Geology topics

E. Lynn Usery

Publications and source records attributed to E. Lynn Usery.

At least 19 recordsLinked to original sources

USGS2024

Transfer learning with convolutional neural networks for hydrological streamline delineation

Hydrological streamline delineation is critical for effective environmental management, influencing agriculture sustainability, river dynamics, watershed planning, and more. This study develops a novel approach to combining transfer learning with convolutional neural networks that capitalize on image-based pre-trained models to improve the accuracy and transferability of streamline delineation. We evaluate the performance of eleven image-based pre-trained models and a baseline model using datasets from Rowan County, North Carolina, and Covington River, Virginia in the USA. Our results demonstrate that when models are adapted to a new area, the fine-tuned ImageNet pre-trained model exhibits superior predictive accuracy, markedly higher than the models trained from scratch or those only fine-tuned on the same area. Moreover, the pre-trained model achieves better smoothness and connectivity between classified streamline channels. These findings underline the effectiveness of transfer learning in enhancing the delineation of hydrological streamlines across varied geographies, offering a scalable solution for accurate and efficient environmental modelling.

North Carolina, Virginia↗

USGS2022

Weakly supervised spatial deep learning for Earth image segmentation based on imperfect polyline labels

In recent years, deep learning has achieved tremendous success in image segmentation for computer vision applications. The performance of these models heavily relies on the availability of large-scale high-quality training labels (e.g., PASCAL VOC 2012). Unfortunately, such large-scale high-quality training data are often unavailable in many real-world spatial or spatiotemporal problems in earth science and remote sensing (e.g., mapping the nationwide river streams for water resource management). Although extensive efforts have been made to reduce the reliance on labeled data (e.g., semi-supervised or unsupervised learning, few-shot learning), the complex nature of geographic data such as spatial heterogeneity still requires sufficient training labels when transferring a pre-trained model from one region to another. On the other hand, it is often much easier to collect lower-quality training labels with imperfect alignment with earth imagery pixels (e.g., through interpreting coarse imagery by non-expert volunteers). However, directly training a deep neural network on imperfect labels with geometric annotation errors could significantly impact model performance. Existing research that overcomes imperfect training labels either focuses on errors in label class semantics or characterizes label location errors at the pixel level. These methods do not fully incorporate the geometric properties of label location errors in the vector representation. To fill the gap, this article proposes a weakly supervised learning framework to simultaneously update deep learning model parameters and infer hidden true vector label locations. Specifically, we model label location errors in the vector representation to partially reserve geometric properties (e.g., spatial contiguity within line segments). Evaluations on real-world datasets in the National Hydrography Dataset (NHD) refinement application illustrate that the proposed framework outperforms baseline methods in classification accuracy.

ACM Transactions on Intelligent Systems and Techno↗

USGS2021

GeoAI in the US Geological Survey for topographic mapping

Geospatial artificial intelligence (GeoAI) can be defined broadly as the application of artificial intelligence methods and techniques to geospatial data, processes, models, and applications. The application of these methods to topographic data and phenomena is a focus of research in the US Geological Survey (USGS). Specifically, the USGS has researched and developed applications in terrain feature extraction, hydrographic network extraction, and semantic modeling. This article is a documentation of the recent work and current state of research and development. The article helps define the accomplishments and directions of research and applications in fields of GeoAI for topographic mapping within the USGS and more broadly.

Transactions in GIS↗

USGS2021

Extensibility of U-net neural network model for hydrographic feature extraction and implications for hydrologic modeling

Accurate maps of regional surface water features are integral for advancing ecologic, atmospheric and land development studies. The only comprehensive surface water feature map of Alaska is the National Hydrography Dataset (NHD). NHD features are often digitized representations of historic topographic map blue lines and may be outdated. Here we test deep learning methods to automatically extract surface water features from airborne interferometric synthetic aperture radar (IfSAR) data to update and validate Alaska hydrographic databases. U-net artificial neural networks (ANN) and high-performance computing (HPC) are used for supervised hydrographic feature extraction within a study area comprised of 50 contiguous watersheds in Alaska. Surface water features derived from elevation through automated flow-routing and manual editing are used as training data. Model extensibility is tested with a series of 16 U-net models trained with increasing percentages of the study area, from about 3 to 35 percent. Hydrography is predicted by each of the models for all watersheds not used in training. Input raster layers are derived from digital terrain models, digital surface models, and intensity images from the IfSAR data. Results indicate about 15 percent of the study area is required to optimally train the ANN to extract hydrography when F1-scores for tested watersheds average between 66 and 68. Little benefit is gained by training beyond 15 percent of the study area. Fully connected hydrographic networks are generated for the U-net predictions using a novel approach that constrains a D-8 flow-routing approach to follow U-net predictions. This work demonstrates the ability of deep learning to derive surface water feature maps from complex terrain over a broad area.

Alaska↗

USGS2021

Spatial data reduction through element -of-interest (EOI) extraction

Any large, multifaceted data collection that is challenging to handle with traditional management practices can be branded ‘Big Data.’ Any big data containing geo-referenced attributes can be considered big geospatial data. The increased proliferation of big geospatial data is currently reforming the geospatial industry into a data-driven enterprise. Challenges in the big spatial data domain can be summarized as the ‘Big Vs’ – variety, volume, velocity, veracity and value. Big spatial data sources can be considered in two broad classes, active and passive, as each is impacted to varying degrees. Some of these challenges may be alleviated by reducing unprocessed, or minimally processed, (raw) data to features, which we refer to as the extraction of Elements of Interest (EOI). In fact, many applications require EOI extraction from raw data to enable their basic employment. This chapter presents current state-of-the-art methods to create EOI from some types of georeferenced big data. We classify the data types into two realms: active and passive. Active data are those collected specifically for the purpose to which they are applied. Passive data are those collected for purposes other than those for which they are utilized, included those ‘collected’ for no particular purpose at all. The chapter then presents use cases from both the active and passive spatial realms, including the active applications of terrain feature extraction from digital elevation models and vegetation mapping from remotely-sensed imagery and passive applications like building identification from VGI and point-of-interest data mining from social networks for land use classification. Finally, the chapter concludes with future research needs.

Book chapter↗

USGS2021

An attention U-Net model for detection of fine-scale hydrologic streamlines

Surface water is an irreplaceable resource for human survival and environmental sustainability. Accurate, finely detailed cartographic representations of hydrologic streamlines are critically important in various scientific domains, such as assessing the quantity and quality of present and future water resources, modeling climate changes, evaluating agricultural suitability, mapping flood inundation, and monitoring environmental changes. Conventional approaches to detecting such streamlines cannot adequately incorporate information from the complex three-dimensional (3D) environment of streams and land surface features. Such information is vital to accurately delineate streamlines. In recent years, high accuracy lidar data has become increasingly available for deriving both 3D information and terrestrial surface reflectance. This study develops an attention U-net model to take advantage of high-accuracy lidar data for finely detailed streamline detection and evaluates model results against a baseline of multiple traditional machine learning methods. The evaluation shows that the attention U-net model outperforms the best baseline machine learning method by an average F1 score of 11.25% and achieves significantly better smoothness and connectivity between classified streamline channels. These findings suggest that our deep learning approach can harness high-accuracy lidar data for fine-scale hydrologic streamline detection, and in turn produce desirable benefits for many scientific domains.

North Carolina↗

USGS2020

Semantically enabling map projections knowledge

Map projections are an area of cartography with a firm mathematical foundation for their creation and display providing a basis for a knowledge representation. Using only variations on a single equation set, an infinite number of projections can be created, but less than 100 are in active use. Because each projection preserves specific characteristics, such as area, angles, global look, or a compromise of properties, classifications of map projections have been developed to aid in knowledge representation. These classifications are used for decision-making. They help select the correct projection for the map use. They assist users with determining the correct orientation, standard parallels and meridians. The classifications also inform the user how to adjust the selection based on size, extent, and latitude. Semantics can be used to automate map projections knowledge into a knowledge base that can be accessed by humans and machines. This work details a semantic representation of map projections knowledge and provides a simple example of a use case that exploits the knowledge base.

Cartography and Geoinformation↗

USGS2020

Improving geospatial query performance of an interoperable geographic situation-awareness system (IGSAS) for disaster response

Disaster response operations require fast and coordinated actions based on the real-time disaster situation information. Although Volunteered Geographic Information (VGI) or crowdsourced geospatial data applications have demonstrated to be valuable tools for gathering real-time disaster situation information, they only provide limited utility for disaster response coordination because of the lack of compatibility and interoperability. VGI based on Geospatial Semantic Web (GSW) technologies has the potential to overcome the incompatibility and heterogeneity problems. However, GSW-based VGI often has poor performance due to complex geometric computation. The objective of this research is to explore how to use optimization techniques to improve performance of an interoperable geographic situation-awareness system (IGSAS) based on GSW technologies for disaster response. We conducted experiments to evaluate various client-side optimization techniques for improving performance of an IGSAS prototype for flooding disaster response in New Haven, Connecticut. Our experimental results show that the developed prototype can greatly reduce the runtime costs of geospatial semantic queries through on-the-fly spatial indexing, tile-based rendering, efficient algorithms for spatial join, and caching, especially for those spatial-join geospatial queries that involve a large number of spatial features and heavy geometric computation.

Transactions in GIS↗

USGS2020

A system design for implementing advanced feature descriptions for a map knowledge base

A prototype system to explore Linked Data that semantically integrates geospatial data in various formats from different publication sources with data from The National Map of the U.S. Geological Survey is presented. The focus is on accessing advanced feature descriptions for data from The National Map with data coreferenced from other sources. The prototype uses Geoserver to access The National Map data, which are converted to Resource Description Framework triples using Karma and stored in the Marmotta triplestore. Marmotta uses a Postgres relational database as a backend for the project and queries to the Marmotta triplestore are converted to structured query language and executed by Postgres. Triples retrieved are linked with same_as relationships to external data sources. The links to these sources provide additional attributes and relationships of the data from The National Map. Visualization of the results is provided using Leaflet and workflows for all parts of the system are defined. A use case for the system is provided to access structures and names information from The National Map for the Washington, D.C., area and link these to Geonames data, with visualization of the graphical and tabular results.

Scientific Investigations Report↗

USGS2019

Problems of Large Spatial Databases

Large spatial databases often labeled as geospatial big data exceed the capacity of commonly used computing systems as a result of data volume, variety, velocity, and veracity. Additional problems also labeled with V’s are cited, but the four primary ones are the most problematic and focus of this chapter (Li et al., 2016, Panimalar et al., 2017). Sources include satellites, aircraft and drone platforms, vehicles, geosocial networking services, mobile devices, and cameras. The problems in processing these data to extract useful information include query, analysis, and visualization. Data mining techniques and machine learning algorithms, such as deep convolutional neural networks, often are used with geospatial big data. The obvious problem is handling the large data volumes, particularly for input and output operations, requiring parallel read and write of the data, as well as high speed computers, disk services, and network transfer speeds. Additional problems of large spatial databases include the variety and heterogeneity of data requiring advanced algorithms to handle different data types and characteristics, and integration with other data. The velocity at which the data are acquired is a challenge, especially using today’s advanced sensors and the Internet of Things that includes millions of devices creating data on short temporal scales of micro seconds to minutes. Finally, the veracity, or truthfulness of large spatial databases is difficult to establish and validate, particularly for all data elements in the database.

Book chapter↗

USGS2019

Topographic mapping evolution: From field and photogrammetric data collection to GIS production and Linked Open Data

Whither the topographic map? Topographic mapping historically has been approached as a map factory operation through the period 1879-1990. During this time, data were field and photogrammetrically collected; cartographically verified and annotated creating a compilation manuscript; further edited, generalized, symbolized, and produced as a graphic output product using lithography, or more recently, through digital means. Adoption of geographic information systems (GIS) as the primary production process for topographic maps, including digital database preparation (1975-2000) and product generation operations (2001-present), has led to faster and more standardized production in a semi-automated process. However, the topographic product has remained the same static graphic. Global Navigation Systems (GNS) began in the post 1990s, led to publicly and commercially produced location-based information traditionally provided by surveyors for topographic maps. Advances in GIS technology, computer processing, memory, and storage devices, along with GNS spawned new location systems and led to ubiquitous, consumer-based cartography through commercial entities on the World Wide Web (Web). This global availability of cartography has provided consumer access and the ability to produce topographic types of map products previously supplied only by traditional National Mapping Agencies (NMAs). Information provided by location-based services made available through connected databases has led to completely new business models based on cartography and geospatial data. A new form of topographic map as an interactive, linked knowledge base is now being created. The appearance of the Semantic Web and Linked Open Data allows the map to become an interactive knowledge base. In this current theory and implementation of topographic mapping, the map is a graphics-based interface to a triplestore knowledge base which includes a topographic feature ontology, semantics and relations, and instance data with geometry and topology available. The topographic map graphic becomes an interactive link to the knowledge base and additional linked data through the Linked Open Data cloud.

Cartographic Journal↗

USGS2019

U.S. Geological Survey accomplishments in cartography 2015-2019

The U.S. Geological Survey (USGS), the United States' official national topographic mapping organization, is building and maintaining geographic databases for fundamental base geographic layers of land cover, structures, boundaries, hydrography, geographic names, transportation, elevation, and orthoimagery as The National Map. Data from the 3D Elevation Program, the National Hydrography Dataset and other national programs provide public domain, authoritative, accurate, and reliable data for The National Map, and data are served to United States government organizations and the public. Products of The National Map include viewable and downloadable data for all data layers, derivative products including US Topo, and web services of the data. The US Topo product is automatically generated from national map databases and produces topographic maps every three years for all 48 of the contiguous United States, Hawaii, and the United States territories.

Conference Paper↗

USGS2018

GNIS-LD: Serving and visualizing the Geographic Names Information System Gazetteer as linked data

In this dataset description paper we introduce the GNIS-LD, an authoritative and public domain Linked Dataset derived from the Geographic Names Information System (GNIS) which was developed by the U.S. Geological Survey (USGS) and the U.S. Board on Geographic Names. GNIS provides data about current, as well as historical, physical, and cultural geographic features in the United States. We describe the dataset, introduce an ontology for geographic feature types, and demonstrate the utility of recent linked geographic data contributions made in conjunction with the development of this resource. Co-reference resolution links to GeoNames.org and DBpedia are provided in the form of owl:sameAs relations. Finally, we point out how the adapted workflow is foundational for complex Digital Line Graph (DLG) data from the USGS National Map and how the GNIS-LD data can be integrated with DLG and other data sources such as sensor observations.

Conference Paper↗

USGS2018

The map as knowledge base

This paper examines the concept and implementation of a map as a knowledge base. A map as a knowledge base means that the visual map is not only the descriptive compilation of data and design principles, but also involves a compilation of semantic propositions and logical predicates that create a body of knowledge organized as a map. The digital product of a map as knowledge base can be interpreted by machines, as well as humans, and can provide access to the knowledge base through interfaces to select features and other information from the map. The design of maps as a knowledge base involves technical approaches and a system architecture to support a knowledge base. This paper clarifies how a map as a knowledge base differs from earlier map theory models by investigating the knowledge-based concepts of implementation through logical modelling, a knowledge repository, user interfaces for information access, and cartographic visualization. The paper ends with proof of concepts for two types of cartographic data query.

International Journal of Cartography↗

USGS2018

Understanding map projections

It has probably never been more important in the history of cartography than now that people understand how maps work. With increasing globalization, for example, world maps provide a key format for the transmission of information, but are often poorly used. Examples of poor understanding and use of projections and the resultant maps are many; for instance, the use of rectangular world maps in the United Kingdom press to show Chinese and Korean missile ranges as circles, something which can only be achieved on equidistant projections and then only from one launch point (Vujakovic, 2014).

Book chapter↗

USGS2017

A reference landform ontology for automated delineation of depression landforms from DEMs

Abstract The landform reference ontology (LFRO) is being developed to formalize ontological distinctions underlying naïve geographic cognition and reasoning about landforms. The LFRO taxonomy is currently based only on form-based distinctions. In this significantly revised version, several new categories have been added to explicate ontological distinctions related to material-spatial dependence and physical support. Nuances of common natural language landform terms and implications for their mapping are discussed.

Conference Paper↗

USGS2017

The logic of selecting an appropriate map projection in a Decision Support System (DSS)

There are undeniable practical consequences to consider when choosing an appropriate map projection for a specific region. The surface of a globe covered by global, continental, and regional maps are so singular that each type distinctively affects the amount of distortion incurred during a projection transformation because of the an assortment of effects caused by distance, direction, scale , and area. A Decision Support System (DSS) for Map Projections of Small Scale Data was previously developed to help select an appropriate projection. This paper reports on a tutorial to accompany that DSS. The DSS poses questions interactively, allowing the user to decide on the parameters, which in turn determines the logic path to a solution. The objective of including a tutorial to accompany the DSS is achieved by visually representing the path of logic that is taken to a recommended map projection derived from the parameters the user selects. The tutorial informs the DSS user about the pedigree of the projection and provides a basic explanation of the specific projection design. This information is provided by informational pop-ups and other aids.

Book chapter↗

USGS2016

A multidimensional representation model of geographic features

A multidimensional model of geographic features has been developed and implemented with data from The National Map of the U.S. Geological Survey. The model, programmed in C++ and implemented as a feature library, was tested with data from the National Hydrography Dataset demonstrating the capability to handle changes in feature attributes, such as increases in chlorine concentration in a stream, and feature geometry, such as the changing shoreline of barrier islands over time. Data can be entered directly, from a comma separated file, or features with attributes and relationships can be automatically populated in the model from data in the Spatial Data Transfer Standard format.

Open-File Report↗