Dendrochronologia 30 (2012) 57–60
Contents lists available at ScienceDirect
Dendrochronologia
journal homepage: www.elsevier.de/dendro
Technical note
detrendeR – A Graphical User Interface to process and visualize tree-ring data
using R
Filipe Campelo a,∗ , Ignacio García-González b , Cristina Nabais a
a
b
Centro de Ecologia Funcional, Departamento de Ciências da Vida, Universidade de Coimbra, 3001-401 Coimbra, Portugal
Departamento de Botánica, Escola Politécnica Superior, Universidade de Santiago de Compostela, Campus de Lugo, 27002 Lugo, Spain
a r t i c l e
i n f o
Article history:
Received 16 September 2010
Accepted 19 January 2011
Keywords:
ARSTAN
Chronology
Detrending
GUI
R language
a b s t r a c t
In this paper, we present the package detrendeR, a Graphical User Interface to facilitate the visualization
and analysis of dendrochronological data, using the R computing environment. This package offers an easy
way to perform most of the traditional tasks in dendrochronology: detrending, chronology building and
graphical presentation of time series. The advantage of detrendeR, compared with the program ARSTAN,
is the graphical interface that provides the user with an easy way to use R language, rich in graphics and
handling routines, with no need to type commands. The detrendeR uses a simple and familiar dialogbox interface and it can read Tucson decadal-format files (*.rwl and *.crn) as well as plain text files. In
addition, detrendeR has the ability to test temporal changes of the common signal using moving intervals.
The detrendeR should make it easier to perform detrending and chronology building of tree-ring series,
taking advantage of the R statistical programming environment.
© 2011 Istituto Italiano di Dendrocronologia. Published by Elsevier GmbH. All rights reserved.
Introduction
Chronology building and quality assessment are two of the most
important tasks in dendrochronological research. For many years
dendrochronologists have been using ARSTAN, which produces
chronologies from tree-ring series after detrending, and computes
tree-ring indices (Cook, 1985). An autoregressive model can also
be applied either to the index series before averaging all of them,
or to the final chronology, if still showing persistence, and chronology quality is assessed on a common interval including most of
the series. Though very powerful, ARSTAN runs in a Command Line
Interface (CLI), which poses some difficulties to new users. Chronology building and quality assessment can also be carried out by
other command line environments such as Matlab (Meko, 2002)
and more recently by R, with the introduction of recent packages
such as dplR (Bunn, 2008; Bunn, 2010) and bootRes (Zang, 2009). R
is an interpreted programming language with a run-time environment with graphics, a debugger, access to certain system functions
and the ability to run commands stored in script files, allowing the
user to create specific functions and routines to solve their own
problems, being therefore easily extensible (R Development Core
Team, 2009). Moreover, R is open source and thus available under
the GNU license agreement. Several hundred packages are available on CRAN and other sites for free download to R users and even
new statistical methods are often first “published” as R packages
before being adapted to commercial statistical software. R uses a
CLI meaning that commands should be entered into the R console
window to perform specific tasks. This is the preferred interface
for experienced users, with a good knowledge of the R language,
because it allows direct control on calculations and it is flexible.
For introductory, educational and sporadic use of the R language, a
Graphical User Interface (GUI) is particularly attractive. GUIs allow
the user to interact with the computer in more ways than typing,
and therefore the learning time is typically shorter as the user does
not need to remember commands, also decreasing the syntax and
typing errors. In fact, the R language has already some GUIs, like the
Rcmdr (Fox, 2005) and de ade4TkGUI (Thioulouse and Dray, 2009)
packages.
The main purpose of this paper is to introduce a new tool called
detrendeR, which combines GUI with R. This tool uses the most
required statistical tools for detrending tree-ring width series, as
well as chronology building and characterization. The main window of the program supplies a group of menus, buttons, and dialog
boxes to read, manipulate, analyze and visualize data, without the
need to type any command into the R console window. In addition, users with strong knowledge of R language can also type R
commands directly in the R console in association with detrendeR.
detrendeR
∗ Corresponding author.
E-mail address: fcampelo@ci.uc.pt (F. Campelo).
To correctly work under Windows, the detrendeR GUI requires
the single-document interface (SDI) to R, so that the detrendeR and
1125-7865/$ – see front matter © 2011 Istituto Italiano di Dendrocronologia. Published by Elsevier GmbH. All rights reserved.
doi:10.1016/j.dendro.2011.01.010
58
F. Campelo et al. / Dendrochronologia 30 (2012) 57–60
Fig. 1. The detrendeR window at startup.
R console windows will float freely on the desktop. To open the
detrendeR GUI the user should first install the detrendeR package.
Once you have loaded the package, you can have the detrendeR
main window visible (Fig. 1) by typing in the R console window
the command:
>detrender()
To demonstrate how detrendeR works, we have used Schulman’s Mesa Verde Douglas fir (Pseudotsugamenziesii [Mirb.]
Franco) data, from the International Tree-Ring Data Bank
(Schulman, 1963) and also included within the dplR package, as
the co021 dataset (Bunn, 2010). In the examples throughout this
paper, we assume that the active dataset is the co021, read in
the co021 dataset by typing the following code in the R console
window:
>data(co021)
Most functions of dentrendeR can be easily used from the main
window. This window contains the menu bar, and three toolbars
below, each comprising several buttons (Fig. 1). The three toolbars, from up to down, are aimed at: (i) managing the data set of
tree-ring series to be processed; (ii) provide general information
about the active data set; (iii) apply the statistical functions related
to detrending and chronology building. The order of the different
bars corresponds to the consecutive steps the user should follow
to build a chronology.
The File menu provides functions to read and save files, to exit
detrendeR and to Quit R:
- “Read file” allows to read a data file and store the information in
a new dataset. The option “clipboard” can be used to read a data
table just copied from a spreadsheet program (like Excel).
- The items “Read rwl” and “Read crn” open a dialog box that allows
to read data files in the Tucson measurement (*.rwl) and in the
Tucson chronology format (*.crn).
- The items “Save rwl”, “Save crn” and “Save csv” allows to save the
active dataset into different formats (*.rwl, *.crn, *.csv).
- “Save Workspace. . .” saves the current workspace to the specified
file. The saved objects can be read back from the file later by using
the function load.
- “Quit detrendeR” closes the detrendeR window.
- “Quit R” opens a dialog box to ask if the environment should be
saved before terminate the current R session.
Using the File menu, datasets can be opened from different files,
and a large variety of file types are supported, including Tucson
measurement (*.rwl) and Tucson chronology format (*.crn). To read
a data file or data from the clipboard into R, select File → Read file.
This operation brings up a dialog box, as shown in Fig. 2. The default
name of the dataset is the name of the file to be opened, but the
user is allowed to change it. In R the names of datasets must start
with a letter and consist entirely of letters, digits, periods (.) and
underscores ( ). You should also remember that R is case-sensitive
and embedded blanks are not allowed in a dataset name. The active
Fig. 2. Reading data from a text file or from the clipboard.
dataset can also be saved into several formats (*.rwl, *.crn, *.csv)
using the items Save rwl, Save crn and Save csv in the File menu.
The Tools menu allows to define settings and to launch detrendeR in batch mode:
- “Define settings” brings up the dialog box shown in Fig. 3. The user
can use this dialog box to define tree mask, type of detrending and
other settings.
- “Batch mode” releases a dialog box identical to the one used to
define settings, but in this case the user can choose the files to be
processed by pressing the button “Ok”.
The active data set is shown in the first toolbar, immediately
below the menu bar. This bar has two buttons, the first indicating the name of the active dataset, or the label “<Please select a
dataset>” or “<No active dataset>” if no dataset is selected (Fig. 1).
The user can load several datasets in memory, and change the
active dataset just by clicking the flat button with the active dataset
name. However, at any given time, only one dataset is active. Once
selected, the subsequent functions are only applied to the active
dataset. The second button, Delete, allows the user to remove the
active dataset and/or other datasets from the R environment.
The second toolbar contains four buttons related to general
information about the active dataset (Fig. 1). The “Information”
button displays the series identification, first and last year, and
length of the series. The button “TreeIds” provides the tree mask of
the active dataset. “Missing rings” indicates the existence of missing rings within the series. The last button “RwlInfo” computes
some common descriptive statistics on individual series, such as
the correlation with the master chronology, mean, median, standard deviation, mean sensitivity and first-order autocorrelation,
and prints them to the R editor window, using the RwlInfo function.
F. Campelo et al. / Dendrochronologia 30 (2012) 57–60
59
Fig. 4. The Detrending options dialog box.
Fig. 3. The dialog window used to define settings for detrendeR.
The lowest toolbar constitutes the core of the program, and contains four buttons that brings up different dialog boxes providing a
variety of mathematical and statistical functions for trend removal,
chronology building and assessing the statistical confidence of a
chronology.
The button “Detrending” extends a drop-down menu with
two additional commands: “1 step” and “2 steps”, that open the
Detrending options dialog box window (Fig. 4). In the first case, a
single detrending method is applied to the selected dataset and two
new datasets are added to the R environment, the sufixes “.cv1” and
“.in1” are placed after the name of the original dataset to identify
the curve and the index datasets, respectively. In the second case,
four new datasets are produced by a two-step detrending, having
the suffixes “.cv1”, “.cv2”, “.in1” and “.in2”.
detrendeR provides four different detrending methods: modified negative exponential, cubic smoothing spline, simple linear
regression and through the mean. The spline algorithm used was
the Andrew Bunn’s fsscap function from the dplR package (Bunn,
2008). In the dialog box window “Detrending options” (Fig. 4)
the checkbox “Interactive detrending” allows the user to verify
how well the detrending curve fits each series, and use different methods for different series (Fig. 5). The smoothing spline has
the parameter ‘bandwidth’ to modify the trend elimination. Large
bandwidths lead to a stiff trend line while a small bandwidth adapts
smoothly to the time series, the effect of different bandwidth can
be easily observed by applying the interactive detrending.
The interactive detrending window can be closed at any time by
pressing the button “Close without saving” or the button “Close and
Save changes”. The statistics of the detrended series are displayed
in the R console window, by applying the RwlInfo function to the
index series.
The button “AR model” can be used to remove the autocorrelation from each series, using the R function ar. The maximum order
to be applied during the univariate autoregressive process is chosen by the user, but the selected order for each series is determined
by the first minimum Akaike Information Criterion. The resulting
series without persistence will be stored into a new dataset, hav-
Fig. 5. Interactive detrending of individual tree-ring series.
60
F. Campelo et al. / Dendrochronologia 30 (2012) 57–60
ing the suffix “.res” placed after the name of the input dataset by
default.
The button “Chrono” can be used to produce chronologies
using the internal function Chron. This function combines treering index or raw-width series into a mean value chronology
by averaging each year using an arithmetic mean or a biweight
robust estimate of the mean. If the series used to produce the
chronology are detrended (standardized) a Standard chronology will be produced. A Residual or a Prewhitened chronology
is produced when the averaged series are residuals from the
autoregressive model of the detrended series. Usually, this chronology shows a strong signal without persistence, however if some
persistence remains, an autoregressive model can be applied to
remove it.
The last button, “EPS”, produces several statistics that indicate
the common signal to all series using the mean correlation between
trees (rbt) and the Expressed Population Signal (EPS). By pressing this button the EPS analysis dialog box is launched and allows
performing three analyses simultaneously, using the EPS.value
function, a changed version of the rwi.stats function from the
dplR package (Bunn, 2008). This function provides a variety of
statistics, such as the mean within- and between-tree correlation
(rwt, rbt) and the EPS (for a better explanation of the algorithm see
Briffa and Jones, 1990). In the“Common interval” analysis only the
period where all series are represented are used to calculate the rbt
and EPS values. The user can choose a certain time span and determine the rbt and EPS values for that period. The output is printed
in the R console window. The analysis can also be performed for
a specified length (“Window length”) and slide this window with
regular steps (“Lag”).
There are two ways to end the detrendeR session. The user
can select File → Quit detrendeR and will be asked whether the R
workspace should be saved. The R session will be kept working
and the detrendeR can be started later by writing the command
detrender() in the R console window. The user can also select
File → Quit R and, in this case, the program will ask whether to
save the R workspace (i.e., the data that R keeps in memory). This
allows the user to maintain different saved workspaces for different
projects.
Conclusions
The detrendeR performs some of the functionalities offered by
ARSTAN and dplR but under GUI, using the R open-source statistical computing environment. As other softwares detrendeR can
easily analyze temporal changes of the common signal using moving intervals. Other functions can be added and we encourage other
researchers to participate in the development of detrendeR.
Availability
The detrendeR package is available as an add-on package in R.
Interested users can download and install R from the Comprehensive R Archive Network website: http://cran.r-project.org/. Within
R, detrendeR can be installed and loaded via:
>install.packages(“detrendeR”)
>library(detrendeR)
References
Briffa, K., Jones, P.D., 1990. Basic chronology statistics and assessment. In: Cook,
E.R., Kairiukstis, L.A. (Eds.), Methods of Dendrochronology: Applications in the
Environmental Sciences. Kluwer Academic Publishers, Dordrecht, pp. 137–152.
Bunn, A., 2008. A dendrochronology program library in R (dplR). Dendrochronologia
26, 115124.
Bunn, A., 2010. Statistical and visual crossdating in R using library. Dendrochronologia 28, 251–258.
Cook, E.R., 1985. A time series analysis approach to tree ring standardization. Ph.D.
Dissertation. University of Arizona, Tucson.
Fox, J., 2005. The R commander: a basic-statistics graphical user interface to R.
Journal of Statistical Software 14, 142.
Meko,
D.,
2002.
Tree-Ring
MATLAB
Toolbox.
http://www.mathworks.com/matlabcentral/fileexchange/1347-tree-ringmatlab-toolbox.
R Development Core Team, 2009. R: A Language and Environment for Statistical
Computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3900051-07-0, http://www.R-project.org.
Schulman, E., 1963. Schulman Old Tree No 1 Data Set. IGBP PAGES/World Data Center
for Paleoclimatology Data Contribution Series 1983-CO021.RWL. NOAA/NCDC
Paleoclimatology Program, Boulder, Colorado, USA.
Thioulouse, J., Dray, S., 2009. ade4TkGUI: ade4 Tcl/Tk Graphical User Interface. R
package version 0.25. http://CRAN.R-project.org/package=ade4TkGUI.
Zang, C., 2009. bootRes: the bootRes Package for Bootstrapped Response and Correlation Functions. R package version 0.1. http://www.Rproject.org.