EGU, VIENNA 2008 - Poster XY0226

NEW SOFTWARE FOR RAPID, SEMI-AUTOMATED ANALYSIS OF LAMINATED CLIMATE-ARCHIVES:

THE WinGeol Lamination Tool
1* 2 1
M. Meyer , R. Faber and C. Spötl
www.terramath.com 1
Institut für Geologie und Paläontologie, Leopold Franzens Universität Innsbruck, Austria.
2
TerraMath,Heiligenstädterstrasse 107/30, 1190 Vienna, Austria (robert.faber@terramath.com)
* Current address of corresponding author: School of Earth and Environmental Sciences,
University of Wollongong, Northfields Avenue, Australia (meyer@uow.edu.au)
Der Wissenschaftsfond

Case studies A mfw 22 / mfd 25 / area off / noisred off /

smoothing 0 / buffer 0 / greyscale
1mm B mfw 8 / mfd 3 / area on / noisred on /
smoothing 5 / buffer 8 / red

Introduction We tested our software on several laminated samples from four different total hit rate: 100%
accuracy: 100%
climate archives : total hit rate: 94% accuracy: 82% 500µm

grayscale
intensity
Geological and biological archives showing an annually laminated internal structure are currently top priority in Example A: Tree rings of a silver fir. The greyscale profile is 200
200

characterized by plateau-like features intersected by sharp drops in the 150 50

palaeoclimate research, as they are recognized as very high-resolution archives of environmental change. Also, greyscale values. All laminae are detected accurately by a mfw of 22 and x-axis x-axis

a mfd of 25 pixel. No further tuning of the detection parameters is needed.

the annual origin of laminations can be validated by absolute age dating or by using instrumental data for the most .
C
Example B: Lacustrine pearl oyster. Lamination in the profile view is mfw 6 / mfd 10 / area on / noisred on / smoothing 2 / buffer 2 / blue

recent period. Microscopic laminae may span several hundreds to thousands of years and frequently reveal a characterized by high amplitude variations best seen in the red colour
channel. The total hit rate is good (94%) and the accuracy of 82% implies
high degree of internal growth variability. Quantitative examination of laminations using transmitted-light or that two out of ten semi-lamina boundaries need adjustment.
.
total hit rate: 80% accuracy: 98% 1cm

epifluorescence microscopy is thus a tedious task and may be partly automated. We developed software - Example C: Banded lake sediments (Lisan Fm). The fb-algorithm 160

intensity
color
yields a total hit rate of 80% and a high accuracy of 98%. The low hit rate is 50

(WinGeol Lamination Tool, Meyer et al., 2006) using C++ capable of semi-automatically detecting and mainly caused by clusters of very closely spaced faintest laminae which
are sometimes misinterpreted as individual lamina. In these cases manual x-axis

measuring individual lamina thicknesses in archives showing large internal growth variability. The Lamination readjustment is required.

Tool enables the operator to efficiently and quantitatively examine laminae down to the micron scale and it was
successfully tested on a variety of annually banded samples, including lake sediments.
Complex lamination 150

60
130

A laminated stalagmite from Ernesto Cave (Italy) records the climate 10

deterioration during the Little Ice Age (LIA) where speleothem deposition LIA (1850 AD)
nearly came to halt (asterisk, Frisia et al., 2003).
data segments

The Workflow The Algorithm We compare our semi-automatically derived results from the WinGeol
A no-data link

Bottom
segment
segments segment
lamination tool (Figure A-F) with the manually measured thickness curve

Top
2
of Frisia et al. (2003. Figure G). Two photo-mosaics, obtained with 1
Fig. C Fig. D
transmitted light microscopy (Fig. A) and epifluorescnce microscopy
(Fig. B), were analysed. B

Bottom
RGB or greyscale image
(WinGeol workspace)

whole lamina Fig. F

Top
LOAD DIGITIZE DEFINE Lamination is dominated by thick organic-rich dark bands prior to 1850 3 Fig. E
DIGITAL IMAGE 1mm
POLYLINE SEGMENT TYPES AD while extremely faint organic layers were deposited thereafter (Fig. 4
A). Gaps along the transect line (no-data segments) arise from bad
EVALUATE PARAMETERS semi- lamina preservation. Sharp peaks and high amplitudes prevail in the
ALONG TEST PROFILE LINE lamina transmitted-light images (A, C, D) and a rather smooth RGB signal with
(optional)
polyline line lower amplitudes is observed in the UV images (B, E, F). We used four 70 160
polylines (1 - 4 in Fig. A and B) and tuned the detection parameters of the
SET PARAMETERS
algorithm for each of these four lamination-types. 10 10
MIN FEATURE WIDTH REDUCE .

(mfw) NOISE The the semi-autmoatically derived lamina record (black line, Figure G)
MIN FEATURE DEPTH compares well to the manually measured lamina thicknesses (white C 100 µm D 100 µm
(mfd) USE AREA lamina boundaries
SELECT BAND max min curve). Lamina thickness for this cave are supposed to be a proxy for
(WinGeol profile window)

(R, G, B or RGB) SMOOTH 255 local temperature (black dotted line. Frisia et al., 2003).
RGB or greyscale porfile

feature
BETWEEN
EXTREMA width max
color or greyscale intensity

FIND LAMINA BOUNDARIES BUFFER G adjusted lamina boundaries - epifluorescence image

ALGORITHM feature
(Fig: The Alogrithm) Tuning
depth
Parameters
total hit rate ~ 90% total hit rate ~ 80% mfw 0.018 / mfd 8 / area off / noisred off / mfw 0.02 / mfd 12 / area on / noisred on /
accuracy ~ 97% smoothing 2 / buffer 25 / blue smoothing 1 / buffer 3 / blue
accuracy ~ 100%
minor readjustments readjustments
(µm) (T, °C)
estimated boundary:
CHECK ADJUST RECALCULATE ASCII min + (max-min) 180 15
E 100 µm F 100 µm
RESULT BOUNDARIES BOUNDARIES EXPORT
2 mean elevation between 140 14
two lamina boundaries 0
100 13
60 12
A digital image is loaded into the WinGeol workspace and a vector The algorithm compares the RGB values of neighbouring pixel and 20
layer generated on which the user digitizes a polyline by defining data-,
11
calculates minima and maxima along the RGB curve (or greyscale curve in 2000 1950 1900 1850 1800 (y.A.D.)
no-data- and link-segments in order to maintain control over laminated the case of greyscale images). The horizontal and vertical distances mfw 0.013 / mfd 2 / area on / noisred on/ mfw 0.02 / mfd 3 / area on/ noisred on/
smoothing 1 / buffer 25 / blue smoothing 1 / buffer 5 / blue
and non-laminated sequences. Data segments cover laminated between successive minima and maxima are memorized as feature widths
sequences, link segments connect two data segments with each other and feature depths. Next, the RGB curve is screened for features above a
and no-data segments are used where lamination is obscured or absent. certain (user defined) minimum width and minimum depth. Features
greater than the minimum feature width (mfw) and the minimum feature
The RGB curve (pixel values for the Red, Green and Blue channel) is
depth (mfd) are recognized as valid and a lamina boundary is placed half Laminae counts and thickness data can be obtained even for archives
obtained from the digital image along the course of the polyline. The
lamina detection algorithm (see next Fig.) runs along the datasegments
only (using the RGB curve as computational base). A number of
way between the minimum and the maximum.
Conclusions with high internal growth variability and faint laminae and multiple and
parallel profile lines can be measured rapidly thus enhancing the
Essentially, the algorithm places boundaries at lightdark and darklight
parameters can be adjusted by the user (e.g. smooting, buffering of robustness of the final results.
transitions and thus detects “semi-laminae” from which real lamina The WinGeol lamination tool is capable of accurately detecting
polyline etc.) to enhance the algorithm efficiency and optimize the final
results. The algorithm suggests lamina boundaries which can be thicknesses (whole lamina) are obtained. We thus morphologically lamina boundaries within a wide range of laminated archives.
In addition, to lamina thicknesses, thickness variations of semi-
evaluated by the user and individual boundaries can be shifted, deleted screen the RGB curve for features above a certain treshold size, rather
than following the conventional approach of calculating a curve function. Lamination analysis using the WinGeol lamination tool is a semi- laminae (e.g. winter and summer layers constituting one laminae i.e.
or missing boundaries can be added as required, thus providing the one year) are retrieved as well and can be used e.g. as a seasonality
highest degree of flexibility to the operator. The lamina detection process is thus straightforward and robust. automated process and offers a high degree of control to the
operator. proxy.

REFERENCES:
Meyer, M.C., Faber, R., and Spötl, C. (2006): The WinGeol Lamination Toolnew software for rapid, Frisia, S., Borsato, A., Preto, N. and McDermott, F. (2003): Late Holocene annual growth in three Alpine stalagmites records
semi-automated analysis of laminated climate archives. The Holocene. 16/5, 1-9. the influence of solar activity and the North Atlantic Oscillation on winter climate. Earth and Planetary Science Letters 216, 411 24.