Search USGS⌕ Search

SEARCH · Search USGS

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 757 records · Page 42Linked to original sources
USGS2016

Bayesian estimation of magma supply, storage, and eruption rates using a multiphysical volcano model: Kīlauea Volcano, 2000–2012

Estimating rates of magma supply to the world's volcanoes remains one of the most fundamental aims of volcanology. Yet, supply rates can be difficult to estimate even at well-monitored volcanoes, in part because observations are noisy and are usually considered independently rather than as part of a holistic system. In this work we demonstrate a technique for probabilistically estimating time-variable rates of magma supply to a volcano through probabilistic constraint on storage and eruption rates. This approach utilizes Bayesian joint inversion of diverse datasets using predictions from a multiphysical volcano model, and independent prior information derived from previous geophysical, geochemical, and geological studies. The solution to the inverse problem takes the form of a probability density function which takes into account uncertainties in observations and prior information, and which we sample using a Markov chain Monte Carlo algorithm. Applying the technique to Kīlauea Volcano, we develop a model which relates magma flow rates with deformation of the volcano's surface, sulfur dioxide emission rates, lava flow field volumes, and composition of the volcano's basaltic magma. This model accounts for effects and processes mostly neglected in previous supply rate estimates at Kīlauea, including magma compressibility, loss of sulfur to the hydrothermal system, and potential magma storage in the volcano's deep rift zones. We jointly invert data and prior information to estimate rates of supply, storage, and eruption during three recent quasi-steady-state periods at the volcano. Results shed new light on the time-variability of magma supply to Kīlauea, which we find to have increased by 35–100% between 2001 and 2006 (from 0.11–0.17 to 0.18–0.28 km 3 /yr), before subsequently decreasing to 0.08–0.12 km 3 /yr by 2012. Changes in supply rate directly impact hazard at the volcano, and were largely responsible for an increase in eruption rate of 60–150% between 2001 and 2006, and subsequent decline by as much as 60% by 2012. We also demonstrate the occurrence of temporal changes in the proportion of Kīlauea's magma supply that is stored versus erupted, with the supply “surge” in 2006 associated with increased accumulation of magma at the summit. Finally, we are able to place some constraints on sulfur concentrations in Kīlauea magma and the scrubbing of sulfur by the volcano's hydrothermal system. Multiphysical, Bayesian constraint on magma flow rates may be used to monitor evolving volcanic hazard not just at Kīlauea but at other volcanoes around the world.

Hawai'i↗
USGS2022

Using active source seismology to image the Palos Verdes Fault damage zone as a function of distance, depth, and geology

Fault damage zones provide a window into the non-elastic processes of an earthquake. Geological and seismic tomography methods have been unable to measure damage zones at depth with sufficient spatial sampling to evaluate the relative influence of depth, distance, and lithological variations. Here, we identify and analyze the damage zone of the Palos Verdes Fault offshore southern California using two 3D seismic reflection datasets. We apply a novel algorithm to identify discontinuities attributed to faults and fractures in large seismic volumes and examine the spatial distribution of fault damage in sedimentary rock surrounding the Palos Verdes Fault. Our results show that damage through fracturing is most concentrated around mapped faults and decays exponentially to a distance of ∼2 km, where fracturing reaches a clearly defined and relatively undamaged background for all examined depths and lithologies (450 m to 2.2 km). This decrease in fracturing with distance from the central fault strand exhibits similar functional form to outcrop studies. However, here we extend analysis to distances seldom accessible (∼10 km lateral distance). Separating the data by geologic units we find that the damage decay and background level differs for each unit, with the older and deeper units having higher levels of background fracturing and shallower exponential decays of fracturing with distance from the fault. Surprisingly, these differences in damage decay and background level trade-off result in a consistent damage zone width regardless of lithology or depth. We find that the damage zone has similar decay trends on both sides of the fault. When examining the damage zone at shorter (4 km vs 17 km) along strike distances, the damage zone has a more complex decay trend and at least two strands are resolvable.

California↗
USGS2023

Snag dynamics and surface fuel loads in the Sierra Nevada: Predicting the impact of the 2012–2016 drought

Forest die-backs linked to extreme droughts are expected to increase as the climate dries and warms. An example is the 2012-2016 hotter drought in California that induced widespread tree mortality in the Sierra Nevada, California. The sudden increase in snags (i.e., standing dead trees) raised immediate concerns about their impact on wildfire hazard and longer-term questions about their impact on ecosystem structure and function. We quantified the likely progression of snag fall and fuel succession following the recent extensive mortality event in the southern Sierra Nevada mixed conifer forest. Our results used data from a long-term demography study to project trends in surface fuel loads at three study sites in Yosemite and Sequoia Kings Canyon National Parks. In the short term (2017-2021), fine woody debris and litter + duff significantly increased across all three sites (>145% and >55%, respectively); coarse woody debris increased significantly at one site (48.6%); and total fuel loads increased significantly at two of the three sites (38% and 69%). Snag longevity increased with size, with the relationship varying by species. Yellow pine was a notable outlier: size played a small role in influencing its fall rates. Overall, species-specific snag fall rates in the southern Sierra Nevada were 20% to 40% slower than previously reported. By 2040, projected median cumulative inputs of biomass from future snag fall range from 49.4 Mg ha-1 to 136.1 Mg ha-1across our three sites, which exceeds the amounts currently present (47.17-89.97 Mg ha-1) and is well above estimates of historical coarse woody debris amounts in the Sierra Nevada (17.7 Mg ha -1). These results provide a robust empirical basis to refine the snag fall algorithm in vegetation simulation models. Options to manage the impact of extreme number of snags and their large surface combustible biomass include salvage operations and prescribed burning, with both methods having operational, financial, and legal limitations that need to be considered.

Forest Ecology and Management↗
USGS2015

Machine learning for predicting soil classes in three semi-arid landscapes

Mapping the spatial distribution of soil taxonomic classes is important for informing soil use and management decisions. Digital soil mapping (DSM) can quantitatively predict the spatial distribution of soil taxonomic classes. Key components of DSM are the method and the set of environmental covariates used to predict soil classes. Machine learning is a general term for a broad set of statistical modeling techniques. Many different machine learning models have been applied in the literature and there are different approaches for selecting covariates for DSM. However, there is little guidance as to which, if any, machine learning model and covariate set might be optimal for predicting soil classes across different landscapes. Our objective was to compare multiple machine learning models and covariate sets for predicting soil taxonomic classes at three geographically distinct areas in the semi-arid western United States of America (southern New Mexico, southwestern Utah, and northeastern Wyoming). All three areas were the focus of digital soil mapping studies. Sampling sites at each study area were selected using conditioned Latin hypercube sampling (cLHS). We compared models that had been used in other DSM studies, including clustering algorithms, discriminant analysis, multinomial logistic regression, neural networks, tree based methods, and support vector machine classifiers. Tested machine learning models were divided into three groups based on model complexity: simple, moderate, and complex. We also compared environmental covariates derived from digital elevation models and Landsat imagery that were divided into three different sets: 1) covariates selected a priori by soil scientists familiar with each area and used as input into cLHS, 2) the covariates in set 1 plus 113 additional covariates, and 3) covariates selected using recursive feature elimination. Overall, complex models were consistently more accurate than simple or moderately complex models. Random forests (RF) using covariates selected via recursive feature elimination was consistently the most accurate, or was among the most accurate, classifiers between study areas and between covariate sets within each study area. We recommend that for soil taxonomic class prediction, complex models and covariates selected by recursive feature elimination be used. Overall classification accuracy in each study area was largely dependent upon the number of soil taxonomic classes and the frequency distribution of pedon observations between taxonomic classes. Individual subgroup class accuracy was generally dependent upon the number of soil pedon observations in each taxonomic class. The number of soil classes is related to the inherent variability of a given area. The imbalance of soil pedon observations between classes is likely related to cLHS. Imbalanced frequency distributions of soil pedon observations between classes must be addressed to improve model accuracy. Solutions include increasing the number of soil pedon observations in classes with few observations or decreasing the number of classes. Spatial predictions using the most accurate models generally agree with expected soil–landscape relationships. Spatial prediction uncertainty was lowest in areas of relatively low relief for each study area.

New Mexico, Utah, Wyoming↗
USGS2016

POLARIS: A 30-meter probabilistic soil series map of the contiguous United States

A new complete map of soil series probabilities has been produced for the contiguous United States at a 30 m spatial resolution. This innovative database, named POLARIS, is constructed using available high-resolution geospatial environmental data and a state-of-the-art machine learning algorithm (DSMART-HPC) to remap the Soil Survey Geographic (SSURGO) database. This 9 billion grid cell database is possible using available high performance computing resources. POLARIS provides a spatially continuous, internally consistent, quantitative prediction of soil series. It offers potential solutions to the primary weaknesses in SSURGO: 1) unmapped areas are gap-filled using survey data from the surrounding regions, 2) the artificial discontinuities at political boundaries are removed, and 3) the use of high resolution environmental covariate data leads to a spatial disaggregation of the coarse polygons. The geospatial environmental covariates that have the largest role in assembling POLARIS over the contiguous United States (CONUS) are fine-scale (30 m) elevation data and coarse-scale (~ 2 km) estimates of the geographic distribution of uranium, thorium, and potassium. A preliminary validation of POLARIS using the NRCS National Soil Information System (NASIS) database shows variable performance over CONUS. In general, the best performance is obtained at grid cells where DSMART-HPC is most able to reduce the chance of misclassification. The important role of environmental covariates in limiting prediction uncertainty suggests including additional covariates is pivotal to improving POLARIS' accuracy. This database has the potential to improve the modeling of biogeochemical, water, and energy cycles in environmental models; enhance availability of data for precision agriculture; and assist hydrologic monitoring and forecasting to ensure food and water security.

Geoderma↗
USGS2016

Spectrally based mapping of riverbed composition

Remote sensing methods provide an efficient means of characterizing fluvial systems. This study evaluated the potential to map riverbed composition based on in situ and/or remote measurements of reflectance. Field spectra and substrate photos from the Snake River, Wyoming, USA, were used to identify different sediment facies and degrees of algal development and to quantify their optical characteristics. We hypothesized that accounting for the effects of depth and water column attenuation to isolate the reflectance of the streambed would enhance distinctions among bottom types and facilitate substrate classification. A bottom reflectance retrieval algorithm adapted from coastal research yielded realistic spectra for the 450 to 700 nm range; but bottom reflectance-based substrate classifications, generated using a random forest technique, were no more accurate than classifications derived from above-water field spectra. Additional hypothesis testing indicated that a combination of reflectance magnitude (brightness) and indices of spectral shape provided the most accurate riverbed classifications. Convolving field spectra to the response functions of a multispectral satellite and a hyperspectral imaging system did not reduce classification accuracies, implying that high spectral resolution was not essential. Supervised classifications of algal density produced from hyperspectral data and an inferred bottom reflectance image were not highly accurate, but unsupervised classification of the bottom reflectance image revealed distinct spectrally based clusters, suggesting that such an image could provide additional river information. We attribute the failure of bottom reflectance retrieval to yield more reliable substrate maps to a latent correlation between depth and bottom type. Accounting for the effects of depth might have eliminated a key distinction among substrates and thus reduced discriminatory power. Although further, more systematic study across a broader range of fluvial environments is needed to substantiate our initial results, this case study suggests that bed composition in shallow, clear-flowing rivers potentially could be mapped remotely.

Geomorphology↗
USGS2025

Favorability mapping for hydrothermal power resource assessments of the Great Basin, USA

The U.S. Geological Survey (USGS) is updating the 2008 assessment of conventional hydrothermal resources for the Great Basin in the western United States. As part of this work, the workflow for hydrothermal resource favorability maps is being modified to integrate modern data-driven machine learning (ML) methods. Improvements include: [1] using new and refined evidence layers (features); [2] using an order of magnitude more training sites (labeled examples); [3] utilizing simple but non-linear supervised ML algorithms; [4] representing positive training sites (wells with measured heat flow) with their ordinal value proportional to the magnitude of convective upflow (i.e., low, high, or very high convective signals instead of past strategies using positive-negative labels); [5] supplementing training sites with additional sites with low convective signals to represent diverse under-sampled areas where hydrothermal systems are unlikely to exist; [6] comparing with competing approaches; and [7] utilizing Monte Carlo cross-validation to estimate and evaluate prediction uncertainty. For the new favorability map, over half of the power-producing systems (i.e., 15 of 28) are predicted in the 99th percentile of most favorable locations (i.e., the highest 1 % of favorability, corresponding to 1 % of the map area), exceeding the performance of past models that have explicitly used power plants as training sites. Previous favorability maps predicted approximately half of the power-producing hydrothermal systems above the 80th percentile (i.e., 20 % of the map area). For the new favorability map, 93 % of power-producing systems (i.e., 26 of 28) are above the 80th percentile. The power-producing systems for which the new model does not perform well are either comparatively small, low-temperature systems or systems also not predicted well by prior modeling approaches, suggesting that these few systems are unusual when compared with most power-producing systems. Focusing research on these known, seemingly different systems may yield new insights and subsequent discovery of new prospects.

California, Idaho, Nevada, Oregon, Utah↗
USGS2024

Hyperspectral (VNIR-SWIR) analysis of roll front uranium host rocks and industrial minerals from Karnes and Live Oak Counties, Texas Coastal Plain

VNIR-SWIR (400–2500 nm) reflectance measurements were made on the surfaces of various cores, cuttings and sample splits of sedimentary rocks from the Tertiary Jackson Group, and Catahoula, Oakville and Goliad Formations. These rocks vary in composition and texture from mudstone and claystone to sandstone and are known host rocks for roll front uranium occurrences in Karnes and Live Oak Counties, Texas. Spectral reflectance profiles, 569 in total, were reduced to 125 representative spectral signatures, which were analyzed using the U.S. Geological Survey's (USGS) Material Identification and Characterization Algorithm (MICA). MICA uses an automated continuum-removal procedure together with a least-squares linear regression to determine the fit of observed sample spectral absorption features to those of reference mineral standards in a spectral library. The reference minerals include various clay, mica, carbonate, ferric and ferrous iron minerals and their mixtures. In addition, absorption feature band-depth analysis was done to identify rock surfaces exhibiting absorption features related to uranium and zeolite minerals, which were not included in the command files used to execute MICA. Rocks from each of the four geologic units produced broadly similar spectral signatures as a result of comparable mineral compositions, but there were some notable differences. For example, Ca- and Na-montmorillonite was matched most frequently to the spectral absorption features in 2-μm (∼2000–2500 nm) wavelengths, while goethite occurred often at 1-μm (∼400–1000 nm) wavelengths. The latter is related to limonitic iron-staining in and around oxidized zones of the uranium roll front as described in previous papers. Rocks of the Jackson Group differed from those of the Catahoula, Oakville and Goliad units in that the former exhibited spectral features we interpret as being due to the presence of lignite-bearing mudstone layers. Goliad rocks exhibit spectral features related to dolomite, gypsum, anhydrite, and an unidentified green clay mineral that is possibly glauconite . Jackson Group rocks also exhibit weak but well-resolved absorption features at 964 and 1157 nm related to either or both zeolite minerals clinoptilolite and heulandite. These zeolite minerals and a few spectra exhibiting hydrous silica absorption features are indicative of alteration of volcanic glass in tuffaceous mudstone and claystone layers. A few sample spectra exhibited strong absorption features at around 1135 nm related to the uranium mineral coffinite. Both the 1135 nm coffinite and 1157 nm zeolite absorption features overlap somewhat, potentially making them difficult to distinguish without additional hyperspectral field, laboratory or remote sensing data. The results of this study were compared to mixtures of minerals described for ore, gangue and alteration minerals in deposit models for sandstone-hosted uranium, sedimentary bentonite and sedimentary zeolite. Use of these spectra can help facilitate mapping of both waste materials from the legacy mining of the above commodities, as well as future exploration and resource assessment activities.

Texas↗
USGS2024

Use of Doppler velocity radars to monitor and predict debris and flood wave velocities and travel times in post-wildfire basins

The magnitude and timing of extreme events such as debris and floodflows (collectively referred to as floodflows) in post-wildfire basins are difficult to measure and are even more difficult to predict. To address this challenge, a sensor ensemble consisting of noncontact, ground-based (near-field), Doppler velocity (velocity) and pulsed (stage or gage height) radars, rain gages, and a redundant radio communication network was leveraged to monitor flood wave velocities, to validate travel times, and to compliment observations from NEXRAD weather radar. The sensor ensemble (DEbris and Floodflow Early warNing System, DEFENS) was deployed in Waldo Canyon, Pike National Forest, Colorado, USA, which was burned entirely (100 percent burned) by the Waldo Canyon fire during the summer of 2012 ( MTBS, 2020 ). Surface velocity, stage, and precipitation time series collected during the DEFENS deployment on 10 August 2015 were used to monitor and predict flood wave velocities and travel times as a function of stream discharge (discharge; streamflow). The 10 August 2015 event exhibited spatial and temporal variations in rainfall intensity and duration that resulted in a discharge equal to 5.01 cubic meters per second (m 3 /s). Discharge was estimated post-event using a slope-conveyance indirect discharge method and was verified using velocity radars and the probability concept algorithm. Mean flood wave velocities – represented by the kinematic celerity ( c k = 2.619 m e t e r s p e r s e c o n d , m / s ± 0.556 p e r c e n t "> 𝑐 𝑘 = 2.619 𝑚𝑒𝑡𝑒𝑟𝑠 𝑝𝑒𝑟 𝑠𝑒𝑐𝑜𝑛𝑑, m/s ± 0.556 𝑝𝑒𝑟𝑐𝑒𝑛𝑡) and dynamic celerity ( c d = 3.533 m / s ± 0.181 p e r c e n t a n d t h e i r u n c e r t a i n t i e s "> 𝑐 𝑑 = 3.533 m/s ± 0.181 𝑝𝑒𝑟𝑐𝑒𝑛𝑡) and their uncertainties were computed. L-moments were computed to establish probability density functions (PDFs) and associated statistics for each of the at-a-section hydraulic parameters to serve as a workflow for implementing alert networks in hydrologically similar basins that lack data. Measured flood wave velocities and travel times agreed well with predicted values. Absolute percent differences between predicted and measured flood wave velocities ranged from 1.6 percent to 49 percent and varied with water slope, hydraulic radius, and depth. The kinematic celerity was a better predictor for steep slopes and wide flood plains associated with the Upper Waldo and Middle Waldo radar streamgages; whereas, the dynamic celerity was a better surrogate for shallow slopes and incised channels such as the Lower Waldo radar streamgage. The method demonstrates the potential extensibility of a post-wildfire warning system by (1) leveraging multiple systems (i.e., weather radar, near-field velocity and stage radars, and rain gages) for accurate and timely warnings of debris and floodflows, (2) establishing an order of operations to site, install, and operate near-field radars and conventional rain gages to record floodflows, forecast travel times, and document geomorphic change in this basin and hydrologically similar basins that lack data, and (3) communicating data operationally with the Colorado Department of Transportation engineering staff, National Weather Service forecasters, and emergency managers.

Colorado↗
USGS2007

Saturn's icy satellites investigated by Cassini-VIMS. I. Full-disk properties: 350-5100 nm reflectance spectra and phase curves

Saturn's icy satellites are among the main scientific objectives of the Cassini-VIMS (Visual and Infrared Mapping Spectrometer) experiment. This paper contains a first systematic and comparative analysis of the full-disk spectral properties of Dione, Enceladus, Epimetheus, Hyperion, Iapetus, Mimas, Phoebe, Rhea and Tethys as observed by VIMS from July 2004 to June 2005. The disk integrated properties (350-5100 nm reflectance spectra and phase curves at 550-2232 nm) and images of satellites are reported and discussed in detail together with the observed geometry. In general, the spectra in the visible spectral range are almost featureless and can be classified according to the spectral slopes: from the bluish Enceladus and Phoebe to the redder Iapetus, Hyperion and Epimetheus. In the 1000-1300 nm range the spectra of Enceladus, Tethys, Mimas and Rhea are characterized by a negative slope, consistent with a surface largely dominated by water ice, while the spectra of Iapetus, Hyperion and Phoebe show a considerable reddening pointing out the relevant role played by darkening materials present on the surface. In between these two classes are Dione and Epimetheus, which have a flat spectrum in this range. The main absorption bands identified in the infrared are the 1520, 2020, 3000 nm H2O/OH bands (for all satellites), although Iapetus dark terrains show mostly a deep 3000 nm band while the 1520 and 2020 nm bands are very faint. In this spectral range, the Iapetus spectrum is characterized by a strong reddening. The CO2 band at 4260 nm and the Fresnel ice peak around 3100 nm are evident only on Hyperion, Phoebe and Iapetus. The phase curves at 550 and at 2232 nm are reported for all the available observations in the 0??-144?? range; Rhea shows an opposition surge at visible wavelengths in the 0.5??-1.17?? interval. The improvement on the retrieval of the full-disk reflectance spectra can be appreciated by a direct comparison with ground-based telescopic data available from literature. Finally, data processing strategies and recent upgrades introduced in the VIMS-V calibration pipeline (flat-field and destriping-despiking algorithm) are discussed in appendices. ?? 2006 Elsevier Inc. All rights reserved.

Icarus↗
USGS2015

Enumeration of Mars years and seasons since the beginning of telescopic exploration

A clarification for the enumeration of Mars Years prior to 1955 is presented, along with a table providing the Julian dates associated with Ls = 0° for Mars Years -183 (beginning of the telescopic study of Mars) to 100. A practical algorithm for computing Ls as a function of the Julian Date is provided. No new science results are presented

Icarus↗
USGS2018

A remote sensing-based model of tidal marsh aboveground carbon stocks for the conterminous United States

Remote sensing based maps of tidal marshes, both of their extents and carbon stocks, have the potential to play a key role in conducting greenhouse gas inventories and implementing climate mitigation policies. Our objective was to generate a single remote sensing model of tidal marsh aboveground biomass and carbon that represents nationally diverse tidal marshes within the conterminous United States (CONUS). We developed the first calibration-grade, national-scale dataset of aboveground tidal marsh biomass, species composition, and aboveground plant carbon content (%C) from six CONUS regions: Cape Cod, MA, Chesapeake Bay, MD, Everglades, FL, Mississippi Delta, LA, San Francisco Bay, CA, and Puget Sound, WA. Using the random forest machine learning algorithm, we tested whether imagery from multiple sensors, Sentinel-1 C-band synthetic aperture radar, Landsat, and the National Agriculture Imagery Program (NAIP), can improve model performance. The final model, driven by six Landsat vegetation indices and with the soil adjusted vegetation index as the most important (n = 409, RMSE = 310 g/m 2 , 10.3% normalized RMSE), successfully predicted biomass for a range of marsh plant functional types defined by height, leaf angle and growth form. Model results were improved by scaling field-measured biomass calibration data by NAIP-derived 30 m fraction green vegetation. With a mean plant carbon content of 44.1% (n = 1384, 95% C.I. = 43.99%–44.37%), we generated regional 30 m aboveground carbon density maps for estuarine and palustrine emergent tidal marshes as indicated by a modified NOAA Coastal Change Analysis Program map. We applied a multivariate delta method to calculate uncertainties in regional carbon densities and stocks that considered standard error in map area, mean biomass and mean %C. Louisiana palustrine emergent marshes had the highest C density (2.67 ± 0.004 Mg/ha) of all regions, while San Francisco Bay brackish/saline marshes had the highest C density of all estuarine emergent marshes (2.03 ± 0.004 Mg/ha). Estimated C stocks for predefined jurisdictional areas ranged from 1023 ± 39 Mg in the Nisqually National Wildlife Refuge in Washington to 507,761 ± 14,822 Mg in the Terrebonne and St. Mary Parishes in Louisiana. This modeling and data synthesis effort will allow for aboveground C stocks in tidal marshes to be included in the coastal wetland section of the U.S. National Greenhouse Gas Inventory. With the increased availability of free post-processed satellite data, we provide a tractable means of modeling tidal marsh aboveground biomass and carbon at the global extent as well.

ISPRS Journal of Photogrammetry and Remote Sensing↗
USGS2009

A global map of rainfed cropland areas (GMRCA) at the end of last millennium using remote sensing

The overarching goal of this study was to produce a global map of rainfed cropland areas (GMRCA) and calculate country-by-country rainfed area statistics using remote sensing data. A suite of spatial datasets, methods and protocols for mapping GMRCA were described. These consist of: (a) data fusion and composition of multi-resolution time-series mega-file data-cube (MFDC), (b) image segmentation based on precipitation, temperature, and elevation zones, (c) spectral correlation similarity (SCS), (d) protocols for class identification and labeling through uses of SCS R 2 -values, bi-spectral plots, space-time spiral curves (ST-SCs), rich source of field-plot data, and zoom-in-views of Google Earth (GE), and (e) techniques for resolving mixed classes by decision tree algorithms, and spatial modeling. The outcome was a 9-class GMRCA from which country-by-country rainfed area statistics were computed for the end of the last millennium. The global rainfed cropland area estimate from the GMRCA 9-class map was 1.13 billion hectares (Bha). The total global cropland areas (rainfed plus irrigated) was 1.53 Bha which was close to national statistics compiled by FAOSTAT (1.51 Bha). The accuracies and errors of GMRCA were assessed using field-plot and Google Earth data points. The accuracy varied between 92 and 98% with kappa value of about 0.76, errors of omission of 2-8%, and the errors of commission of 19-36%. ?? 2008 Elsevier B.V.

International Journal of Applied Earth Observation↗
USGS2011

Monitoring of urban subsidence with SAR interferometric point target analysis: A case study in Suzhou, China

Ground subsidence, mainly caused by over exploitation of groundwater and other underground resources, such as oil, gas and coal, occurs in many cities in China. The annual direct loss associated with subsidence across the country is estimated to exceed 100 million US dollar. Interferometric SAR (InSAR) is a powerful tool to map ground deformation at an unprecedented level of spatial detail. It has been widely used to investigate the deformation resulting from earthquakes, volcanoes and subsidence. Repeat-pass InSAR, however, may fail due to impacts of spatial decorrelation, temporal decorrelation and heterogeneous refractivity of atmosphere. In urban areas, a large amount of natural stable radar reflectors exists, such as buildings and engineering structures, at which radar signals can remain coherent during a long time interval. Interferometric point target analysis (IPTA) technique, also known as persistent scatterers (PS) InSAR is based on these reflectors. It overcomes the shortfalls in conventional InSAR. This paper presents a procedure for urban subsidence monitoring with IPTA. Calculation of linear deformation rate and height residual, and the non-linear deformation estimate, respectively, are discussed in detail. Especially, the former is highlighted by a novel and easily implemented 2-dimensional spatial search algorithm. Practically useful solutions that can significantly improve the robustness of IPTA, are recommended. Finally, the proposed procedure is applied to mapping the ground subsidence in Suzhou city, Jiangsu province, China. Thirty-four ERS-1/2 SAR scenes are analyzed, and the deformation information over 38,881 point targets between 1992 and 2000 are generated. The IPTA-derived deformation estimates correspond well with leveling measurements, demonstrating the potential of the proposed subsidence monitoring procedure based on IPTA technique. Two shortcomings of the IPTA-based procedure, e.g., the requirement of large number of SAR images and assumed linear plus non-linear deformation model, are discussed as the topics of further research.

Jaingsu Province↗
USGS2013

Hierarchical Bayesian spatial models for predicting multiple forest variables using waveform LiDAR, hyperspectral imagery, and large inventory datasets

In this paper we detail a multivariate spatial regression model that couples LiDAR, hyperspectral and forest inventory data to predict forest outcome variables at a high spatial resolution. The proposed model is used to analyze forest inventory data collected on the US Forest Service Penobscot Experimental Forest (PEF), ME, USA. In addition to helping meet the regression model's assumptions, results from the PEF analysis suggest that the addition of multivariate spatial random effects improves model fit and predictive ability, compared with two commonly applied modeling approaches. This improvement results from explicitly modeling the covariation among forest outcome variables and spatial dependence among observations through the random effects. Direct application of such multivariate models to even moderately large datasets is often computationally infeasible because of cubic order matrix algorithms involved in estimation. We apply a spatial dimension reduction technique to help overcome this computational hurdle without sacrificing richness in modeling.

Maine↗
USGS2014

Performance and effects of land cover type on synthetic surface reflectance data and NDVI estimates for assessment and monitoring of semi-arid rangeland

Federal land management agencies provide stewardship over much of the rangelands in the arid andsemi-arid western United States, but they often lack data of the proper spatiotemporal resolution andextent needed to assess range conditions and monitor trends. Recent advances in the blending of com-plementary, remotely sensed data could provide public lands managers with the needed information.We applied the Spatial and Temporal Adaptive Reflectance Fusion Model (STARFM) to five Landsat TMand concurrent Terra MODIS scenes, and used pixel-based regression and difference image analyses toevaluate the quality of synthetic reflectance and NDVI products associated with semi-arid rangeland. Pre-dicted red reflectance data consistently demonstrated higher accuracy, less bias, and stronger correlationwith observed data than did analogous near-infrared (NIR) data. The accuracy of both bands tended todecline as the lag between base and prediction dates increased; however, mean absolute errors (MAE)were typically ≤10%. The quality of area-wide NDVI estimates was less consistent than either spectra lband, although the MAE of estimates predicted using early season base pairs were ≤10% throughout the growing season. Correlation between known and predicted NDVI values and agreement with the 1:1regression line tended to decline as the prediction lag increased. Further analyses of NDVI predictions,based on a 22 June base pair and stratified by land cover/land use (LCLU), revealed accurate estimates through the growing season; however, inter-class performance varied. This work demonstrates the successful application of the STARFM algorithm to semi-arid rangeland; however, we encourage evaluation of STARFM’s performance on a per product basis, stratified by LCLU, with attention given to the influence of base pair selection and the impact of the time lag.

Idaho, Utah, Wyoming↗
USGS2019

Mapping cropland extent of Southeast and Northeast Asia using multi-year time-series Landsat 30-m data using Random Forest classifier on Google Earth Engine

Cropland extent maps are useful components for assessing food security. Ideally, such products are a useful addition to countrywide agricultural statistics since they are not politically biased and can be used to calculate cropland area for any spatial unit from an individual farm to various administrative unites (e.g., state, county, district) within and across nations, which in turn can be used to estimate agricultural productivity as well as degree of disturbance on food security from natural disasters and political conflict. However, existing cropland extent maps over large areas (e.g., Country, region, continent, world) are derived from coarse resolution imagery (250 m to 1 km pixels) and have many limitations such as missing fragmented and\or small farms with mixed signatures from different crop types and\or farming practices that can be, confused with other land cover. As a result, the coarse resolution maps have limited useflness in areas where fields are small (<1 ha), such as in Southeast Asia. Furthermore, coarse resolution cropland maps have known uncertainties in both geo-precision of cropland location as well as accuracies of the product. To overcome these limitations, this research was conducted using multi-date, multi-year 30-m Landsat time-series data for 3 years chosen from 2013 to 2016 for all Southeast and Northeast Asian Countries (SNACs), which included 7 refined agro-ecological zones (RAEZ) and 12 countries (Indonesia, Thailand, Myanmar, Vietnam, Malaysia, Philippines, Cambodia, Japan, North Korea, Laos, South Korea, and Brunei). The 30-m (1 pixel = 0.09 ha) data from Landsat 8 Operational Land Imager (OLI) and Landsat 7 Enhanced Thematic Mapper (ETM+) were used in the study. Ten Landsat bands were used in the analysis (blue, green, red, NIR, SWIR1, SWIR2, Thermal, NDVI, NDWI, LSWI) along with additional layers of standard deviation of these 10 bands across 1 year, and global digital elevation model (GDEM)-derived slope and elevation bands. To reduce the impact of clouds, the Landsat imagery was time-composited over four time-periods (Period 1: January- April, Period 2: May-August, and Period 3: September-December) over 3-years. Period 4 was the standard deviation of all 10 bands taken over all images acquired during the 2015 calendar year. These four period composites, totaling 42 band data-cube, were generated for each of the 7 RAEZs. The reference training data (N = 7849) generated for the 7 RAEZ using sub-meter to 5-m very high spatial resolution imagery (VHRI) helped generate the knowledge-base to separate croplands from non-croplands. This knowledge-base was used to code and run a pixel-based random forest (RF) supervised machine learning algorithm on the Google Earth Engine (GEE) cloud computing environment to separate croplands from non-croplands. The resulting cropland extent products were evaluated using an independent reference validation dataset (N = 1750) in each of the 7 RAEZs as well as for the entire SNAC area. For the entire SNAC area, the overall accuracy was 88.1% with a producer’s accuracy of 81.6% (errors of omissions = 18.4%) and user’s accuracy of 76.7% (errors of commissions = 23.3%). For each of the 7 RAEZs overall accuracies varied from 83.2 to 96.4%. Cropland areas calculated for the 12 countries were compared with country areas reported by the United Nations Food and Agriculture Organization and other national cropland statistics resulting in an R 2 value of 0.93. The cropland areas of provinces were compared with the province statistics that showed an R 2 = 0.95 for South Korea and R 2 = 0.94 for Thailand. The cropland products are made available on an interactive viewer at www.croplands.org and for download at National Aeronautics and Space Administration’s (NASA) Land Processes Distributed Active Archive Center (LP DAAC): https://lpdaac.usgs.gov/node/1281 .

International Journal of Applied Earth Observation↗
USGS2020

Surface soil temperature seasonal variation estimation in a forested area using combined satellite observations and in-situ measurements

Surface soil temperature is the soil temperature from the surface to 10 cm in depth. Surface soil temperature plays a significant role in agricultural drought monitoring, ecosystem energy transfer modeling, and global carbon cycle evaluation. Studies have been proposed to estimate surface soil temperature, but surface soil temperature monitoring within forested areas still poses a significant challenge. In this study, we proposed a surface soil temperature retrieval method using combined satellite observations and in-situ measurements for the Great Dismal Swamp (GDS). The GDS is a U.S. protected area managed and protected by the U.S. Fish and Wildlife Service. It is located along the boundary of Virginia and North Carolina, with maple gum, Atlantic white cedar, and pine pocosin as the main forest cover types. Ground-based surface soil temperature measurements were collected for these forest types from May 2015 to April 2017. Both the Land Remote Sensing Satellite (Landsat) Thermal Infrared Sensor (TIRS) and the Moderate Resolution Imaging Spectroradiometer (MODIS) carry two thermal infrared (TIR) channels. The TIR channels with similar corresponding wavelengths were first fused using an improved fusing model to generate high resolution TIR measurements. Then the enterprise algorithm was applied to calculate land surface temperature (LST) from the fused TIR bands. An improved soil temperature retrieval method was applied to generate surface soil temperature based on LST and vegetation index (VI) within the study area for the three forest types. In-situ measurements were used to build the surface soil temperature retrieval method, and results were then validated. The normalized difference vegetation index (NDVI) and enhanced vegetation index (EVI) were integrated separately as VIs in the model to monitor surface soil temperature. The R 2 for retrieved surface soil temperature through satellite observations was 0.76, and the RMSE was 1.96 &#x2103; ">℃ ℃ when NDVI was integrated in the model; the R 2 was 0.78, and the RMSE was 1.85 &#x2103; ">℃ ℃ when EVI was used.

North Carolina, Virginia↗