Download the full version of the textbook now at textbookfull.

com

Geospatial Technologies for All Ali

Mansourian

https://textbookfull.com/product/geospatial-
technologies-for-all-ali-mansourian/

Explore and download more textbook at https://textbookfull.com

 Recommended digital products (PDF, EPUB, MOBI) that
you can download immediately if you are interested.

Geospatial Technologies for Urban Health Yongmei Lu

https://textbookfull.com/product/geospatial-technologies-for-urban-
health-yongmei-lu/

textbookfull.com

Geospatial Technologies for Crops and Soils Tarik Mitran

https://textbookfull.com/product/geospatial-technologies-for-crops-
and-soils-tarik-mitran/

textbookfull.com

Geospatial Technologies for Land Degradation Assessment

and Management R S Dwivedi

https://textbookfull.com/product/geospatial-technologies-for-land-
degradation-assessment-and-management-r-s-dwivedi/

textbookfull.com

Luhmann and Socio Legal Research An Empirical Agenda for

Social Systems Theory 1st Edition Celso Fernandes
Campilongo (Editor)
https://textbookfull.com/product/luhmann-and-socio-legal-research-an-
empirical-agenda-for-social-systems-theory-1st-edition-celso-
fernandes-campilongo-editor/
textbookfull.com
Language Assessment: Principles and Classroom Practices
3rd Edition H. Douglas Brown

https://textbookfull.com/product/language-assessment-principles-and-
classroom-practices-3rd-edition-h-douglas-brown/

textbookfull.com

The Cambridge Handbook of Intercultural Communication

Cambridge Handbooks in Language and Linguistics Guido
Rings
https://textbookfull.com/product/the-cambridge-handbook-of-
intercultural-communication-cambridge-handbooks-in-language-and-
linguistics-guido-rings/
textbookfull.com

The Theatre Experience 14th Edition Edwin Wilson

https://textbookfull.com/product/the-theatre-experience-14th-edition-
edwin-wilson/

textbookfull.com

Money: Understandings and Misunderstandings Mohammad

Ashraf

https://textbookfull.com/product/money-understandings-and-
misunderstandings-mohammad-ashraf/

textbookfull.com

China s New Navy 1st Edition Xiaobing Li

https://textbookfull.com/product/china-s-new-navy-1st-edition-
xiaobing-li/

textbookfull.com
Brutal The Driven World 1st Edition S. Valentine

https://textbookfull.com/product/brutal-the-driven-world-1st-edition-
s-valentine/

textbookfull.com
Lecture Notes
in Geoinformation and Cartography

Ali Mansourian · Petter Pilesjö

Lars Harrie · Ron van Lammeren Editors

Geospatial
Technologies
for All
Selected Papers of the 21st AGILE
Conference on Geographic Information
Science
Lecture Notes in Geoinformation
and Cartography

Series editors
William Cartwright, Melbourne, Australia
Georg Gartner, Wien, Austria
Liqiu Meng, München, Germany
Michael P. Peterson, Omaha, USA
The Lecture Notes in Geoinformation and Cartography series provides a contempo-
rary view of current research and development in Geoinformation and Cartography,
including GIS and Geographic Information Science. Publications with associated
electronic media examine areas of development and current technology. Editors from
multiple continents, in association with national and international organizations and
societies bring together the most comprehensive forum for Geoinformation and
Cartography.
The scope of Lecture Notes in Geoinformation and Cartography spans the range
of interdisciplinary topics in a variety of research and application fields. The type of
material published traditionally includes:
• proceedings that are peer-reviewed and published in association with a
conference;
• post-proceedings consisting of thoroughly revised final papers; and
• research monographs that may be based on individual research projects.
The Lecture Notes in Geoinformation and Cartography series also includes various
other publications, including:
• tutorials or collections of lectures for advanced courses;
• contemporary surveys that offer an objective summary of a current topic of
interest; and
• emerging areas of research directed at a broad community of practitioners.

More information about this series at http://www.springer.com/series/7418

Ali Mansourian ⋅ Petter Pilesjö
Lars Harrie ⋅ Ron van Lammeren
Editors

Geospatial Technologies
for All
Selected Papers of the 21st AGILE
Conference on Geographic Information
Science

123
Editors
Ali Mansourian Lars Harrie
Department of Physical Geography Department of Physical Geography
and Ecosystem Science/GIS Centre and Ecosystem Science/GIS Centre
Lund University Lund University
Lund Lund
Sweden Sweden

Petter Pilesjö Ron van Lammeren

Department of Physical Geography Laboratory of Geo-information Science
and Ecosystem Science/GIS Centre and Remote Sensing
Lund University Wageningen University and Research
Lund Wageningen, Gelderland
Sweden The Netherlands

ISSN 1863-2246 ISSN 1863-2351 (electronic)

Lecture Notes in Geoinformation and Cartography
ISBN 978-3-319-78207-2 ISBN 978-3-319-78208-9 (eBook)
https://doi.org/10.1007/978-3-319-78208-9
Library of Congress Control Number: 2018935230

© Springer International Publishing AG, part of Springer Nature 2018

This work is subject to copyright. All rights are reserved 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, express 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.

Printed on acid-free paper

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

The Association of Geographic Information Laboratories in Europe (AGILE) has

held annual conferences focusing on Geographic Information Science for more than
two decades. The themes of the conferences have varied in response to changing
research trends. In this way, the annual conference has always remained relevant
and current and is an important meeting point for European as well as
non-European researchers in the field. The annual AGILE conference is now widely
held to be the most important academic Geographic Information Science conference
in Europe.
The 21st AGILE conference took place in Lund, Sweden, on 12–15 June, 2018.
The theme of the conference was Geospatial Technologies for All. The conference
accepted full papers, short papers and posters and included a workshop day prior
the main conference. For more than a decade, the full papers from the annual
conference have been published as a book by Springer International Publishing AG.
This year there were 46 full papers submitted to the conference, of which 19 were
accepted for publication in this volume, its title taken from the conference theme.
We would like to thank all of the authors for their contributions, which are a critical
prerequisite for maintaining the quality of AGILE conference. We would also like
to thank all the reviewers in the Scientific Committee for carefully reviewing the
papers, which enabled us to make this selection.
The papers in this volume are divided into four parts. Part I deals with Natural
Resource Management and Earth Observation and includes five studies about how
geospatial technologies can contribute to a sustainable development of our envi-
ronment. During recent years’ AGILE conferences, there have been many sub-
missions in the field of Volunteered Geographic Information (VGI) and
Participatory GIS. This trend has continued this year, and the six papers in part two
of this volume represent that work. Part III is called Spatiotemporal Data Modelling
and Data Mining, and it includes six papers that describe new methods for
understanding spatiotemporal phenomena. A notable reflection here is the
increasing importance of the temporal dimension, as seen in several submissions;
GIScience is about more than just space. The final Part IV of this volume deals with
the issue of Quality and Uncertainty of Geographic Information, a key issue for

v
vi Preface

everyone that wants to make a decision based on spatial analysis. To sum up, we
think that the 19 contributions published in this volume give a broad overview
of the field at this time and do a superb job in exemplifying the title of this book:
Geospatial Technologies for All.
We are grateful to everyone that has helped us with editing this volume, as well
as creating the programme for the 21st AGILE conference. The AGILE Council
was of key assistance in the scientific organization of the conference. We also
would like to thank our colleagues at Lund University for all their help with the
various large and small tasks that enable a conference like this to succeed. We
would also like to thank Springer International Publishing AG for their helpful
cooperation, and for continuing to provide the opportunity to publish the selected
full papers in their academic series, Springer Lecture Notes in Geoinformation and
Cartography.

Lund, Sweden Ali Mansourian

Lund, Sweden Petter Pilesjö
Lund, Sweden Lars Harrie
Wageningen, The Netherlands Ron van Lammeren
February 2018
Organizing Committee

Scientific Programme Committee

Ali Mansourian, Lund University, Sweden (Chair)

Petter Pilesjö, Lund University, Sweden
Lars Harrie, Lund University, Sweden
Ron van Lammeren, Wageningen University and Research

Local Organizing Committee

Petter Pilesjö (Chair)

Lars Harrie (Workshop Chair)
David Tenenbaum
Eva Andersson
Andreas Persson
Roger Groth
Karin Larsson
Micael Runnstrom
Mitch Selander

Scientific Committee

Ana Paula Afonso, University of Lisbon, Portugal

Fernando Bacao, New University of Lisbon, Portugal
Marek Baranowski, Institute of Geodesy and Cartography, Poland
Melih Basaraner, Yildiz Technical University, Turkey
Giedrė Beconytė, Vilnius University, Lithuania

vii
viii Organizing Committee

Itzhak Benenson, Tel Aviv University, Israel

Lars Bernard, Technical University Dresden, Germany
Michela Bertolotto, University College Dublin, Ireland
Ralf Bill, Rostock University, Germany
Sandro Bimonte, IRSTEA, France
Thomas Blaschke, University of Salzburg, Austria
Arnold Bregt, Wageningen University and Research, The Netherlands
Thomas Brinkhoff, Jade University Oldenburg, Germany
Pedro Cabral, New University of Lisbon, Portugal
Sven Casteleyn, University Jaume I of Castellon, Spain
Christophe Claramunt, Naval Academy Research Council, France
Serena Coetzee, University of Pretoria, South Africa
Lex Comber, University of Leeds, UK
Joep Crompvoets, KU Leuven, Belgium
Isabel Cruz, University of Illinois at Chicago, USA
Sytze de Bruin, Wageningen University and Research, The Netherlands
Cidalia Fonte, University of Coimbra, Portugal
Anders Friis-Christensen, European Commission, Joint Research Centre, Italy
Jerome Gensel, University of Grenoble, France
Michael Gould, Esri and University Jaume I, Spain
Carlos Granell, University Jaume I of Castellón, Spain
Henning Sten Hansen, Aalborg University, Denmark
Lars Harrie, Lund University, Sweden
Francis Harvey, University of Leipzig, Germany
Roberto Henriques, New University of Lisbon, Portugal
Gerard Heuvelink, Wageningen University and Research, The Netherlands
Stephen Hirtle, University of Pittsburgh, USA
Hartwig Hochmair, University of Florida, USA
Joaquín Huerta, University Jaume I of Castellon, Spain
Bashkim Idrizi, Mother Teresa University, Republic of Macedonia
Mike Jackson, University of Nottingham, UK
Bin Jiang, University of Gävle, Sweden
Didier Josselin, University of Avignon, France
Derek Karssenberg, Utrecht University, The Netherlands
Tomi Kauppinen, Aalto University, Finland
Marinos Kavouras, National Technical University of Athens, Greece
Dimitris Kotzinos, University of Cergy-Pontoise, France
Petr Kuba Kubicek, Masaryk University, Czech Republic
Patrick Laube, Zurich University of Applied Science, Switzerland
Robert Laurini, University of Lyon, France
Francisco J. Lopez-Pellicer, University of Zaragoza, Spain
Malgorzata Luc, Jagiellonian University, Poland
Ali Mansourian, Lund University, Sweden
Bruno Martins, University of Lisbon, Portugal
Filipe Meneses, University of Minho, Portugal
Organizing Committee ix

Peter Mooney, Maynooth University, Ireland

João Moura Pires, New University of Lisbon, Portugal
Beniamino Murgante, University of Basilicata, Italy
Javier Nogueras-Iso, University of Zaragoza, Spain
Juha Oksanen, Finnish Geospatial Research Institute, Finland
Toshihiro Osaragi, Tokyo Institute of Technology, Japan
Frank Ostermann, University of Twente, The Netherlands
Volker Paelke, Hochschule Ostwestfalen-Lippe, Germany
Marco Painho, New University of Lisbon, Portugal
Petter Pilesjö, Lund University, Sweden
Poulicos Prastacos, FORTH, Greece
Hardy Pundt, Harz University of Applied Sciences, Germany
Ross Purves, University of Zurich, Switzerland
Viktor Putrenko, National Technical University of Ukraine, Ukraine
Martin Raubal, ETH Zürich, Switzerland
Wolfgang Reinhardt, Bundeswehr University Munich, Germany
Claus Rinner, Ryerson University, Canada
Jorge Rocha, University of Minho, Portugal
Armanda Rodrigues, New University of Lisbon, Portugal
Maribel Yasmina Santos, University of Minho, Portugal
Tapani Sarjakoski, Finnish Geospatial Research Institute, Finland
L. Tiina Sarjakoski, Finnish Geospatial Research Institute, Finland
Sven Schade, European Commission—DG JRC, Belgium
Christoph Schlieder, University of Bamberg, Germany
Monika Sester, Leibniz University of Hanover, Germany
Takeshi Shirabe, Royal Institute of Technology, Sweden
Jantien Stoter, Delft University of Technology, The Netherlands
Maguelonne Teisseire, IRSTEA, France
Fred Toppen, Utrecht University, The Netherlands
Nico Van de Weghe, Ghent University, Belgium
Ron van Lammeren, Wageningen University and Research, The Netherlands
Jos van Orshoven, KU Leuven, Belgium
Danny Vandenbroucke, KU Leuven, Belgium
Lluis Vicens, University of Girona, Spain
Luis M. Vilches-Blázquez, Pontifical Xavierian University, Spain
Kirsi Virrantaus, Aalto University, Finland
Vít Voženílek, Palacky University Olomouc, Czech Republic
Monica Wachowicz, University of New Brunswick, Canada
Gudrun Wallentin, University of Salzburg, Austria
Robert Weibel, University of Zurich, Switzerland
Stephan Winter, University of Melbourne, Australia
F. Javier Zarazaga-Soria, University of Zaragoza, Spain
Alexander Zipf, Heidelberg University, Germany
Jonas Ardö, Lund University, Sweden
Dirk Burghardt, TU Dresden, Germany
Visit https://textbookfull.com
now to explore a rich
collection of eBooks, textbook
and enjoy exciting offers!
x Organizing Committee

Cécile Duchêne, IGN, France

Lars Eklundh, Lund University, Sweden
Sara Irina Fabrikant, University of Zurich, Switzerland
Jan-Henrik Haunert, University of Bonn, Germany
Fredrik Lindberg, Gothenburg University, Sweden
Martijn Meijers, Delft University of Technology, Netherlands
Peter van Oosterom, Delft University of Technology, Netherlands
Andreas Persson, Lund University, Sweden
Micael Runnström, Lund University, Sweden
David Tenenbaum, Lund University, Sweden
Contents

Part I Natural Resource Management and Earth Observation

Evaluating Spatial Data Acquisition and Interpolation
Strategies for River Bathymetries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Robert Krüger, Pierre Karrasch and Lars Bernard
PLANTING: Computing High Spatio-temporal Resolutions
of Photovoltaic Potential of 3D City Models . . . . . . . . . . . . . . . . . . . . . . 27
Syed Monjur Murshed, Amy Lindsay, Solène Picard
and Alexander Simons
Interpolation of Rainfall Through Polynomial Regression
in the Marche Region (Central Italy) . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
Matteo Gentilucci, Carlo Bisci, Peter Burt, Massimilano Fazzini
and Carmela Vaccaro
An Artificial Stream Network and Its Application on Exploring
the Effect of DEM Resolution on Hydrological Parameters . . . . . . . . . . 75
Haicheng Liu
Spatio-Temporal Analysis of Mangrove Loss in Vulnerable
Islands of Sundarban World Heritage Site, India . . . . . . . . . . . . . . . . . 93
Biswajit Mondal and Ashis Kumar Saha

Part II Volunteered Geographic Information and Participatory GIS

3D Georeferencing of Historical Photos by Volunteers . . . . . . . . . . . . . . 113
Timothée Produit and Jens Ingensand

xi
xii Contents

Patterns of Consumption and Connectedness in GIS

Web Sources . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
Andrea Ballatore, Simon Scheider and Rob Lemmens
Charting the Geographies of Crowdsourced Information
in Greater London . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
Andrea Ballatore and Stefano De Sabbata
Supporting the Monitoring of Cheetahs in Kenya
by Mobile Map Apps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
Jan Jedersberger, Christian Stern, Mary Wykstra and Gertrud Schaab
Introducing Spatial Variability to the Impact Significance
Assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189
Rusne Sileryte, Jorge Gil, Alexander Wandl and Arjan van Timmeren
Development of System for Real-Time Collection, Sharing,
and Use of Disaster Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211
Toshihiro Osaragi and Ikki Niwa

Part III Spatiotemporal Data Modelling and Data Mining

A Top-Down Algorithm with Free Distance Parameter
for Mining Top-k Flock Patterns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233
Denis Evangelista Sanches, Luis O. Alvares, Vania Bogorny,
Marcos R. Vieira and Daniel S. Kaster
Optimization and Evaluation of a High-Performance
Open-Source Map-Matching Implementation . . . . . . . . . . . . . . . . . . . . . 251
Karl Rehrl, Simon Gröchenig and Michael Wimmer
A Personalized Location-Based and Serendipity-Oriented
Point of Interest Recommender Assistant Based on Behavioral
Patterns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 271
Samira Khoshahval, Mahdi Farnaghi, Mohammad Taleai
and Ali Mansourian
Identifying and Interpreting Clusters of Persons with Similar
Mobility Behaviour Change Processes . . . . . . . . . . . . . . . . . . . . . . . . . . 291
David Jonietz, Dominik Bucher, Henry Martin and Martin Raubal
Mixed Traffic Trajectory Prediction Using LSTM–Based
Models in Shared Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 309
Hao Cheng and Monika Sester
Contents xiii

A Framework for the Management of Deformable

Moving Objects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327
José Duarte, Paulo Dias and José Moreira

Part IV Quality and Uncertainty of Geographic Information

A Positional Quality Control Test Based on Proportions . . . . . . . . . . . . 349
Francisco Javier Ariza-López, José Rodríguez-Avi
and Virtudes Alba-Fernández
Animation as a Visual Indicator of Positional Uncertainty in
Geographic Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 365
Carsten Keßler and Enid Lotstein
 Part I
Natural Resource Management and Earth
Observation
Evaluating Spatial Data Acquisition
and Interpolation Strategies for River
Bathymetries

Robert Krüger, Pierre Karrasch and Lars Bernard

Abstract The study implements a workflow to evaluate the effects of different data
sampling methods and interpolation methods, when measuring and modelling a
river bathymetry based on point data. Interpolation and sampling strategies are
evaluated against a reference data set. The evaluation of the results includes criti-
cally discussing characteristics of the input data, the used methods and the trans-
ferability of the results. The results show that the decision for or against a particular
sampling method and for a specific setting of the parameters can certainly have a
great influence on the quality of the interpolation results. Further, some general
guidelines for the acquisition of bathymetries are derived from the study results.

Keywords Spatial interpolation ⋅

Riverbed modelling ⋅ Spatial sampling
Water frame work directive ⋅
Bathymetry

1 Introduction and Motivation

It is almost two decades after the European Water Frame Directive (WFD) entered
into force and required the European Commission and the Member States to
develop a monitoring and reporting system to provide evidence on the progress
made towards sustainable water use and (long term) protection of the available
water resources in Europe (European Union 2000). Geoinformation technologies
have always been core to the improvement of the data acquisition processes as well
as to the data processing and integration efforts. Today, the Water Information
System for Europe (WISE1) provides the European overview on the current state of
the WFD implementation and progress towards the general WFD goal that all
surface waters and groundwaters will be in good ecological status by 2027.

1
http://water.europa.eu/.

R. Krüger (✉) ⋅ P. Karrasch ⋅ L. Bernard

Chair of Geoinformatics, Technische Universität Dresden, Dresden, Germany
e-mail: robert.krueger@tu-dresden.de

© Springer International Publishing AG, part of Springer Nature 2018 3

A. Mansourian et al. (eds.), Geospatial Technologies for All,
Lecture Notes in Geoinformation and Cartography,
https://doi.org/10.1007/978-3-319-78208-9_1
4 R. Krüger et al.

The assessment of the ecological status of the inland water builds on a compi-
lation of indicator sets about the biological, chemical and hydro morphological
states of the water resources. Considering recent developments in geodata acqui-
sition and processing techniques, this paper proposes and evaluates new methods
for the acquisition of the hydro morphological characteristics of rivers, in order to
design more time and cost-efficient monitoring methods.
Assessing hydro-morphological characteristics of rivers requires information on
river depths and their variances in longitudinal and cross profiles of a river. For this
purpose, various methods for obtaining river bathymetry are available. Bathymetry
can be obtained by traditional survey or GPS methods if the river is (Milne and Sear
1997; Casas et al. 2006; Merwade et al. 2008). Alternatively, bathymetry can be
gathered by combining survey data collected during low water levels with sonar
data collected during high water levels (Casas et al. 2006). For shallow waters, it is
possible to obtain bathymetry data by airborne bathymetric LiDAR systems
(Hilldale and Raff 2008), while for deeper rivers multi-beam sonar is an option.
Both systems can generate high resolution bathymetries, but generate high costs
due to the expensive equipment (Conner and Tonina 2014). In the recent past,
various examples show that parameters to represent water body structures can be
obtained with inexpensive/low-cost unmanned boat systems (Fig. 1).
As such Systems typically only have single-beam echo sounders, which can
gather single depth information in the immediate environment of the boats or are
expanded by a step motor, which allows the echo sounder to turn the beam in one
direction (usually transversely to the direction of travel). However, all these
methods have in common that they do not allow a uniform and gapless recording of
the channel profile. Such a complete measurement can be used for determining the
depth variances. As the acquired datasets contain several gaps, spatial interpolation
methods are required to derive bathymetry coverage.
Current GIS and statistic packages offer number of well-established statistical
(e.g. Kriging) and non-statistical spatial (e.g. Inverse Distance Weighting, TIN)
interpolation methods to derive surfaces (Pebesma 2004). Additionally, considering
the different possible parameterizations for each of these methods, a plethora of
different interpolation results can be easily achieved. A qualitative assessment of the
results can be done on the basis of the known fundamental advantages and

Fig. 1 Unmanned boat systems carrying laser scanners, cameras (RGB, infrared, 360°) and echo
sounders (single beam)
Evaluating Spatial Data Acquisition and Interpolation … 5

disadvantages of the methods. A quantitative assessment, on the other hand, can

only be done using real measurements and comparing them with the interpolation
results. Additional to error analysis, Monte Carlo approaches or cross-validations
are often used as an instrument to assess the performance and stability of an
interpolation model (Pebesma 2004). With regard to the example of analyzing
channel profiles, recommendations should help in choosing the most appropriate
method, parameterization and the strategy for data sampling.
In a study testing several interpolation methods for calculating bathymetry from
cross-sections, Glenn et al. (2016) concluded that the result is mainly influenced by
the cross-section spacing (acquisition strategy) and the coordinate system used.
Earlier, Heritage et al. (2009) stated, that the layout of survey strategy is more
important than the choice of the interpolation methods. In a recent study, Santillan
et al. (2016) tested a small set of acquisition strategies for the interpolation of river
bathymetries and summarized that the survey configuration had a bigger effect on
the interpolation accuracy than the choice of interpolation method. Nevertheless,
the authors argued, that their results are limited to the survey configurations used,
the small study area and the low resolution of the data.

2 Implementing a Workflow to Compare Bathymetry

Acquisition and Interpolation Methods

The study presented here implements an evaluation workflow as depicted in Fig. 2.

Based on an available reference channel profile, different acquisition strategies are
simulated. The extracted depth information from the reference data set forms the

Fig. 2 Overview of the evaluation study workflow

6 R. Krüger et al.

input to different spatial interpolations. The interpolation results are evaluated

against the reference data set using different statistical measures.

2.1 Creation of the Reference Dataset

To do the qualitative and quantitate analyses of the different interpolation methods,

the points used for the interpolation are taken from a well-known surface. There are
two possibilities to obtain such a surface: First, using a measured (real) channel and
second by the creation of a synthetic channel. Both options have their pros and
cons. Taking measured channel as a reference provides the opportunity to use a
real-world scenario for benchmarking the different interpolation methods. On the
other hand, this somewhat limits the analysis about specific characteristics of a river
channel. The use of a synthetic channel offers the possibility to create specific
channel characteristics and to test the interpolation methods on a wider range.
Nevertheless, the conclusions drawn from such interpolations strongly rely on how
well the synthetic channel resembles the characteristics of a real one. Thus, for this
study, a dataset of real channel is used. By choosing different subsections of the
channel, the influence of different river parameters (sinuosity, depth variance) is
addressed.
The dataset used for this study contains bathymetric data of the Ohio River
(USA) and covers the first 200 km downstream from Pittsburgh (US Army Corps
of Engineers 2016). It results from several campaigns by ship-based sonar between
August 2006 and September 2010. The dataset contains point elevation data of the
channel in Cartesian x, y, h-coordinates, and is stored in several text-files. For the x,
y-coordinates the UTM zone 17N NAD83 is used as spatial reference (US Survey
feet). For the h-coordinate NAVD88 (US Survey feet) is used. Depending on the
subset, the density (grid with 5, 10, 20 ft spacing) as well as the layout of the points
(grid and cross-sections respectively) differs. Thus, considering the width of the
Ohio River channel in the study area (800–1300 ft) the data set provides a quite
high resolution.
The workflow to prepare the reference data has been implemented in ESRI
ArcGIS 10.3. In a first step, the areas with the highest point density (5 ft point
spacing) are selected from the dataset and imported into ArcGIS. After a visual
inspection, four subsets with different characteristics have been selected (Fig. 3).
While subsets 1, 2 and 4 are relatively straight (sinuosity about 1), subset 3 has a
higher sinuosity (sinuosity = 1.46). These subsets also show different ranges of
height variances, with 2.5 ft for subset 2–8.2 ft for subset 4.
As shown in prior works, a flow-oriented and curvilinear coordinate system is
expected to yield better results (Merwade et al. 2006; Glenn et al. 2016). Thus the
approach of Merwade et al. (2005) is adopted, which allows the transformation of
points from the cartesian x, y-coordinate system into the s, n-coordinate system. The
transformation process requires an arbitrary line within the river channel as a
starting point. In the original approach, the thalweg derived from existing
Evaluating Spatial Data Acquisition and Interpolation … 7

Fig. 3 River subsets used in this study (water depth)

bathymetry data, is used. As it is assumed, that the rivers’ bathymetry is the not yet
known target of the study, the centerline is used. Using satellite imagery (ArcGIS
Basemap Imagery) the centerline was extracted.

2.2 Data Acquisition Strategies

For all interpolations executed in this study, the input values (i.e. measurements) are
extracted from the reference surface. Different acquisition strategies, i.e. different
layouts for extracting points have been applied. In general, these strategies can be
categorized as orthogonal cross-sections or cruising trajectories.
Cross-sections are measurements taken in lines orthogonal to the flow direction
of a river. A point spacing of 10 ft is chosen in n-direction (along the cross-section).
Starting from 10 ft the spacing between the cross-sections is increased by doubling
the value for each test case. The maximum interval is 2560 ft, so that depending on
the subset, there are between five (subset 2) and nine (subset 3) cross-sections in the
input data. A further increase would result in too few cross-sections for the
interpolation.
Cruising trajectories for boat measurement campaigns are typically realized as
either parallel to the river centerline or as zig-zag trajectory. The smallest
wave-length of the zig-zag trajectory is set to 160 ft, as decreasing the wavelength
more would effectively lead to a duplication of the cross-section geometries. Further
wavelengths are generated by doubling the value for each test case.
Along the cruising trajectories and cross-sections, points from the reference
surface are extracted with a spacing of 10 ft.
In order to evaluate the effect of an applied acquisition strategy, 56 acquisition
strategies for each of the subsets (Fig. 3) are created. Each strategy defines the path
for the data acquisition and consists of different cross sections or different trajec-
tories (Fig. 4).
Visit https://textbookfull.com
now to explore a rich
collection of eBooks, textbook
and enjoy exciting offers!
8 R. Krüger et al.

Fig. 4 Examples of different acquisition strategies: cross-sections (a, b), zig-zag (once) (c),
zig-zag (twice) (d, e), zig-zag + centerline (f), zig-zag + 2 parallel trajectories (g), zig-zag + 4
parallel trajectories (h)
Evaluating Spatial Data Acquisition and Interpolation … 9

2.3 Spatial Interpolation Methods

Five interpolation methods are evaluated: Inverse Distance Weighting, Simple

Kriging, Ordinary Kriging and Radial Basis Functions. The parameters are slightly
changed for each execution, in order to conduct the sensitivity analysis. The pur-
pose is to find the optimal parameters for each of the methods. Further, it is
investigated which parameters have the greatest impact on prediction quality. The
sensitivity analysis is conducted only for a selection of strategies (Table 1) to stay
with a manageable number of interpolations.

2.3.1 Inverse Distance Weighting (IDW)

IDW is a deterministic method where unknown points are approximated by a

distance-decayed weighted average of the values of the nearby points (Shepard
1968; Rase 2016). The exponent if the inverse distance weight ðpÞ is the parameter,
which defines the significance of the nearby points.
For the IDW algorithm, three parameter settings are tested. First the exponent p
of the weighing function is varied. Second, the number of used data points is
modified. According to Vande Wiele (2001), beside the number of points also the
layout of the points has an influence on the interpolation result. Since, data points
evenly surrounding the location of the interpolated point typically lead to better
results than using clustered data points, a quadrant search for data points is used,
enforcing a more evenly spread point pattern. For each quadrant N points are used.
And third, the anisotropy ratio ar is altered (Merwade et al. 2006). Instead of using
a circular search neighborhood, an elliptical one with the axes aligned to the flow
direction, is used. The anisotropy ratio ar is the ratio of the main and minor axis of
the ellipse. The main axis is set parallel to the flow direction (s-coordinate).

2.3.2 Radial Basis Functions (RBFs)

RBFs are also deterministic methods. The resulting surface of interpolated points
must fulfill two requirements: Pass through the input data points, and be as smooth
as possible. The resulting surface can, in contrast to the IDW method, contain

Table 1 Strategies used for sensitivity analysis

Acquisition strategy Points Density (Pts./105 ft2)
Strategy 1 cross-section – 20 ft 83405 5146
Strategy 2 Zig-Zag (twice) 1280 ft + centerline 11760 726
Strategy 3 cross-section – 160 ft 10488 647
Strategy 4 Zig-Zag (twice) 5120 ft + 4 parallel trajectories 9550 589
Strategy 5 cross-section – 1280 ft 1221 75
10 R. Krüger et al.

values which are smaller or bigger than the minima and maxima of the input points.
The distance dependent weights are obtained by solving a system of n linear
equations.
There exists a number of RBFs—for this study, four RBFs provided by the
ArcGIS Geostatistical Analyst are used (ESRI 2001; Rase 2016): the Multiquadric
Function; the Inverse Multiquadric Function; the Spline With Tension Function and
the Completely Regular Spline Function.
For all four functions, a sensitivity analysis is conducted to compare the func-
tions to determine the influence of two parameters. Like for the IDW method, a
quadrant search neighborhood is used and the number of points N is varied. For the
interpolations in the s, n-coordinate system the anisotropy ratio ar is ascertained
according to Merwade et al. (2006) by computing the variance ratio of the variance
of zi across the flow and along the flow. For the four subsets, ar results to ar = 3.

2.3.3 Ordinary Kriging (OK)

Kriging and its variants (Krige 1966) are geostatistical methods, which take the
spatial correlation between the measured points into account, when calculating the
weights for the points used in the interpolation. The spatial correlation is quantified
by a semivariogram model as a function of the distance (lag) between a pair of data
points.
Different Kriging variants have been developed in the past. In this study Ordi-
nary and Simple Kriging are used. Ordinary Kriging assumes an unknown, but
constant mean in the data and therefore focuses on the spatial component. Simple
Kriging assumes that the mean is constant and known. More detailed information
on Kriging can be found in Cressie (1993).
Kriging methods are difficult to automate, as a suitable model for the data must
be chosen. The ArcGIS implementation supports the user in finding the right
parameters for the model, but the final choice for the right model is up to the user.
As the ArcGIS Geostatistical Analyst is including 11 model types, an analysis for
these in combination with a sensitivity analysis is not feasible. Thus, the different
models are analyzed and models with Nugget-parameter not close to zero are
removed, which is preferable for Digital Elevation Models without gaps. The visual
selections are quantified by cross validation. The spherical and circular models
produce a slightly worse RMSE score compared to the tetraspherical and pentas-
pherical models, but they are less computational expensive and are thus selected for
sensitivity analysis. Once suitable models are chosen, the Kriging process can be
partly automated in ArcGIS.
Two parameters have a strong impact on the variogram—size and count of lags.
The product of count and size of the lags should be at maximum around 50% of the
greatest distance in the dataset (ESRI 2001): Subset 2 has the shortest length of
about 13500 ft, hence 100 lags of 50 ft width are chosen.
For the Kriging algorithm, the number of points in the search neighborhood are
varied and the quadrant search neighborhood is used. In contrast to the other
Evaluating Spatial Data Acquisition and Interpolation … 11

algorithms, ar is determined by ArcGIS from the data (direction and both axis).
Again, this parameter is only used in the s, n-coordinate system. Further it is
examined if removing the spatial trend in the data (linear in flow direction,
non-linear perpendicular to the direction of flow) prior to the interpolation has an
effect on the interpolation result.

2.3.4 Simple Kriging (SK)

For Simple Kriging, the main difference compared to the workflow for OK, is that
the data is standardized after the trend removal through a Normal Score Trans-
formation. Afterwards, all steps are similar to the interpolation with the OK algo-
rithm. Accordingly, the influence of the four parameters N, ar , trend removal and
variogram model is examined. A cross-validation for subset 2 yields the same
qualitative order of the models as for OK. Due to the longer processing times
compared to OK, the two models with the highest RMSE in SK (circular and
spherical) are not used.

2.3.5 TopoGrid Algorithm

The TopoGrid algorithm is first described by Hutchinson (1989) and is a variation

of the Thin-Plate-Spline method to create hydrologic correct DEMs. This objective
is fulfilled by integrating a drainage-system into the interpolation. The Topogrid
algorithm does not need any parameters to be set. Some parameters can be varied
by the user, but have standard settings depending on the input data. In this study,
the standard settings are used, and thus no sensitivity analysis is conducted. The
Topogrid algorithm is taken as a benchmark to compare an approach not requiring
parameter settings with the algorithms above, which typically all require parameter
settings.

3 Assessing the Bathymetry Interpolation Methods

Behaviors

To assess the quality of the interpolated riverbed surfaces, different accuracy

measures are calculated. The most widely used accuracy measure for the quality of
Digital Elevation models is the Root Mean Square Error (RMSE) (Aguilar et al.
2005). As the RMSE is sensitive to large outliners the Mean Average Error
(MAE) is used in this study as well. For both RMSE and MAE 0 indicates no error.
The third measure used is the Pearson Correlation coefficient (R), which measures
the linear correlation for the predicted and the observed values between 1 (best) and
0 (worst). Thus, it evaluates the spatial fit of the prediction.
12 R. Krüger et al.

For the following sensitivity analysis, the RMSE is calculated for each set of
parameters. As the parameters are tested for five different acquisition strategies, the
RMSE is normalized for each of them. Subsequently, those normalized RMSEs are
averaged for each parameter set. Finally, the parameter sets with the lowest average
RMSE are selected for the comparison of the interpolation and acquisition strategies
in the next chapter.

3.1 Assessing Inverse Distance Weighting (IDW)

For the IDW method, three parameters ðN, p, ar Þ are varied. Table 2 shows the ten
selected parameter sets for the main analysis, as well as the RMSE for the five
tested strategies. The sets with the lowest RMSE for each of the tested strategies are
chosen. Further, the parameter sets with the lowest averaged RMSE for the five
strategies are selected.
As can be seen in the table, the strategies with the highest point densities yield the
lowest RMSE. Further can be seen, that that the RMSE increases as the point density
decreases. Figure 5 shows that using ar > 1 (considering anisotropy) and p < 4 yields
a lower RMSE unless the point density is very high. Figure 6 shows that a high
number of points N produces the lowest RMSE for low point densities and low
values for N yield the lowest RMSE for high point densities. Figure 7 shows that the
RMSE is very sensitive to variations of N for p ≤ 2 and high point densities.

3.2 Assessing Radial Basis Functions (RBFs)

For the RBF methods, five RBFs are tested and two parameters ðN, ar Þ are varied.
Table 3 shows the three selected parameter sets and functions for the main analysis,
as well as the RMSE for the five tested strategies.
Table 2 Selected parameter sets for IDW
Parameter RMSE (ft)
N ar p Strategy 1 Strategy 2 Strategy. 3 Strategy 4 Strategy 5 RMSEnorm
2 2 2 0.363 2.269 2.023 3.412 4.446 0.9217
8 2 3 0.403 2.258 2.026 3.346 4.274 0.9234
6 2 3 0.394 2.262 2.026 3.362 4.318 0.9234
10 2 3 0.409 2.263 2.028 3.336 4.251 0.9244
8 3 3 0.430 2.288 2.022 3.184 4.269 0.9251
4 3 2 0.479 2.249 1.997 3.181 4.310 0.9386
2 1 2 0.356 2.319 2.029 3.684 4.609 0.9476
4 2 1 0.633 2.211 2.046 3.275 4.194 0.9948
8 5 1 1.113 2.393 2.240 3.039 4.094 1.1801
10 5 1 1.232 2.435 2.309 3.051 4.081 1.2322
Random documents with unrelated
content Scribd suggests to you:
“I have thought that possible, but even then, he could easily write to
me in confidence, and tell me where he is,” said the girl.
“Where does Ruthen live?” I enquired.
“In Whitehall Court,” and she gave me the number.
“You have no idea what his profession may be?”
“Like Stanley—he is independent.”
“Audley is a rich man, isn’t he?” I asked.
“No doubt. When we first met he gave me some very expensive
presents merely because I happened to look after a girl he knew who
was suffering from pneumonia. He’s an awfully generous boy, you
know.”
“The fact is, Miss Day, I am doing all I can to discover Stanley
Audley. Can you tell me any other facts—anything concerning his
other friends?”
“He had another friend named Graydon, living at the same chambers
in Half Moon Street, a rather stout, round-faced man. But he has
also left London, I understand.”
“Graydon!” I ejaculated. So it seemed that the pair exchanged
names when occasion required. At Half Moon Street Audley was
Graydon, but outside, he took the name of the man who lived on the
floor below!
What could have been the motive?
I afterwards took my pretty companion to the theatre, and, later, she
took me to Ham-Bone Club, where we danced till nearly two.
From members there, I gleaned several facts concerning Stanley
Audley. He was apparently a rich young “man-about-town,” but
surrounded, as all wealthy young men are, by parasites who
sponged upon his generosity. Of these Harold Ruthen was
undoubtedly one.
Days passed, and although I went hither and thither, making
inquiries in all likely quarters, I could obtain no further knowledge.
Stanley Audley had disappeared. I felt more convinced than ever
that Thelma possessed knowledge she feared to disclose.
In my perplexity, I thought, at last, of old Dr. Feng. Perhaps he would
be able to help me. I wrote to him in care of his solicitor and received
a prompt reply asking me to go and see him at an address in
Castlenau, Barnes.
The house was just across Hammersmith Bridge. The anonymous
letter I had received had been posted, I remembered, at
Hammersmith. It was a queer coincidence.
Doctor Feng’s house, I found, was of a large, old-fashioned
detached residence which, a century ago, had probably been the
dwelling-place of some rich City Merchant who drove each morning
into London in his high dog-cart, his “tiger” with folded arms seated
behind him.
A maid conducted me to the front sitting-room, a large, well-
furnished apartment, where a big fire blazed.
“Well, Yelverton!” exclaimed the old doctor, rising, and putting out his
hand. “And how are you? I went to see my sister down at Mentone,
but the weather on the Riviera was simply abominable—a mistral all
the time. So I came back and took up my quarters here. Comfortable
—aren’t they? Sit down. It’s real good to see you again!”
I stretched myself in a deep comfortable chair beside the fire, and we
chatted for a time about Mürren.
“I wonder where Humphreys is?” he remarked. “He wasn’t a bad
sort, was he? And how about your temporary bride—the ‘Little Lady,’
as you called her!”
“Well, doctor,” I said, “that is really what I came to see you about.
The whole affair is a tangle and I wondered if you could help me. I
have found out a lot of things about Stanley Audley that are certainly
most disconcerting and mysterious.”
He passed a box of cigars. “Have a smoke over it,” he said, “if I can
help you I will. But first tell me what happened after I left Mürren.”
“A lot,” I replied. “You know Thelma’s husband left for London. Well,
he never came back.”
“The young cad,” said the doctor. “But, after all, I more than half
expected it.”
“Why?” I asked.
“Well,” he said, hesitatingly, “shall we say his sudden departure was
rather suspicious? To put it plainly the excuse was a bit thin. Would
any firm let an employee start on a honeymoon and three days later
find he was the man for an important appointment such as Audley
spoke of? Of course, such a thing might happen, but a more
probable excuse would have carried more conviction. To me it
suggested a story made up suddenly, in default if anything better, to
explain a departure forced upon him by some much less welcome
reason. However, I had no reason for saying this at the time and,
after all, I might have been wrong. But as things have turned out it
seems I was right and I am very sorry for his wife. After all, whatever
her husband may be, she is a charming girl—much too good for him,
anyhow. But go on, tell me what you have found out.”
I frankly told him, and as he smoked he sat back listening
thoughtfully without a word of comment.
At last, when I had concluded, he asked—
“Have you seen Harold Ruthen?”
“Not yet. He is an enemy of Thelma’s.”
“What makes you think that?” he asked, whereupon I told him of the
curious conversation I had overheard.
He bit his lip and smiled mysteriously, but said nothing. It was,
however, plain that what I had described greatly interested him.
“And little Mrs. Audley will tell you nothing—eh? She refuses. She is
evidently hiding some secret of her husband’s. Don’t you think so?”
“To me, she seems in deadly fear lest I should discover her
husband.”
“Oh! I quite agree, Yelverton,” the old man said. “There’s more
behind this curious affair than we’ve hitherto suspected. A man
doesn’t leave his young wife in the hands of a stranger without some
strong and very doubtful motive. Depend upon it that you were
marked down as the victim.”
“Not by Thelma!” I protested.
“No, she has been your fellow victim.”
“But the motive of it all?” I asked in dismay. “What is your opinion,
doctor?”
“The same that I formed when you first told me of your offer of help
—that you’ve been a silly idiot, Yelverton. Didn’t I point out at the
time the risks you were running?”
“Yes, you did,” I replied, “but I still intend—at all hazards—to get to
the bottom of the affair.”
Feng hesitated, and then, looking me straight in the face, said very
seriously—
“If you take my advice you will drop the whole affair.”
“Why?” I asked, in surprise.
“Because those men who lived at Half Moon Street and their friends
are evidently a very queer lot. In any case you ought to cease
visiting Mrs. Audley.”
I paused, recollecting that strange warning I had received, of which I
had not told him.
“But, after all,” I protested, “we are very good friends. Surely I ought
to help her by finding her husband?”
“When she probably knows where he is all the time!” scoffed Feng. “I
don’t see what good you will do that way.”
“Anyhow,” I said shortly, “I’m not going to see her left in the lurch like
this if I can help it.”
“Really, Yelverton, I don’t see what good you think you can do. We
both believe she knows where he is. If that is so why should you
interfere? Of course, what you tell me about the girl Day is very
interesting and may throw a good deal of light on Stanley Audley’s
character. But, after all, men change their minds and if Audley
preferred Thelma to Marigold, there was no reason why he should
not have asked her to marry him.”
“None the less, take my advice, drop the whole thing. You haven’t
the shadow of a legal right to interfere. The men who lived in Half
Moon Street, quite obviously a shady lot, have fled, evidently
frightened of something and apparently your temporary bride is as
frightened as they are. I don’t see why you should run any risk in the
matter.”
“But what earthly risk do I run?” I asked. “Surely I am capable of
looking after myself.”
“Considerably more risk than you imagine, unless I am very much
mistaken,” he replied gravely.
I wondered for a moment whether my mysterious warning had come
from the doctor himself. But what could he know about the affair? I
could not read anything in his inscrutable face, but his manner
certainly suggested that he was in deadly earnest, and, to my
intense surprise, he suddenly let fall a remark, quite unintentionally, I
believed, that, I realized with a curious suspicion, showed that he
knew Thelma and her mother were living at Bexhill. Here was indeed
a new complication. I made no sign that I had noticed his slip, but sat
as if thinking deeply, as indeed I was.
How, and for what purpose, had he obtained that information. He
had professed not to know what had happened after he had left
Mürren.
The idea flashed through my mind that he and Thelma were acting in
collusion to “call me off,” but this seemed so absurd that I dismissed
it at once.
“Now, look here, Yelverton,” he said presently. “You’ve not told me
everything.”
“Yes I have,” I protested.
“You haven’t told me that you’ve fallen deeply in love with little Mrs.
Audley. That is why I warned you—and still warn you—of rocks
ahead.”
“I did not think that necessary,” I said with some heat. “That is surely
my own affair!”
“Certainly,” he said, dryly, in the paternal tone he sometimes
assumed. “But remember my first view of the situation was the
correct one. I thought you extremely indiscreet to accept the trust
you did. It was a highly dangerous one—for you.”
“But you agreed afterwards that I did the right thing,” I argued.
“You acted generously in the Little Lady’s interests, but you have
certainly fallen into some extraordinary trap. That’s my point of view,”
he answered. “In any case, you are in love with a wife whose
husband is absent. That is quite enough to constitute a very grave
danger to both of you. So, if I were you I’d keep away from her. Take
my advice as an old man.”
His repeated warning angered me, and I fear that I did not attempt to
conceal my impatience. At any rate I took my leave rather abruptly,
and as I walked in the direction of Hammersmith Bridge I felt more
than ever puzzled at his attitude, and more than ever determined not
to deviate from the course upon which I had embarked.
 CHAPTER IX
CROOKED PATHS

One cold evening I returned from the office after a heavy day which
had been devoted to the successful settlement of a very complicated
and serious action for libel against a provincial newspaper which we
represented.
As I entered my room, Mrs. Chapman, in her spotless black dress—
just as she always wore when my father was alive—followed me in,
saying—
“Oh! Mr. Rex. A gentleman called about three o’clock. He wouldn’t
leave a card. He gave his name as Audley—Mr. Stanley Audley. He
repeated it three times, and told me to be sure to recollect the name.
He said he was extremely sorry you were not at home, but you were
not to worry about him in the least.”
I started, staring blankly at her.
“Wouldn’t leave a card? Wouldn’t he call again?”
“He seemed to be in a very great hurry, sir. He said he had come
from abroad to see you, but couldn’t wait and said he was very sorry.
Only I was to give you his urgent message.”
“What was he like?”
“Well, sir, he was a round, rather red-faced gentleman. He was
evidently greatly disappointed at not meeting you, but he impressed
upon me the message that he was all right, and that you were not to
worry about him.”
This was indeed a surprise.
It was evident that my caller was the man who had lived on the first
floor in Half Moon Street, and was the friend of the Stanley Audley
who had married Thelma!
What did that amazing visit portend? It worried me. Why should a
reassuring message be given to me by a man who was not the
person in whom I was interested, and whom I had never met? The
whole affair was becoming more and more obscure and mysterious.
As a solicitor I had been brought into contact with more than one
queer affair, but the Audley mystery was beyond anything in my
experience.
“Couldn’t he call again, Mrs. Chapman?” I asked.
“No, sir. He said he had come to see you just for a moment, and that
he was sorry that he couldn’t wait. He had a taxi outside.”
“Thanks, Mrs. Chapman. I’m sorry I was not at home to see him. Did
you give him my office address?”
“I did, sir. But he said he had no time to go round to Bedford Row,
and that you would no doubt understand.”
Understand! What could I understand? I was more bewildered than
ever.
Next day I called again upon Belton, in Half Moon Street, and
questioned him more closely about his recent “Box and Cox”
tenants. But he could tell me nothing more than he had already. Mr.
Graydon and Mr. Audley were close friends. That was all.
“Tell me something about their visitors,” I asked. “Did Mr. Graydon,
the gentleman who lived above, have many?”
“No, sir. Very few. Several of them I knew quite well when I was in
service—gentlemen from the clubs. One a Canadian millionaire,
came often, but Mr. Graydon never had any lady visitors except that
young lady we spoke about a short time ago—the lady whose
photograph you showed me, Miss Shaylor.”
“And Mr. Audley, who lived below?”
“Oh, he had quite a lot of callers—both ladies and gentlemen. He
was older than Mr. Graydon, and seemed to have quite a big circle of
acquaintances. They used to play bridge a lot.”
“Now, tell me, Mr. Belton. What is your private opinion about your
tenants?”
“Well, sir, as you are a solicitor”—he had gained that knowledge from
my card,—“I can speak quite frankly. Now that they are gone I don’t
mind saying I held them both in suspicion. They had plenty of money
and paid well, but I don’t think they were on the straight. That’s my
firm opinion and my wife thinks the same.”
“What first aroused your suspicion?”
“Their card parties. They weren’t always square. I’m sure of it. Mr.
Audley had an invalid friend, an old man named Davies, who came
about three times, and when he came woe betide those who played.
I kept my eyes and ears open when I served their drinks, and I’m
sure I am not mistaken.”
“An invalid!” I exclaimed. “What kind of man was he?”
“Oh! he was very lame, was Mr. Davies, sir. An old man, but as keen
as mustard on poker.”
“Did Mr. Graydon play?” I asked.
“Very little, sir.”
“Did he ever meet this Mr. Davies?”
“I think not, sir. Because on the first occasion Mr. Davies came I
recollect that Mr. Graydon was away in Norway. The next time he
came, Mr. Graydon was away in Paris. No,” he went on, “as far as I
can recollect Mr. Graydon never met Mr. Davies.”
“Then this Mr. Davies was a person to be avoided?” I suggested.
“Distinctly so, sir. He was a shrewd and clever gambler, and I feel
certain that he was in league with Mr. Audley. Indeed, I know that on
the morning after one of their sittings they divided up a thousand
pounds between them. It had been won from a man named Raikes,
a manufacturer from Sheffield.”
“So they shared the spoils?” I said. “But tell me more about this
interesting invalid.”
“Well, sir. He was a grey-bearded man of about sixty I should think,
and he walked with difficulty with two sticks. He seemed to lisp when
he spoke.”
It struck me at once that the ex-butler’s description would have fitted
old Mr. Humphreys very closely, except that Humphreys did not lisp.
I had no reason for thinking that Humphreys could have known
Graydon, but he might have done so and he certainly was a very
keen poker player.
“Had he a rather scraggy, pointed beard and did he wear in his tie a
blue scarab pin?” I asked.
“No,” was Belton’s prompt reply, “he had a round beard and I never
saw him wearing a scarab pin.”
Now old Mr. Humphreys always wore an antique pin of that
description; I never saw him without it. He was immensely proud of it
and used to declare it was a mascot that brought him good luck. He
had a wonderful story of how he obtained it from some old Egyptian
tomb. So the chance of Mr. Davies and old Humphreys being
identical seemed a coincidence almost too peculiar to be true. Yet I
could not get rid of a suspicion that they were one and the same
person.
“You are quite certain that he never met the young gentleman you
knew as Mr. Graydon?” I asked Belton.
“I’m quite certain of that, sir. One day Mr. Audley asked me not to say
that Mr. Davies had been there, and asked that I would keep his
visits a secret from young Graydon as he did not wish them to meet.
There was, I remember, a lady named Temperley, who sometimes
came with Mr. Davies. She was a stout, dark-eyed, over-dressed
woman whom I put down as a retired actress. She had a young, thin
rather ugly daughter, a girl with a long face, and protruding teeth.
Both mother and daughter seemed to be on terms of close friendship
with Mr. Davies.”
“Davies was an invalid. How did he get up these stairs?”
“With difficulty, sir. I used to help him up, and sometimes Mr. Audley
helped me,” was the ex-butler’s reply. “At poker he was marvelous.
I’ve seen poker played in several families in whose service I’ve been,
but I never saw a finer player. He was more like a professional than
an ordinary player for amusement.”
“And your tenant, Mr. Audley?”
“He was a fine player, of course. He used to have friends in at night
and sometimes they would play till dawn.”
“And did Mr. Graydon never play?” I asked.
“Very seldom; the parties usually took place when he was away.”
It was quite evident that Stanley Audley, alias Graydon, was a
person of mystery and his friends were as mysterious as himself.
After a moment’s reflection I decided to take Belton fully into my
confidence and tell him the whole story.
“Now, look here, Belton,” I said, “you may be able to help me
considerably. I will tell you the whole story so far as I know it, and
perhaps you will be able to remember further facts that may help.”
So I related to him everything that had happened since I first met
Stanley Audley and his bride at Mürren.
Belton listened in silence. When I had finished he asked me one or
two questions.
“Well, sir,” he said at last, “I think you had better see my wife. She
may know something more.”
He fetched Mrs. Belton and briefly outlined to her the facts I had
given him.
“You see, Ada,” he said, “the gentleman who called himself Audley
here, was not the Mr. Audley who married the daughter of
Commander Shaylor. Mr. Graydon is her husband. Isn’t it a puzzle?”
“It is,” replied his wife. Then, after I had made my explanation I
begged her to tell me any further fact which might be of service in my
inquiry. She hesitated for a moment and at last said:
“Don’t you recollect, Jack, that Mr. Graydon, before he came to us,
lived at Seton’s, in Lancaster Gate. He was very friendly with Mr.
Seton, who you remember was butler to old Lord Kenhythe at
Kenhythe, in Kirkcudbrightshire. You went there one shooting
season from Shawcross Castle, to oblige his lordship.”
“Oh! yes, of course!” exclaimed her husband. “Really, Ada, you’ve a
long memory!”
“Well, I was head-housemaid once at Shawcross Castle. You forget
that! But, don’t you recollect that young Mr. Graydon was very
friendly with Mr. Seton. I don’t know why he left there and came to
us, but I fancy it was because there was such a row at a party he
had there, and he wouldn’t apologize, or something like that.”
“Ah! I remember it all now, of course, Ada,” exclaimed the woman’s
husband. “Yes, you’re right—perfectly right! If there’s one man in
London who knows about Mr. Graydon it’s Mr. Seton.”
He gave me the address of Lord Kenhythe’s ex-butler, and an hour
later I called at a large private hotel facing Hyde Park, near
Lancaster Gate, with a scribbled card from Belton.
The man who received me was a tall, very urbane person with small
side-whiskers. He took me into his private parlor in the basement,
where I told him the object of my visit.
“Yes, sir. I know Mr. Philip Graydon. A very estimable young
gentleman.”
“Who is he?”
“Well, his father was the great Clyde shipbuilder, whose works are at
Port Glasgow—the firm of Graydon and Hambling. When his father
died, about two years ago, he left him a quarter of a million.”
“You know him well?”
“I did, sir. His father used to shoot with his lordship regularly, and Mr.
Philip often came with him.”
I briefly told him that I was making inquiries into certain very curious
circumstances, and said—
“I want your private opinion, Mr. Seton. Is there anything peculiar
concerning Mr. Graydon? I ask this because on his marriage he took
the name of Audley.”
“His marriage! I didn’t know he’d married, sir.”
“Yes. And he is missing. It is on behalf of his wife, who is a friend of
mine, that I’m making these inquiries.”
“Mr. Graydon married!” he repeated. “Pardon me, sir, but whom did
he marry?”
“A young lady named Shaylor.”
“Ah!” he ejaculated. “Yes, I know. He was very fond of her—very
fond! Her mother is a widow in very straitened circumstances, I’ve
heard. But do you say he’s missing?”
“Yes. He disappeared while they were on their honeymoon in
Switzerland.”
“And where is his wife now?”
“With her mother in Bexhill. But tell me, Mr. Seton, Mr. Graydon as
you call him, was with you for some months, wasn’t he?”
“For nearly a year and a half, sir.”
“And during that time did a man named Audley ever visit him?”
“Yes, a round-faced man who lived at Belton’s. He visited Mr.
Graydon first about six weeks before he left me to go and live at
Belton’s.”
“Why did he leave you?”
“Well, he had a bachelor party one night—they were very noisy and I
remonstrated with him, and—well, he’s only young, sir—and the fact
is he insulted me. So I gave him notice. But we’re still the best of
friends,” said the ex-butler.
And then Seton sprang on me perhaps the greatest surprise of my
life.
“Now I know your reason for wanting to see Mr. Graydon,” he said. “I
may as well tell you he is here now.”
“Here!” I gasped excitedly, “do you mean he is staying here?”
“Yes, sir,” was the reply, “he’s in number eighteen. He came here
yesterday quite unexpectedly.”
At last I had run Thelma’s mysterious husband to earth!
“He came in half-an-hour ago,” Seton went on, “and I gave him a
letter which came for him by express messenger. I know he’s
upstairs. If you would like to see him, I will send up.”
“No, thanks,” I said. “Under the circumstances I think I would prefer
to go up unannounced if you have no objection.”
“Not in the least,” replied Seton. “Number Eighteen is on the second
floor.”
So I eagerly ascended the wide, thickly-carpeted stairs. I had no very
clear idea as to how I should approach the man I had known as
Stanley Audley, but I was determined to demand an adequate
explanation of why he had married Thelma under an assumed name
and so cruelly deserted her, and, if necessary, to back my demand
by a threat of legal proceedings.
 CHAPTER X
IN ROOM NUMBER EIGHTEEN

On the second landing I rapped at the door of room Number 18,

feeling considerable pleasure at the thought of giving my whilom
friend an unwelcome surprise.
There was no reply, but I fancied I heard a movement inside. I
listened eagerly.
I knocked again. Yes. I felt sure someone was within, but my knock
met with no response.
A third time I knocked and more loudly, but to no avail. I tried the
door—it was locked.
Five times I hammered with my fist, but there being no answer I
descended the stairs and found Mr. Seton.
“But he must be up there if his door is locked,” he said. “He never
takes his key but always leaves it on the peg here,” and he indicated
a board on the wall in a little box-like room off the hall where visitors
left their keys. To each key was attached a bulky ball of wood, in
order that the key should not be carried away accidentally in the
pocket.
With the landlord I reascended the stairs and Seton knocked at the
door, calling his guest by name. But there was still no response.
“Do you know, I believe I heard somebody inside when I first
knocked,” I remarked.
Seton bent and peered through the keyhole.
“At any rate the door is locked on the inside,” he said.
Then he thundered at the door, after which we both listened. There
was no sound, but I thought I detected the smell of burning paper.
All the other guests were apparently out at the time, for the noise we
made attracted only the servants.
“Baker!” Seton cried to a man who was in his shirtsleeves and wore
an apron of green baize, “we must force this door. There’s a crow-
bar down in the cellar. Go and get it.”
As the man addressed ran downstairs, the ex-butler turned to me
with a scared expression upon his face, saying——
“This is very peculiar, sir. Why has he locked himself in like this? Did
you really hear a noise?”
“Yes. I am sure I did, yet with the roar of the traffic out in the road, I
really couldn’t quite swear to it,” was my reply.
“What I heard was like a man bustling about hurriedly, and yet trying
to make no noise.”
“Surely he can’t have fainted--or--or committed suicide!” Seton
remarked.
For a few minutes we stood outside the door utterly mystified, until
the porter brought us a rusty bar of iron about three feet long, curved
and flattened at the end—a very serviceable crow-bar.
This, Seton inserted between the door and the jamb, close to the
lock, and then drew it back slowly. The woodwork groaned, creaked
and cracked and with a sudden jerk the wood round the mortice lock
tore away and the door flew open.
We stood amazed. The room was empty.
In a few seconds we had searched the big old-fashioned wardrobe
and had looked beneath the bed and behind the curtains. But
nobody was there. And, moreover, while the key was still in the door
on the inside the window was closed and latched!
The fireplace was a small one with a flue through which not even a
small boy could pass. In the grate were smoldering ashes of
something, apparently a coat that had been hastily burned. There
was an odor of consumed petrol, and it occurred to me at once that
some clothing had been hurriedly saturated from a bottle of motor-
spirit and set fire to—for the room was still heavy with smoke.
Seton crossed to the window and saw at once that it had not been
opened. I glanced out and down. From the narrow window-sill there
was a sheer drop to the paved basement forty or fifty feet below with
not even a stackpipe by which an active man might have escaped.
“Well, this is extraordinary,” cried Seton. “How could Mr. Graydon
possibly get out of the room and leave it still locked on the inside?”
Seton bent suddenly over the fireplace. “Well, we may as well see
what he was burning,” he said as he picked up a half charred piece
of paper that had apparently been crumpled up hastily and thrown
into the grate. He smoothed it out and looked at it in amazement.
It was a portion of a fifty-pound Bank of England note! It was partly
burned but quite enough was left to identify it without any possibility
of a mistake.
“Well,” I exclaimed, “burning fifty-pound notes is certainly a new kind
of pastime. What on earth can it mean?”
“I can’t imagine,” replied Seton. “And how can Mr. Graydon have
gone? Certainly not through the door or the window.”
“And before he went,” I added, “he burnt a coat or something of the
kind and a fifty-pound note!”
In front of the window was a small early Victorian escritoire. Upon it
were several loose sheets of paper from a new writing-pad, an ink-
stained envelope, and a couple of bills from a local chemist.
Seton opened two or three of the drawers and from one of them
drew a folded wad of papers. “More notes!” he ejaculated, as he felt
with his fingers the crisp familiar crackle.
There were three notes for fifty pounds each, obviously quite new.
Clearly Graydon, in his hurry, had forgotten that they were there.
“It seems to me,” I said to Seton, “that Graydon must have been
frightened by something and had to get away quickly.”
“Frightened, but of what?” Seton asked. “I saw him only half-an-hour
before you came, and he seemed all right then.”
“Do you think my visit might have frightened him?” I asked.
“Well, sir, I don’t know. But why did he burn a fifty-pound note and
how did he get out? That’s what puzzles me. I could have
understood it if he had locked his door on the outside.”
“It beats me, anyhow,” I said, looking round the room. I noticed
Graydon’s two suitcases stood open and some of his clothes were
hanging in the wardrobe. Why, and above all how had he vanished
so suddenly? But for the fact that he had actually called to see me—
showing that he certainly was not afraid of meeting me—I might well
have thought that he would be alarmed on recognizing my voice. But
he had evidently not done so and must have thought I was someone
else whom he urgently desired to avoid.
Those fifty-pound notes puzzled Lord Kenhythe’s ex-butler as
completely as they did myself. Men do not usually go about burning
fifty-pound notes. We knew that the young fellow who, in
Switzerland, had posed as a hard-working electrical engineer
welcoming the prospect of a “rise,” was on the contrary, a rich young
man. But that he should burn bank-notes of such value or leave
them discarded as he had done, was simply inexplicable on any
hypothesis we could frame.
I was deeply chagrined. I had come within an ace of capturing the
truant bridegroom and yet he had eluded me. Could it really, I asked
myself, have been the same man? Again I carefully described to
Seton the man I had known as Stanley Audley. He was emphatic in
his assertion that it was Philip Graydon, the man who had been in
that very room barely half-an-hour before. And as if to make
assurance doubly sure, I found on one of his suitcases a label of the
Kürhaus Hotel at Mürren and another put on at Mürren station,
registering this case through to Victoria.
There could not be the slightest doubt as to the mystery man’s
identity as Thelma’s husband.
“Look here!” said Seton, suddenly, as he held up a towel he had
taken from the rail. It was stained with blood. The hand basin was
half full of water deeply tinged with blood.
“Evidently he had cut himself badly,” was Seton’s comment.
“Perhaps,” I said, “but is this his own blood or someone else’s?”
“Surely, sir, you don’t suspect he has been guilty of a crime?” gasped
Seton.
I pointed to the charred fragments of the coat. “It might be so,” I
rejoined.
A few moments later, however, on making a closer search of the
room we found in the waste-paper basket a broken medicine bottle
and on the edge of a piece of glass was a blood stain. It told its own
tale—he had cut his hand upon the glass. Further, close beside the
dressing-table were three or four dark spots. I touched one, and
found it to be blood.
“I wonder why he destroyed his coat?” Seton remarked. “He’s gone
away leaving everything behind.”
“But how did he get out?” I persisted. “The door and window were
both fastened and there is no fanlight.”
We again carefully examined the lock. It was intact, it had been
locked from the inside and the key was still there.
Together we went carefully through the fugitive’s belongings, but
found nothing of interest. They were merely clothes of good quality
or the wardrobe of a fashionable young man. From the pocket of the
suitcase that bore the label “B. O. B.”—or Bernese Oberland Bahn—
I took out three one-pound Treasury notes. But we found not a scrap
of writing of any sort. There was some burnt paper in the fireplace,
suggesting that with the coat he had destroyed all documents that
might give a clue to his identity. The broken bottle smelt of petrol and
apparently he had kept the spirit ready for use if he wanted quickly to
destroy anything.
Our search concluded, Seton had all the things removed to an
unoccupied room and locked the door.
“The Bank will pay the halfnote,” said Seton. “I shall pay the lot in
and hold the money until Mr. Graydon turns up again. He has plenty
of money, of course, and may not have missed it. There is no doubt
some explanation. I cannot believe, knowing Mr. Graydon as I did,
that there can be anything very seriously wrong.”
“But why should the note be burned?” I queried.
“It might have been accidentally among the other papers he
destroyed, sir. Don’t you think so?”
This, of course, was possible. For a long time we sat in Seton’s room
discussing the strange affair. At first Seton thought he ought to tell
the police, but I urged him not to do so. It would get into the papers, I
argued, and that was the last thing desirable for a high-class private
hotel such as his. I did not want a public scandal that must involve
Thelma in most unpleasant publicity.
“I wonder whether he had an inkling that you’d called, sir?”
suggested Seton. “Perhaps he saw you from one of the front
windows and then rushed up and prepared to bolt.”
“But why should he? I have acted towards him only as a friend and I
see no reason why he should take such extreme steps to avoid me.
Besides, he actually called at my flat.”
“Yes, I had forgotten that,” Seton admitted. “But still, I think
something must have frightened him—and frightened him badly, too.
He wouldn’t have cut his hand in opening the bottle of petrol, burned
his clothes and papers, and got away so swiftly if there wasn’t some
very strong motive for doing so. What’s your opinion?”
“The same as yours, Seton,” I answered. “But the affair is full of
remarkable circumstances. How did he get out of that locked room?
He was certainly in there when I first knocked.”
“My own belief,” said Seton, “is that he must have started to destroy
his things as soon as you knocked. He was certainly in a great hurry
for he smashed the neck of the petrol bottle when he found he could
not get the cork out—it’s still in the neck of the broken bottle—and
cut his hand in doing so.”