Professional Documents
Culture Documents
Full Ebook of Spatiotemporal Analytics Jay Lee Online PDF All Chapter
Full Ebook of Spatiotemporal Analytics Jay Lee Online PDF All Chapter
https://ebookmeta.com/product/online-learning-analytics-data-
analytics-applications-1st-edition-jay-liebowitz/
https://ebookmeta.com/product/industrial-ai-applications-with-
sustainable-performance-1st-edition-jay-lee/
https://ebookmeta.com/product/developing-informed-intuition-for-
decision-making-data-analytics-applications-1st-edition-
liebowitz-jay/
https://ebookmeta.com/product/graph-powered-analytics-and-
machine-learning-with-tigergraph-driving-business-outcomes-with-
connected-data-1st-edition-victor-lee/
Off Balance Painted Bay 1 1st Edition Jay Hogan Hogan
Jay
https://ebookmeta.com/product/off-balance-painted-bay-1-1st-
edition-jay-hogan-hogan-jay/
https://ebookmeta.com/product/tamarillo-tart-southern-
lights-2-1st-edition-jay-hogan-hogan-jay/
https://ebookmeta.com/product/powder-pavlova-southern-
lights-1-1st-edition-jay-hogan-hogan-jay/
https://ebookmeta.com/product/real-analysis-jay-cummings/
https://ebookmeta.com/product/flat-whites-chocolate-fish-
southern-lights-3-1st-edition-jay-hogan-hogan-jay/
Spatiotemporal Analytics
This book introduces readers to spatiotemporal analytics that are extended
from spatial statistics. Spatiotemporal analytics help analysts to quantita-
tively recognize and evaluate the spatial patterns and their temporal trends
of a set of geographic events or objects. Spatiotemporal analyses are very
important in geography, environmental sciences, economy, and many other
domains. Spatiotemporal Analytics explains with very simple terms the
concepts of spatiotemporal data and statistics, theories, and methods used.
Each chapter introduces a case study as an example application for an in-
depth learning process. The software used and the provided codes enable
readers not only to learn the analytics but also to use them effectively in
their projects.
Jay Lee
Designed cover image: Image from Shutterstock (Image ID: 1115087306)
© 2023 selection and editorial matter, Jay Lee; individual chapters, the contributors
Reasonable efforts have been made to publish reliable data and information, but the author and
publisher cannot assume responsibility for the validity of all materials or the consequences of
their use. The authors and publishers have attempted to trace the copyright holders of all material
reproduced in this publication and apologize to copyright holders if permission to publish in this
form has not been obtained. If any copyright material has not been acknowledged please write
and let us know so we may rectify in any future reprint.
Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced,
transmitted, or utilized in any form by any electronic, mechanical, or other means, now known or
hereafter invented, including photocopying, microfilming, and recording, or in any information
storage or retrieval system, without written permission from the publishers.
For permission to photocopy or use material electronically from this work, access www.
copyright.com or contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive,
Danvers, MA 01923, 978-750-8400. For works that are not available on CCC please contact
mpkbookspermissions@tandf.co.uk
Trademark notice: Product or corporate names may be trademarks or registered trademarks and
are used only for identification and explanation without intent to infringe.
DOI: 10.1201/9781003304395
Typeset in Times
by codeMantra
v
vi Contents
Index ......................................................................................................255
Editor
Dr. Jay Lee received his PhD in geography from the University of Western
Ontario in 1989. Since then, he has taught GIS and quantitative methods in
geography at Kent State University. Dr. Lee’s research interests stem from
integrating operations research and spatial analysis. He coauthored two edi-
tions of Statistical Analysis with GIS. Since 2017, Dr. Lee has worked on
extending spatial analytics to spatiotemporal analytics. This book represents
the research outcomes from the Applied Geography Laboratory at Kent State
University from 2017 to 2022.
vii
Contributors
Zhuo Chen Jay Lee
Department of Geography Department of Geography
Kent State University Kent State University
Kent, Ohio Kent, Ohio
Tao Hu Bo Wan
Department of Geography School of Computer Science
Oklahoma State University China University of Geosciences
Stillwater, Oklahoma Wuhan, China
ix
x Contributors
Ortis Yankey
Department of Geography
and Environmental
Science
University of Southampton
Southampton, United Kingdom
1 Introduction to
Spatiotemporal
Analytics
Jay Lee
Kent State University
CONTENTS
1.1 From Spatial Analytics to Spatiotemporal Analytics ....................... 4
1.2 Spatial Dependency and Spatiotemporal Dependency among
Geographic Events or Objects .......................................................... 5
1.3 Space–Time Dependency.................................................................. 6
1.4 Concluding Remarks......................................................................... 9
References ................................................................................................. 9
Over the last decade or so, several newly developed spatiotemporal analyt-
ics have been proposed for users to analyze the spatial processes of a set
of geographic objects or events. These analytics not only enable users to
observe and analyze the different spatial patterns of events or objects at
different times, but they also provide users ways to quantitatively assess
how the different time-specific spatial patterns evolve over time into a spa-
tial process. Figure 1.1 shows the relationship between a series of time-
observed spatial patterns linked to form a spatial process. In many ways,
understanding spatial patterns is not the ultimate goal of a scientific inquiry
when investigating data that have both spatial and temporal elements. It is
often our interest in finding not only spatial patterns of a certain phenom-
enon, but also how such patterns change over time.
For linking a sequence of time-specific spatial patterns of events or
objects into a spatial process, conventional approaches often burden view-
ers of the sequence of displays (or maps) the task of imagining how those
displays relate to each other and what that relevance may suggest. This is
a weakness that the current geographical information systems (GIS) have
because users are limited to only a handful of analytics when analyzing
spatiotemporal processes. For example, Figure 1.2 shows sequences of
maps for reported cases of dengue fever in Kaohsiung City, Taiwan, from
2003 to 2007. Either reported as raw counts of case numbers or against
DOI: 10.1201/9781003304395-11
2 Spatiotemporal Analytics
FIGURE 1.2 Reported dengue fever cases in Kaohsiung City, Taiwan, 2003–2007.
3
4 Spatiotemporal Analytics
REFERENCES
Bertazzon, S. (2003). Spatial and temporal autocorrelation in innovation diffusion
analysis. Computational Science and Its Applications – ICCSA 2003, 23–32.
Cressie, N. and C. K. Wilkie (2011). Statistics for Spatiotemporal Data. Hoboken,
NJ: John Wiley & Sons, Inc.
Diggle, P. J., A. G. Chetwynd, R. Haggkvist, and S. E. Morris (1995). Second-
order analysis of space-time clustering. Statistical Methods in Medicine 4,
124–136.
Gao, S., Y. Liu, Y. Wang, and X. Ma (2013). Discovering spatial interaction com-
munities from mobile phone data. Transactions in GIS 17(3), 463–481.
Gonzalez, M. C., C. A. Hidalgo, and A. L. Barabasi (2008). Understanding indi-
vidual human mobility patterns. Nature 453, 779–782.
10 Spatiotemporal Analytics
CONTENTS
2.1 Introduction .....................................................................................13
2.2 Review of Relevant Literature .........................................................16
2.3 Analytical Methods .........................................................................18
2.3.1 Centrography of Spatiotemporal Points ................................18
2.3.1.1 Spatiotemporal Mean Center ...................................19
2.3.1.2 Weighted Spatiotemporal Mean Center...................19
2.3.1.3 Changes in Spatiotemporal Mean Centers ............. 20
2.3.2 Dispersion of Spatiotemporal Points.................................... 22
2.3.2.1 Standard Spatiotemporal Distance ......................... 22
2.3.2.2 Standard Spherical Volume .................................... 23
2.4 Application Example ...................................................................... 24
2.5 Software and Usage ........................................................................ 27
2.5.1 Hardware/Software Requirements ....................................... 28
2.5.2 Software Usage for Spatiotemporal Mean Centers .............. 29
2.5.3 Software Usage for Standard Spatiotemporal Distance ........31
2.6 Concluding Remarks....................................................................... 32
References ................................................................................................33
2.1 INTRODUCTION
The first step in analyzing a large volume of data is often the calculation of
their statistical summaries. The summaries provide quantitatively digested
information that offers a glance of the data, sometimes as an initial assess-
ment. This is especially useful when dealing with large quantities of data.
In addition, such summaries calculated from multiple datasets enable ana-
lysts to do a general comparison between/among the datasets. Following
DOI: 10.1201/9781003304395-2 13
14 Spatiotemporal Analytics
the terminology used in Lee and Wong (2001), we refer to these summary
statistics as those measuring the central tendency of the data. In this era of
big data, such statistical summaries seem to be even more important than
before.
In a similar way, measuring the central tendency of a set of spatiotem-
poral objects or events that may be represented as points allows analysts
to understand how the points concentrate around their central point, or
the centroid of the set of points, and how the other points deviate from
the central point. Such measurements of central tendency are not only a
quantitative way to understand and compare spatiotemporal datasets, they
also enable analysts to visually, or qualitatively, map the distributions of
multiple datasets for comparisons.
Statistical summaries of a set of data usually include values of the max-
imum, minimum, average, median, mode, standard deviation, and vari-
ance of the data values. In terms of measuring the variation of data values
beyond these descriptive statistics, one may calculate the skewness, kurto-
sis coefficients of a set of data values.
For example, consider the exam scores students achieve for a spatio-
temporal statistics class. Assuming there are n students, their scores are
xi , i = 1,2,3,…, n. In this case, the scores are our data values. The mean
score would tell us how the group performs on average. If mean scores of
other similar datasets (from other class groups) are available, the calculated
mean score would let us compare how this group of students performs as
compared to last year’s group or another group of students taking the same
course/exam with a different instructor, and so on. The mean score can be
calculated with the following equation:
∑
n
xi
x= i=1
(2.1)
n
It should be noted, however, that the mean calculated from a dataset that
has a very much larger or smaller value from others in the dataset would
typically not be a good representative of the entire dataset. Therefore, the
measured central tendency is often accompanied by some measurements
of averaged deviation of the data values from their mean. The statistics for
such averaged deviation include standard deviation and variance, and both
of these give the degree to which data values in a dataset deviate from their
mean. The smaller the values of standard deviation and variance, the more
concentrate the values are around their mean in a dataset. Alternatively,
large values of standard deviation and variance indicate that the data val-
ues are more disperse from their mean.
Spatiotemporal Centrography and Dispersion 15
∑
n
( xi − x )2
S= i=1
(2.2)
n
∑
n
( xi − x )2
V=S 2
= i=1
(2.3)
n
because variance of a set of data values is the squared value of their stan-
dard deviation.
Geographic objects or events are often represented as points, lines, or
polygons. These three types of objects are often referred to as forming
the geometry of geographic elements. Points represent objects that do not
occupy any areas in space. Examples include a tree, a water hydrant, or
a power pole by a street. Lines are used to represent objects whose pri-
mary property of our concern are their lengths and directions, such as seg-
ments of rivers, street centerlines, or boundaries of administrative units.
Schools, hospitals, or buildings on a large-scale map are often represented
as polygons because they do occupy space on the ground. These objects are
depicted by polygons whose boundaries delineate their area extents.
Geographic scale affects how objects are represented on a map, or in a
dataset of geographic information. An example is a city that occupies the
entire map sheet at a large geographic scale, e.g., one inch on the map rep-
resenting 100 m on the ground. On a map of the scale of 1 inch on the map
representing 10,000 km on the ground (or a small-scale map), however, that
same city may become just a point on the map.
Geographic objects or events, when represented by points, can be defined
by a pair of coordinates, (x, y), in space. In practice, the coordinates may be
longitudes and latitudes. Alternatively, coordinates of a user-defined coor-
dinate system can also be used to define locations of geographic objects
or events. It should be noted that distances between actual longitude and
latitude coordinates of certain locations on the surface of the Earth may be
different, depending on the map projections and the coordinate systems in
which the coordinates are defined.
In a two-dimensional space, the mean center (or spatial mean) is the
average location of a set of locations given by a set of geographic objects
or events being represented as points. Once the coordinate system of these
16 Spatiotemporal Analytics
points is defined, the mean center can be easily calculated by using the fol-
lowing equation:
∑ ∑yi
n n
xi
( x , y ) = i=1
,
i =1
(2.4)
n n
where
x , y are the coordinates of the mean center,
xi , yi are the coordinates of the i-th point, and
n is the number of points.
the data values distribute, how concentrated or how dispersed they distrib-
ute, or if the distribution has any directional bias (Dickinson, 1963). Such
statistics, along with the computation of standard distance, can be found in
Bachi (1963).
Levine (1996), in a software package, crimestat (https://nij.ojp.gov/
topics/articles/crimestat-spatial-statistics-program-analysis-crime-
incident-locations#about), that he and his group developed, provides tools
for spatial statistics. Later, Lee and Wong (2001) extended the one-dimen-
sional descriptive statistics into a two-dimensional descriptive statistics for
spatial statistics with accompanying computer scripts (Avenue) for users
to apply spatial statistics. Their discussion included the measurements for
central tendency and dispersion of points defined in space. Later, Burt
et al. (2009) also discussed similar measures for central tendency and dis-
persion of spatially defined locations. The introduction of spatial statistics
was soon noticed by commercial vendors of GIS software. For example,
Scott and Janikas (2010) provided a summary of tools that ArcGIS devel-
oped and offered.
Spatial descriptive statistics were widely applied in mapping and analy-
sis of geographic information as first presented by Dickinson (1963). Others
also extended the classic statistics into spatial statistics in their applica-
tions. For example, Wong (1999) reviewed centrographic measures as
examples for implementing spatial statistics in GIS. Scott and Warmerdam
(2005) used spatial statistics to analyze crime events. For public health
applications, Lu (2011) applied spatial statistics to identify and visualize
18 Spatiotemporal Analytics
∑ ∑ ∑ ti
n n n
xi yi
( xmc , ymc , tmc ) = ( x , y , t ) = i=1
, i =1
, i =1
(2.5)
n n n
where
( xmc , ymc ,tmc ) or ( x , y , t ) is the coordinates of the mean center,
xi , yi , ti are the coordinates of point pi, the i-th point in the dataset, and
n is the number of points in the dataset.
FIGURE 2.2 A set of spatiotemporal points (triangles) and their mean center
(circle).
20 Spatiotemporal Analytics
where
wi is the weight associated with point pi.
FIGURE 2.3 The movement (or distance) between two mean centers. Triangles
are spatiotemporal points and mean centers are represented by circular points.
one to appreciate the changes in the positions of points and changes in the
distributions.
By detecting and visualizing the movements of mean centers over time,
one can better understand the scope and the trend of behavioral patterns of
the spatiotemporal points. Note that mean centers can be calculated from
multiple sets of spatiotemporal points. The distances, or the magnitudes of
changes in the positions of mean centers, can give hints for how different
the point sets are. So, a spatiotemporal deviation distance between two
mean centers can be calculated as
( xa − xb ) + ( ya − yb ) + (ta − tb )
2 2 2
Dab = (2.7)
where
Dab is the distance between spatiotemporal mean centers a and b,
a, b represent two different scopes of coordinate ranges, or two different
time periods, etc., and
( xa , ya ,ta ) and ( xb , yb ,tb ) are coordinates of the two mean centers.
Figure 2.3 shows the distance (or movement) between mean center a and
mean center b by a straight line. In this figure, points of the two sets are
represented by triangles of two gray shades and the two mean centers are
the two circles.
22 Spatiotemporal Analytics
V = ( x b − x a , yb − ya , tb − ta ,) (2.8)
∑ ∑ ∑
n n n
wi ( xi − x wmc ) + wi ( yi − ywmc ) + wi ( ti − twmc )
2 2 2
SD = i=1 i =1 i =1
∑
n
wi
i=1
(2.11)
where
( xwmc, ywmc, twmc ) is the coordinate triplet of the weighted spatiotemporal
mean center.
FIGURE 2.4 Standard sphere example. Square is the spatiotemporal mean cen-
ter, solid color circles are inside the sphere, and the empty circles are outside of
the sphere.
Given that the spatial units and temporal units are not the same and are
not directly compatible, we suggest that all coordinates be standardized
prior to the calculation of the deviation, such as the following steps:
Spatiotemporal Centrography and Dispersion 27
FIGURE 2.9 Standard sphere of the 2013 burglary events in Portland, Oregon.
{
Given a set of n spatiotemporal points, P = pi | ( xi , yi , zi )
where i = 1, 2,…, n} coordinates can be standardized as ( xi′, yi′,ti′ ), using
xi
xi′ = (2.12)
x max − x min
yi
yi′ = (2.13)
ymax − ymin
ti
ti′ = (2.14)
tmax − tmin
and using the tools. For software code and example data sets, please email
spatiotemporal.analytics@gmail.com with name, affiliation information, a
copy of purchase receipt and allow up to a week to receive links to down-
load Chapter02.zip. Once downloaded, the .zip files can be uncom-
pressed or restored to any folder on your computer hard drive, such as C:\
sttools\centrography.
This tool supports multiple time formats. Users can select the format
they prefer. In addition to the common year, month, and day, it is possible
to just sequence the events into a numerical order, such as 1, 2, 3, …. Please
note, however, that the sequential order should be in integers only. For this
purpose, data may need to be pre-processed so that temporal data can be
expressed as sequential integers. For example,
TABLE 2.1
Parameters in MeanCenter.bat
Parameter Description
Input point file The name of the input file. The file should
contain spatial coordinates and time
stamps of each spatiotemporal point
X field Name of the x-coordinate
Y field Name of the y-coordinate
Time field Name of the temporal coordinate (time
stamp)
Weight field (optional) Name of the field that contains weights.
This is optional, not required
Time format The format of time, choose among month/
day/year, month-day-year, year/month/
day, year-month-day, or time series in
integers
Time granularity Day, week, month, or year
Output file Folder path and name of output file
Base map (optional) The base map (shapefile format) for
display. This is optional
Create txt file Check to create output in txt format (for
other applications)
Create shapefile Check to create shapefile of the calculated
points
Plot 3D Check to create 3D visualization of the
spatiotemporal centrography
fact very important so as not to derive distorted results from spatial analysis.
Such distortions can be seen in many spatial analyses when the time stamps
of geographic events are not considered. Take, for an example, the hotspots
of burglary events, if not considering when they occurred, can be very differ-
ent from those derived from calculation with temporal data.
A final note on the spatiotemporal analytics discussed in this chapter is
that, selecting appropriate spatial and temporal units for the analysis can be
very important because, like results from spatial analysis at different geo-
graphical resolutions, different results from spatiotemporal analysis may
be expected from using different spatial and temporal units. Unfortunately,
there is not a single criterion that can be used to guide all analyses. Users
should carefully consider the implications of using particular spatiotempo-
ral analytical units in their work.
REFERENCES
Bachi, R. (1963). Standard distance measures and related methods for spatial anal-
ysis. Papers in Regional Science, 10(1), 83–132.
Bujdoso, Z., J. Penzes, S. Madaras, and L. David (2015). Analysis of the spa-
tial trends of Romanian tourism between 2000–2012. Geographia Technica,
10(2), 9–19.
Burt, J. E., G. M. Barber, and D. L. Rigby (2009). Elementary Statistics for
Geographers. Guilford Press.
Dickinson, G. C. (1963). Statistical Mapping and the Presentation of Statistics.
London: E. Arnold.
Lee, J., and D. W. Wong (2001). Statistical Analysis with ArcView GIS. John Wiley
& Sons.
Levine, N. (1996). Spatial statistics and GIS: Software tools to quantify spatial
patterns. Journal of the American Planning Association, 62(3), 381–391.
Lu, T. C. (2011). Cross-correlation networks to identify and visualize disease
transmission patterns (Doctoral dissertation, University of Washington).
Melendez-Pastor, I., E. I. Hernandex, J. Navarro-Pedreno, and I. Gomez (2014).
Socioeconomic factors influencing land cover changes in rural areas: The
case of the Sierra de Albarracin (Spain). Applied Geography, 52, 34–45.
Plane, D. A., and P. A. Rogerson (2015). On tracking and disaggregating center
points of population. Annals of the Association of American Geographers,
105(5), 968–986.
Scott, L. M., and M. V. Janikas (2010). Spatial Statistics in ArcGIS. Redlands, CA:
ESRI Press.
Scott, L., and N. Warmerdam (2005). Extend Crime Analysis with ArcGIS Spatial
Statistics Tools. Redlands, CA: Esri Press.
Thapar, N., D. Wong, and J. Lee (1999). The changing geography of population
centroids in the United States between 1970 and 1990. The Geographical
Bulletin, 41(1), 45.
Wong, D. W. (1999). Several fundamentals in implementing spatial statistics in
GIS: Using centrographic measures as examples. Geographic Information
Sciences, 5(2), 163–174.
Another random document with
no related content on Scribd:
löydetäkään: kissa on käynyt toimittamassa siankin saman tien kuin
edellisetkin.
Nousi siitä nyt tuuminki, mihin tästä olisi paras pakoon mennä ja
mistä suojapaikka saada, kun iltakin oli jo lähenemässä. Kissahan
siihen neuvon tiesi. »Mihinkäpä tästä muualle», sanoi hän,
»mentäisiin kuin sinne Katilan autiotuvalle, jossa on ennenkin
paossa oltu. Eikös tämä kukko hyväkäskin mennä touhottanut sinne
sitä maailman loppuansa pakoon? Niin ettei muuta kuin pian tielle,
että saadaan yöksi kattoa päämme päälle». Ärtyisesti kissa näin
toimitti kaikki matkaan ja niin lähdettiin mennä nykyttelemään,
ensimmäisenä härkä, sitten ruuna, sitten sika ja kukko sekä
viimeisenä pahantuulinen Kääpälän kissa, mauruten ja pitäen pahaa
ääntä.
Ennen vanhaan — alotti härkä — oli tämä Immilä niin köyhä talo,
ettei siinä ollut sanottavasti muuta kuin sija vain. Emäntä kuoli,
lapset menivät mierolle, vain ukko ja yksi härkä jäivät henkiin. Ei ollut
ukostakaan enää paljon mihinkään, vanhaksi kun oli käynyt, mutta
härkäpä sen sijaan olikin vasta uhkea ja komea. Se kun heinikossa
syödä syyhyrytteli ja purra pryyhyrytteli, niin kieli sillä kiersi heinää
kuin kesäorava, häntä heilui kuin Hämehen miekka, pää oli kuin
Jumin kurikka, silmämäljärit olivat suuret kuin renkaat ja sarvet — ne
vasta suuret ja terävät olivat. Ainoana tehtävänä oli Immilän ukolla
nyt tätä härkäänsä hoitaa ja syötellä sekä siinä välissä omienkin
hammasten rakoon murusta haalia.
Illan tullen, juuri kun hän oli kotiin lähtemässä, tunsikin hän, ettei
nyt ollut kaikki oikein hänen läheisyydessään. Hän katsoi tarkemmin
ja eikös ollutkin ilmestynyt siihen mörykkä, itse karhu nimittäin,
katselemaan häntä ja hänen syöntiään. Ei härkä ohtoa kuitenkaan
pelännyt, vaan syödä rouskutteli vain niinkuin ei olisi mitään
tapahtunut ja odotteli, milloin mahtaisi metsän kuningas antautua
niinkuin puheisille. Eipä aikaakaan, niin jo kysyy karhu:
Mutta siitä olikin odottamaton seuraus. Sen sijaan että härkä olisi
tuosta pahoin säikähtänyt ja ruvennut karhulta armoa pyytämään, se
puhaltautuikin vimmattuun laukkaan, ja kun iso härkä oikein laukata
kuopaisee, on sillä selkä kuin mikäkin kiikku ilman kiinnipidintä, niin
että siinä vasta pysytteleminen on. Mörykkä huomasi kohta
tehneensä hullun tempun, ei tiennyt, kuinka siinä parhaiten pidellä
kiinni, kun töppösetkin tahtoivat tahrautua ja ilkeästi tarttua
pikeytyneeseen tervaan kiinni. Hänellä rupesi pian maailma
kieriskelemään silmissä perin pahasti ja hän koetti ärjähdellä härälle,
mutta ei se auttanut. Tämä vain meni yhä huimemmasti, myrisi kuin
maan alta ja heitteli takapuoltaan korkealle ilmaan, niin että karhu
koppina siinä selässä leiskahteli. Mörykkää hirvitti. Hän arveli
heittäytyä tiepuoleen, mutta ei uskaltanut tehdä sitäkään, kun puut,
maa, kivet ja kannot ihan viivoina vilahtelivat ohitse, niin että olisi
saattanut niskansa taittaa, jos niihin olisi paiskautunut. Hän rupesi jo
rukoilemaan härkää. »Heitä minut tuolle kivelle!» pyysi hän, »tahi
tuolle kannolle tahi mättäälle, kuule!», mutta ei härkä muuta kuin
entistäkin hurjemmin laukata kaapaisee. Samaa kyytiä jatkui aina
Immilän tanhualle saakka, jossa vaari ihmeissään oli katsomassa
härän tuloa. Mutta nähtyään puolipökerryksissä olevan karhun
härkänsä selässä ei ukko turhia kysellyt, muuta kuin paiskasi
kontion, ennen kuin se virkosi, kiireesti lujaan karsinaan, löi vankat
salvat eteen ja silitteli hyvillään härkänsä selkää, kun se tällaiset
antimet oli metsästä tuonut. Hyvillään härkäkin muljautteli
silmämäljäreitään ja heilautteli Hämehen miekkaansa, niin että eri
viuhkeen piti käydä siinä tanhualla iltasuitsun ääressä.
»Kaksipa kalosarvustani,
Kaksipa hörppäkorvastani,
Kaksipa silmämäljäriä!
Pää kuni Jumin kurikka,
Häntä kuin Hämehen miekka,
Kieli kuin kesäorava!»
Ohtosen! ainoiseni,
Mesikämmen kaunoiseni,
En sua kiellä kiertämästä,
Enkä käymästä epeä.
Eihän mitään sen kummempaa kuin että kettu liuvari oli sattunut
näkemään kotieläinten kulkueen ja seurannut varovasti perässä
nähdäkseen, minne he oikein olivat menossa. Huomattuaan heidän
menevän Katilan autiotupaan ja kaikesta päättäen aikovan jäädä
sinne yöksi oli ketulle juolahtanut sukkela tuuma mieleen. Tästähän
nyt karhu saisi korvauksen niistä kuolleista hevosista, joista hän sille
oli äsken valehdellut, ja siinä sivussa saisi hänkin, kettu itse, maistaa
hiukan tuoretta lihaa. Tuumasta toimeen! Hän läksi kiireesti mennä
litvittelemään sinne päin, jossa arvasi karhun oleskelevan, eikä
aikaakaan, kun jo löysikin hänet yksinään pimeässä korvessa
nuuhkailemassa. Mutta hän oli nyt pahalla tuulella, kun olivat selkä ja
sivut kipeät sen soiton opettelemisen jäljeltä, ja muljautti vihaisesti
kettuun päin. »Mitä kureja sinulla nyt on?» kysyi hän istahtaen
mättäälle. »Ei, eno kuoma, nyt joudetakaan istumaan!» sanoi kettu,
»nyt onkin vasta maku pirtissä.» — »Missä pirtissä?» kysyi karhu