Remote Sensing For Characterization Of

Geohazards And Natural Resources Estelle

Chaussard download

https://ebookbell.com/product/remote-sensing-for-
characterization-of-geohazards-and-natural-resources-estelle-
chaussard-58449694

Explore and download more ebooks at ebookbell.com

Here are some recommended products that we believe you will be
interested in. You can click the link to download.

Remote Sensing For Malaria Monitoring And Predicting Malaria From

Operational Satellites 1st Ed Felix Kogan

https://ebookbell.com/product/remote-sensing-for-malaria-monitoring-
and-predicting-malaria-from-operational-satellites-1st-ed-felix-
kogan-22502428

Remote Sensing For Sustainable Forest Management 1st Edition Steven E

Franklin

https://ebookbell.com/product/remote-sensing-for-sustainable-forest-
management-1st-edition-steven-e-franklin-2613870

Remote Sensing For Hydrocarbon Exploration Andreas Laake

https://ebookbell.com/product/remote-sensing-for-hydrocarbon-
exploration-andreas-laake-37318860

Remote Sensing For Biodiversity And Wildlife Management Synthesis And

Applications Steven E Franklin

https://ebookbell.com/product/remote-sensing-for-biodiversity-and-
wildlife-management-synthesis-and-applications-steven-e-
franklin-4136676
Remote Sensing For Geoscientists Image Analysis And Integration Third
Edition 3rd Edition Gary L Prost

https://ebookbell.com/product/remote-sensing-for-geoscientists-image-
analysis-and-integration-third-edition-3rd-edition-gary-l-
prost-4748316

Remote Sensing For Climate Change Xander Wang

https://ebookbell.com/product/remote-sensing-for-climate-change-
xander-wang-50812898

Remote Sensing For Sustainability Qihao Weng

https://ebookbell.com/product/remote-sensing-for-sustainability-qihao-
weng-5700596

Remote Sensing For Landscape Ecology Monitoring Modeling And

Assessment Of Ecosystems Second Edition Robert C Frohn

https://ebookbell.com/product/remote-sensing-for-landscape-ecology-
monitoring-modeling-and-assessment-of-ecosystems-second-edition-
robert-c-frohn-6840022

Remote Sensing For Gis Managers Stanley Aronoff

https://ebookbell.com/product/remote-sensing-for-gis-managers-stanley-
aronoff-6874498
Springer Remote Sensing/Photogrammetry

Estelle Chaussard
Cathleen Jones
Jingyi Ann Chen
Andrea Donnellan Editors

Remote Sensing
for Characterization
of Geohazards and
Natural Resources
Springer Remote Sensing/Photogrammetry
The Springer Remote Sensing/Photogrammetry series seeks to publish a broad
portfolio of scientific books, aiming at researchers, students, and everyone interested
in the broad field of geospatial science and technologies. The series includes peer-
reviewed monographs, edited volumes, textbooks, and conference proceedings. It
covers the entire area of Remote Sensing, including, but not limited to, land, ocean,
atmospheric science and meteorology, geophysics and tectonics, hydrology and
water resources management, earth resources, geography and land information,
image processing and analysis, satellite imagery, global positioning systems, archae-
ological investigations, and geomorphological surveying.
Series Advisory Board:
Marco Chini, Luxembourg Institute of Science and Technology (LIST), Belvaux,
Luxembourg
Manfred Ehlers, University of Osnabrueck
Venkat Lakshmi, The University of South Carolina, USA
Norman Mueller, Geoscience Australia, Symonston, Australia
Alberto Refice, CNR-ISSIA, Bari, Italy
Fabio Rocca, Politecnico di Milano, Italy
Andrew Skidmore, The University of Twente, Enschede, The Netherlands
Krishna Vadrevu, The University of Maryland, College Park, USA
Estelle Chaussard • Cathleen Jones •
Jingyi Ann Chen • Andrea Donnellan
Editors

Remote Sensing for

Characterization of
Geohazards and Natural
Resources
Editors
Estelle Chaussard Cathleen Jones
Structures and Geohazards Research Radar Science and Engineering
FM Global Jet Propulsion Laboratory, California
Norwood, OR, USA Institute of Technology
Pasadena, CA, USA
Jingyi Ann Chen
Aerospace Engineering and Engi. Andrea Donnellan
Mechanics Instrument Systems
The University of Texas at Austin Jet Propulsion Laboratory, California
Austin, TX, USA Institute of Technology
Pasadena, CA, USA

ISSN 2198-0721 ISSN 2198-073X (electronic)

Springer Remote Sensing/Photogrammetry
ISBN 978-3-031-59305-5 ISBN 978-3-031-59306-2 (eBook)
https://doi.org/10.1007/978-3-031-59306-2

© The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland
AG 2024
This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse
of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and
transmission or information storage and retrieval, electronic adaptation, computer software, or by similar
or dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

If disposing of this product, please recycle the paper.

Preface

In the past two decades, remote sensing data has rapidly evolved technically,
enabling entirely new applications and collaborations between previously siloed
disciplines. Many textbooks have focused either on a single technique or on a single
discipline, often ignoring the potential that interdisciplinary works offer. This book
encourages readers to think outside the box of their discipline by providing a single
venue for learning the basics of state-of-the-art remote sensing techniques, together
with their interdisciplinary applications.
The value of interdisciplinary research in pushing science forward, accelerating
discovery, and adding value to disciplinary work has long been emphasized
in reports by the National Aeronautics and Space Administration (NASA), the
National Academy of Sciences (NAS), and the National Science Foundation (NSF).
In 2010, the NSF’s Introduction to Interdisciplinary Research read that “important
research ideas often transcend the scope of a single discipline.” One parable I often
share to illustrate the value of interdisciplinary work is the story of the “Blind men
and an elephant.” Blind persons are each holding onto a different part of an elephant,
but no individual can comprehend the elephant as a whole. Only by accounting for
other truths or a totality of truth is one’s limited perception challenged. Viewing a
scientific problem through a disciplinary lens is limiting. For example, atmospheric
signal is called “noise” when geodetic data is used to quantify ground deformation,
but the same signal informs tsunami wave propagation. Multipath reflections at GPS
antennas, previously labeled as “errors,” are now used to derive soil moisture and
snow depths. Interdisciplinary studies bypass such a self-imposed limited view of
data and allows for problem-focused research—identifying the elephant rather than
its parts.
In this book, through a geoscientist’s perspective, the reader will understand the
benefits of using remote sensing techniques to address interdisciplinary problems
with a high societal impact: identifying the drivers of geohazards (including
seismic, volcanic, landslide, and land subsidence hazards) and developing new
methods for tracking natural resources. A review of remote sensing methods used in
geohazards and natural resources sciences is presented, with appropriate referencing
for readers wishing to further their technique-specific learning. Detailed examples

v
vi Preface

of interdisciplinary applications of these remote sensing techniques convey recent

ground-breaking discoveries as well as future opportunities.
Why geohazards and natural resources?
Recent works have demonstrated that the study of geohazards and natural
resources have directly benefited from one another. Hydraulic fracturing is the
most straightforward example of the direct interactions between natural resources
and geohazard processes. The disposal of wastewater associated with natural gas
extraction has led to Oklahoma being more seismically active than California
over the past decade. Much is in the process of being learned about earthquakes
through the study of these induced events since the associated stress changes can
be constrained more accurately than for traditional earthquakes. Another example
of natural resources–geohazards interactions was discovered when exploring the
impact of the 2012–2015 California drought on water resources. Researchers found
that hard rocks and soft sediments both responded to the changes in water, but in
opposite ways. In the Central Valley, the ground was subsiding as soft sediments
of aquifer systems were compacting due to the decrease in pore water pressure
associated with increased groundwater pumping. Such ground deformation data is
now in turn used to assess and monitor water resources. In contrast to the valley,
the Sierra Nevada mountain range uplifted during the drought due to the reduced
snow and surface water loads. This uplift at a few millimeters per year was enough
to influence the seismicity on the nearby San Andreas fault, demonstrating that in
some cases even small stress changes are enough to influence earthquakes. Beyond
such process-based interactions, the joint study of natural resources and geohazards
also strengthens our ability to manage ongoing stresses on resources and prepare for,
withstand, and recover from geohazards. This book thus highlights socially relevant
scientific opportunities, challenges, and potential future directions.

Norwood, OR, USA Estelle Chaussard

Pasadena, CA, USA Cathleen Jones
Austin, TX, USA Jingyi Ann Chen
Pasadena, CA, USA Andrea Donnellan
6 December 2023
Contents

Enhancing Stewardship of Earth Through Remote Sensing . . . . . . . . . . . . . . . . 1

Roland Bürgmann

Part I Remote Sensing Techniques for Geohazards and Resource

Monitoring
The Global Navigation Satellite System (GNSS): Positioning,
Velocities, and Reflections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Ronni Grapenthin
Interferometric Synthetic Aperture Radar (InSAR) . . . . . . . . . . . . . . . . . . . . . . . . . 53
Pablo J. González
Lidar for Geohazard and Natural Resource Characterization . . . . . . . . . . . . . 75
Olaf Zielke
Optical Geodesy and the Measurement of Ground Deformation
by Image Correlation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
James Hollingsworth, Simon Daout, Marie-Pierre Doin,
and Manon Cantraine
The Gravity Recovery and Climate Experiment (GRACE) . . . . . . . . . . . . . . . . . 131
Bert Wouters and Ingo Sasgen
Thermal Remote Sensing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
Jordi Cristóbal, Rudiger Gens, and Anupma Prakash
Multibeam Echosounder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
Daniele Casalbore

Part II Geohazards—Volcano Monitoring

Remote Sensing of Volcano Deformation and Surface Change . . . . . . . . . . . . . 173
Michael P. Poland

vii
viii Contents

Gas and Thermal Emissions of Volcanoes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 205

Andrea Gabrieli and Robert Wright
Modeling of Remote Sensing Data: Common Practices, State
of the Art, and Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
Kimberly DeGrandpre and Zhong Lu

Part III Geohazards—Earthquake and Seismic Hazards

Fault Structure from Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273
Romain Jolivet
Rapid Characterization of Damages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289
Sadra Karimzadeh and Masashi Matsuoka
The Seismic Cycle: From Observations to Models of Fault Slip . . . . . . . . . . . . 305
William Barnhart and Estelle Chaussard

Part IV Geohazards—Land Subsidence

Land Subsidence Hazards: A Case Study of Mexico City . . . . . . . . . . . . . . . . . . . 329
Enrique Cabral-Cano, Darío Solano-Rojas, Enrique A. Fernández-Torres,
and Luis Salazar-Tlaczani
Quantifying Subsidence in Tropical Peatlands . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 347
Alison M. Hoyt, Estelle Chaussard, Sandra S. Seppalainen,
and Charles F. Harvey
Mining/Post-mining Surface Deformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 359
Daniel Raucoules and Michael Foumelis
Sinkholes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 371
Cathleen Jones
Natural Compaction of Sediments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 389
Pietro Teatini, Cristina Da Lio, Claudia Zoccarato, and Luigi Tosi
Coastal Flooding and Structure Stability. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 405
Gonéri Le Cozannet and Pietro Teatini

Part V Geohazards—Landslides
Landslide Hazards . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 417
Adam M. Booth
Underwater Mass Wasting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 443
Sebastian Krastel, Morelia Urlaub, and Felix Gross
Contents ix

Part VI Natural Resources—Freshwater Resources

Observations of Confined Aquifer Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 463
Jingyi Ann Chen and Estelle Chaussard
Large-Scale Terrestrial Water Storage Changes Sensed by Geodesy . . . . . . 473
Yuning Fu, Brian F. Thomas, and Estelle Chaussard
Bridging the Scale Gap Between Ground Deformation
and Gravity: Tools for Sustainability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 493
Pascal Castellazzi, Laurent Longuevergne, and Wei Feng

Part VII Natural Resources—Gas and Hydrocarbon Exploitation

Carbon Capture and Storage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 509
Michela Vellico and Estelle Chaussard
Hydraulic Fracturing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 521
Julia Kubanek

Part VIII Future of Remote Sensing Methods

Future of Remote Sensing for Geohazards and Resource Monitoring . . . . . 533
Tim J. Wright

Part IX Application of Remote Sensing in Other Geoscience

Fields
Space Geodetic Sensing of Atmospheric Water Vapor and Its
Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 555
James Foster
Oceans . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 577
Isabelle Dadou, Gael Alory, and Habib B. Dieng
Cryospheric Applications of Remote Sensing: Snow Water Equivalent . . . 603
Jonathan C. Ryan

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 623
Enhancing Stewardship of Earth
Through Remote Sensing

Roland Bürgmann

Earth is humanity’s only home and environment, and living on and with Earth
requires increasingly accurate knowledge of the many hazards posed by our planet
and the finite natural resources we rely on to prosper. Proper stewardship of Earth
starts with understanding of the different Earth system components, including the
atmosphere, the hydrosphere, and the solid Earth, as well as their often-complex
interactions. As the world’s population has roughly doubled over the last 50 years
and is expected to grow by nearly two billion persons in the next 30 years (United
Nations 2023), more people are being exposed to a variety of geohazards and depend
on a limited supply of natural resources that are increasingly challenging to access.
Characterization of natural hazards and resources requires knowledge of materials
and processes above, at, and below the Earth’s surface. Keeping up with the need to
recognize, characterize, and quantify these hazards and resources requires compre-
hensive and accurate data. Increasingly, remote sensing from satellite, airborne, and
ship-based platforms has become the tool of choice to comprehensively monitor
the Earth. In this book, Remote Sensing for Characterization of Geohazards and
Natural Resources, edited by Estelle Chaussard, Cathleen Jones, Jingyi Ann Chen,
and Andrea Donnellan, the focus is on a variety of remote sensing techniques and
their often integrated and cross-disciplinary use to study geohazards and natural
resources at the surface and shallow subsurface of our planet.
This book is organized into four major parts: (1) seven chapters summarizing key
remote sensing methods, (2) 14 chapters discussing four major types of geohazards,
(3) five chapters addressing remote sensing applied to the study of natural resources,
and (4) four chapters assessing the future of remote sensing applied to geohazards,
natural resources, and related Earth systems. This makes for a comprehensive and

R. Bürgmann (✉)
Department of Earth and Planetary Science, University of California, Berkeley, CA, USA
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 1

E. Chaussard et al. (eds.), Remote Sensing for Characterization of Geohazards and
Natural Resources, Springer Remote Sensing/Photogrammetry,
https://doi.org/10.1007/978-3-031-59306-2_1
2 R. Bürgmann

up-to-date compendium of the current state of knowledge about this topic. It is

notable that many of the authors are leading early- and mid-career scientists, and
this is evident in the timeliness of these contributions.
Geological hazards (geohazards) are the result of active geologic processes,
including, but not limited to volcanoes, earthquakes, landslides, and land subsi-
dence, which are highlighted in this volume. None of these processes act in isolation,
and there is increasing realization that we need to consider cascading multi-hazards
involving two or more of these or other geohazards (e.g., avalanches, tsunamis,
meteorite impacts, forest and coal fires, sea level rise, floods, and droughts), as well
as manmade hazards (e.g., induced seismicity, aquifer depletion, and slope failures).
Geohazards are increasingly costly and deadly, and average global economic losses
over the last few decades have been assessed at ~$300 billion per year (e.g., Ward
et al. 2020). Proper risk assessment, also considering exposure and vulnerability,
starts with accurate characterization of the underlying hazards. Characterizing
and mitigating these diverse hazards requires a wide variety of observations and
analyses. Thus, the study of geohazards needs to take a multi-hazard perspective
and must be cross-disciplinary in nature. Thanks to a diverse global fleet of remote
sensing satellites, enhanced computing capabilities, and modern data analysis
techniques, we now have the capability to monitor many hazards globally, allowing
for improved hazard assessment and rapid response to natural disasters.
Remote sensing also plays an invaluable role for finding, inventorying, monitor-
ing, and protecting natural resources on Earth, including renewable (water, plants,
etc.) and non-renewable (minerals, ores, fossil fuels, etc.) resources. There are a
vast number of natural resources, and here the focus is on the use of remote sensing
to characterize renewable groundwater and non-renewable hydrocarbon systems.
Of course, if we take more of our renewable resources than are being naturally
replenished, their use also becomes unsustainable, and this has become an urgent
concern with water. Arguably, the most essential natural resource for human survival
may well be fresh water, and the use of remote sensing for the evaluation and
management of global groundwater resources has become increasingly important.
Of course, unsustainable withdrawal of groundwater can also represent a natural
hazard, in the form of land subsidence leading to damage to infrastructure and
enhanced coastal flooding. This is just one example of how our exploration
of natural resources also becomes entangled with our need to mitigate natural
hazards.
A wide range of remote sensing techniques have proven of great value for the
characterization and mitigation of geohazards and the discovery and assessment
of natural resources. The first section of this book introduces seven particularly
important observational techniques, each of which contributes in unique ways to
characterizing the shape and make-up of and the dynamic processes at the Earth’s
surface. Many of these methods rely on Earth-orbiting satellites, but some are
operating on airborne platforms, or (when exploring the depths of the Earth’s water
bodies) on ships. Importantly, these methods enhance each other, and thus the later
chapters that are focused on a specific geohazard often emphasize the value of
optimally employing multiple techniques. Later chapters that describe a variety
Enhancing Stewardship of Earth Through Remote Sensing 3

of geohazards and natural resources introduce additional observational systems not

covered in separate chapter.
The Global Navigation Satellite System (GNSS) and Interferometric Synthetic
Aperture Radar (InSAR) have been used to investigate deformation associated with
geohazards for about 40 and 30 years, respectively. What is striking about the
chapters on GNSS by Ronni Grapenthin (chapter “The Global Navigation Satellite
System (GNSS): Positioning, Velocities, and Reflections”) and InSAR by Pablo
Gonzales (chapter “Interferometric Synthetic Aperture Radar (InSAR)”) is how
much these space geodetic systems have evolved since their early years and early
review papers (e.g., Dixon 1991; Bürgmann et al. 2000). Not only have the accuracy,
spatiotemporal resolution, and quantity of position and deformation measurements
using these systems dramatically improved, GNSS and InSAR are now being used
in innovative ways that had not even been considered in the 1990s. For example,
GNSS reflectometry allows for sensing of the local environment around a GNSS
station, thus providing valuable information about temporal changes of near-surface
soil moisture, vegetation height and density, water level of lakes and the oceans,
and snow depth. These capabilities expand the variety of geohazards that can be
addressed using GNSS.
It is evident that while a wide range of remote sensing methods have been
important to study geohazards and natural resources, one of the most powerful
and commonly employed observational tools is Synthetic Aperture Radar (SAR).
Since InSAR was first applied to measure the surface deformation of the 1992
Landers earthquake in astonishing detail (Massonnet et al. 1993), the method
has substantially matured. Not surprisingly, there has been a recent surge in
SAR missions by both government agencies and commercial endeavors, including
sensors operating at a variety of radar wavelengths, image configurations, and signal
polarization, several of which involve large constellations of identical spacecraft. As
has been the case with GNSS, SAR-derived amplitude and phase observations have
also been used to measure non-deformation changes in surface properties that can be
related to geomorphic and hydrological processes, geohazard damage proxies, and
variations in biomass. Chapters on LiDAR (Light Detection And Ranging), optical
image geodesy, global gravity measurements from the GRACE (Gravity Recovery
and Climate Experiment) and GRACE Follow-on satellites, thermal remote sensing,
and sonar observations using multibeam echosounder systems provide similarly
forward-looking discussions of the diverse capabilities, future challenges, and
opportunities associated with these technologies.
This book only includes one chapter (chapter “Multibeam Echosounder” by
Casalbore on multibeam echosounder imaging of the seafloor) specifically targeting
offshore methods; however, we should expect increasing efforts and development
of new technologies targeting hazards and natural resources below the oceans (e.g.,
Bürgmann and Chadwell 2014). This mostly requires systems relying on acoustic
signals, but also includes the deployment of optical fiber, gravimeters, and pressure
sensing equipment. Soon, it may be timely to develop a book focused on recent
developments in remote sensing of the world’s oceans and the geohazards and
resources they present.
4 R. Bürgmann

Geohazards rarely occur in isolation and generally involve multiple processes,

thus requiring multiple and complementary methods to properly study and assess.
The 14 chapters focused on hazards associated with volcanoes, earthquakes, land
subsidence, and landslides highlight the promise of such an integrated approach
to hazard assessment based on remote sensing. Chapter “Remote Sensing of
Volcano Deformation and Surface Change” by Michael Poland on remote sensing
of volcano deformation and surface change exemplifies this promise; describing
the integrated use of GNSS together with InSAR, optical imagery, and LiDAR
acquired from airborne, ground-based, and satellite platforms to observe and assess
a variety of volcanic pre-, co-, and post-eruption deformation and surface-change
processes. Poland also emphasizes the need for relevant technologies representing
below-water remote sensing analogs to study submarine volcanoes. This chapter is
nicely complemented by the contribution from Andrea Gabrieli and Robert Wright
(chapter “Gas and Thermal Emissions of Volcanoes”) on the use of remote sensing
to assess gas and thermal emissions of volcanoes and the comprehensive chapter
“Modeling of Remote Sensing Data: Common Practices, State of the Art, and
Limitations” by Kimberly DeGrandpre and Zhong Lu describing approaches to
modeling of remote sensing data that are needed to properly interpret the diverse
observations.
There are three chapters dedicated to summarizing the use of remote sensing
for identifying and characterizing active fault structures representing earthquake
hazards (Jolivet, chapter “Fault Structure from Space”), for elucidating the different
components of the earthquake cycle (Barnhard and Chaussard, chapter “The Seismic
Cycle: From Observations to Models of Fault Slip”), and for assessing the damage in
the aftermath of earthquakes and, by extension, other natural disasters (Karimzadeh
and Matsuoka, chapter “Rapid Characterization of Damages”). Jolivet makes a
strong case for the need to inventory the three-dimensional geometry of active faults
aided by thorough analysis of the surface geology and tectonic geomorphology
provided by complementary optical, radar, and LiDAR observations from satellite
and airborne platforms. The expression of subsurface faulting in the geology and
topography that can be observed at the surface is complex, but a combination
of geological insight, integration of complementary geophysical methods, and
modeling approaches has allowed for the development of comprehensive geometric
fault models in many plate boundary zones of the world. Nonetheless, this job
is not completed anywhere, and we continue to be surprised by earthquakes that
rupture faults we knew little about, with the 2019 Ridgecrest earthquake sequence
being a recent case in point (e.g., Thompson Jobe et al. 2020). Active faults on
the ocean seafloor represent an important frontier, and much remains to be learned
from enhanced remote sensing of structures below the world’s oceans that are more
difficult to access.
Once an earthquake strikes, its surface rupture and associated deformation field
more completely illuminate the structure of the causative faults, and InSAR, GNSS,
and other remote sensing information in combination with mechanical modeling
can be used to describe the geometry and slip of the activated faults quite precisely
(chapter “Fault Structure from Space”). Barnhard and Chaussard (chapter “The
Enhancing Stewardship of Earth Through Remote Sensing 5

Seismic Cycle: From Observations to Models of Fault Slip”) expand on how

model inversions of space-geodetic data allow for imaging not just the geometry
of faults, but also details of the deformation sources and the underlying dynamic
processes involved in the different phases of the earthquake cycle. Advances in
our understanding of geohazards come not just from improved remote sensing
capabilities, but also from the enhanced sophistication of data analysis and modeling
methodologies. Of course, as Barnhard and Chaussard emphasize, it is crucial that
uncertainties, resolution limits, biases, and tradeoffs are carefully considered when
evaluating and interpreting the derived models.
There are equally informative chapters on remote sensing approaches to studying
hazards associated with subsidence due to natural systems (sinkholes, sediment
compaction, aquifer depletion) and anthropogenic activities (syn- and post-mining
deformation, tunnels). Given the wide range of spatial and temporal scales and
rates of deformation associated with this wide variety of land subsidence processes,
different combinations of sensors and processing methodologies are often indicated
to optimally study these systems (e.g., LiDAR, radar altimetry, InSAR, optical
imaging). In her chapter on sinkholes, Jones (chapter “Sinkholes”) also points to
the value of complementary methods, such as ground-penetrating radar, electrical
resistivity surveys, and gravimetry, that enable subsurface imaging complementing
surface observations. Natural processes, manmade activities, and climate change
often combine to exacerbate such land subsidence hazards (e.g., see chapters by
Jones (chapter “Sinkholes”) and Teatini et al. (chapter “Natural Compaction of
Sediments”), but the focus in this book is mostly on natural geohazards.
Active landslides are one of the costliest geohazards, leading to thousands of
fatalities and loss of infrastructure around the world, each year. Booth (chapter
“Landslide Hazards”) provides a clear overview about the ways in which remote
sensing observations have enabled the investigation of landslide hazard through
detailed mapping to establish comprehensive landslide inventories, deformation
monitoring of slow-moving landslides, and time-series analysis and modeling
allowing for estimates the evolving hazard of landslide failures, and improved
understanding of the causes of catastrophic failures. This improved understanding
may allow for the development of forecasting or even landslide early-warning
approaches (Dai et al. 2020), but such efforts are hindered by the long acquisition
intervals of remote sensing images and our still limited understanding of the physics
of landslide destabilization processes (Lacroix et al. 2020). Booth convincingly
argues that complementing remote sensing observations with large numbers of low-
cost, in situ subsurface sensors would allow for comprehensive real-time monitoring
and enhance landslide forecasting and warning capabilities, thus mitigating land-
slide risk. Krastel et al. (chapter “Underwater Mass Wasting”) provide a view of the
challenges and recent progress in using acoustic remote sensing techniques to study
subaqueous landslides, which are as scientifically interesting and societally relevant
(i.e., they can generate tsunamis and damage seafloor cables) as their subaerial
counterparts.
Earth’s natural resources are limited and thus it is of utmost importance to
carefully assess the extraction, use, and reuse of these resources to maximize
6 R. Bürgmann

their benefit to humanity. The most valuable and essential natural resource is
clearly water, which humans need to survive on a daily basis. However, billions
of people are currently impacted by limited water resources, often associated with
the unsustainable use of groundwater. Chapters by Chen and Chaussard (chapter
“Observations of Confined Aquifer Systems”) and Fu et al. (chapter “Large-Scale
Terrestrial Water Storage Changes Sensed by Geodesy”) describe how InSAR,
GNSS, and GRACE gravity data can be used to track changes in surface water and
groundwater storage from surface deformation and gravity changes, independent of
the availability of more direct observations. Mechanical models of both the elastic
deformation due to changes in mass loading and of poroelastic deformation due to
changes in subsurface water storage further increase the value of the deformation
data for water monitoring [(e.g., Khorrami et al. 2023)]. When geodetic data
are combined with other hydrological data, it is possible to determine additional
information about subsurface aquifers that can provide the basis for the development
of sustainable groundwater extraction, storage, and recharge practices. Castellazzi
et al. (“Bridging the Scale Gap Between Ground Deformation and Gravity: Tools
for Sustainability”) emphasize the need to understand not just the changes of
water storage in individual compartments of the water cycle, but also the dynamic
processes involved in the fluxes of water between them. Optimal integration of the
existing space-geodetic systems will be essential to overcome the challenges posed
by their individual observational limitations in the spatiotemporal resolution and
accuracy. Continuation and further improvements of space-geodetic technologies,
such as the GRACE Follow-On mission and the NISAR (NASA-ISRO SAR) radar
mission and their proposed advanced successors, will be essential to ensure the
global transition to long-term water sustainability. It is essential that governments
and agencies tasked with the management of groundwater systems around the world
will be able to take full advantage of these valuable data sets in their routine
decision-making and water use practices.
As is the case for the assessment of groundwater changes, remote sensing is
also of value to monitor changes in oil and gas fields, including space-geodetic
observations of deformation due to the extraction or injection of fluids and remote
detection of gas emissions into the atmosphere. InSAR measurements of extraction-
related deformation and other near-surface changes can be used to optimize and
maximize the extraction of hydrocarbons from subsurface reservoirs (e.g., Ferretti
2020). They are equally valuable for the assessment of secondary environmental
effects and geohazards, as well as monitoring of efforts aimed at injecting and
storing of waste fluids and CO2 associated with fossil fuels. The chapter by
Kubanek (chapter “Hydraulic Fracturing”) is focused on the use of remote sensing
to assess the effects of hydraulic fracturing (also known as fracking) and associated
wastewater injection, including its environmental impact and associated secondary
hazards such as induced seismicity. Given the anthropogenic changes in atmospheric
composition and associated climate change resulting from the burning of these fossil
fuels, a new challenge lies in optimally capturing and storing the CO2 that is the
primary product of this process. Vellico and Chaussard (chapter “Carbon Capture
and Storage”) discuss several remote sensing observations (e.g., change detection
Enhancing Stewardship of Earth Through Remote Sensing 7

in hyperspectral optical and infrared imagery, LiDAR backscatter intensities, and

space-geodetic deformation measurements) as complementary monitoring tools to
help recognize gas leakage and to verify that CO2 permanently stays in storage.
Clearly, remote sensing also plays an increasingly important role when it comes
to assessing, responding to, and mitigating the damages and losses due to all types
of natural disasters, and thus the free and timely production and availability of
such damage proxy data is of great importance to society. Post-processed SAR
and optical imagery are of particular value and can provide detailed maps of
damage intensity following natural or anthropogenic disaster (e.g., chapter “Rapid
Characterization of Damages” by Karimzadeh and Matsuoka; Rao et al. 2023).
Karimzadeh and Matsuoka emphasize the practical value of community projects,
such as NASA’s Advanced Rapid Imaging and Analysis (ARIA) and the Copernicus
Emergency Management Service (EMS), which are aimed at the timely production
and free distribution of data products in support of post-disaster response and
recovery.
The emphasis of this book is on geological hazards and natural resources
in the terrestrial environment; however, there are three chapters introducing the
application of remote sensing for studies of the Earth’s atmosphere, oceans, and
cryosphere. Foster’s comprehensive chapter (chapter “Space Geodetic Sensing of
Atmospheric Water Vapor and Its Application”) on remote sensing of water vapor
in the atmosphere using various space-geodetic measurements makes for a great
example of the situation where one scientist’s most troublesome source of noise
(turbulent and stratified atmospheric delays of GNSS and InSAR signals used
to study Earth deformation) is another scientist’s valuable signal, enabling the
mapping and improved understanding of the processes and dynamics of moisture
transport in the atmosphere. These data are of great value for improved under-
standing of weather systems and atmospheric processes needed for meteorological
modeling and forecasting over a range of time scales. In turn, the increasingly
accurate and high-resolution determination of the refractive structure of the atmo-
sphere will indeed improve the accuracy of our space-geodetic remote sensing
methods.
Similarly, Dadou et al. (chapter “Oceans”) introduce remote sensing sensors
collecting data over a wide range of frequencies that are useful for studies of
the oceans. Complementing in situ observations, these sensors provide valuable
information about the elevation, roughness (reflecting wind and waves), temper-
ature, salinity, and color (reflecting marine chemistry and biomass) parameters.
Finally, Ryan (chapter “Cryospheric Applications of Remote Sensing: Snow Water
Equivalent”) examines the use of remote sensing to determine variations in the
presence, extent, and character of snow. Quantifying snow cover, often expressed
as snow water equivalent, represents a particularly challenging target for remote
sensing, as the density and depth of snow are difficult to capture without in situ
information. LiDAR, optical image stereometry, INSAR, and GNSS reflectometry
are currently the best tools available, but future missions promise to greatly improve
our ability to measure changes in snow cover at high accuracy and spatiotemporal
resolution.
8 R. Bürgmann

What is the “Future of Remote Sensing for Geohazards and Resource Moni-
toring” (Wright, chapter “Future of Remote Sensing for Geohazards and Resource
Monitoring”)? Wright provides a personal perspective on recent and future devel-
opments in the rapidly growing arena of remote sensing platforms and science, with
a focus on optimally taking advantage of the growing and diverse big-data streams
and democratizing the use of remote sensing data. It is exciting to learn about some
of the improvements and innovative enhancements to remote sensing platforms and
methodologies in this chapter, which will likely revolutionize the field over the next
decade. Another emphasis is on a throughgoing theme of this volume, pointing
to the importance for interdisciplinary collaboration and advanced data analysis
methods and modeling to fully capture, synthesize, and translate the remote sensing
data into information that decision-makers can act on. This effort will benefit from
increasingly accurate, prompt, long-lasting, and freely available data streams and
tools like machine learning and artificial intelligence, which are needed to properly
digest the global firehose of information about the many geohazards and natural
resources that will continue to shape mankind’s future.

References

Bürgmann R, Chadwell CD (2014) Seafloor geodesy. Annu Rev Earth Planet Sci 42:509–534.
https://doi.org/10.1146/annurev-earth-060313-054953
Bürgmann R, Rosen PA, Fielding EJ (2000) Synthetic aperture radar interferometry to measure
Earth’s surface topography and its deformation. Annu Rev Earth Planet Sci 28:169–209
Dai K et al (2020) Entering the era of earth observation-based landslide warning systems:
a novel and exciting framework. IEEE Geosci Remote Sens Mag. https://doi.org/10.1109/
MGRS.2019.2954395
Dixon TH (1991) An introduction to the global positioning system and some geological applica-
tions. Rev Geophys 29:249–276
Ferretti A (2020) Satellite InSAR data: reservoir monitoring from space. Earthdoc, London. https:/
/doi.org/10.3997/9789462820036
Khorrami M, Shirzaei M, Ghobadi-Far K, Werth S, Carlson G, Zhai G (2023) Groundwater volume
loss in Mexico City constrained by InSAR and GRACE observations and mechanical models.
Geophysical Research Letters, 50, e2022GL101962. https://doi.org/10.1029/2022GL101962
Lacroix P, Handwerger AL, Bièvre G (2020) Life and death of slow-moving landslides. Nat Rev
Earth Environ 1:404–419. https://doi.org/10.1038/s43017-020-0072-8
Massonnet D, Rossi M, Carmona C, Adragna F, Peltzer G, Feigl K, Rabaute T (1993) The
displacement field of the landers earthquake mapped by radar interferometry. Nature 364:138–
142
Rao A, Jung J, Silva V, Molinario G, Yun S-H (2023) Earthquake building damage detection based
on synthetic-aperture-radar imagery and machine learning. Nat Hazards Earth Syst Sci 23:789–
807. https://doi.org/10.5194/nhess-23-789-2023
Thompson Jobe JA, Philibosian B, Chupik C, Dawson T, Bennett SEK, Gold R, DuRoss C,
Ladinsky T, Kendrick K, Haddon E et al (2020) Evidence of previous faulting along the 2019
Ridgecrest, California, earthquake ruptures. Bull Seismol Soc Am 110:1427–1456. https://
doi.org/10.1785/0120200041
Enhancing Stewardship of Earth Through Remote Sensing 9

United Nations (2023). https://www.un.org/en/global-issues/population. Last accesses 21 July

2023
Ward PJ, Blauhut V, Bloemendaal N, Daniell JE, de Ruiter MC, Duncan MJ, Emberson R, Jenkins
SF, Kirschbaum D, Kunz M, Mohr S, Muis S, Riddell GA, Schäfer A, Stanley T, Veldkamp
TIE, Winsemius HC (2020) Review article: natural hazard risk assessments at the global scale.
Nat Hazards Earth Syst Sci 20:1069–1096. https://doi.org/10.5194/nhess-20-1069-2020
 Part I
Remote Sensing Techniques for
Geohazards and Resource Monitoring
The Global Navigation Satellite System
(GNSS): Positioning, Velocities, and
Reflections

Ronni Grapenthin

1 Introduction

After more than 20 years of development (e.g., Easton 1974) and conceptual testing
by the U.S. Department of Defense, the Global Positioning System (GPS) achieved
full constellation in 1993. The system, in which satellites broadcast civil and
military signals modulated onto sinusoidal carrier signals, lends itself to many uses
in Earth science spanning solid Earth geophysics, surface processes, atmospheric
science, and space weather. The ability to achieve millimeter positioning precision
by tracking the carrier phase (e.g., Counselman et al. 1980; Counselman and Goure-
vitch 1981; Hoffmann-Wellenhof et al. 2008; Misra and Enge 2011) in addition to
utilizing the much less precise ranging codes nothing less but revolutionized the
field.
GPS enabled direct measurements of plate motions (e.g., Feigl et al. 1993) on
a global scale at an affordable cost per receiver (e.g., Segall and Davis 1997).
Since then, polar motion measurements improved with GNSS (e.g., Herring et al.
1991; Desai and Sibois 2016), tectonic plate motion velocity models are being
refined (e.g., Argus and Heflin 1995; Argus et al. 2010), micro-plates (e.g., Jansma
et al. 2000; Wallace et al. 2004; Apel et al. 2006) and terranes (e.g., Fletcher
and Freymueller 1999; Elliott et al. 2010) have been identified or characterized,
further constraining terrestrial dynamics (Copley et al. 2011) and informing on
seismic hazards (e.g., Newman 1999; Bilham et al. 2001). In addition to capturing
deformation during earthquake ruptures (e.g., Nikolaidis et al. 2001; Larson et al.
2003; Grapenthin and Freymueller 2011), and mapping the recorded surface
deformation back to slip on finite fault surfaces (e.g., Simons et al. 2011; Galetzka

R. Grapenthin ()
Geophysical Institute, University of Alaska Fairbanks, Fairbanks, AK, USA
e-mail: [email protected]

© The Author(s), under exclusive license to Springer Nature Switzerland AG 2024 13

E. Chaussard et al. (eds.), Remote Sensing for Characterization of Geohazards and
Natural Resources, Springer Remote Sensing/Photogrammetry,
https://doi.org/10.1007/978-3-031-59306-2_2
14 R. Grapenthin

Fig. 1 Horizontal velocity solutions from Herring et al. (2016) for the Network of the Americas
(NOTA) spanning the contiguous USA (a), Alaska (b), and the Caribbean (c). Uncertainties at the
95% confidence level are plotted at the arrow tips but are not visible at this scale. West of 100.◦ W
only about 15% of the available stations are shown, the background color in panel (a) indicates the
1.◦ .× 1.◦ station density. (Source: Herring et al. 2016)

et al. 2015), another important contribution to seismic hazard assessment arising

from GNSS observations is the recording of plate boundary processes such as
interseismic strain build-up (e.g. Wang et al. 2001, Fig. 1), slow slip (e.g., Dragert
et al. 2001; Rogers and Dragert 2003), and the amount of coupling between the
subducting and overriding plates (e.g., Freymueller et al. 2008; Xiao et al. 2021).
Near volcanoes, we can resolve subsurface magma migration (e.g., Cervelli et al.
2006; Elsworth et al. 2008; Hreinsdóttir et al. 2014), co-eruptive magma extrusion-
driven deformation (e.g., Sigmundsson et al. 2015), and even piston-like motion due
to caldera collapse (Gudmundsson et al. 2016; Neal et al. 2019). Beyond tectonic
and magmatic applications, we capture crustal deformation due to dynamics of the
cryosphere and hydrosphere (e.g., Heki 2001; Grapenthin et al. 2006; Amos et al.
2014; Borsa et al. 2014; Argus et al. 2014), including the resolution of hemispheric
mass exchange due to seasonal winter loading of the continents (Blewitt et al.
2001), measurements of glacial isostatic adjustment (e.g., Sella et al. 2007; Thomas
et al. 2011) and its acceleration due to climate change response of the cryosphere
(Compton et al. 2015), and inference of Earth rheology from such observations (e.g.,
Grapenthin et al. 2006; Argus et al. 2021).
Subdaily (e.g., Nikolaidis et al. 2001), high-rate (e.g., Larson et al. 2003;
Galetzka et al. 2015), and real-time GNSS (e.g., Grapenthin et al. 2014b; Melgar
et al. 2019; Melbourne et al. 2021) applications in geophysics have been devel-
GNSS: Positioning, Velocities, and Reflections 15

oped over the last two decades. The major limitation to enable this technology,
particularly in the geophysically most interesting, remote locations, is the design
of sustainable power supply and telemetry systems that support the transfer of the
large data volumes resulting from high-rate (1 sps) to very high-rate (up to 50 sps)
observations. In regions where the engineering challenges can be met, GNSS can be
used in real-time hazard analysis (e.g., Grapenthin et al. 2014b; Melgar et al. 2019)
and early warning (e.g., Murray et al. 2018) as was successfully demonstrated for
the 2014 M.w 6.0 South Napa earthquake in California (Grapenthin et al. 2014a)
and the 2019 M.w 7.1 Ridgecrest earthquake (Melgar et al. 2019; Melbourne et al.
2019), also in California. However, real-time GPS-only positioning precision is at
the centimeter level, even when using high-quality dual-frequency receivers. The
lower magnitude threshold for GNSS to resolve any meaningful displacements
depends very much on the distance from the hypocenter. For crustal faults with
nearby GNSS stations, recording of events in the magnitude 5 range is possible
(e.g., Geng et al. 2013).
A recent development lowering the position noise has been the combination
of accelerometers and high-rate GNSS positions to generate seismogeodetic data
streams that provide positions at accelerometer frequencies (Bock et al. 2011).
This requires colocation of accelerometers at geodetic-quality GNSS stations and
resolves earthquake displacements at the temporal resolution of the accelerometer.
The technique depends on the alignment of positioning solutions from GNSS
and accelerometer data, which are downweighted in the combination to suppress
accelerometer drift.
The propagation of the satellite signal through the ionosphere, the troposphere,
and its reflection off the ground before reaching the antenna resulted in the
development of several non-positioning applications that use GNSS as a remote
sensing tool. For instance, GNSS is used to characterize total electron content of
the ionosphere (e.g., Mannucci et al. 1998), resulting in applications to not only
monitor space weather, but also propagation of acoustic and gravity waves due to
earthquakes (e.g., Calais and Minster 1995), volcanic eruptions (e.g., Heki 2006),
explosions (e.g., Fitzgerald 1997), and tsunamis (e.g., Artru et al. 2005). Meng et al.
(2019) provide an excellent, more extensive overview of theory and measurement
techniques of upper atmosphere perturbations. GNSS can furthermore be used to
characterize the distribution of precipitable water content in the troposphere (e.g.,
Bevis et al. 1992), detect and characterize volcanic ash plumes (e.g., Houlié et al.
2005; Grapenthin et al. 2013; Larson 2013; Larson et al. 2017a; Grapenthin et al.
2018a), and determine local snow depth, soil moisture, vegetation water content,
or decadal changes of permafrost around the GNSS monument (e.g., Larson 2016,
2019; Liu and Larson 2018). At coastal sites, ocean tides (Larson et al. 2013) and
storm surges (Peng et al. 2019) have been estimated, turning GNSS into a tide
gauge that can be decoupled from surface deformation and is registered in a global
reference frame. This astonishing richness in applications of a single observation
system can be explained through the observation models described below.
Several other satellite positioning constellations have been developed and estab-
lished, such as the European Galileo, the Russian GLONASS, and the Chinese
16 R. Grapenthin

BeiDou, which are all globally operating, while the Indian NAVIC and the Japanese
QZSS operate regionally (see, e.g., Hoffmann-Wellenhof et al. 2008; Misra and
Enge 2011, for details). Combined, these constellations form the Global Navigation
Satellite System (GNSS). Multi-GNSS analysis approaches can use a large number
of signals from these different systems and promise significant noise reduction for
both classic static analysis and kinematic or real-time applications (e.g., Geng et al.
2018). Effective positioning estimation approaches leveraging the strength of all
available signals are still a very active area of research (e.g., Montenbruck et al.
2014; Liu et al. 2017; Geng et al. 2019; Zheng et al. 2019). For the purposes of this
chapter, however, I will focus on the legacy GPS constellation that transmits on two
frequencies L1 (1575.42 MHz) and L2 (1227.60 MHz).

2 How GNSS Works

The GPS constellation requires a minimum of 24 satellites, orbiting the Earth at

20,350 km. The satellites are distributed on 6 orbital planes that are inclined at
55.◦ (Fig. 2). This design results in repeated ground tracks for each GPS satellite
at about 11 hour 58 minute periods (see Agnew and Larson 2006, for more precise
repeat time calculations). While 24 operational satellites are required to guarantee
a minimum of 4 visible satellites, more satellites are in orbit to strengthen the
constellation and add redundancy (30 operational satellites on 9 October 2019,
https://www.navcen.uscg.gov/). Monitoring of satellite health, orbits, and other
tasks to maintain the system is performed by a system of ground stations that are
globally distributed such that each satellite is always in view of at least two ground
stations.
The GPS satellites broadcast signals on at least two radio frequencies: Link 1
(L1, 1575.42 MHz) and Link 2 (L2, 1227.60 MHz). The carrier signals at these
frequencies are derived from a 10.23 MHz atomic clock on board of the satellites.

Fig. 2 GPS constellation

model. The minimum
constellation requires 4
satellites on each of the 6
orbital planes (gray lines) that
are inclined 55.◦ to each other.
(Source: GPS.gov)
GNSS: Positioning, Velocities, and Reflections 17

Legacy GPS provides just one signal on L1, the coarse/acquisition (C/A) signal, to
be used for several-meter precision civil positioning applications. However, phase
tracking of the carrier signals on L1 and L2 enables the very precise (mm-precision
for static applications) uses of GPS. As operational GPS satellites approach their
end of life, they are replaced with newer generations, offering opportunities to
modernize the system. New demands on navigation and interoperability with other
systems and general advances in technology result in the addition of new signals and
even new transmission bands. Notable is L2C, a new civilian, unencrypted signal
that is currently available on 19 satellites and has a higher signal-to-noise ratio than
the C/A signal. L2C remains preoperational (availability not yet achieved on 24
GPS satellites) as of spring 2023 (https://www.gps.gov/systems/gps/modernization/
civilsignals/). The availability of two unencrypted civilian signals will enable the
mitigation of ionospheric delay without the need to track and resolve ambiguities
for the carrier phase (see below), which will bring a significant improvement
of consumer positioning applications as dual-frequency receivers become more
affordable. Additionally, many satellites now also broadcast on L5 (1176.45 MHz),
a dedicated safety-of-life signal in a protected frequency band (unlike L2), which
will provide an additional civilian-use signal at higher power once operational status
is achieved. Two civilian signals in a protected frequency band (L1 C/A, and L5) will
enable robust precision navigation (due to ionospheric delay mitigation) for aircraft
and other sensitive equipment requiring high position precision.

3 Positioning in a Nutshell

The main applications of GNSS relate to positioning and position changes, and the
provision of precise global timing. The advance that came with the availability of
precise timing is that positioning could move from the measurement of angles to
the measurement of distances. GNSS works by determining the distance between
receiver and satellites and then solving for the position that puts the receiver where
the various distances from all tracked satellites intersect. One of the satellites,
though, will be used to correct the much less precise clock of the ground receiver.
These four unknowns (3 position values in 3D space and clock correction) require
as least four satellites in view, which became the main design criterion for the
GPS constellation. However, because GNSS signals traverse the ionosphere and
troposphere, are reflected off of the ground, and interfere with signals arriving
directly at the antenna from the satellites, precise positioning requires treatment
of these error terms. Here, I will first introduce reference systems used in GNSS
analysis, after which I will present the observation models that link satellite
observations to the desired estimates for position and time. Following this, I will
discuss the treatment of the error terms to further enhance the position precision.
18 R. Grapenthin

3.1 Reference Systems

Two Cartesian coordinate systems are necessary to realize GNSS positioning. Both
are well-defined, which allows transformations of points from one reference system
to the other (see, e.g., Hoffmann-Wellenhof et al. 2008). The first coordinate system
is Earth-centered space-fixed, which is necessary to express satellite motions around
a Sun-orbiting Earth. The origin is defined at Earth’s center of mass, the z-axis
corresponds to the average direction of Earth’s rotation axis (Celestial Intermediate
Pole, CIP), the x-axis points to the vernal equinox in an equatorial plane, and the
y-axis is selected to make the coordinate system right-handed (e.g., Misra and Enge
2011). Varying speeds around the Sun as well as precession and nutation of the
axis of rotation pose potential issues in defining a stable coordinate system, but
these processes are well-understood and can be embedded in the realization of the
reference system.
The second, and for our purposes more prominent coordinate system, is Earth-
centered Earth-fixed (ECEF). It rotates with the Earth, which means the user
position is fixed. While it has a formal definition, it is realized through a set of
points and their velocities to account for tectonic plate motion (Bock and Melgar
2016). One such time-variable realization of a reference frame is the World Geodetic
System 1984 (WGS84, Decker 1986), common to consumer-grade GPS applications
and maintained by the US National Geospatial Agency. Another reference system,
the International Terrestrial Reference Frame (ITRF), is updated more frequently
as extended time series at ground stations allow for increased precision, analysis
approaches improve, and more stations globally allow for tighter constraints on the
reference frame. The most recent version is ITRF14 (Altamimi et al. 2016), which
for the first time also includes non-linear station motions induced by annual seasonal
variations and post-seismic deformation at sites near large earthquakes (Altamimi
et al. 2016). The next generation, ITRF2020, is based on a new analysis strategy
and includes, for instance, updates to the post-seismic models; a full article is in
preparation (Altamimi et al. 2023).
As Cartesian coordinates are not very intuitive to convey a position and its
change on the Earth’s surface, we can define a smooth reference model in the
form of an ellipsoid, with the same origin as the ECEF system. The z-axis is the
axis of revolution of the ellipsoid, and, for instance, WGS84 defines the ellipsoid
semi-major axis and its flattening. Once the ellipsoid is defined, we can transform
from .[X, Y, Z] coordinates to latitude, longitude, and height on the ellipsoid (e.g.,
Hoffmann-Wellenhof et al. 2008; Bock and Melgar 2016).
At this point it is crucial to understand that the absolute height values for GNSS
are given above a reference ellipsoid (as defined by, e.g., WGS84). This is a smooth
oblate simplification of the Earth’s shape. Traditionally, however, height values
have been given with respect to mean sea level (orthometric height), which is
expressed through the geoid (e.g., Earth Gravitational Model 2008 (EGM2008),
Pavlis et al. 2012). This is an undulating equipotential surface that varies with
the position-dependent gravitational potential of the Earth. Differences between
GNSS: Positioning, Velocities, and Reflections 19

reference ellipsoid and geoid can be tens of meters. Hence, it is important to note
the reference frame in which heights are given, particularly when comparisons to
topographic heights are made, which are generally with respect to mean sea level.
Once we build up time series of GNSS positions at a single site, we are generally
interested in change over time. This change is more intuitive when position solutions
.[Xi , Yi , Zi ] at epoch i relative to an initial position .[X0 , Y0 , Z0 ] at epoch 0 are

rotated into a local north–east–up (NEU) system (e.g., Bock and Melgar 2016):
⎡ ⎤ ⎡ ⎤⎡ ⎤
ΔNi − sin(φ) cos(λ) − sin(λ)) sin(φ) cos(φ) Xi − X0
. ⎣ ΔEi ⎦ = ⎣ − sin(λ) cos(λ) 0 ⎦ ⎣ Yi − Y0 ⎦ (1)
ΔUi cos(λ) cos(φ) cos(φ) sin(λ) sin(φ) Zi − Z0

where .φ and .λ are the geodetic latitude and longitude of the site, respectively.
The dominant signal (in the horizontal component) of GNSS time series is
generally the steady-state rigid tectonic plate motion. Removal of this signal is
often desired to highlight short-term transients or unmodeled temporal signals,
for instance, interseismic strain build-up along the plate boundaries (Fig. 1). Plate
velocities are determined from geodetic observations at locations of the continent
that are presumed stable and are expressed as angular velocity around an Euler Pole,
which represents a translation on a sphere. Recent global plate velocity models are,
for instance, GEODVEL by Argus et al. (2010) or the model by Kreemer et al.
(2014). North America-centric ones such as NA12 (Blewitt et al. 2013) or NAM14
(Herring et al. 2016) are also available.

3.2 Pseudorange Model

With the importance of reference frames to positioning addressed, we can move

toward the mathematical models behind GNSS positioning. The range describes
the geometric distance between two points, in our case a satellite and a receiver.
This could, for instance, be inferred by measuring the transit time, .τ , of a signal
that travels from satellite to receiver at the speed of light, c, if the signal contains
a timestamp for the send time and the receiver notes the arrival time. However, the
GNSS receiver has an imprecise clock, and the signal travel path is affected by path
delay effects due to ionosphere and troposphere and other error sources, resulting
in longer travel than the pure geometric distance would suggest. Hence, we call the
range observable provided by a GNSS receiver a pseudorange to a satellite.
The pseudorange from receiver u to satellite s, .ρ (s) (in length units), can be
expressed as a superposition of the true geometric range .r (s) to satellite s and the
known error sources (Misra and Enge 2011):

ρ (s) = r (s) + c(δtu − δt (s) ) + I + T + ϵ

. (2)
20 R. Grapenthin

where c remains the speed of light, .δtu is the receiver clock bias, .δt (s) is clock bias
of satellite s (the broadcast ephemeris typically results in accuracy to a few meters;
more precise products are available or special processing strategies can remove this
error term; see below), and .I, T are ionospheric and tropospheric delays. The last
term, .ϵ, captures unmodeled effects, such as multipath, measurement errors, etc.
(see Sect. 3.4). Note that subscripts (e.g., u) reflect receiver specific values, while
superscripts identify individual satellites; these are not powers of .(s)!
Substituting the geometric range between satellite and receiver in Earth-centered
Earth-fixed Cartesian coordinates into Eq. 2 and linearizing the result via Taylor
series expansion about an approximate initial position and expected receiver clock
bias .(x0 , y0 , z0 , te0 ), in vector notation, we get
⎡ ⎤
Δx
⎡ ⎤ ⎢ Δy ⎥
Δρ (s) = ∂ρ (s) ∂ρ (s) ∂ρ (s) ∂ρ (s) ⎢ ⎥
.
∂x ∂y ∂z ∂te ⎣ Δz ⎦ + ϵ (3)
Δte

Here .Δρ (s) is the difference between the measured pseudorange (observed by
the receiver) and the expected geometric range between the satellite position and
the a priori position. The terms .[Δx, Δy, Δz, Δte ] are the difference between
the actual receiver position and the initial approximation. Adding these values to
the approximated position will yield an improved absolute position estimate. To
simplify Equation 3, all error terms have been absorbed into .ϵ for the time being.
If we solve the partial derivatives in Eq. 3 by applying the chain rule and
appropriate substitutions, we are left with
⎤
⎡
Δx
⎡ ⎤
x0 −x (s) y0 −y (s) z0 −z(s) ⎢ ⎥
Δρ (s) = c ⎢ Δy ⎥ + ϵ (4)
.
ρ0
(s)
ρ0
(s)
ρ0
(s) ⎣ Δz ⎦
Δte

(s)
where the .(x (s) , y (s) , z(s) ) remains the position of satellite s and .ρ0 is the approx-
imated distance between the receiver’s approximated initial position and satellite s,
whose position we assume to be known here. Assuming that we have n satellites in
view, each of which giving us a pseudorange measurement .ρ (1) , . . . , ρ (n) , we can
set up a linear system of equations in matrix–vector notation:
⎡ ⎤
x0 −x (1) y0 −y (1) z0 −z(1)
⎡ (1) ⎤
c
⎢ ⎥ ⎡ Δx ⎤
(1) (1) (1)
Δρ ρ0 ρ0 ρ0
⎢ x0 −x (2) y0 −y (2) z0 −z(2) ⎥
⎢ Δρ (2) ⎥ ⎢ c⎥
⎢ ⎥ ⎢ ρ0
(2)
ρ0
(2)
ρ0
(2) ⎥⎢ ⎥
⎢ Δy ⎥ + ϵ.
.⎢ . ⎥= ⎢
⎢
⎥
.. ⎥ ⎣ Δz ⎦
(5)
⎣ .. ⎦ .. .. ..
⎢ . . . .⎥
Δρ (n) ⎣ ⎦ Δte
x0 −x (n) y0 −y (n) z0 −z(n)
(n) (n) (n) c
ρ0 ρ0 ρ0
GNSS: Positioning, Velocities, and Reflections 21

d = Gm + ϵ (6)

Equation 6 represents a shorthand of Eq. 5, where the matrix G contains the

partial derivatives, d is a vector holding the pseudorange differences, and m is
a vector of unknown differences between actual and approximate position and
receiver clock error. Given G and d, we can solve this linear system of equations for
m with least-squares techniques (e.g., for general least-squares solutions see Aster
et al. 2018; Lichten 1989, is a reference for GNSS-specific analyses) to minimize
the sum of squared residuals, for instance, using the normal equations:

m = (GT G)−1 GT d
. (7)

We could also introduce a weight matrix W to, for instance, reduce the impact of
satellites at low elevation angles on the solution as they provide noisier signals due
to longer signal paths through Earth’s atmosphere:

m = (GT W G)−1 GT W d
. (8)

where W can be diagonal and contains, for instance, the reciprocal variances of the
measurements . σ12 . Using the inverse of the full data covariance matrix as weight
matrix W is a more rigorous approach as this also accounts for correlations between
the measurements in space and time (e.g., Bock and Melgar 2016).
Once we have a solution .m = [Δx, Δy, Δz, Δte ], we can add it to the a priori
values to get an updated absolute position:
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
xnew x0 Δx
⎢ ynew ⎥ ⎢ y0 ⎥ ⎢ Δy ⎥
.⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎣ znew ⎦ = ⎣ z0 ⎦ + ⎣ Δz ⎦ (9)
tenew te 0 Δte

and iterate until improvements are small.

3.3 Carrier Phase Model and Ambiguity Resolution

A more precise measure of a receiver position can be achieved by tracking the carrier
phase of the signal. Misra and Enge (2011) describe in detail several methods of how
a receiver tracks the carrier phase, the details of which are beyond the scope of this
chapter. Assuming we have measurements of the phase observable in units of cycles,
the observation model equation for satellite s is Misra and Enge (2011)

Фs = λ−1 [r (s) + IФ + TФ ] + f (δtu − δt (s) ) + N (s) + ϵФ

(s) (s)
. (10)
22 R. Grapenthin

where .λ is the carrier signal wavelength and f is the carrier frequency (.f = c/λ,
where c is the speed of light), r remains the geometric range between satellite and
receiver, and .IФ and .TФ are the ionospheric and tropospheric propagation delays,
respectively. The clock errors are again captured by .δtu and .δt s for receiver and
satellite clocks, respectively. An important new term in this model is .N s , the integer
ambiguity for satellite s.
The solution strategy for finding the geometric range between satellite and
receiver remains similar to that in Sect. 3.2. However, before this can be done, we
need to find the correct value for N, the integer ambiguity. This term represents the
number of cycles that the signal has gone through before the receiver started tracking
the signal. One can imagine that when the satellite first appears on the horizon, the
receiver captures the factional phase of the signal and keeps adding or removing
full cycles as it maintains lock to the satellite. Given the sinusoidal nature of the
carrier signal, the total number of cycles required to travel from satellite to receiver
is unknown. All we know is that it must be an integer number of cycles. Several
strategies have been proposed to solve this problem, commonly captured under the
term ambiguity resolution.
A number of approaches exist to resolve integer ambiguities. One instructive
analytical method uses the dual-frequency measurements of the phase on L1 and L2
and combines them into a widelane measurement, .Ф12 (e.g., Misra and Enge 2011):
r
.Ф12 = Ф1 − Ф2 = + N12 + ϵФ12 (11)
λ12

The resulting longer wavelength, .λ12 = c/(fL1 − fL2 ) = 0.862 m, of the

combined signal reduces the uncertainty in the integer ambiguity estimate, but also
amplifies the noise in the signal, which is the reason we do not use this combination
for positioning. .N12 is the difference of the integer ambiguities on L1 and L2 and
can be estimated as (e.g., Misra and Enge 2011)
⎡ ⎤
ρ1
N̂12
. = Ф12 − (12)
λ12 roundoff

Misra and Enge (2011) determine that the standard deviation of this estimate
is about 1.2 cycles, suggesting that this could be reduced to less than 0.5 cycle
with uncorrelated measurements over 10 epochs, and even more with additional
measurements. Once we have an acceptable estimate for the widelane integer
ambiguity, we can use this to estimate the L1 and L2 integer ambiguities by solving
the (simplified) measurement models for the range (e.g., Misra and Enge 2011):

r = λ1 (Ф1 − N1 − ϵФ1 )
.

r = λ2 (Ф2 − N2 − ϵФ2 )

and equating them:

GNSS: Positioning, Velocities, and Reflections 23

λ1 (Ф1 − N1 − ϵФ1 ) = λ2 (Ф2 − N2 − ϵФ2 ).

. (13)
λ2 λ2
N1 − N2 = Ф1 Ф2 + ϵ (14)
λ1 λ1

Since .N1 − N2 = N12 , we can solve Eq. 14 to estimate .N1 and .N2 (e.g. Misra and
Enge 2011):
⎡ ⎤ ⎡ ⎤
λ2 −1 λ2 λ2
.N̂1 = 1 − Ф1 − Ф2 − N12 . (15)
λ1 λ1 λ1
⎡ ⎤−1 ⎡ ⎤
λ2 λ2
N̂2 = 1 + −Ф1 + Ф2 + N12 (16)
λ1 λ1

Equations 15 and 16 have not been corrected for the impact of the ionosphere,
however. The next section below will explore how to eliminate such nuisance
terms. It should be obvious that this approach will benefit significantly from the
upcoming triple-frequency observations. However, ambiguity resolution through
widelane estimation is by no means a guarantee as it relies on good data quality
(Misra and Enge 2011).
Following this analytical approach, I want to briefly discuss search solutions
that resolve the integer ambiguities for all satellites simultaneously. Hoffmann-
Wellenhof et al. (2008), Misra and Enge (2011), and Bock and Melgar (2016)
discuss these methods at greater depth. Most prominent is the Least-squares
AMBiguity Decorrelation Adjustment method (LAMBDA method, Teunissen 1993,
1995). The basic principle of the LAMBDA algorithm is to first find a least-squares
float positioning solution that disregards the integer property of the ambiguities and
estimates floating point ambiguities. The next step then decorrelates the ambiguities
and hence reduces the search space to allow for a mapping to integer values in the
transformed space. Lastly, the integer ambiguity estimates are used to find a fixed
positioning solution.

3.4 Precise Positioning Strategies and Error Terms

When introducing pseudorange and carrier phase positioning and the respective
observation models (Eqs. 2 and 10), we have neglected to address any of the error
terms. If our goal is to achieve subcentimeter positioning, we need to correct the
delay introduced by the ionosphere, dry and wet delays imposed on the signal in
the troposphere, and in some cases direct and indirect arrivals of the signal from the
satellite at the receiver (the so-called multipath, MP).
Treatment of these nuisance terms is intimately related to the data processing
strategy used to determine the GNSS position. Below, I introduce the two dominant
strategies: network processing (Dong and Bock 1989; Blewitt 1989) and precise
point positioning (Zumberge et al. 1997), and discuss how the nuisance terms are
24 R. Grapenthin

treated or eliminated. It is important to keep in mind that these terms complicate the
positioning problem, but they can be and have been exploited as signals to constrain
the properties of the medium that the GNSS signals propagate through.

3.4.1 Differential or Network Processing

In differential positioning, observables from several stations and satellites are

differenced to eliminate some of the nuisance parameters from the observation
model. The simplest example is that of a single station pair. If we difference the
observations from one satellite recorded at these two stations, we can eliminate
the satellite clock error. Differencing observations from two satellites at the same
receiver eliminates the receiver clock error. Forming the double difference of these
two results will eliminate both clock errors. For the carrier phase measurement from
satellite k recorded at receivers u and r, we can write the two observation models
and form the single difference (see Eq. 10):

1 (k)
φu(k) =
. ∗ (ru(k) + Iu(k) + Tu(k) ) + f ∗ (δtu − δt (k) ) + Nu(k) + ϵφ,u
λ
1 (k)
φr(k) = ∗ (rr(k) + Ir(k) + Tr(k) ) + f ∗ (δtr − δt (k) ) + Nr(k) + ϵφ,r
λ
(k)
φur = φu(k) − φr(k)
1 (k)
= ∗ (rur
(k)
+ Iur
(k)
+ Tur
(k)
) + f ∗ δtur + Nur
(k)
+ ϵφ,ur (17)
λ

Here, the satellite clock error, .δt (k) , is common to the observations and effectively
cancels in the difference. If we make the assumption that the baseline is short, for
example, on the order of tens of kilometers, the ionosphere and troposphere errors
may also be small as the signal likely traverses the same atmosphere to reach both
stations. However, in many instances, significantly different local weather patterns
or substantial elevation differences (e.g., mountain and valley) may prohibit this
assumption:

(k)
rur (k)
.
(k)
φur = + f ∗ δtur + Nur
(k)
+ ϵφ,ur (18)
λ
If we form a similar single difference for the same receivers, u and r, but a
different satellite, l, we get
(l)
φur
. = φu(l) − φr(l)
(l)
rur (l)
= + f ∗ δtur + Nur
(l)
+ ϵφ,ur (19)
λ
GNSS: Positioning, Velocities, and Reflections 25

We can now use Eqs. 18 and 19 to form the double difference and eliminate the
receiver clock errors captured in .δtur :
(kl)
φur
. = φur
(k)
− φur
(l)

= (φu(k) − φr(k) ) − (φu(l) − φr(l) )

(kl)
rur (kl)
= + Nur
(kl)
+ ϵφ,ur (20)
λ
At this stage, we are still left with the need to estimate at least the integer
ambiguities. While the atmosphere and ionosphere may cancel in the single
differences, we can treat them by applying models in precision position estimation.
Common models for the atmosphere include the global pressure and temperature
model and the global mapping function (GPT/GMF) (Boehm et al. 2006b, 2007),
the Vienna Mapping Function (Boehm et al. 2006a, (VMF1)), or GPT2 (Lagler et al.
2013). If no ionosphere-free linear combinations of the carrier phase observables are
used (e.g., Hoffmann-Wellenhof et al. 2008; Misra and Enge 2011; Bock and Melgar
2016), ionospheric corrections from NASA’s Jet Propulsion Laboratory (JPL) or the
International GNSS Service (IGS) can be applied. Furthermore, to achieve precise
positioning solutions, solid Earth tide and ocean tidal loading (OTL) effects, driven
by the gravitational attraction of Sun and Moon, must be considered. As summarized
by Bock and Melgar (2016), solid Earth tides can reach up to 1 m and the elastic
response of the Earth to OTL can achieve 10 cm. While models applied in the
processing can especially remove the solid Earth effects (e.g., International Earth
Rotation and Reference Systems 2010 conventions Gérard and Luzum 2011), great
care must be taken when correcting for OTL as the coefficients (e.g., FES2004,
Lyard et al. 2006) must be inferred for the same reference system as orbit parameters
if estimated separately (e.g., center of mass of the solid Earth, CE versus center of
mass of the Earth system CM, Fu et al. 2012); otherwise, systematic errors will be
introduced in the time series. Herring et al. (2016) provide a very detailed account
of the specifics of the EarthScope Plate Boundary Observatory (PBO, Silver et al.
1998, now called NOTA—Network of the Americas) double difference processing.

3.4.2 Precise Point Positioning

While the double difference approach works very well, the increasing number of
available GNSS stations results in significant computational burden. Approaches
to mediate this analyze smaller sized subnets individually and merge the resulting
solutions (Herring et al. 2016). This can be somewhat cumbersome.
A different strategy processes each station individually and is called precise point
positioning or PPP (Zumberge et al. 1997). The immediately obvious drawback
is that the satellite clock errors cannot be removed from the solution. Instead,
this approach requires external products to provide these corrections. While these
products require a network solution, the globally distributed network of stations can
26 R. Grapenthin

be much smaller than the number of stations we are usually generating positioning
solutions for (Zumberge et al. 1997). It is important that the models, including the
antenna phase center models, used in the generation of the external products (such as
clock corrections) are the same as the ones applied in the PPP positioning solutions
or the result may include systematic position errors. Ionospheric and tropospheric
delays and solid Earth and ocean tides can be mediated through the same models
described above for double differencing. Herring et al. (2016) describe in detail the
PPP processing strategy for the NOTA network.

3.5 Applications

Figure 3 shows examples of daily station positioning solutions for stations AC12
(UNAVCO Community 2008) on Chernabura Island and AC53 (UNAVCO Com-
munity 2006) in Willow, both in Alaska, in a local NEU coordinate system. Each
dot represents the position estimate for a single day, and uncertainties are shown as
gray bars (barely visible for AC12 due to the large displacements). The solutions
were generated by the National Science Foundation’s Geodetic Facility for the
Advancement of Geoscience (NSF-GAGE) facility as described by Herring et al.
(2016). I removed very few large outliers. The time series have the rigid plate motion
of the North American plate removed (NAM14). The linear trends in the horizontal
components are consistent with plate boundary processes: strain build-up due to the
convergence of Pacific and North American plates. In the stable plate interior, this
time series would be flat in the east and north components. Thus, when analyzed
over large regions, these long-term trends and their uncertainties can be estimated
and displayed in the form of velocity maps (e.g., Fig. 1). These highlight the spatial
coherence of, for instance, large-scale plate boundary processes. Notably, at AC12
(Fig. 3, left column), the linear trends in the horizontal components terminate in
three steps due to the 22 July 2020 M7.8 Simeonof megathrust earthquake (Crowell
and Melgar 2020; Liu et al. 2020; Ye et al. 2021; Xiao et al. 2021), the 19 October
2020 M7.6 Sand Point strike-slip earthquake (Grapenthin et al. 2024), and the 29
July 2021 M8.2 Chignik megathrust earthquake (e.g., Elliott et al. 2022; Liu et al.
2022). Clearly, AC12 was much closer to the slip regions of the Simeonof and Sand
Point earthquakes, and the much larger Chignik earthquake occurred further to the
east. AC53 (Fig. 3, right column) shows longer-term variations in the horizontal
component in addition to the south and down step due to the 2018 M7.1 Anchorage
earthquake (West et al. 2019). This multiyear motion opposing the interseismic
strain build-up is due to a slow slip event that released the energy equivalent of
an M7.5 earthquake between 2008 and mid-2013 with a fault slip rate of about
5 cm/yr (Fu and Freymueller 2013; Fu et al. 2015b). The vertical component shows
seasonal effects, annual cycles of upward motion in summer and fall and downward
motion during winter and spring. This is caused by the snow loading and unloading
in annual cycles with time-varying amplitudes due to variable precipitation. Short-
GNSS: Positioning, Velocities, and Reflections 27

Fig. 3 GPS time series in east, north, and vertical components for continuous GNSS station AC12
on Chernabura Island, Alaska, and AC53 in Willow, Alaska. The time series are in a stable North
America reference frame (NAM14, Herring et al. 2016). Data processed by the NSF-GAGE facility
as described by Herring et al. (2016). (left, AC12) The horizontal components show clear linear
trends consistent with NNW motion caused by the convergence of the Pacific and the North
American plates. Three steps are marked and indicate offsets due to M.w 7.8 Simeonof, M.w 7.6
Sand Point, and M.w 8.2 Chignik earthquakes. These manifest differently in the north, east, and
vertical components and clearly depend on distance to and mechanism of the earthquake. Very
clear after the M.w 7.8 Simeonof earthquake is also the onset of post-seismic deformation, here
labeled as afterslip, indicating slow stress-driven slip on the fault interface around the main rupture.
(right, AC53) Most obvious in the horizontal components is interseismic motion interrupted by a
multiyear slow slip event (Fu and Freymueller 2013), and the 2018 M7.1 Anchorage earthquake
(West et al. 2019). The vertical component shows the characteristic sinusoidal signal due to
winter snow loading (Heki 2001) in annual cycles with time-varying amplitudes due to variable
precipitation. We also see winter outliers that are processing artifacts caused by snow or ice on the
antenna impacting the signal travel path from satellite to antenna

term series of outliers in places with snow are often observed when the antenna is
covered in snow or ice.
It is clear from the figure that the time series are complex with multiple processes
superimposed on each other. This is after the well-understood processes, such as
Earth and ocean tides, have been removed. Identification, analysis, and modeling of
these processes have furthered our understanding of Earth processes significantly
over the last decades. With continuous time series now approaching, and in some
instances exceeding, 20 years, much more subtle and longer-term transient signals
will be resolved. In Fig. 3, for instance, we see clear evidence for post-seismic
deformation beginning after the Simeonof earthquake prior to the Sand Point
event. Right after the earthquake, this indicates rapid afterslip on parts of the fault
28 R. Grapenthin

Fig. 4 Vertical GNSS displacements across the Western United States from March 2011 to March
2014. Squares mark GNSS station locations with colors indicated recorded vertical displacements.
The gray patch shows the location of California’s Central Valley where stations were excluded as
they move in the opposite directions due to soil compaction during water extraction. (From: Borsa
et al. 2014)

surrounding the main slip area. After some time, visco-elastic relaxation of the upper
mantle will set in. For large earthquakes, this can continue for decades (e.g., Suito
and Freymueller 2009).
The recognition that seasonal signals in continuous GPS time series are linked
to precipitation came about in the early 2000s. Heki (2001) first connected the
seasonal sinusoids in the GPS observations of the dense, continuous GEONET
in Japan to the substantial snow loading in the mountains, which induces winter
subsidence and summer uplift of the elastic crust. With the establishment of broad
and dense continuous GNSS instrumentation in the Western United States through
the NOTA network, similar observations of seasonal loading and inference of the
respective water equivalent snow loads have been made there (e.g., Fu et al. 2012;
Argus et al. 2014; Fu et al. 2015a). Figure 4 shows a longer-term application of
these new hydrogeodetic observations as Borsa et al. (2014) visualize the drought
evolution in the Western USA with the vertical displacements from March 2011 to
March 2014. At the beginning of spring, we would expect widespread subsidence
due to snow loading as observed in 2011. However, starting in 2012 and very clear
in 2013 and 2014, the GNSS sites recorded increasing uplift consistent with lacking
precipitation. This uplift is particularly evident in the Sierra Nevada and Costal
Ranges of California, for which Amos et al. (2014) also showed long-term uplift due
to substantial long-term ground water extraction in the San Joaquin Valley (southern
part of gray patch in Fig. 4).

4 High-Rate and Real-Time Positioning

All of our discussion so far has focused on static positioning: finding the most
precise position of a station for a given day of data. Because of the high precision,
GNSS: Positioning, Velocities, and Reflections 29

time series of subsequent static solutions can reveal subtle, months-long processes
such as magma recharge into a volcanic edifice that induces less than 2 cm of
deformation (e.g., Dixon et al. 1997; Cervelli et al. 2006; Fournier et al. 2009;
Grapenthin et al. 2013). However, many processes, such as earthquakes or volcanic
eruptions, work on much shorter time scales and benefit from kinematic GNSS
solutions on subdaily time scales. Here, we can broadly distinguish two main
approaches (either work in network or PPP mode): post-processed epoch-by-epoch
solutions and true real-time processing. Before explaining these, I have to touch on
a peculiar pitfall that arises when converting between GPS and UTC times.

4.1 Leap Seconds

Something we could ignore so far, but have to pay close attention to now, is the
fact that GPS time is running ahead of UTC time by a time-variable number of
seconds. UTC time is adjusted to stay in tune with the Earth’s general slowdown in
rotation and rotational irregularities in order to keep it aligned with mean solar time
at Greenwich. To that end, leap seconds are occasionally introduced to UTC, 27
seconds since its establishment in 1970. GPS time, introduced in 1980, was initially
aligned with UTC time, but it does not have a requirement to maintain the same
alignment with Earth’s rotations. Hence, no leap seconds were ever introduced to
GPS time, which since January 2017 is now 18 seconds ahead of UTC time.
Why is this important? Most times, such as earthquake origin times (https://
earthquake.usgs.gov/earthquakes/map/), are given as UTC times. If any meaningful
comparison of GNSS signals and physical processes is desired, the two time systems
need to be aligned. This is easily achieved by subtracting the correct number of leap
seconds from the GPS timestamps. As leap seconds are introduced irregularly, this
requires a look-up table, however.

4.2 Epoch-by-Epoch Postprocessing

The kinematic use case of position estimation for each epoch of observations under
the assumption that the receiver is moving is not that different from the static use
case and is, in fact, one of the main application areas for GNSS positioning. In
geophysics, Nikolaidis et al. (2001) were among the first to show dynamic motion
due to a seismic wave using 30 s observations of the 1999 M.w 7.1 Hector Mine
earthquake. Larson et al. (2003) used 1 sps data to resolve dynamic motion induced
by the 2002 M.w 7.9 Denali earthquake, and Galetzka et al. (2015) resolved the
dynamic slip propagation along the fault during the M.w 7.8 Gorkha earthquake from
5 sps data.
The main complication that arises here is positioning noise due to signal mul-
tipath (e.g., Georgiadou and Kleusberg 1988, Fig. 5) and the resulting interference
30 R. Grapenthin

Fig. 5 Multipath geometry

where the ground is a
reflector. Antenna height and
satellite-elevation angle (e)
determine the additional path
length of the reflected signal
(red line) that result in
time-varying interference
with the direct signal (blue)
measured by the GPS
antennas. The inset shows
interference examples. (From:
Larson 2016)

of the direct and indirect signals at the antenna. This is a slowly varying process
with periods from minutes to hours and amplitudes on the order of centimeters
(Georgiadou and Kleusberg 1988). For the GPS constellation, these errors repeat
approximately every 11 hours and 58 minutes, which is one revolution period of the
GPS satellites. Multipath is generally not modeled and removed directly for static
positioning under the assumption that it averages out with the use of an entire day
of data and this average signal is repetitive on subsequent days (sometimes low-
elevation satellite data are masked out in particularly bad multipath environments).
For subdaily kinematic positioning, this assumption no longer holds and multipath
should be treated if the signal of interest is of similar magnitude. Since antenna
design cannot fully shield against surface reflected signals, several methods for
multipath reduction exist. Axelrad et al. (1996), for instance, propose a technique to
correct multipath in double differenced phase observations. On a time series level,
sidereal filtering is an effective technique where, for instance, low-pass-filtered
observations from the previous day (without a geophysical signal) are removed from
the current day with the noisy signal (e.g., Larson et al. 2007, Fig. 6). However, for
this to yield best results, the use of the correct orbit repeat times is recommended to
calculate the time shift before the signal subtraction (Choi et al. 2004; Agnew and
Larson 2006) as the GPS constellation does not exactly repeat after 11 hours and 58
minutes (Agnew and Larson 2006).

4.3 Real-Time Complications

Moving from the epoch-by-epoch postprocessing case to a real-time setting poses

a few additional challenges. Obviously, a hardened telemetry system is required to
deliver the data in a timely fashion with minimum data loss. This likely increases
GNSS: Positioning, Velocities, and Reflections 31

Fig. 6 Example of repeated noise due to multipath (Day 1 top trace, Day 2 third trace) in 1 Hz
position data. Forward shifting of the Day 1 data by the average orbit repeat time and low-pass
filtering yields the Day 1 LP trace, which is subtracted from Day 2 to result in the fourth trace, a
much cleaner time series. (From: Larson et al. 2007)

the power requirements, which poses additional engineering challenges at remote

locations.
On a technical level, however, the main requirement for real-time processing
is that it cannot fall behind the data delivery. This means, each epoch must be
processed before the next arrives. This puts an upper limit on any iterative processes
that run during position estimation, potentially reducing precision. Additionally, any
smoothing of the time series will have to be limited due to time constraints. Further,
for network processing, we are also somewhat limited by the number of sites that can
be processed simultaneously as the computational time increases by .N 3 (Zumberge
et al. 1997), where N is the number of stations. To avoid this while still benefiting
from the elimination of nuisance terms through double differencing, simple baseline
processing has proven adequate for many applications. Here, a single roving station
is processed relative to a base station that is assumed static. Since mathematically,
this is equivalent to subtracting motion at the base station from the roving station,
this relationship can be expressed in the mathematical models analyzing the
positioning data, and absolute positioning may not be required. Grapenthin et al.
(2014b) have demonstrated that this can be an effective real-time GNSS analysis
strategy for earthquake early warning in a crustal fault setting, and Grapenthin et al.
(2014a) (see Sect. 4.4) captured and analyzed the 2014 M.w 6.0 California Napa
strike-slip earthquake this way in real time. This strategy is applicable to many
processes that induce local to regional deformation, such as landslide monitoring
or volcanic deformation.
Large subduction zone earthquakes, on the other hand, could pose a significant
challenge for such a differential processing strategy as short, and in particular,
32 R. Grapenthin

trench parallel baselines may not resolve the translation of the entire network well
enough above the noise (see Sect. 4.4). Hence, these events are best dealt with
using a PPP processing scheme. Just like before, in the static positioning case,
this poses additional challenges as the satellite clock errors need to be corrected.
Troposphere and ionosphere delays also require dedicated treatment to achieve
positioning precision to centimeter levels. Real-time orbit information and clock
corrections are now distributed by the IGS to allow for PPP positioning in real
time (http://rts.igs.org). Commonly, these products are provided in 30 s resolution,
which we can interpolate due to the generally smooth nature of the variations.
Commercial solutions are available, too. Trimble, for instance, broadcasts globally
its own corrections and performs PPP positioning on site for its receivers to stream
centimeter-level positioning solutions from the field (Trimble RTX). Real-time
processing of a global network of stations may allow the dedicated user to generate
their own corrections, too. As these clock corrections are calculated and distributed
in real time, delays in the range of 10–20 s are to be expected for the IGS real-time
service (e.g., Hadas and Bosy 2015). Instantaneous velocity-derived displacements
(see Sect. 5) may be able to bridge this latency gap.
During our work on the real-time GNSS integration into an earthquake early
warning system (Grapenthin et al. 2014b), Ingrid Johanson (USGS-HVO) suggested
an experiment to test the real-time processing system with large antenna offsets. We
set up 7 tripods on the roof of McCone Hall, the building housing the Berkeley
Seismological Laboratory and the Department of Earth and Planetary Science, on
the University of California Berkeley campus (Fig. 7a). Four tripods had antennas
mounted and connected to GPS receivers. One setup served as base station (BASE),
and the other three were supposed to experience a rapid large offset. We carried
the antennas to the remaining three tripods that were 1, 3, and 6 m away from the
initial location (Fig. 7a). We collected data for an hour, ducked below the ground
plane of the antenna, and quickly carried it to the receiving tripod to collect data
for another hour. The processing was based on trackRT, the real-time component
of GAMIT/GLOBK (Herring et al. 2010), with parameter settings similar to the
real-time network running at the time (Grapenthin et al. 2014b). The generated
displacements for each station were relative to the static base station BASE.
The experiment yielded the key insight that real-time processing should not be
optimized to static observations. As the motion of the kinematic GPS stations is
modeled as a random walk, a key parameter is the process noise that constrains
how much random walk noise is assumed for a time step. Figure 7b shows a
comparison between post-processed “true” displacements (top), and low processing
noise (middle) and high process noise (bottom) real-time displacement time series
for the 6 m offset experiment. It is obvious that too tightly constrained process noise
results in a prolonged ramp rather than a near instantaneous step function. Notable
remains the “overshoot” in the high process noise time series (Fig. 7b, bottom). This
is due to the identification and insertion of cycle slips for all satellites at the time
of the offsets by the processing software. Figure 7c shows the results for all three
offsets with high process noise and (one form) of cycle slip detection suppressed.
While some satellites are still flagged with a cycle slip, enough remain to produce a
GNSS: Positioning, Velocities, and Reflections 33

Fig. 7 (a) Experiment setup, tripods without antennas are 1, 3, and 6 meters from origin tripods,
photo: R. √
Grapenthin. (b) Six-meter offset time series. Post-processed
√ on top, low process noise
(0.002 .m/ s) in the middle, high process noise (0.2 .m/ s) at the bottom. (c) All three distances
with high process noise and suppressed cycle slip detection. Time series courtesy of Ingrid
Johanson, USGS-HVO

well-constrained solution and clean step function matching the post-processed time
series.

4.4 Applications

In this section I present some recent examples to showcase the potential of high-rate
and real-time GNSS solutions to characterize, for instance, earthquakes or volcanic
eruptions.
First, I analyze 5 sps observations to generate kinematic time series for the M.w 7.8
Gorkha, Nepal, earthquake (Galetzka et al. 2015). I use final IGS orbits and IGS08
antenna phase center models (Dow et al. 2009), remove solid Earth and ocean tides,
and to model the atmosphere, I use the GPT2 global pressure and temperature model
(Lagler et al. 2013). I estimate the kinematic position solution as a random walk
process.
The resulting time series for the two sites KKN4 and NAST are shown in Fig. 8.
The time series begin at the earthquake origin time (2015-04-25 06:11:25 (UTC))
and have properly taken into account that GPS time was 16 seconds ahead of UTC in
2015. The data show very clearly that it takes more than 20 seconds for the S-wave
to travel from the hypocenter to these sites. NAST is obviously further away from
the hypocenter than KKN4, given the later arrival time of dynamic motion, and
shows much longer lasting dynamic excitation. This is due to its location within the
Kathmandu basin, which traps and amplifies seismic energy (Galetzka et al. 2015).
The very dense high-rate GNSS network of Japan captured the 2011 M.w 9.0
Tohoku-oki earthquake (Simons et al. 2011). Grapenthin and Freymueller (2011)
34 R. Grapenthin

Fig. 8 Displacements at two GNSS sites during the 25 April 2019 Gorkha earthquake in Nepal
(see Fig. 12 for location). The sites clearly show different responses. KKN4 is located on bedrock,
whereas NAST is located in the Kathmandu Basin, which traps significant amounts of seismic
energy, resulting in the long-term excitation at the site

animated the timeseries for each station in map view to visualize the spatial
propagation of the dynamic wavefield and the build-up of the static displacements
in both horizontal and vertical components. Figure 9 shows a snapshot about 3 min
30 seconds after the origin time. The co-seismic static offsets have fully developed
at this time, and the S-wave and surface waves are traversing southern and northern
Japan. Amazingly, these dynamic features can be highlighted if we subtract the final
static offsets from the top row in Fig. 9, which yields the bottom row, showing S-
wave and Love-wave patterns in the horizontal and Rayleigh wave in the vertical
component.
The 2014 M.w 6.0 South Napa earthquake in California was the first earthquake
captured and analyzed for slip distribution and magnitude in real time (Grapenthin
et al. 2014a). Figure 10 shows a double difference solution for this event. All data,
including the finite fault slip model, that went into this figure were in fact available
26 s after the event, as indicated in the figure. The first GNSS-based magnitude
solution was available 24 s after the event. Grapenthin et al. (2014a) reported that a
software error resulted in an artificial delay of the processing by 10 seconds. With
8 seconds S-wave travel time from the hypocenter to the nearest GNSS station and
6 seconds data acquisition and processing latency, the map and model shown in
Fig. 10 could have been available after just 14 seconds.
GNSS: Positioning, Velocities, and Reflections 35

Fig. 9 The elastic wave field about 3:30 minutes following the rupture initiation of the 11 March
2011 M.w 9.0 Tohoku-oki earthquake as sensed by about 1200 high-rate GNSS sites. The left panels
show horizontal displacements, and the right panels show vertical displacements. Lower row has
the permanent co-seismic displacements subtracted to highlight the dynamic features. The S-wave
packets are clearly visible in SE Japan in the horizontal panels, and the Love wave is slightly north
of that. In the vertical panels we can see the Rayleigh wave (after Grapenthin and Freymueller
2011)

More recently, the 2019 Ridgecrest earthquake sequence provided another test of
the capabilities of real-time GNSS for rapid earthquake analysis and early warning
in a densely instrumented region. Melgar et al. (2019) perform a detailed com-
parison of archived real-time and post-processed positioning solutions, concluding
that the real-time GNSS performance during the Ridgecrest sequence manifested
its utility as a monitoring and early warning tool. Hence, Melbourne et al. (2019)
36 R. Grapenthin

Fig. 10 2014 Napa earthquake real-time solution produced 26 s after origin time (event location
given by star). The thin black lines (left, middle panels) indicate the network of baselines for
which differential solutions were generated during this event. The left panel shows baseline offsets
adjusted relative to site P256 (large red dot). The blue vectors are static horizontal offsets from
rapid daily time series (courtesy of UNR). The red vectors indicate real-time offsets that are the
difference between about 5 min of averaged pre-event data and 3 s of averaged postevent data.
The middle panel shows the finite slip model (vertical fault is rotated into map view) at 26 s after
the event inferred from the red vector offsets in the left panel. White to yellow colored baselines
indicate model resulting misfit to the data. Pink colors indicate slip amplitude. The right panel
shows at the top the time series of GPS-inferred magnitude, and the black circle indicates an initial
seismic real-time estimate. The bottom four panels show north (blue) and east (black) displacement
time series for the bold, similarly colored baselines in the middle panel. Crosses mark the offsets
inferred along these baselines (time shift between GPS solutions and offsets is due to 6 s data
acquisition and processing latency). (From: Grapenthin et al. 2014a)

rightfully point out the significant potential of real-time GNSS to augment existing
hazard monitoring tools.
A point in case is the 2018 Kı̄lauea eruption in Hawaii, my last example. Neal
et al. (2019) present incredible GNSS time series, reproduced here in Fig. 11,
capturing the history of the caldera collapse at the summit. This is only rivaled
by observations of about 65 m of GNSS-observed total subsidence at the caldera
center during the 2015 Bardarbunga collapse (Gudmundsson et al. 2016). Kı̄lauea
GNSS stations NPIT and CALS provide a finely resolved history of collapse and
broadening of the affected area. Neal et al. (2019) count 62 discrete collapse events,
up to 8.5 m per event recorded at NPIT, resulting in a deepening of the caldera by
more than 500 m in places. These observations contribute to a detailed model of
onset of the caldera collapse (Anderson et al. 2019) and the geometry of caldera
bounding faults (Segall et al. 2019), thus helping to understand the mechanics
driving the collapse.
GNSS: Positioning, Velocities, and Reflections 37

Fig. 11 Timeline of the activity Kı̄lauea’s summit area during the 2018 eruption showing the GPS
displacements at NPIT, CALS, radial tilt at UWD, hourly earthquake counts and representative
pictures of the activity. The inset map shows station locations and caldera morphology, and HMM
marks Halema’uma’u crater. NPIT is located close to the initial collapse focus, and CALS is closer
to the edge of the new rim, experiencing significant deformation later and in a less discrete, step-
wise fashion. The tiltmeter spikes and spikes in earthquake counts show the near-daily collapse
events. (From: Neal et al. 2019)

5 Instantaneous Velocities

A different approach to utilize both phase and range observables is to use the
difference of subsequent measurements to infer instantaneous receiver velocities
(e.g., Grapenthin et al. 2018b), which is well known in geomatics (e.g., Misra
and Enge 2011; Gaglione 2015). Colosimo et al. (2011) first applied this method
to the geosciences and coined their algorithm VADASE (Variometric Approach
for Displacement Analysis Stand-alone Engine), which they used to estimate
earthquake waveforms and co-seismic displacements. As we will see below, this
technique, when applied to carrier phase observations, works at surprisingly high
precision even with just single-frequency observations. This is due to the elimination
of some of the error terms in the observation models (Eqs. 2 and 10). Assuming
continuous tracking (i.e., no cycle slips), the constant integer ambiguity, N, is
eliminated. If the observations are differenced over short time intervals, we can
also neglect the effects of ionospheric and tropospheric delays, and solid Earth
and ocean tides. Moreover, because of generally smooth satellite trajectories, we
can use broadcast orbits to remove the satellite velocity from the observations, and
thus we are left with the receiver velocity as the primary product (e.g., Misra and
Enge 2011; Grapenthin et al. 2018b). Parameswaran et al. (2023) show that these
GNSS velocities are almost identical to velocity records of colocated strong motion
instruments for the data from the 2021 M8.2 Chignik earthquake. Here, we follow
the derivation by Grapenthin et al. (2018b).
38 R. Grapenthin

5.1 Observation Model

An observation model can be derived by differentiating the phase observation

model (Eq. 10) with respect to time. In practice we can employ a finite difference
approximation using subsequent carrier phase observations, .ΔФs (Misra and Enge
2011; Gaglione 2015; Grapenthin et al. 2018b):

ΔФs = (vs − vu ) ∗ 1s + ḃ + δϵФ

. (21)

where .(vs −vu )∗1s is the change in range between satellite and receiver between the
two observations in terms of satellite velocity, .vs , and receiver velocity, .vu , which
the unit vector, .1s , projects onto the satellite–receiver line of sight. The difference of
satellite and receiver clock bias is captured by .ḃ, and .δϵФ represents the differenced
error terms. By using orbit information (broadcast or higher-quality post-processed
products), we can remove the average satellite velocity and its clock drift over the
two epochs. This yields the average Doppler shift, .D s , of satellite s induced by
receiver motion at velocity .vu :

D s = −1s ∗ vu + ḃu + δϵФ

. (22)

While some GNSS receivers produce this observable directly, tests show that it
is not as precise as the phase differenced approach (e.g., Gaglione 2015). If we have
such observations from at least 4 satellites, we again can set up a linear system of
equations:
⎡ ⎤
vu
.D = G + δϵФ (23)
ḃu

where .D is a vector of Doppler shift observations and .G is the system matrix that
contains unit vectors to project the receiver velocities .vu = [vx , vy , vz ]T onto the
line of sight to the satellite. We can now estimate receiver velocity, .vu , and receiver
clock bias, .ḃu , with standard least-squares techniques (e.g., Aster et al. 2018).
Note that Eq. 23 has no frequency dependency; it can contain single-frequency
observations or combinations of observables from different bands. The resulting
velocity will be relative to an Earth-centered Earth-fixed reference frame, which is
straightforward to rotate into a local north–east–up reference frame (see Sect. 3.1
and Bock and Melgar 2016).

5.2 Application

Figure 12 shows how instantaneous velocities can characterize an earthquake.

Here, we again use 5 Hz GNSS observations of the 2015 M.w 7.8 Gorkha, Nepal,
GNSS: Positioning, Velocities, and Reflections 39

83° 84° 85° 86° 87°

75 km
DNSG
CHLM

28° GRHI
KKN4
SYBC
NAST

SNDL

RMTE
27˚
north cm/s

20
10
0
−10
−20
0 125
seconds from OT

Fig. 12 North instantaneous GNSS velocities derived from 5 Hz GNSS following the M.w 7.8
Gorkha, Nepal, earthquake in 2015. Scale at the bottom is in cm/s; maximum velocities are on
the order of 80 cm/s. NAST in Kathmandu Basin shows clear basin reverberation effects compared
to the bedrock site KKN4

earthquake from Galetzka et al. (2015) to generate L2-only velocities from final
IGS orbits (Dow et al. 2009) and plot only the north component. Most striking
is the difference in response at KKN4 and CHLM, the two sites closest to the
epicenter (black star), compared to NAST. The latter is located in the Kathmandu
Basin and shows a significantly longer response due to seismic energy trapped and
reverberating within the basin.
To convince ourselves that the, in this case single-frequency, instantaneous veloc-
ities are a precise estimate of ground motion, we can compare them to the previously
generated (Fig. 8) post-processed kinematic positioning solutions at NAST and
KKN4 (Fig. 13). The east component of the instantaneous velocities is shown in
black in the left panels, while the red lines are the differentiated displacements from
Fig. 8. These two velocity records track each other very well. The main difference
is some higher- frequency noise in the single-frequency instantaneous velocities,
which results in the drift errors common to seismic data integrated to displacements.
We can see this in the right panels in Fig. 13 where we have done just this and
compare the integrated instantaneous velocities to GPS displacements. Fitting and
removing a polynomial from the time series up to the earthquake origin time allows
us to detrend and remove the drift from the integrated velocities such that these
displacements track the positioning derived displacements very well. Differencing
the averages over .∼20 seconds of pre-event and postevent values for integrated
velocities and absolute displacements yields misestimations of 8.2 and 14.1 cm of
horizontal and .−3.6 and .−11.0 cm of vertical displacements at NAST and KKN4,
40 R. Grapenthin

Fig. 13 Comparison of east velocities and displacements at stations NAST (top) and KKN4
(bottom) for the Gorkha earthquake (see Fig. 12). (left) L2 instavels (black) and differentiated
displacements inferred from positioning solutions (red) with excellent match. Maximum inferred
horizontal peak ground velocity (pgv.h , using east and north) and their time with respect to origin is
given in title bars. (right) L2 instavels integrated to displacements (black, fit to pre-earthquake drift
removed) and positioning solutions (red). Note different time scales between columns. Instavels
capture the characteristics of this earthquake very well without application of atmospheric or other
models

respectively, which are 6–8% of the error from the total co-seismic offset. This is
similar to the errors Melgar et al. (2019) observe when comparing real-time and
post-processed kinematic positioning solutions for the 2019 Ridgecrest event. It
is important to reiterate that the instantaneous velocities were derived without a
very precise a priori position, ionosphere, or troposphere corrections, ambiguity
resolution, and with just single-frequency data (albeit L2), while the absolute
displacements are a post-processed precision product (see Sect. 4). This quality is
achievable in real time, simple to implement for on-receiver processing, and with
potential for applications in hazard mitigation and characterization where change in
position is sufficient and absolute position time series are often not necessary.

6 GNSS Reflectometry

GNSS reflectometry is a remote sensing technique, developed over the last two
decades, that focuses on the analysis and interpretation of GNSS signals reflected off
of the Earth’s surface. Some initial applications related changes in correlation power
to sea surface roughness (Garrison et al. 1998), determined ocean wind speed and
direction (Armatys et al. 2000), or sea ice properties from the reflected signal power
GNSS: Positioning, Velocities, and Reflections 41

(Komjathy et al. 2000). While some of these applications used colocated upward
and downward facing GNSS antennas, more recent applications exploit direct and
indirect signal interference patterns for single, upward facing antenna installations,
hence enabling the use of these techniques at any suitable, existing GNSS site and all
of its historic data. These unconventional environmental sensing techniques provide
access to up to an .∼1000 m.2 footprint of observations around a GNSS antenna at
.∼2 m height.

Instead of using the phase and range observables discussed above, Larson et al.
(2008) suggest to use the signal-to-noise ratio (SNR, ratio of signal power to
measurement noise) observable routinely recorded by GNSS receivers. However,
SNR reporting across manufacturers is not standardized, and measurements are
sometimes discretized at 1 db-Hz intervals (Bilich et al. 2007), which can render the
data unsuitable for the methods described here. While the SNR measurements are
not useful for positioning, they record the same interference signal between direct
and indirect signal that is recorded in the other observables. Larson and Small (2014)
have demonstrated for the entire PBO GNSS network that interference between
direct and ground-reflected satellite signals offers a means to sense the environment
for changes in soil moisture (predominantly phase shift in interference pattern;
Larson et al. (2007)), vegetation height/water content (amplitude of interference
pattern smaller with higher vegetation; Small et al. (2010)), as well as snow
depth or ocean tides (frequency change of the interference pattern, frequency
increases as antenna height decreases; Larson et al. (2009, 2013)). Compared to
conventional positioning, SNR interference observations can be analyzed along each
satellite ground track individually without requiring joint analysis of observations
for multiple satellites. Thus, the footprint sensed by an individual GPS station can
be visualized with antenna height and satellite-elevation -dependent Fresnel zones
(e.g., Figure 2 in Larson 2016). Soil moisture estimates are an average along the
satellite ground track and should be averaged over multiple satellite tracks.

6.1 SNR Model

The multipath interference between planar-reflected (Georgiadou and Kleusberg

1988) and direct SNR can be modeled as (Larson 2016)
⎧ ⎫
4π h
SNR = A(e) cos
. sin(e) + φ (24)
λ

where A is the amplitude, h is the height of the antenna phase center above the
ground, .λ is the signal wavelength (e.g., GPS L1 or L2), e is the satellite-elevation
angle, and .φ is the phase offset. After converting SNR observations from db-Hz into
a linear scale (volts/volts), a lower order polynomial is removed to detrend the data
(signal strength increases as satellites rise to nadir, Fig. 14). From the remaining
interference pattern, phase and amplitude can be calculated through least-squares
42 R. Grapenthin

Fig. 14 SNR data for GPS L1 (a) and GPS L2 (b). The magenta polynomial fit represents the
direct signal, which should be removed to utilize the SNR data for environmental sensing. The
multipath signal is strongest at low elevation angles (blue axes labels). If phase, amplitude, or
frequency of the multipath change over time, the properties of the reflector must have changed
accordingly. (From: Larson 2016)

estimation (e.g., Chew et al. 2016). Larson et al. (2008) observed strong correlations
between .φ and in situ soil moisture measurements, and Chew et al. (2014) showed
in a modeling study that phase, .φ, is most affected by soil moisture changes.
Chew et al. (2015) showed in a different modeling study that while soil moisture
also affects the interference signal amplitude, A, it is actually mostly affected by
vegetation permittivity and height, which also affects the phase. In summary, the
signal we are interested in is affected not only by soil moisture, but also surface
roughness (including slope) and vegetation. Chew et al. (2016) present an algorithm
for soil moisture retrieval from GPS SNR measurements that removes vegetation
effects by estimating vegetation impact from the signal amplitude, predicting its
impact on phase using the models from Chew et al. (2015), and removing it from the
measurements. Given the nature of the GPS signals, the soil moisture measurements
are sensitive to volumetric near-surface soil moisture to a depth of at most 5 cm,
important for driving evaporation.

6.2 Applications

The ability to infer changes in the reflector height from SNR changes has probably
gotten the most attention. The initial application was snow depth estimation (see
below), but since then additional processes changing the reflector height around the
GPS antenna have been explored such as decadal changes of permafrost (Liu and
Larson 2018) and notably ocean tides (Larson et al. 2013, 2017b) with an extension
to storm surges (Peng et al. 2019). The latter simply require a GNSS monument with
partial view of the ocean and SNR-based reflector height analysis only for satellite
signals from this azimuth. Since every satellite can be evaluated independently, this
is easily achievable. Measurements of sea level with GNSS are very attractive as
these are automatically tied into a global reference frame. The positioning solutions
provide any changes in land surface elevation in this reference frame, which allows
Random documents with unrelated
content Scribd suggests to you:
 CHAPTER II

THE WRECKED SCHOONER

The great boat lay almost against the road. As the buckboard sped
by she loomed above it in the gathering dusk, menacing and
mountainous. Her broken bowsprit swung over the wagon and
creaked in the breeze that had just sprung up. Directly below the
bowsprit was a carved figurehead, larger than life and clearly
outlined against the dull gray of the ship. Sea and rain had washed
away the figure’s paint and worn the wood bone-white. It
represented a demon nailed to the battered prow, its wide ugly grin
and blank eyes peering almost into Ann’s face as the buckboard
passed beneath. Ann was on the side of the wagon which was closer
and could have touched the face if she had reached out her hand to
do so. Helen gave a little shriek of fright at sight of the thing and
Ann felt the cry echoing in her brain as if she had been the one who
called out.

Instinctively she dodged back against Jo, and felt that his muscles
were tense against the tightened reins in his hands.

Jerry needed no urging; with his back flattened down he ran,

swinging his heavy feet swiftly as he mounted the hill toward the
house. Ann glanced up from the strong brown hands holding the
reins and saw that Jo was staring straight ahead as though he had
not looked at the figurehead as he went by and was determined not
to turn and look back at it afterward.

They were past, but as they went up the hill the evening wind
suddenly grew stronger and sighed through the weatherworn boards
that covered the schooner’s hull, and the rattling of their loose ends
was like the sound of clapping hands.
 What was this old boat, and why did it impress them so? And yet
Ann did not feel like asking Jo about it. She wished that her father
would say something to quiet this fear that had come over her so
suddenly. She never before had felt anything like this strange
impression that the schooner was more than just a plain ordinary
boat cast up on a narrow strip of beach.

As though Mr. Seymour had read her mind he asked Jo, “Where
did that schooner come from? She wasn’t here last summer when I
was down.”

“No, sir.” Jo had trouble in making his stiff lips move. “She came in
on a blizzard the winter past and stove up on the pond rocks.”

“Whose boat was she? What is her name?”

“She had no cargo on board,” said Jo slowly, as if he did not wish

to say anything about it. “She had no log either. And the waves were
so heavy that her name plate was gone and never came ashore.”

“But wasn’t there somebody on board to tell you who she was?”

“A man had no chance to live in the sea the day she came in,”
explained Jo. “Four of the crew were washed ashore the next day,
but they carried no papers and nobody claimed them. None of the
folks wanted to bury them down in the village churchyard so pop
and I put them up back of the barn where grandpop lies. It didn’t
seem right not to give them a bit of ground to lie in, even though we
didn’t know what brought them in here.”

Mrs. Seymour exclaimed indignantly, “I never heard of anything so

inhuman! Do you really mean that the people in the village refused
to bury those poor shipwrecked sailors in the cemetery? Jo! Not here
in a civilized land?”

“You couldn’t blame the folks,” apologized Jo.

 But evidently Mrs. Seymour was quite positive that she could, and
Ann agreed with her most thoroughly.

Jerry had stopped running. He was going uphill and besides they
were almost home now, but Jo had time to say, “Nobody ever
claimed the boat. I guess nobody owns her. And not even the sea
wants her you can make that out by the way it threw her away up
here by the road, just as if it wanted to be free of her. Only the flood
tides reach her now.”

They had reached the house as Jo talked, and he jumped down

from his seat with his face still grim and set. And then everything
changed, for the house door was flung open with a flood of
lamplight over the doorstep and there stood Fred Bailey, Jo’s father.

“Come right in,” he called, striding to meet them. “Don’t mind that
stuff, Mr. Seymour. We’ll take it in for you.”

Ann liked Fred Bailey almost as much as she had liked Jo. As soon
as she saw him standing there, tall and thin and gangling in his
rough clothes, a fisherman and a farmer, all thoughts of the strange
wrecked ship were forgotten. Here was some one who made her feel
at home, some one who was strong and trustworthy and honest as
the good brown earth and the mighty cliffs.

Mr. Seymour had rented the Bailey house and Jo and his father
had moved into the barn for the summer. So presently, when the
baggage had been brought in and when Mr. Bailey had shown Mrs.
Seymour where things were in the pantry and the kitchen and the
woodshed and where the linen and blankets were kept, he and Jo
went off to their summer quarters leaving the Seymours alone.

Provisions had been sent from the village store and Ann and her
mother found the shelves well stocked with all kinds of food, with
big barrels of sugar, flour, and potatoes stored under the shelf in the
pantry. After they had studied the workings of the kerosene stove
they cooked the first meal over it, and Ann loved just such an
opportunity to show how much she knew about cooking. Ben was
ready to admit that she could boil potatoes expertly when she didn’t
forget and let the water boil away. As there was plenty of water this
time, and as Mrs. Seymour knew how to cook the steak deliciously in
a hot pan, and as Fred Bailey had left them a batch of soft yellow
biscuits, the hungry travelers were very well off indeed this evening.

Mr. Seymour was already gloating over the work he meant to do

this summer. “That boat is a find I didn’t expect. I’ll start sketching
her the first thing in the morning. Just think of having a cottage with
a wrecked schooner right in the front yard.”

“I don’t like that boat,” said Helen. Her lips twisted as though she
were going to cry. “It has such big round eyes that stare at you.”

Her mother laughed. “You must have been sleepy when you
passed the boat. That was only the figure of a man cut out of wood.
The eyes didn’t belong to anybody who is actually alive.”

“I don’t know about that, mother,” Ben said soberly. “I saw the
eyes, too, and I was wide-awake, for I pinched myself to make sure.
Those eyes made little holes right through me when they looked
down at me. They were looking at me, really, and not at Helen.”

“They were looking at me!” Helen insisted. “And I don’t like that
ship! I want to go home to Boston.”

Mr. Seymour looked at her in astonishment. “Come, come, my

dear child, you mustn’t let a thing like that frighten you. It is strange
and grotesque but that only makes it more interesting. I’ll tell you
about figureheads. The sailors think of the ship’s figurehead as a
sort of guardian spirit that watches over the boat and protects it
during storms. Even if it were alive it wouldn’t hurt you because it
was created only to protect. But it isn’t alive, Helen, it is made out of
wood. I’ll go with all of you to-morrow and let you touch it and then
you will never be afraid of it again.”
 “Do they always put figureheads on big boats, father?” asked Ann.
She would not have been willing to admit that she, too, had those
eyes upon her and had thought they seemed very much alive.

“No, not always,” Mr. Seymour explained. “Sometimes the portion

over the cutwater of a ship is finished off with scrollwork, gilded and
painted. Modern steamers don’t have them now, very often, but the
deep-sea men who are on a sailing vessel months at a time like to
feel that they have a figurehead to watch and care for them while
they are asleep. The owners decide what it will be, and give
directions to the builders. That is, if they name a boat after a man
they will carve a statue of him for the bow, or else they will choose a
saint or an old-time god, like Neptune, who was once supposed to
rule over the sea. Sometimes they will have a mermaid, because
mermaids are gay and dancing and will make the ship travel more
swiftly; no sea could drown a mermaid. When a sailing ship makes a
safe passage through storm and peril and brings the sailors home
happy and well, they are very likely to believe that the figurehead
has had as much to do with it as the captain with his real knowledge
of navigation and charts.”

“It is a mascot, then?” said Ben.

“Yes, a sort of mascot,” his father assented. “And some of the old
figureheads are beautifully made, real works of art. When he retired,
many a sea captain took the figurehead from his ship and nailed it
over the door of his home, for he felt a real affection for it. Perhaps
he thought that since Neptune had taken such good care of the ship
at sea he was entitled to the same enjoyment and rest ashore that
the captain had earned.”

Mr. Seymour seemed to feel that everything was clear now, but
Ann was not satisfied.

“This ship did not get home safely,” she said in a half whisper.
 “No, it didn’t,” her father assented. He was perfectly frank in
admitting that even the best of figureheads failed when storms were
too heavy or when sailors made mistakes in calculating the force of
wind and currents. “But that would not be the fault of the
figurehead. I am sure we shall learn that the captain lost track of
where he was and came in too close to shore.”

Ann’s doubts showed in her face. “But the crew and cargo have
disappeared.”

“You mustn’t be superstitious, Ann. There is always a logical

explanation for everything that seems strange and unnatural. There
must be a good reason why that boat had no cargo and probably we
shall learn all about her this summer before we go back to Boston.
Some of the people about here may know more than they care to
admit and have purposely kept it secret from Jo and Mr. Bailey.”

“Wouldn’t it be fun if we could find out all about her!” Her father’s
calm confidence had reassured Ann; her father must be right and
she didn’t want to be silly and timid. Never before had she felt the
least bit afraid of anything.

Ben had been thinking. “Just exactly what does it mean to be

superstitious, dad?” he asked.

“If you try to make yourself believe that the wooden figure out
there is alive, or if you are willing to accept any one else’s belief in
such nonsense, you will be superstitious and not intelligent. For
instance, you may think you see something, or hear something, and
not be able to explain what it is immediately. If instead of working to
learn a true explanation you remember the incident as it first
impressed you——”

“Like thinking a mouse at night is a burglar,” Ann interrupted.

“That is it exactly,” said Mr. Seymour. “Take that figurehead of a

demon on the boat; we passed by it just at twilight when it couldn’t
be seen as plainly as in full sunlight, and because the face was
leaning toward us, with shadows moving over it, it gave you the
impression that the thing was alive and watching you. To-morrow
when the sun comes out you will go back to look at it and see that it
is only a wooden statue, while if we should go home to-night, as
Helen wishes, you children would remember it all your lives as
something evil. And in that case you would be permitting yourselves
to grow superstitious instead of taking this as an opportunity for the
exercise of honest thinking and intelligent observation.”

“Is Jo superstitious?” asked Ben abruptly.

“Jo is too sensible to be superstitious,” answered his father.

“But Jo is afraid of that boat! I saw his face when we went past.
And even Jerry was afraid. He ran.”

Mr. Seymour glanced quickly across the table to where his wife sat
between Ann and Helen. Ann saw the look that passed between him
and her mother and realized that they both were worried. They did
not want Helen and Ben to go on thinking about the boat, nor did
they want the children to know that they, too, had felt the
strangeness of that gray broken boat and that grinning face.

Ann believed with her father that this was nothing more than an
old wooden sailing vessel thrown on the shore by a great storm.
Where had it come from, and for what port was it bound? Where
were the families who were waiting for their men to come home to
them? Were there children who thought that their father would
come back in a few weeks, now that good weather had made the
seas safe? Were there mothers who believed that their sailor sons
would soon be home? How anxious they must be, waiting all this
time since last winter. Something ought to be done about letting
them know the truth. It was tragic, and it was romantic, too.

And if there was a mystery attached to the ship that mystery could
be explained by a detective or by any one else who had the courage
and determination to find out what was at the bottom of this
strangeness. Her father had said there was a reason for everything
that was queer and uncanny. If only she were brave enough to face
that grinning demon! Should she be sensible, or should she let
herself be weak and unintelligent? Intelligent, that was what father
wanted them all to be, it was his favorite expression, “Be intelligent.”

The others began to chatter about other things while they were
finishing supper and washing the dishes afterward, but although Ann
took part in the work and the jokes and laughter and all the
anticipations of a great time to-morrow, she could think in the back
of her mind of nothing but the ship. If Jo would help them, she and
Ben would try to find out all about the wreck. It would be much
more fun than hunting imaginary Indians and bears in the woods.

After supper had been cleared away and the sweet old kitchen put
in order, all the Seymours trooped through every room in the house,
patting the wide soft feather beds that stood so high from the floor
that a little flight of steps was needed to climb into them.

“A tiny stepladder beside my bed!” exclaimed Helen. “What fun! I

love this house.”

The unaccustomedness of the quaint old furniture, the wide floor

boards polished with age, the small-paned windows, the bulky
mahogany chests of drawers that smiled so kindly as they waited for
the children’s clothes to be unpacked, all these things crowded the
ship out of Helen’s mind. She went to bed perfectly happy.

“Don’t you fall out,” called Ben from his room, “because if you
should you’d break your leg, probably, you’re so high.”

“I couldn’t fall out,” Helen called back. “You wait until you try your
bed. It seemed high before I got in, but I sank away down and
down into a nest; I think I’ll pretend I am a baby swan to-night with
billows of my mother swan’s feathers all about me to keep me warm.
I never slept in such a funny bed, but I like it!”
 And then Helen’s voice trailed off into silence.

In each room the Seymours found a lamp trimmed and filled ready
for use, with its glass chimney as spotlessly clear as the glass of a
lighthouse.

“How kind the Baileys are!” exclaimed Mrs. Seymour gratefully. “I

don’t feel as if we were renting this house; Jo and his father seem
like old friends already.”

This time it was Ann and her father who exchanged a quick
glance, a flash of understanding and satisfaction. Impulsively Ann
threw her arms around her mother’s neck and kissed her. Her
mother should have a chance to rest here, if Ann’s help could make
it possible, dear mother who still looked so pale and tired after the
long weeks of nursing Helen and bringing her back to health.

“I knew that you’d like the Baileys,” said Mr. Seymour.

“Jo is an unusually nice boy, isn’t he, father?” Ann had already
grown attached to him.

“He certainly is,” Mr. Seymour agreed heartily. “And I know that
you will like him even better as you become better acquainted. His
father couldn’t get along without Jo. He does a man’s work on the
farm and helps bring in the lobsters every morning.”

“I’m going to be just like him,” Ben called from his bed in the next
room. Jo’s sturdy strength and the simple unconscious way the boy
used it had fired Ben’s imagination.

“Nothing could make me happier than to have you as well and

strong as he is, when we go away next fall,” answered Mr. Seymour.

With supper and the lamplight and the homely charm of the old
house, the atmosphere of uncanny strangeness had vanished, but
after Ann had blown out her lamp, just before she was ready to
climb the steps to her bed, she went to the window and peered
through the darkness toward the wrecked ship.

And as she looked a flickering light passed across the deck.

She must be mistaken. It was a firefly. No, there it was again, as

though a man walked carrying a swinging lantern with its wick no
bigger than a candle flame. He passed the bow, and the glow swung
across the figure of the demon.

Was it Jo or his father? That was Ann’s first thought, but she
wanted to make sure. From a second window in her room, across a
corner, she could see the windows of the barn which the Baileys had
made into a living room, and she leaned far out to see clearly. Jo
was there. He was talking to some one at the back of the room.

If Jo and his father were talking together, who could be prowling

around the boat? She crossed the room to look again at the
schooner. And as she watched, the bright pin prick of light
disappeared; the lantern had been carried behind some opaque
object that hid it.

“What’s up, Ann?” Ben stirred restlessly in the adjoining room. “It
will be morning before you get to bed.”

“Oh, I was looking out of the window. The stars are so bright in
Maine!”

“Ann! What do you think about that ship? I feel as if ghosts lived
on her.”

Ann climbed her little flight of steps and slid down between upper
sheet and feathers.

“Nonsense,” she called to Ben. “Ghosts don’t carry lanterns.”

 “What?” Ben’s voice sounded much more awake. “What did you
say, Ann?”

“I said I don’t believe in ghosts.”

Ann slid farther into her feather nest and promptly went to sleep.
 CHAPTER III

HOW THE BOAT CAME ASHORE

Vaguely Ann heard a bell ringing. She thought that she was
lobstering with Jo and that Jo was pulling up a bell in one of the
heavy lobster pots. They were bobbing about on waves as high as
mountains.

“It is seven o’clock! No farmer stays in bed late, you know.”

It was Mrs. Seymour’s voice.

How could her mother have come away out to sea? Ann sat up in
bed, not awake yet. And then she saw the sun pouring in through
the open windows. Her mother was standing in the hall between
Ann’s room and Ben’s, swinging an old ship’s bell that she must have
found somewhere in the house.

“In one minute, mother!”

How queer to wash in a huge bowl in her room instead of in a

bathroom! And how lovely to dry oneself while standing on a braided
mat before the washstand with the sun pouring down on one’s back
and legs! Bloomers and middy had miraculously appeared from her
baggage; some fairy had been at work while Ann was sleeping.

The smell of breakfast tweaked her hungry nose and she scurried
madly with her dressing, for Ben and Helen would eat everything in
sight if they felt half as starved as she did.

The kitchen seemed altogether different in the daytime. It had

grown smaller without the flickering shadows from the lamps. The
ceiling was low and Mr. Seymour bumped his head as he came
through the doorway; he would have to remember to stoop.

The big kitchen stove hummed merrily with the sweet smell of
wood smoke seeping up through the lids, a delicate fragrant thread
of gray that curled and disappeared. Mrs. Seymour explained that
Mr. Bailey built the fire for her; he had come early to show her how
to make it. Just as she spoke he appeared in the doorway again with
a foaming milk pail in his hand. His face was unsmiling but his blue
eyes were alight.

“So much milk for us?” inquired Mrs. Seymour.

“Drink it down, free as water,” he answered. “That’s what puts the

color in children’s cheeks. Get your milk pans ready.”

“Hello,” said Ann. “Isn’t this a fine morning?”

“Morning? Morning?” said Mr. Bailey. “This be the middle of the

forenoon.”

Ann saw that his eyes were laughing at her although his face
never moved a muscle. “What time is morning up here?” she
demanded.

“Oh—about half past three, these days. That’s dawn.”

“Do we have to get up at half past three?” cried Ben.

“Well, you do if you want to keep up with Jo,” answered his father.

“Where’s Jo now?” Ben asked, getting up from his chair.

“He’s hoein’ corn,” said Mr. Bailey. “Got two rows done already.
He’s not one to lie in bed, not Jo.”

“May I hoe with him? I’d like to, really.”

 Fred Bailey looked at Ben’s mother. She nodded permission and
Ben was off like a shot.

“Won’t you sit down and have a cup of coffee with us,” asked Mrs.
Seymour, “to celebrate our first morning?”

“I don’t know but what I might,” said Fred Bailey. “Only don’t
leave that pail o’ milk out there by the door for a minute.” And he
picked it up and handed it to Ann. “It’ll be tipped over the second
you take your eyes off it.”

“Your barn cats come over this far for milk?” inquired Mr. Seymour
laughing. “They can smell a good thing from a long distance.”

“It ain’t no cats that dump it out on me,” said Fred soberly. “And I
think that I’d better warn you, first thing. It’s the spirits, the spirits
from the ship. They pester me almost to death, dumping out the
milk from pails, and they tear up the packages left beside the door.
You don’t want to leave nothin’ about.”

“You think that ship is haunted?” Mrs. Seymour poured out a big
cup of coffee.

Helen had gone already and Ann hoped that neither of her parents
would notice that she had stayed. She made as little noise as
possible with the milk pans and then came and sat down quietly. She
saw her mother’s eye wander toward her but she smiled pleadingly,
hoping that her mother would know she could not be frightened by
any story about ghosts.

Fred was evidently glad to talk, once he had started on the

subject. “I shouldn’t wonder but what something was aboard that
boat that shouldn’t be there. I know this much—I’ve been bothered
uncommon ever since she came ashore, and not by human beings.”

“How did she happen to be wrecked?” Mr. Seymour was as eager

as Ann for the story, now that he felt sure that a story existed.
 “She struck last winter in January,” began Fred, settling himself
more comfortably in his chair. “It was during the worst storm we’ve
had in these parts in the last hundred years.”

“It must have been a howler,” commented Mr. Seymour.

Mr. Bailey nodded soberly. “You’re right, I never saw nothin’ like
it,” he said. “The storm had been brewing for days and we could feel
it coming long before it struck us up here; there was warning
enough in the Boston paper. Then the sea grew flat and shining
without a hint of a whitecap on her. The wind was so strong it just
pressed right down and smothered the waves, and it blew straight
off the land. It never let up blowing off the land all through the
storm, and that was one of the queer things that happened.

“We had three days o’ wind, and then the snow broke, all to once,
as though the sky opened and shook all its stuffing right out on us.
With the coming o’ the snow the wind eased up a bit an’ let the
water churn on the top of the sea until it was as white as the falling
snow. Finally I couldn’t tell where the water ended and the snow
began.

“The wind driving the sleet was cruel. Whenever Jo or I ventured

out it cut our faces and made them raw and bleeding. At times the
wind lifted the house right off its stone foundations and shook it,
and I feared it would be blown clear over the bluff and set awash in
the sea.”

“How terrible!” exclaimed Mrs. Seymour.

“It was all of that,” Fred agreed. “The second day of the snow I
thought the wind hove to a mite, it seemed more quiet. I went to
the window to see if the snow had let up. It had—but not in any way
I ever had seen it in all my fifty years of life on this bluff. It was as if
a path had been cut through the flying storm, straight and clear with
the wind sweeping through, so that I could see beyond the bluff
over the water. It was then I had my first glimpse of it, riding over
the waves and coming ashore dead against the gale. It was such a
thing as no mortal ever saw nowadays. I thought I was losing my
wits to see a boat coming toward me, riding in to shore against the
wind and while the tide was running out. I just couldn’t believe what
my eyes were telling me, for no boat that I ever heard tell of had
struck on this section of the coast. Nature built here so that they
can’t come in, what with Douglas Head stretching out to the north
and making a current to sweep wrecks farther down; they strike to
the north or the south of us, but never here.”

“To see a ship coming in and be powerless to help it!” exclaimed

Mr. Seymour as Fred paused for a sip of coffee and a bite of
doughnut. “There was nothing that you could do?”

“Not a thing. I was alone with Jo, and even if we had been able to
get out a small boat we couldn’t have done nothin’. She was coming
in too fast. So we bundled up, Jo and I, and went out to stand by on
the shore.”

“Into that storm?” Anne demanded. She had drawn close to her
mother’s chair during the story and now she stood tense against it.
She could almost see the two figures, Fred so tall and Jo a little
shorter, as they ventured out into the wind that threatened to blow
them into the water. How the cutting sleet must have hurt, and how
cold they must have been as they stamped their feet on the ice-
covered rocks and beat their hands to keep from freezing!

“Nothing else to do but try to save the men as they washed

ashore, now was there?” Fred asked gently, and Ann shook her
head. She knew that if she had been there she would have gone
with them and borne the cold as best she could.

“We waited and watched,” Fred continued. “And all that time the
narrow path stayed in the storm, swept clear of the driving snow.
And the boat came nearer with no sails set and on even keel. When
she struck she cried like a living thing.
 “We couldn’t see a man aboard. We waited all day and when night
closed in I sent Jo down to the village for help, and I listened alone
all night for the cry of some one washed to the beach; but no one
came.

“When dawn broke Jo came back with ten or twelve men. They
hadn’t known a thing about the wreck in the village nor we
shouldn’t, either, if it hadn’t been for that path in the storm; the
snow was falling too thick for any one to see through it. Well, that
morning the storm was over and the sun burst out. And there she
lay, almost as you see her now, but farther out. The water was
boiling all about her. The waves were crashing in pretty high but we
thought we could get one of the boats launched at the mouth of the
river and work it round to the ship. So we left Jo to watch the bluff
here and picked my dory to make the trip as she shipped less water
and rode the waves easier. We got her down the river and around
the point and after a couple of attempts we pulled in under the
schooner’s stern and three of us swung aboard while Les Perkins
and Pete Simonds held the dory.

“When we got on the schooner’s deck we found that the sea had
swept her clean of anything that might have identified her. The
name plates looked as if a mighty hand had wrenched them loose
and great cuts showed in the bow and stern where they had been.
There wasn’t a sound but the pounding of the waves along her side.
It made a queer sussh-sussh that didn’t seem to come from where
the water touched her. We broke open the hatches and went down
in her—two by two. Wasn’t a man of us who dast go down there
alone, for you never can tell what you’re going to find in a wrecked
ship’s cabin. We looked all about, but no one was in the place and I
don’t believe that any one was on her when she struck. The crew’s
quarters were in order but the cabin appeared as if there had been a
struggle there, though the sea might have done it, tossing things
about. Then we searched her careful but found no log nor no
papers. Some clothes were scattered here and there but the pockets
were empty and turned wrongside foremost. She had no cargo and
the fire was still a-going in the stove.”

Mr. Bailey had another cup of coffee and drank it silently while the
Seymours waited for the rest of the story.

“Well, that’s how she came in,” he said at last.

“But what makes you think there are spirits on board?” asked Mr.
Seymour. “There must have been something more than you have
told us, to make you believe that.”

“Yes, there is more to it,” admitted Fred, “but if I was to tell ye

you’d think me foolish.”

“We’d never think that, I can assure you,” said Mrs. Seymour
quickly. “If we had been with you on the schooner probably we
should be feeling exactly as you do about her.”

“Perhaps you might, and perhaps you might not. I would think
that the trouble was with me if it hadn’t been for the other men, but
every one of them down to the cove would back me up in what I
say. And I might as well tell you, because if I don’t some one else
will, no doubt.

“We had almost finished searching when I got a sort of feeling

that some one or something was peering at me. I kept looking
around behind me, and then I noticed that the other men were
doing the same thing. There was nothin’ there. We kind of looked at
each other and laughed at first. But soon it was all I could do to
keep from running around the next corner to catch whatever was
behind it. We did our search thorough, but I can tell you I was glad
when Les Perkins pulled the dory under the stern and I could drop
into her. None of us hankered to stay aboard that ship.”

In spite of herself Ann shivered and was glad when her mother
hugged her reassuringly.
 “Two days after that,” Fred continued, “we picked up four men
who had been washed in by the sea. We are God-fearing people up
here and I couldn’t understand why the folks in the village wouldn’t
put those sailors in the churchyard, but some of the people were
foolish and said those men should not be put in consecrated ground,
coming out of the sea like that. I didn’t know quite what to do, and I
suppose I should have taken them out and put them back into the
sea, the way most sailormen are done by when they’re dead. But I
didn’t decide to do that way; I buried them with my own people,
yonder in the field, and they lie there marked by four bits of
sandstone.

“Jo and I have been back on the boat several times, for we felt we
had a duty by her, lying at our door as she does, but we can’t find a
trace of anything to identify her and we both had that feeling that
something there is wrong. Something was watching us all the time
we were on her. So I’ve given up trying to think where she came
from or who sailed on her, for such things a man like me is not
supposed to know. Spirits from the sea no doubt came on board
during the storm and threw the crew overside. But if those spirits
are there now I don’t understand why the sea don’t claim her and
break her up. Sea seems to be shoving her back on the land as
though it wanted to be rid of her.”

“That is a great story, Fred,” said Mr. Seymour. “And I can

sympathize with the way you felt; it must have taken a great deal of
courage to go back to her when you and Jo looked her over. And you
have never seen anything move on the boat?”

Ann wanted to tell about the light she had seen there last night,
but that was her discovery and she so hoped to be the one to solve
the mystery! She said not a word about it.

“Nary a sight of anything have we ever had,” Fred answered.

Welcome to our website – the perfect destination for book lovers and
knowledge seekers. We believe that every book holds a new world,
offering opportunities for learning, discovery, and personal growth.
That’s why we are dedicated to bringing you a diverse collection of
books, ranging from classic literature and specialized publications to
self-development guides and children's books.

More than just a book-buying platform, we strive to be a bridge

connecting you with timeless cultural and intellectual values. With an
elegant, user-friendly interface and a smart search system, you can
quickly find the books that best suit your interests. Additionally,
our special promotions and home delivery services help you save time
and fully enjoy the joy of reading.

Join us on a journey of knowledge exploration, passion nurturing, and

personal growth every day!

ebookbell.com