seasons when the dam is spilling over the top, and sediment tends to evacuate during the fallow seasons when it is flowing through the sluice gates at the base of the dam. The sediment load calculations using the rules-based model criteria indicated that between 14,200 and 380,000 tons of suspended sediment moved through the Shoshone River around Willwood Dam during the irrigation seasons of 2019, 2020, and 2021; 380,000 tons of suspended sediment were transported during the cool, wet year of 2019, and 14,200 tons of suspended sediment were transported in 2020, which was relatively dry. During fallow seasons 2019, 2020, and 2021, which had fewer complete records, between 1,140 and 106,000 tons of suspended sediment was estimated to have moved through the river. For all seasons except fallow season 2022, the models estimated that more sediment was released from the dam than entered the dam, but the modeled mean loads at each streamgage were nearly always within the prediction intervals of each other, making the sediment balance indeterminant. Examination of suspended-sediment loads during irrigation seasons indicated that between 65 and 85 percent of fine sediment was transported during annual high flows and storm events, with the remainder transported during steady, lower streamflows. Examination of suspended loads during fallow seasons indicated that deliberate sediment releases through Willwood Dam accounted for between 39 and 67 percent of the total sediment moved during the fallow seasons. Deliberate sediment releases from Willwood Dam had estimated net exports of between 1,360 and 22,400 tons. Between August 2017 and July 2023, suspended-sediment concentration and bedload sediment samples were collected from 9 tributaries to the Shoshone River during 137 sampling events, including stable and precipitation-runoff conditions. During irrigation season precipitation events, the mean total sediment yields ranged from 0.33 to 9.51 tons per day per square mile; during fallow season precipitation events, the mean total yields ranged from 0.04 to 0.95 ton per day per square mile. The mean total sediment yield per unit area across all samples at each tributary site ranged from 0.26 to 3.08 tons per day per square mile. Bedload was a minor fraction of the total load, constituting a mean of 4 percent across all samples; 3 and 6 percent for events and nonevents, respectively, during irrigation season; and 3 and 1 percent for events and nonevents, respectively, during the fallow season. With the exception of one tributary, Dry Creek, these mean yield values were within the range of watershed-scale background sediment yield values estimated from reservoir surveys and previous suspended-sediment studies. Imagery from irrigation seasons 2012, 2015, 2017, 2019, and 2022 was used to determine the planimetric backwater extent of the pool area in the Shoshone River behind Willwood Dam to identify any changes in sediment storage. Active river channel widths in the Shoshone River upstream from Willwood Dam were all similar between years except 2015, which was determined to be statistically different from all other years. Bathymetric data taken in the pool behind Willwood Dam during three different surveys between November 2017 and April 2022 indicated no statistically significant differences in bed elevations between the years. Results from the planimetric and bathymetric survey data provide multiple lines of evidence indicating that sediment did not accumulate behind the dam within the error of the methods used. Examination of how precipitation affects sediment transport in the Shoshone River upstream from Willwood Dam indicated that accumulated rainfall from the natural runoff events captured during the study period varied from a trace to as much as 4.26 inches, with associated predicted suspended-sediment loads varying from 112 to 232,000 tons of suspended sediment. The behavior of the sediment loads relative to accumulated precipitation did not appear to change depending on irrigation or fallow season. A model of suspended-sediment concentrations relative to the 2-day accumulated precipitation indicated that suspended-sediment concentrations in the Shoshone River upstream from Willwood Dam increased exponentially for accumulations of 0.3 inch or more; such storms accounted for 10 percent or less of precipitation events observed during the 1981 to 2018 period of record. The gaps in records, precision of the instrumentation, and large variation in grain sizes in suspended-sediment mixtures downstream from the dam made closing the sediment budgets for most seasons unattainable. The biggest recent change in sediment storage measured using the planimetric area of deposits behind Willwood Dam took place between 2015 and 2017. The main event between these two measurements was the installation of new Willwood Canal gates in October 2016, which resulted in the large unplanned sediment release. Because the sediment budgets were nearly always indeterminate and the planimetric and bathymetric data indicated little change in the bed and bank material, it is likely that the change in sediment storage behind the dam during the study period was small relative to the precision of the statistical models and other uncertainties. This body of evidence suggests that, averaged during the 3-year study period, no major changes in storage took place, and that the current operations may be keeping storage at near-equilibrium. This condition could have been initiated because the middle sluice gate has now been operational since 2014, and the sediment release in October 2016 evacuated a large amount of legacy sediment from storage. Although the uncertainties are large, sluicing events allow for controlled releases of sediment that contributed to the near equilibrium conditions observed over an annual basis during this study.

Wyoming↗