Access to this full-text is provided by De Gruyter.
Content available from Holzforschung
This content is subject to copyright. Terms and conditions apply.
Wood Growth/Morphology
Max L. Flaig, Jens Berger, Stephanie Helmling, Andrea Olbrich, Heinz J. Schaffrath, Daniel Zahn
and Bodo Saake*
Chemotaxonomic and anatomic wood species
identification in bleached pulp: blind test and
method validation
https://doi.org/10.1515/hf-2024-0025
Received April 4, 2024; accepted July 3, 2024;
published online August 1, 2024
Abstract: This paper presents a comparative analysis of
the blind test outcomes of two independent methods for
the identification of tropical wood species in pulp and pa-
per products. Both, the established anatomical and the rela-
tively new chemotaxonomic method support the European
Deforestation Regulation 2023/1115 (EUDR), which aims to
ensure that only legally harvested timber that has not
contributed to deforestation is traded in the EU. The blind
test involved 570 decisions on 15 test sheets of 37
self-manufactured mixed tropical hardwood pulps and an
industrial beech pulp, used as a matrix. Both detection
techniques demonstrated robust performance with over
80 % hit rates. The results show that the synergies and
combination of the strengths of both methods can be utilized
and lead to even better combined performance. In order
to establish the chemotaxonomic identification method as a
complement to the conventional anatomy-based method,
statistical analyses were performed to assess its intermedi-
ate precision between three different GC-MS systems. In
most cases, the method gave consistent results independent
of the instrument used.
Keywords: CITES; database; GC-MS; mixed tropical hard-
wood; pulp and paper; wood anatomy
1 Introduction
Pulp and paper are primarily derived from renewable
wood resources. The raw material plays a pivotal role in
sustainable paper production. To address concerns related
to biodiversity conservation and the preservation of crucial
rainforest ecosystems, it becomes imperative to substantiate
the identification of tropical wood species in pulp and pa-
per products, along with the recognition of Convention
on International Trade in Endangered Species of Wild
Fauna and Flora (CITES)-listed timber species in global trade.
In alignment with the European Union’s Deforestation
Regulation 2023/1115 (EUDR), the placement of unlawfully
harvested timber on EU internal markets is explicitly
prohibited (European Union 2023). This regulation replaces
the European Timber Regulation (EUTR) that has been in
effect since 2013 and is essentially an additional extension to
deforestation-free agricultural supply chains (Köthke et al.
2023). But still the regulation’s purview encompasses timber,
timber products, as well as pulp and paper, beginning at
the point of their initial introduction to the EU markets.
In light of the substantial consumption of wood
resources in paper manufacturing (Windhagen et al. 2019),
the ability to ascertain the composition of paper products
assumes paramount significance. An essential objective
shared by both anatomical and chemotaxonomic method-
ologies is to systematically bolster the enforcement of the
EUDR. Adherence to this regulation necessitates market
participants to exercise due diligence.
To date, the lack of accessible reference samples of
relevant species and analytical methodologies designed for
the verification of manufacturer-supplied data has been
constantly shrinking. Forensic identification techniques for
*Corresponding author: Bodo Saake, Institute of Wood Science,
Chemical Wood Technology, University of Hamburg, Haidkrugsweg 1,
22885 Barsbüttel-Willinghusen, Germany, E-mail: bodo.saake@uni-
hamburg.de
Max L. Flaig and Jens Berger, Institute of Wood Science, Chemical Wood
Technology, University of Hamburg, Haidkrugsweg 1, 22885 Barsbüttel-
Willinghusen, Germany. https://orcid.org/0009-0000-0585-1734 (M.L. Flaig)
Stephanie Helmling and Andrea Olbrich, Johann Heinrich von Thünen
Institute (TI), Institute of Wood Research, Leuschnerstr. 91, 21031 Hamburg-
Bergedorf, Hamburg, Germany
Heinz J. Schaffrath, Papierfabrikation und Mechanische
Verfahrenstechnik, TU Darmstadt, Alexanderstr. 8, 64283 Darmstadt,
Germany
Daniel Zahn, Physikalische Analytik und Materialprüfung,
Sterilisierverpackung, ISEGA Forschungs- und Untersuchungsgesellschaft
mbH, Zeppelinstraße 3, 63741 Aschaffenburg, Germany
Holzforschung 2024; 78(9): 487–502
Open Access. © 2024 the author(s), published by De Gruyter. This work is licensed under the Creative Commons Attribution 4.0 International License.
wood, can broadly be categorized as anatomical, chemical,
and genetic methodologies (Flaig et al. 2023; Low et al. 2022;
Lowe and Cross 2011; Schmitz et al. 2020). Solid wood is
generally easier to identify using different methods with
similarly good results (Ravindran and Wiedenhoeft 2020).
Even some endangered tree species can already be classified
using deep learning methods (Zheng et al. 2024). Wood-based
products such as pulp, paper and particle board tend to be
more difficult to identify than solid wood (Sieburg-Rockel
and Koch 2020). The selection of the most appropriate tech-
nique depends on the specific identification inquiry in
question (Braga et al. 2020; Schmitz et al. 2019, 2020). Notably,
for the genus-level identification of pulp and paper products,
exclusively anatomical and chemotaxonomic approaches
prove applicable. Wood anatomy provides robust perfor-
mance, delivering accurate results when applied on com-
mercial paper samples at the genus taxonomic level. It also
benefits from a well-established and extensive database
of reference materials (InsideWood 2004; Richter and Dall-
witz 2000; Wheeler 2011), which can be supported by deep
learning techniques (Nieradzik et al. 2024). Chemotaxonomy
based on pulp extractives/chromatographic fingerprint
data represents a relatively nascent field, with its database
currently containing a modest 38 reference extracts from 37
tropical plant pulps and 1 beech pulp (Flaig et al. 2023). For
the validation of the chemotaxonomic identification method
of wood taxa in pulp and paper, it is crucial that the method
is independent of the GC-MS measurement system used. In
general, the variability between repeated measurements
is referred to as precision. The intermediate precision
lies between the reproducibility with maximum variability,
defined as “the closeness of agreement between indepen-
dent results obtained with the same method on identical test
material but under different conditions (different operators,
different apparatus, different laboratories and/or after
different intervals of time)”, and the repeatability with
minimum variability under constant conditions (Gold 2019;
ISO 1994).
Across the entirety of the paper manufacturing process,
which encompasses stages ranging from pulping to various
bleaching procedures, the chemicals employed in these
processes effectively obliterate the DNA originally present
in solid wood. Although anatomical identification through
the examination of morphological attributes of wood vessels
has been well-established (Helmling et al. 2016, 2018; Ilves-
salo-Pfäffli 1995), complications can arise, especially when
fibers undergo modification through refining - a common
occurrence in paper production. This refining process
typically results in the destruction of residual vessel
elements (Helmling et al. 2018). Furthermore, some wood
genera exhibit limited distinctive anatomical structural
features, making them susceptible to confusion with other
genera, including some that are taxonomically distant
(Gasson 2011). A significant advantage of the chemotaxo-
nomic approach lies in its immunity to the mechanical
alterations of anatomical features, as can occur during fiber
refining. In return identifying wood genera in pulp through
chemotaxonomic methods is challenging due to the minimal
extractive content. In industrial pulping and bleaching
processes, extensive efforts are made to eliminate wood
extractives to enhance paper quality and streamline
production. Moreover, the composition of the extractives in
pulps differs from that in solid wood.
Typically, chemotaxonomy, or fingerprint analysis, has
been employed to classify and distinguish plants based on
their secondary compounds (Alaerts et al. 2014). While prior
chemotaxonomic research primarily focused on solid
wood, this study conducts chemotaxonomic and anatomical
analysis on the pulp and paper derived from tropical
and temperate wood genera (including the three monocot
plants Bamboo, Coconut and Oil palm).
2 Materials and methods
Obtained from various mainly commercial sources (lacking
documented origin), a collection of 37 distinct solid mixed
tropical hardwood (MTH) samples (Table 3, first column),
each weighing around 2 kg, was amassed. The relevant
genera and species were chosen in cooperation with the
NGOs Greenpeace and World Wide Fund for Nature (WWF).
The focus of the wood selection was on the presence of
woods in South East Asia, as this region and its wood species
were found to be of particular relevance in the research
carried out as part of the ‘Detection of Tropical Hardwood in
Paper - Chemotaxonomy and Anatomy for the Identification
of Mixed Tropical Hardwood’project (Saake et al. 2021).
Wood anatomists, aided by the resources of the Xylothek at
the Thünen Institute in Hamburg, Germany, carried out the
morphological identification of the genera of these wood
samples.
2.1 Blind test
Subsequently, each specific reference wood belonging to a
different genus/subgenus/species underwent a pulping and
bleaching process. The wood samples were cut into 3 ×3cm
chips and kraft pulped using a 7-L M/K digester (M/K Systems
INC., Williamstown, USA). Pulping involved 550–1,000 g
wood chips, 4:1 liquor-to-wood ratio, and 22–25 % NaOH
with 35 % sulfidity. The process included 90 min at 165 °C.
488 M.L. Flaig et al.: Blind test of two wood taxa identification methods
The produced pulps underwent slot screening and a 5-stage
bleaching sequence, targeting 90 % ISO brightness. Bleach-
ing stages included oxygen (O), complexing agent (Q),
oxygen-enhanced peroxide (OP), chlorine dioxide (D), and
peroxide stages (P). This laboratory-scale kraft pulping
mirrors industrial standards, yielding modern ECF pulps.
A detailed overview of pulping and bleaching methods
as well as a discussion on the chosen parameters is given
by Flaig et al. (2023).
The fully bleached kraft pulps were then utilized for the
production of the blind pulp mixture samples. Therefore, the
reference pulps (Table 3, first column) were entrusted to an
external paper research institute (Hochschule München).
There, the blind samples were created with specific com-
positions without revealing these to the participating
working groups. A total of 15 blind samples (Figure 1a) were
mixed according to the guidelines for the production of the
blind samples by the external institute. From each blind
sample round test sheets (Figure 1b) were produced.
2.2 Guidelines for the production of the
blind samples
–The beech pulp served as the matrix pulp into which the
other pulps were mixed.
–All pulps were included at least once in a blind sample
with a minimum of 5 %, but other proportions such as
10 % and 20 % were also used.
–As the four subgenera of Shorea are difficult to
distinguish, they were included in the samples also as
mixtures.
–Due to its CITES II protection status, at least four samples
containing Gonystylus spp. were prepared.
–Gonystylus spp. can easily be confused with Durio
spp. and Lophopetalum spp. and were therefore
included in the samples separately and in mixtures.
–Swintonia spp. was used when Durio spp. was used
because the Durio spp. pulp was contaminated with
traces of Swintonia spp.
–Until the evaluation of the blind test, the information on
the composition was kept secret.
All participating entities, including Thünen Institute (TI),
Technical University of Darmstadt (TUDa), ISEGA and Uni-
versity of Hamburg (UHH), were informed about these
guidelines. The test sheets were divided and made available
to the participating institutes. Additionally, the TUDa and
ISEGA received the ‘Atlas of vessel elements’in digital format
(Helmling et al. 2018), along with descriptions of unpublished
references for six pulps.
2.3 Chemotaxonomy
2.3.1 Grinding
The blind test sheets were milled in a CryoMill (Retsch
GmbH, Haan, Germany), a liquid nitrogen-cooled ball mill,
at a frequency of 25 Hz. Each sample was milled five times
for 1 min, with cooling applied for 0.5 min between cycles
according to Flaig et al. (2023).
2.3.2 Extraction
Freshly distilled n-hexane (VWR International, LLC., Radnor,
USA), equivalent to GC-grade, was used. Extraction was
performed using a Soxtherm SOX 6 apparatus (C. Gerhardt
GmbH, Königswinter, Germany) with an automated
MULTISTAT control system. The ground blind test sheets
(5 g per cellulose extraction thimble) were extracted with
Figure 1: Blind test: (a) exemplary blind pulp mixture sample; (b) test
sheets.
M.L. Flaig et al.: Blind test of two wood taxa identification methods 489
140 ml of solvent. The extraction parameters described in
detail by Flaig et al. (2023) resulted in a total extraction time
of 4 h 7 min. Extracts were then adjusted to a final volume
of 50 ml.
2.3.3 Sample preparation
The extracts were applied to the pyrolysis crucibles for
TD-GC-MS analysis by evaporating the solvents of the applied
extract solutions. This process was repeated with adjusted
volumes until the desired dry mass of 90 ±5 µg per crucible
was obtained.
For the analysis of the system independence of the
chemotaxonomic method (Section 3.2), six randomly
selected mixed blind samples and seven randomly selected
pure samples were uniformly prepared and analyzed as
described above and below. The samples were prepared
and analyzed by the same operator in the same laboratory,
with more than one year elapsing between measurements
on GC-MS system A and systems B and C. From each sample
extract, eight pyrolysis crucibles were prepared and
measured.
2.3.4 TD-GC-MS systems and parameters
The TD-GC-MS analysis systems A and B consisted of a
Frontier Laboratories Ltd. micro furnace Double-Shot
Pyrolyzer (Py-2020iD) with an Auto-Shot Sampler (AS-1020E).
They were coupled to Agilent Technologies Inc. GC-MS: system
A (6890 N/5973 N), system B (6890 N/5973 inert). System C
consisted of a Frontier Laboratories Ltd. Multi-Shot Pyrolyzer
(EGA/Py-3030D) with an Auto-Shot Sampler (AS-1020E)
connected to an Agilent Technologies Inc. GC-MS (8890/5977B).
The TD temperature was 325 °C, while the interface was
set to 330 °C. For polyethylene (PE) pyrolysis, the pyrolizer
operated at 500/600 °C. The GC inlet and the MS interface
temperature were maintained at 320 °C. A 30 m ×0.25 mm i.
d. and 0.25 µm film thickness low polarity column (ZB-5HT
7HG-G015-11, Phenomenex Inc., Torrance, USA) was
used with helium as carrier gas, split ratio set to 20:1, and a
constant flow rate of 1 ml/min.
The GC oven temperature started at 45 °C for 2 min, then
increased to 340 °C with decreasing heating rates and was
maintained at this temperature for 30 min. The total analysis
time was 134.17 min. The GC oven temperature settings
followed Flaig et al. (2023). An electron impact ionization
energy of 70 eV was used for mass spectral detection.
The measurement scanned a range of 29–700 m/zin total
ion current (TIC) mode with a threshold of 100. Data was
recorded with a scan frequency of 2.22 scans/second
and sample rate [N=2].
2.3.5 Data processing
Each blind test pulp extract underwent a minimum of two
analyses. One representative chromatogram from each set
was selected for subsequent analysis. The chromatogram
was subjected to pretreatment procedures such as rounding
and smoothing of mass traces, peak deconvolution, inte-
gration and retention index (RI) calculation and finally
matched against the reference database of single-variety
wood pulp extracts (DB) following Flaig et al. (2023).
For the analyses of the system independence of the
chemotaxonomic method (Section 3.2), the preprocessing
method for the raw GC-MS data and the DB settings were
slightly adjusted compared to the parameters of Flaig et al.
(2023) used for the blind test (Section 3.1). Table 1 shows
the updated DB settings and Table 2 a short version of
the updated preprocessing steps. The full version of the
preprocessing operations in tabular form is provided as
Supplementary Table S1 and in OpenChrom Method (.OCM)
file format for use with the software as Supplementary
File S1.
The process of identifying a pretreated unknown mixed
or pure sample extract chromatogram containing informa-
tion on the deconvoluted peaks with individual mass traces,
RI and peak areas involved aligning the chromatogram using
an RI corridor of ±40, followed by searching for peak
matches using the NIST algorithm (Stein and Scott 1994)
based on mass spectral match quality. A minimum match
factor (MF) of 70 % was used for peak comparisons due to
potential impurities and low intensity mass traces of small
peaks in the mixtures. The cosine algorithm (Alfassi 2004)
Table :Updated pulp extract reference database settings.
Name Description Setting
Min match factor (MF) Matches below the given value
are ignored
.%
Min reverse match factor Matches below the given value
are ignored
.%
Mass spectrum
comparator
Comparison algorithm used NIST (identity
normal)
Delta window RI corridor used for peak
comparison
.
Calculate penalties False
Retention index (RI) limit The penalty is applied on all
values outside of this RI
window
.
Penalty level factor Calculates an MF penalty,
depending on RI delta
between peaks
.
Max penalty Maximum penalty applied on
the MF
.%
490 M.L. Flaig et al.: Blind test of two wood taxa identification methods
was used for whole chromatogram comparison and the
OpenChrom®software (version: 1.5.0.20231117-1619, Lablicate
GmbH, Hamburg, Germany) was utilized for all preprocessing
and DB operations.
2.4 Anatomy
2.4.1 Maceration
For the purpose of comparing unknown paper samples and
analyzing the blind test samples in this paper, anatomical
references were established. To accomplish this, samples of
all reference woods were macerated according to Franklin
(1945). The maceration made the cell structure, especially the
wood vessel elements and their individual features, more
visible under a microscope.
2.4.2 Sample preparation
The samples were prepared analogous to the description of
the preparation of fibrous materials in the ‘Atlas of vessel
elements’(Helmling et al. 2018). For the hardwoods, vessel
elements were systematically separated from the samples,
prepared on microscope slides, and embedded. Since soft-
wood species do not contain vessel elements, the reference
permanent slides for Cunninghamia lanceolata were pre-
pared from its tracheids. Subsequent visual documentation
was performed using a light microscope. Initially, micro-
scopic images of at least 36 vessel elements from the
Table :Updated raw data preprocessing operations (short version); underlined parts were adjusted.
Name Description Settings
Scan cleaner (remove empty) This filter removes empty scans
High pass ions This filter preserves the n highest ions per
scan
{"Number Highest":}
Ion round method This filter sets the system ion round method
settings
{"Ion round Method":"MINUS_. (incl.) to PLUS_. (excl.)"}
Nominalize (unit mass) This filter condenses the scan to nominal
mass
Ion remover filter This filter removes all specified ions from a
mass spectrum
{"Mode":"INCLUDE","Ions":" "}
SNIP This baseline detector uses the SNIP
algorithm.
{"Window Size":,"Number of Iterations":}
Chromatogram baseline
subtract filter
This filter enables to remove the baseline
from the chromatogram
Chromatogram selection
(select range)
This filter selects the chromatogram range { …,"Start RT (Minutes)":.,…,"Stop RT (Minutes)":.}
Scan duplicator This filter duplicates series of scans {"Duplicated scan (Merge Traces)":true}
Savitzky-Golay smoothing (×) This filter applies the Savitzky-Golay filter {"Filter Ions":true,"Order":,"Width":}
First derivative Implementation of a first derivative peak
detector
{…,"Window Size":,…,"Optimize baseline (VV)":true,"Min S/N Ra-
tio":.,…,"Threshold":"HIGH","Filter Mode":"EXCLUDE"}
High pass peaks Keep the n-highest peaks {"Filter Option":"HEIGHT","Number Highest":}
MCR-AR [targeted chained] This peak detector provides targeted peak
deconvolution where targets are generated
by first derivate peak detector
{"Settings":{".DefaultProfile":"{\"Local minima scan slope Threshold\":-
.,\ …:,\ …:,\ …:}",".DefaultProfile":"{\"Minimum analysis
segment Width\":}",".DefaultProfile":"{\"Local Maxima scan Moving
average Window Size\":,\ …:,\ …:,\ …:.}",".DefaultProfi-
le":"{\"Percentage Height after Peak\":.,…:.,\ …:,\ …":,\ …:,\
…":.,\ …:-,\ …:}"} …}
Peak integrator trapezoid This extension point tries to implement a
ChemStation similar peak integrator
{"Include Background":false,"Area Constraint":true,"Traces to
Integrate":""}
High pass peaks Keep the n-highest peaks {"Filter Option":"AREA","Number Highest":}
Retention index calculator
(embedded)
Calculates the retention indices for scans and
peaks in the chromatogram
Example: { …,"Extrapolate (Right)":true, …,"Retention index
Marker":". |.|C(Octene);. |.|C(Nonene);
…;. |.|C (Henpentacontene)","Extrapolate (Left)":true}
NIST (extern) This plugin uses the NIST library to identify
peaks
{"NIST Folder (MSSEARCH)" …,"Min. Match Factor":.,"Min. Reverse
match Factor":.,…,"Number of Targets":}
OpenChrom chromatogram
(*.ocb)
Reads and writes OpenChrom
chromatograms
{"File Name":"{chromatogram_name}","Export Folder":" …}
M.L. Flaig et al.: Blind test of two wood taxa identification methods 491
hardwoods were taken in various focal planes. The images of
the vessel elements were examined for their qualitative
characteristics and systematically evaluated. The charac-
teristics of the vessel elements that were analyzed and
documented are: vessel element description, tails, perfora-
tion plates, intervessel pits, vessel-ray pits, pits to axial
parenchyma cells, areas without any pits, tyloses, vessel
length and width plus their l/w ratio, intervessel pit borders
in vertical diameter, pit apertures, fiber length, fiber width
and fiber wall thickness.
2.4.3 Analysis
Different numbers of slides were made from each test
sheet; TUDa made 5, TI 10 and ISEGA 10. The analysis
involved comparing the structural characteristics of the
vessel elements found in the blind test samples with the
descriptions and images in the ‘Atlas of vessel elements’
(Helmling et al. 2018). Morphological key featuresincluded the
shape of the vessel elements and their tails, the type of vessel
perforation plates, the presence of tyloses or helical thicken-
ings, as well as the arrangement and shape of intervessel
pits and vessel-ray pits. Using histometric measurements,
parameters such as vessel length and width, as well as the size
of the intervessel pit apertures and borders were determined
from the images and compared with the descriptions in the
atlas.
3 Results and discussion
3.1 Blind test
From the 37 self-manufactured single variety bleached kraft
pulps and the industrial beech matrix pulp, 15 blind samples
of defined composition were externally produced as
described in Section 2.1 without informing the authors of the
composition. The aim for the participating working groups
was to correctly identify the composition of the test sheets.
In the blind test, 15 ×38 = 570 decisions had to be made by
every participating testing laboratory. The results of three
institutes using the anatomical method and one institute
using the chemotaxonomic method are shown in Table 3
for every wood taxon.
In addition to the results for each taxon, Table 5 shows
the results for all members of the family Dipterocarpaceae
and separately for the subgenera of Shorea alone, which also
belong to the family. They are particularly important
because of their dominance in the tropical rainforests of
Southeast Asia, with over 200 species (Bansal et al. 2019).
3.1.1 Chemotaxonomy
The chemotaxonomic analysis of the test sheets was per-
formed on the GC-MS system A- the same system on which
Table :Blind test results for every institute and genus/subgenus/
species.
Taxon Chemotaxonomy Anatomy
UHH ∑TUDa ISEGA TI
% right %
right
%
right
%
right
%
right
Acacia spp.
Alniphyllum spp.
Avicennia spp.
Calophyllum spp.
Canarium spp.
Castanopsis spp.
Cocos nucifera
Cunninghamia
lancelota
Dendrocalamus spp.
Dipterocarpus spp.
Durio spp.
Elaeis guineensis
Eucalyptus spp.
Fagus sylvatica
Gonystylus spp.
Heritiera spp.
Hevea spp.
Ilex spp.
Intsia spp.
Koompassia spp.
Lophopetalum spp.
Mangifera spp.
Nyssa spp.
Palaquium sp.
Parashorea spp.
Paulownia
tomentosa
Phellodendron sp.
Pterygota sp.
Rhizophora spp.
Schima spp.
Shorea subg.
Anthoshorea
Shorea subg.
Richetia
Shorea subg.
Rubroshorea
Shorea subg. Shorea
Swintonia spp.
Tectona grandis
Terminalia
tomentosa
Tetramerista spp.
Total hit rate
492 M.L. Flaig et al.: Blind test of two wood taxa identification methods
the DB references were measured. The identification process
relied on the ‘Marker Peak Score’(MPS), peaks unique to a
single reference, and ambiguous peaks occurring in multiple
references. The ‘reverse Similarity Index’(rSI) was also
crucial when comparing an unknown extract to the DB, as
the rSI assesses the extent to which the library peaks of a
single DB reference match the peaks in the mixed sample
(Figure 5). The ‘Matched Chromatogram Area’(MCA),
i.e. the percentage of the total unknown peak area that
was identified as a specific reference (Figure 2c), and the
comparison factors mentioned above, required extensive
analysis to make an informed decision.
The screenshot of the software interface (Figure 2),
displays the outcomes of a query performed on blind sample
07, compared to the reference DB. A visual comparison tool
(Figure 2a) shows the unknown chromatogram at the top
with red unmatched, yellow ambiguous and green marker
peaks compared to the grey DB reference chromatogram
of Dipterocarpus atthebottom.Theretentiontimesofboth
chromatograms are shifted in the figure, but this was
corrected for analysis by RI calculation and alignment. For
Dipterocarpus anrSIof99.2%isreported,alongwith51
matched marker peaks and three ambiguous peaks
(Figure 2b). Furthermore, a substantial portion, specifically
72.6 %, of the peak area within the chromatogram of the
unknown pulp extract was attributed to Dipterocarpus
(Figure 2c), which the authors correctly concluded to be pre-
sent in the mixture after analyzing these distinct findings.
In the end eight woods were identified 100 % correctly
as present or absent in the 15 blind samples and an overall
blind test evaluation rate of 86 % correct decisions was
achieved (Table 3). Nevertheless, there were certain wood
species/genera that were more difficult to identify. These
included Avicennia,Calophyllum,Castanopsis,Eucalyptus,
Hevea,Ilex,Koompassia,Lophopetalum,Nyssa,Pterygota,
Rhizophora and Shorea subg. Anthoshorea. These genera
were never clearly identifiable with the current DB version.
There are several possible reasons for this: One is that the
chemical fingerprints were not always specific enough to be
separated from all other references. Although isolated
markers pointing to these genera were sometimes detected
in some blind samples, the expressions of these markers
were too weak to make clear decisions. Another reason could
be that even if a specific marker substance was detected in a
pure reference extract, its amount/peak size may be too low
in a mixture containing only a small proportion of that
pulp, so that the marker signal falls below the detection
threshold of the MS. In the evaluation of the blind test, the
Dipterocarpaceae family as a whole was distinguished from
representatives of other families by clear marker substances
(Table 4). However, the close relationship between members
of the Dipterocarpaceae family, especially the Shorea
subgenera, was evident in the chromatograms. Neverthe-
less, the Shorea subgenus Richetia was 100 % correctly
identified in the blind test. To refine the identification of
the other Diptocarpaceae members, a separate database
Figure 2: Database query results from the blind sample 07 (containing Dipterocarpus spp.) against the reference DB: screenshot of the software interface:
(a) chromatogram comparison; (b) statistic query results; (c) graphical area distribution (MCA); (d) normalized peak area comparison aligned by RI.
M.L. Flaig et al.: Blind test of two wood taxa identification methods 493
containing only these very similar species could be created
in the future. It may then be possible to narrow down the
matching parameters even further, thereby increasing
the accuracy of the matching results.
The authors acknowledge that the present chemotaxo-
nomic approach requires additional time and involves a
more extensive laboratory process when compared to the
effort required by the anatomical method.
3.1.2 Anatomy
The blind test participants using the anatomical method
prepared colored microscopic specimens from the received
15 test sheets and examined them under a light microscope.
The effectiveness of identifying individual taxa is remark-
able (Table 3). For instance, in the case of the genus Rhizo-
phora, known for its distinctive and less common features
(scalariform perforation plates and scalariform intervessel
pits), all participants correctly identified the genus as pre-
sent or absent in the 15 blind samples with 100 % accuracy.
This was the case for only three taxa in total. However, 20
additional taxa were identified with 100 % accuracy by two
of the three participants, and the third participant made only
one or two incorrect identifications for most of these taxa.
These taxa can be reliably identified by anatomical analysis,
considering the anatomical references and the characteris-
tics of each taxon. In contrast, the subgenera of the genus
Shorea and the genera Dipterocarpus and Parashorea were
described by all participants as challenging to differentiate
because they appear very similar in terms of the dimensions
of the vessel elements and the arrangement of the pits. In
particular, the subgenera Richetia and Rubroshorea could
not be correctly identified with 100 % accuracy by any of
the participants. When evaluating these subgenera and
the genera of the family Dipterocarpaceae separately, the
success rates were lower than the average for all taxa to
be determined (Table 4). Therefore, for the genus Shorea,
as described by Helmling et al. (2018), one should be aware
of the high risk of confusion within the family and should
refrain from making determinations down to the subgenus
level.
Other genera with challenging anatomical differentia-
tion included Mangifera and Swintonia, both belonging to
the Anacardiaceae family. There were also groups of genera,
that were not even closely related, but shared highly similar
anatomy of vessel elements and therefore had a high risk of
being confused. Of particular importance was the group
containing the genera Durio (Figure 3a), Lophopetalum
(Figure 3b) and CITES Annex II protected Gonystylus
(Figure 3c). It is worth noting that the rate of correct iden-
tifications for the genus Gonystylus in the 15 blind samples
was 100 % by two institutes. However, when all three genera
were considered together, the rates were 96 %, 87 % and
100 % for only one participant. The genera Koompassia and
Heritiera also show a striking similarity in vessel anatomy.
In a blind sample containing Koompassia, only one partici-
pant correctly identified this genus, while all participants
initially misidentified it as Heritiera. It is important to note
that real samples may contain other wood species for which
references are not yet available, and these species may
also exhibit high similarity to Gonystylus. The detailed
descriptions and figures of the microscopic images of all
these challenging genera are described in the Atlas of vessel
elements (Helmling et al. 2018). Their similarity can be seen
there. If only pure samples were to be analyzed, identifica-
tion could be made more clearly by the total number of
vessel elements belonging to only one genus. Identification
in mixed samples is therefore more difficult because
each vessel element has to be evaluated individually, if
Table :Comparison of the percentage of correct choices by group: all
taxa, the family Dipterocarpaceae and the subgenera of the genus Shorea.
Group Chemotaxonomy Anatomy
UHH (%) TUDa
(%)
ISEGA
(%)
TI
(%)
All taxa
Dipterocarpaceae
family
Shorea subgenera
Figure 3: Microscopic images of vessel elements: (a) Durio spp.;
(b) Lophopetalum spp.; (c) Gonystylus spp. (Helmling et al. 2018).
494 M.L. Flaig et al.: Blind test of two wood taxa identification methods
possible. If a vessel element with a meaningful pit arrange-
ment is turned sideways on the slide or shows no meaningful
features at all, it cannot be assigned to a single genus.
As TI and ISEGA microscoped 10 slides per sample
and achieved better results than TUDa who only prepared
five slides, this suggests that a larger number of slides can
increase the accuracy of the results.
3.2 System independence of the
chemotaxonomic method
The independence from the analytical instrument used is
crucial for the practical application of the chemotaxonomic
method. Therefore, three parameters (MPS, rSI and MCA)
were analyzed in two different types of samples (mixed
and pure) on three different instruments (TD-GC-MS systems
A, B and C) as schematically illustrated in Figure 4. As in
this investigation the laboratory and operator remained
constant, but the GC-MS systems used were changed, neither
repeatability nor reproducibility was determined, but what
is known as intermediate precision.
From each mixed sample extract two pyrolysis crucibles
were measured on GC-MS system A and three on each of
the systems C and B. All eight chromatograms per sample
underwent identical preprocessing as described in Section
2.3.5 Data processing and were matched against the DB.
Each DB query gave scores for MPS, rSI and MCA for each of
the 38 reference extracts included in the DB. MPS, rSI and
MCA are most important for the identification of wood taxa
and therefore need to be of high precision across different
instruments. The most promising candidates predicted in
the blind test identification were also selected for the sta-
tistical analysis of the system independence. Therefore, a
selection of six blind samples and 1–5 MPS of potentially
containing reference taxa per blind sample chromatogram
were analyzed. A total of 16 MPS times eight chromatograms
per sample equals 128 MPS, divided into the systems A: 32, B:
48 and C: 48 (Figure 4). The same applies to the rSI and MCA
values, i.e. a total of 384 data points of the mixed samples
were analyzed. The results are shown in Tables 5–7.
For each GS-MS system (A, B and C) the mean values of
the three parameters studied are given. As suggested by
(Gold 2019; ISO 1994; Zwanziger and Sorkau 2020) the
intermediate precision variance (Var), intermediate preci-
sion standard deviation (SD) and intermediate precision
relative standard deviation or coefficient of variance (CV)
were calculated to analyze the system independence of the
method. The overall CV is given at the end of each table as
an average of all calculated CV values. In addition, Analysis
of Variance or ANOVA was conducted to assess whether
the means of the three groups (GC-MS systems) were
significantly different from each other. It accomplished
this by partitioning the total variance observed in a dataset
into different components: variance between groups and
variance within groups. If the variance between groups
was significantly greater than the variance within groups,
ANOVA suggested that there were differences among the
means of the groups. The null hypothesis (H0) stated that
there was no significant difference among the group means.
If ANOVA yielded non-significant differences, this
indicated that there was insufficient evidence to reject the
H0. In other words, the observed differences between the
groups were within the range of what could be expected
Figure 4: Schematic illustration of the variance analysis of the system independence of the chemotaxonomic method.
M.L. Flaig et al.: Blind test of two wood taxa identification methods 495
due to random chance alone. This meant that any observed
differences were not statistically significant based on the
chosen significance level (p) of 0.05 and H0 was accepted.
Table 5 shows the mean marker peak scores of the
systems A, B and C and the overall statistics (∑ABC) per
sample and taxon in the mixed samples. The overall CV of
32 % indicates that the variation of the MPS between all
the chromatograms tested was quite high, but interestingly
the F-tests of the ANOVAs were not significant in 13 of the 16
cases tested (p-value > 0.05). This means that the differences
Table :Statistics results for marker peak scores (MPS) of mixed samples (MS).
No. Sample Analyzed taxon GC-MS system Total marker peak scores
ABC ∑ABC
Mean Mean Mean Mean Var SD CV ANOVA
MS- Dipterocarpus .... . . %F(,)=.,p<.
MS- Fagus .... . . %F(,)=.,p=.
MS- Lophopetalum .... . . %F(,)=.,p<.
MS- Acacia .... . . %F(,)=.,p=.
MS- Canarium .... . . %F(,)=.,p=.
MS- Castanopsis .... . . %F(,)=.,p=.
MS- Palaquium .
... . . %F(,)=.,p=.
MS- Schima .... . . %F(,)=.,p=.
MS- S. subg. Richetia .... . . %F(,)=.,p=.
MS- S. subg. Shorea ...
. . . %F(,)=.,p=.
MS- Fagus .... . . %F(,)=.,p<.
MS- Acacia .... . . %F(,)=.,p=.
MS- Elaeis .... . . %F(,)=.,p=.
MS- Terminalia .... . . %F(,)=.,p=.
MS- Cunninghamia .... . . %F(,)=.,p=.
MS- Tectona .... . . %F(,)=.,p=.
Mean .%
Table :Statistics results for reverse similarity index (rSI) of mixed samples.
Sample Analyzed taxon GC-MS system Total reverse similarity index
ABC ∑ABC
Mean Mean Mean Mean Var SD CV ANOVA
MS- Dipterocarpus . . . . . . %F(,)=.,p=.
MS- Fagus . . . . . . %F(,)=.,p=.
MS- Lophopetalum . . . . . . %F(,)=.,p=.
MS- Acacia . . . . . . %F(,)=.,p=.
MS- Canarium . . . . . . %F(,)=.,p=.
MS- Castanopsis . . . . . . %F(,)=.,p=.
MS- Palaquium . . . . . . %F(,)=.,p=.
MS- Schima . . . . . . %F(,)=.,p=.
MS- S. subg. Richetia . . . . . . %F(,)=.,p=.
MS- S. subg. Shorea . . . . . . %F(,)=.,p=.
MS- Fagus . . . . . . %F(,)=.,p<.
MS- Acacia . . . . . . %F(,)=.,p=.
MS- Elaeis . . . . . . %F(,)=.,p=.
MS- Terminalia . . . . . . %F(,)=.,p=.
MS- Cunninghamia . . . . . . %F(,)=.,p=.
MS- Tectona . . . . . . %F(,)=.,p=.
Mean .%
496 M.L. Flaig et al.: Blind test of two wood taxa identification methods
between most of the group means were smaller than the
differences within the groups. The H0 - groups are equal - was
accepted in most cases. Most of the mean values of the MPS
for the three GC-MS systems are therefore not statistically
significantly different, i.e. GC systems A, B and C are statisti-
cally equal, although the total variance is quite high! For
example, the mean MPS of system C of 4.3 for Castanopsis in
the mixed sample 02 (Table 5) is composed of three single
values 4, 3 and 6, giving a CV of 35 % within this group.
The total CV of all individual values of all groups A, B and C
for MS-02 Castanopsis is much smaller - only 23 % (Table 5).
In this case, looking only at the group means; 4.0, 5.3 and 4.3,
they have an even smaller CV of 15 %. Therefore, it makes
sense that the ANOVA also concludes that the variation
within groups is higher than between groups. Raw data and
calculations can be found in the Supplementary Table S2.
As explained in detail by Flaig et al. (2023), the rSI is the
most important similarity index for the identification of
mixed samples in particular. Figure 5 shows that only the
peaks that are present in the DB reference chromatogram
are compared with the unknown peaks from the mixed
sample to calculate the rSI value.
The results in Table 6 clearly show that the precision
of the rSI values in the mixed samples was very accurate
with an overall CV of 6.9 %. There were also no statistically
significant differences in the rSI values for the three
different GC-MS systems, except for ‘MS-12 Fagus’. For this
sample, the p-value for the systematic measurement error
of the different GC-MS systems was 0.0005, well below the
0.05 significance level, which means that H0 had to be
rejected and the alternative hypothesis (H1) - groups differ -
was accepted in this case.
Table 7 shows the results of the MCA analyses for the
mixed samples, i.e. the percentage of the peak area of
the mixed sample that matched a particular reference.
According to the recommended acceptance criteria for
intermediate precision by Little (2016) the overall CV of
12.6 % is excellent and all but one of the ANOVAs are not
Table :Statistics results for matched chromatogram areas (MCA) of mixed samples.
Sample Analyzed taxon GC-MS system Total matched chromatogram area (%)
ABC ∑ABC
Mean Mean Mean Mean Var SD CV ANOVA
MS- Dipterocarpus .... . . %F(,)=.,p=.
MS- Fagus .... . . %F(,)=.,p=.
MS- Lophopetalum .... . . %F(,)=.,p=.
MS- Acacia .... . . %F(,)=.,p=.
MS- Canarium .... . . %F(,)=.,p=.
MS- Castanopsis .... . . %F(,)=.,p=.
MS- Palaquium .... . . %F(,)=.,p=.
MS- Schima .... . . %F(,)=.,p=.
MS- S. subg. Richetia .... . . %F(,)=.,p=.
MS- S. subg. Shorea .... . . %F(,)=.,p=.
MS- Fagus .... . . %F(,)=.,p<.
MS- Acacia .... . . %F(,)=.,p=.
MS- Elaeis .... . . %F(,)=.,p=.
MS- Terminalia .... . . %F(,)=.,p=.
MS- Cunninghamia .... . . %F(,)=.,p=.
MS- Tectona .... . . %F(,)=.,p=.
Mean .%
Figure 5: Graphical explanation of the similarity indices.
M.L. Flaig et al.: Blind test of two wood taxa identification methods 497
significant, confirming the equivalence of the results
from the three instruments. The MCA values given are not
absolute but partial. For instance, for the mixed sample
07 measured on system A, a mean MCA of 28.9 % is given
for Dipterocarpus (Table 7). The total sum of the areas of
all peaks detected in MS-07 and assigned to the reference
Dipterocarpus is actually 82.1 %. This is much higher than
the 28.9 % given in Table 7. The difference is due to the
many ambiguous peaks that are matched by more than
one reference peak from the DB. In this example of MS-07,
the largest peak (RI 3509, Dammaran-3-one, 20,24-epoxy-
25-hydroxy) represents approximately 50 % of the total area
of all peaks in the sample. This ambiguous peak was matched
with Dipterocarpus and six other references. Therefore,
the peak area of 50 % is divided by seven and then only
this 7.1 % portion is included in the sum of the matched
peak areas for Dipterocarpus and each of the other 6 DB
references, resulting in much smaller MCA values listed
in the results tables than the actual total values.
For the pure samples the statistical analyses were per-
formed in the same way as for the mixed samples, but for
MPS, rSI and MCA only one genus per sample was analyzed.
In Table 8, the statistical results for the MPS of the pure
samples reveal high Var, SD and an overall CV of 49.4 %.
The ANOVA analyses results - in all cases p< 0.05 - further
underscore the significant differences between these
three GC-MS systems for MPS in pure samples. System B and
C perform similarly. System A achieves more than twice as
many marker peaks in most cases which can be explained by
the fact that it is the original system on which the database
was created. The results of B and C allow a better judgement
about how universally the database can be used. Therefore,
the intermediate precision of MPS for pure samples may
not be as bad as it appears from Table 8. Furthermore,
wood identification is not only about the MPS number itself.
Even in the by far worst example with a CV of 73 % for the
reference Hevea in pure sample 17, an average of 10.5 MPS
was identified on system B (Table 8). This score still allows
a very clear identification decision in favor of Hevea,as
the query results for almost all other reference woods in the
DB are 0 or in two cases maximum 3 MPS (Supplementary
Figure S1). Since the decision making process is always a
combination of considering not only the MPS but also the
rSI (0.92, Table 9) and MCA (4.6 %, Table 10, 79.4 % in total),
it is obvious that even in this drastic example with the
highest variation between the systems the correct decision
for Hevea is easily made on all systems. In the end, this is
what counts in practice, as demonstrated by the excellent
blind test results (3.1 Blind test).
The overall intermediate precision CV for rSI values in
pure samples stands at 12.6 % (Table 9). Although this is less
favorable than the outcomes of 6.9 % for mixed samples
(Table 6), it still reflects a satisfactory level of reliability.
Table 10 shows the results of the MCA analysis for
the pure samples. Only two of the seven ANOVAs are not
significant but the overall CV of 14.4 % is acceptable showing
that MCA values for pure samples are quite reproducible on
different instruments.
To improve the intermediate precision, some of the
preprocessing parameters of the original method (Flaig et al.
2023) were adjusted prior to this investigation. Rounding
was set to the range of m/z−0.38 to m/z+0.62 according to
Khrisanfov and Samokhin (2022). In addition, the 206 largest
peaks by area were analyzed per reference chromatogram
and the 125 highest mass traces per scan were retained. The
mass traces were smoothed using the ‘Savitzky-Golay’filter
(Savitzky and Golay 1964) four times in a row for better
smoothing results, signal to noise ratio and peak detection.
The mass spectra were compared using the ‘NIST (Identity
Normal)’algorithm, which goes a step beyond the cosine
Table :Statistics results for marker peak scores (MPS) of pure samples (PS).
Sample/taxon GC-MS system Total marker peak scores
ABC ∑ABC
Mean Mean Mean Mean Var SD CV ANOVA
PS- Castanopsis .... . . %F(,)=.,p<.
PS- Intsia .... . . %F(,)=.,p<.
PS- Dendrocalamus .... . . %F(,)=.,p<.
PS- Hevea .... . . %F(,)=.,p<.
PS- Eucalyptus .... . . %F(,)=.,p<.
PS- Rhizophora .... . . %F(,)=.,p<.
PS- Lophopetalum .... . . %F(,)=.,p<.
Mean .%
498 M.L. Flaig et al.: Blind test of two wood taxa identification methods
comparison used in the original DB. This approach, outlined
by Stein and Scott (1994), involves additional comparison of
the highest intensity m/zvalues. The originally used cosine
method exhibits slight weaknesses in distinguishing isomers
and alkanes. Detailed preprocessing operations and DB
setup parameters are given in Section 2.3.5 Data processing.
The latest version of the pulp extract DB is available from the
authors on request.
Even with the updated and improved DB, reproducing
MPS for pure samples proved challenging, with unfavorable
results from ANOVA analyses and an overall CV of 49.4 %.
However, given that real-life samples primarily consist of
mixtures, the focus shifts to the crucial precision of MPS, rSI
and MCA values in mixed samples. For mixed samples, the
overall CV results of MPS (32.0 %), rSI (6.9 %) and MCA
(12.6 %) scores were satisfactory, or rSI and MCA even
excellent according to the definition by Little (2016).
Encouragingly, the ANOVA results for mixed samples were
also highly positive, providing assurance that the three
GC-MS instruments did not exhibit statistically significant
differences.
To reduce the system dependency and increase the
overall robustness of the chemotaxonomic method across
different laboratories, standardized procedures for the
analysis are proposed. These should include protocols for
data processing, analysis, and sample preparation, including
guidelines for extraction methods as described in this paper
and by Flaig et al. (2023). Calibration procedures should
consistently use standard spectra tune (s-tune) and the same
amount of polyethylene for retention index calibration.
Quality control samples with known marker peak signatures
could be included in each batch of analysis to monitor
system variability in real time. To ensure consistency
between laboratories, maintaining and expanding reference
libraries for known genera and including metadata such
as analytical equipment, calibration details, and sample
preparation methods, can facilitate cross-referencing and
verification of results.
Table :Statistics results for reverse similarity index (rSI) of pure samples.
Sample/taxon GC-MS system Total reverse similarity index
ABC ∑ABC
Mean Mean Mean Mean Var SD CV ANOVA
PS- Castanopsis . . . . . . %F(,)=.,p=.
PS- Intsia . . . . . . %F(,)=.,p=.
PS- Dendrocalamus . . . . . . %F(,)=.,p=.
PS- Hevea . . . . . . %F(,)=.,p<.
PS- Eucalyptus . . . . . . %F(,)=.,p=.
PS- Rhizophora . . . . . . %F(,)=.,p<.
PS- Lophopetalum . . . . . . %F(,)=.,p=.
Mean .%
Table :Statistics results for matched chromatogram areas (MCA) of pure samples.
Sample/taxon GC-MS system Total matched chromatogram area (%)
ABC ∑ABC
Mean Mean Mean Mean Var SD CV ANOVA
PS- Castanopsis .... . . %F(,)=.,p=.
PS- Intsia .... . . %F(,)=.,p=.
PS- Dendrocalamus .... . . %F(,)=.,p<.
PS- Hevea .... . . %F(,)=.,p<.
PS- Eucalyptus .... . . %F(,)=.,p<.
PS- Rhizophora .... . . %F(,)=.,p<.
PS- Lophopetalum .... . . %F(,)=.,p<.
Mean .%
M.L. Flaig et al.: Blind test of two wood taxa identification methods 499
3.3 Limitations
It shall be noted that the same pulps used for the references
were used to create the blind samples of unknown compo-
sition. In addition, the knowledge of the participants that
each of the pulps must be present at least once in one of
the blind samples changes the attention to it. The search
for each of the 38 taxa probably leads to more species be-
ing named by the testing laboratories in the blind test than in
real samples.
For both methodologies, a critical factor revolves around
the extent of reference samples present in their databases.
Genera can only be identified if there are references for them.
The expansion of the reference database is therefore very
important for reliable identification. For the anatomical
method, if an unknown sample contains cells for which there
are no references, but there is a declaration of the genus being
processed, a new reference can be created relatively quickly
using the Thünen wood collection. The declaration can then
be checked for plausibility.
Notably, the chemotaxonomic method faces limitations
regarding the determination of MPS, which is currently
dependent on the GC-MS system used. Other factors that
can hamper this method and make the accurate identification
of certain taxa difficult are the low extractive content of the
pulp and the limited number of marker substances available.
Also, its capacity to comprehensively encompass the
full spectrum of potential variations within tree species
is limited. This limitation arises from the well-known
phenomenon of extractives compositions undergoing
alterations in diverse environmental settings (Deklerck et al.
2020; Silva et al. 2018), whereas anatomical features tend
to remain relatively stable (IAWA Committee 1989).
4 Conclusions
A comprehensive blind test was conducted, allowing for a
comparative evaluation of the outcomes produced by two
independent identification methods: chemotaxonomy and
anatomy. All testing laboratories demonstrated a high level
of accuracy in identifying the wood taxa present in the blind
test pulp samples. The anatomical method proved highly
effective in identifying 23 wood taxa 100 % correctly by at
least two laboratories, with particularly notable accuracy in
the case of the genera Acacia and Rhizophora. These results
underscore the reliability of anatomical analysis for specific
taxa, even in cases where differentiation can be challenging.
Using the chemotaxonomic method, the authors identified
eight wood genera with 100 % accuracy. The method and
software proved effective in matching blind samples against
the DB. Key factors for identification included the total rSI
value, the count of marker/ambiguous peaks and the total
matching area. Some challenging genera, such as Eucalyptus,
Avicennia and Koompassia displayed weak marker expres-
sions. The study aimed to compare the individual strengths
of each method and optimize their collective capabilities,
especially in complicated cases such as highly beaten pulp, to
improve overall identification performance. An examina-
tion of Table 3 revealed interesting patterns in the reliability
of identification by anatomy and chemotaxonomy. For
example, the genus Nyssa was correctly identified 98 %
of the time by the anatomically trained participants, but
was unrecognizable by chemotaxonomy. Conversely, taxa
such as Intsia,Parashorea and Shorea subg. Richetia were
confidently identified by chemotaxonomy, while anatomical
identification was more uncertain. This clear contrast
highlighted the complementary nature of the two methods.
By integrating both methods, synergies could be used
to achieve even more robust and accurate results. This
study emphasizes the importance of a comprehensive and
multi-faceted approach to wood species identification.
The systematic validation, in which three parameters
were analyzed in two different sample types on three
different instruments, affirmed the system-independence
and robustness of the chemotaxonomic technique under
certain conditions. Uniform preparation and analysis of
mixed and pure tropical hardwood sample extracts on the
GC-MS systems A, B and C demonstrated the consistent
performance of the method for rSI and MCA values. MPS
values were less reproducible. The poor statistical results
were due to much higher numbers of marker peaks achieved
by system A, while the average results of systems B and C
were quite similar. The outperformance effect of system
A was more or less observable in all investigations, but it was
worst for MPS statistics in pure samples. It was expected that
the original system A, on which all the DB reference samples
were measured, might have an advantage over the other
instruments. However, systems B and C, representing all
other GC-MS instruments, performed well and reproducibly
even for MPS in most pure samples. As real-life samples are
mainly mixtures, the noteworthy intermediate precision
results for rSI (6.9 % CV) and MCA (12.6 % CV) in mixed
samples underline the reliability of our approach and rein-
force its usefulness as a versatile tool for wood taxa identi-
fication in different research settings and laboratories. To
further reduce the inter-instrument variance, it would be
useful in the future to enrich the pulp extract DB not only
with additional wood references from only one instrument,
but also with chromatography data references from other
instruments.
500 M.L. Flaig et al.: Blind test of two wood taxa identification methods
Whilst the first attempt to create a chemotaxonomic
database proved successful with equivalent blind test re-
sults to the established anatomical method, the results of the
validation showed promising results with potential im-
provements to the method. For general application, it would
be useful to identify and refine the parameters associated
with its partial system dependency and use standardized
protocols for sample preparation, calibration, data process-
ing and analysis. Further extension with a diverse range of
samples to account for variation between different wood
provenances is anticipated in future chemotaxonomic
studies. Although more work needs to be done, the combi-
nation of both approaches is particularly valuable when
dealing with taxa that present challenges to either method
alone and contributes to the support of EUDR and sustain-
able forestry.
Acknowledgments: PD Dr. habil. Jürgen Odermatt, who
initiated the research for the chemotaxonomic method and
passed away suddenly in 2019, deserves special thanks. The
authors would also like to thank Prof. Dr. Helga Zollner-Croll
for preparing the blind test samples and Doris Helm, Sergej
Kaschuro, Anne Wettich and Birte Buske for their assistance
with microscopic and laboratory work. For the preparatory
work the authors also thank Othar Kordsachia, Nils
Grützmann and Alina Wassink.
Research ethics: As the research does not involve the use of
humans or animals, the Declaration of Helsinki does not
apply. Therefore, the research ethics statement is not
applicable. The manuscript has not been published previ-
ously and is not under consideration for publication
elsewhere.
Author contributions: MLF, JB, AO and SH conceived and
planned the experiments. MLF and JB carried out the
chemotaxonomic analysis. AO, SH, HJS and DZ carried out
the anatomic analysis. MLF conceived, planned and carried
out the method validation and wrote the original manu-
script. BS revised and edited the manuscript and supervised
the project. All the authors have accepted responsibility for
the entire content of this submitted manuscript and
approved submission.
Competing interests: The authors declare that they have no
conflicts of interest regarding this article.
Research funding: This work was funded by the German
Federal Foundation for Environment (DBU) (Grant/Award
Number: “AZ 34295/01”) in connection with the project
“Detection of Tropical Hardwood in Paper –Chemotax-
onomy and Anatomy for the Identification of Mixed Tropical
Hardwood”. The funding organization played no role in
study design; analysis, and interpretation of data; in the
writing of the report; or in the decision to submit the report
for publication.
Data availability: The raw data and the latest version of the
pulp extract database can be obtained on request from the
corresponding author.
References
Alaerts, G., Pieters, S., Logie, H., van Erps, J., Merino-Arevalo, M., Dejaegher,
B., Smeyers-Verbeke, J., and Vander Heyden, Y. (2014). Exploration and
classification of chromatographic fingerprints as additional tool for
identification and quality control of several artemisia species. J. Pharm.
Biomed. Anal. 95: 34–46.
Alfassi, Z.B. (2004). On the normalization of a mass spectrum for
comparison of two spectra. J. Am. Soc. Mass Spectrom. 15: 385–387.
Bansal, M., Nagaraju, S.K., and Prasad, V. (2019). The origin and
diversification of plant family Dipterocarpaceae. OJG 9: 593–596.
Braga, J., Deklerck, V., Espinoza, E., Groening, M., Koch, G.,
Monteiro Pastore, T.C., Ramananantoandro, T., Schroeder, H.,
Watkinson, C., Wiedenhoeft, A.C., et al. (2020). Scientific methods for
taxonomic and origin identification of timber. Global Timber Tracking
Network. Available at: https://www.researchgate.net/publication/
342003654 (Accessed 15 May 2023).
Deklerck, V., Lancaster, C.A., van Acker, J., Espinoza, E.O., van den Bulcke, J.,
and Beeckman, H. (2020). Chemical fingerprinting of wood sampled
along a pith-to bark gradient for individual comparison and
provenance identification. Forests 11: 1–13.
European Union (2023). Regulation (EU) No 2023/1115 of the European
Parliament and of the Council of 31 May 2023: on the making available
on the Union market and the export from the Union of certain
commodities and products associated with deforestation and forest
degradation and repealing Regulation (EU) No 995/2010. Off. J. Eur.
Union 150: 206–247.
Flaig, M.L., Berger, J., Wenig, P., Olbrich, A., and Saake, B. (2023).
Identification of tropical wood species in paper: a new
chemotaxonomic method based on extractives. Holzforschung 77:
860–878.
Franklin, G.L. (1945). Preparation of thin sections of synthetic resins and
wood-resin composites, and a new maceration method for wood.
Nature 155: 51.
Gasson, P. (2011). How precise can wood identification be? Wood anatomy’s
role in support of the legal timber trade, especially cites. IAWA J 32:
137–154.
Gold, V. (2019). The IUPAC compendium of chemical terminology. International
Union of Pure and Applied Chemistry (IUPAC), Research Triangle Park,
NC.
Helmling, S., Olbrich, A., Tepe, L., and Koch, G. (2016). Qualitative and
quantitative characteristics of macerated vessels of 23 mixed tropical
hardwood (MTH) species: a data collection for the identification of
wood species in pulp and paper. Holzforschung 70: 839–844.
Helmling, S., Olbrich, A., Heinz, I., and Koch, G. (2018). Atlas of vessel
elements: identification of Asian timbers. IAWA J 39: 249–352.
IAWA Committee (1989). IAWA list of microscopic features for hardwood
identification: with an appendix on non-anatomical information. IAWA
Bulletin n.s. 10: 219–332.
Ilvessalo-Pfäffli, M.-S. (1995). Fiber atlas: identification of papermaking fibers.
Springer Verlag, Berlin, Heidelberg.
M.L. Flaig et al.: Blind test of two wood taxa identification methods 501
InsideWood (2004). Database: published on the internet, Available at:
https://insidewood.lib.ncsu.edu (Accessed 19 April 2023).
ISO (1994). Accuracy (trueness and precision) of measurement methods and
results: Part 1: General principles and definitions. No. 5725-1-3:1994.
International Organization for Standardization.
Khrisanfov, M. and Samokhin, A. (2022). A general procedure for rounding
m/z values in low-resolution mass spectra. Rapid Commun. Mass
Spectrom. 36: 1–6.
Köthke, M., Lippe, M., and Elsasser, P. (2023). Comparing the former EUTR
and upcoming EUDR: some implications for private sector and
authorities. Forest Policy and Economics 157: 103079.
Little, T. (2016). Establishing acceptance criteria for analytical methods.
BioPharm Int. 29: 44–48.
Low, M.C., Schmitz, N., Boeschoten, L.E., Cabezas, J.A., Cramm, M., Haag, V.,
Koch, G., Meyer-Sand, B.R., Paredes-Villanueva, K., Price, E., et al.
(2022). Tracing the world’s timber: the status of scientific verification
technologies for species and origin identification. IAWA J 1: 1–22.
Lowe, A.J. and Cross, H.B. (2011). The application of DNA methods to timber
tracking and origin verification. IAWA J 32: 251–262.
Nieradzik, L., Sieburg-Rockel, J., Helmling, S., Keuper, J., Weibel, T., Olbrich,
A., and Stephani, H. (2024). Automating wood species detection and
classification in microscopic images of fibrous materials with deep
learning. Microsc. Microanal. 38: 1–17.
Ravindran, P. and Wiedenhoeft, A.C. (2020). Comparison of two forensic
wood identification technologies for ten Meliaceae woods: computer
vision versus mass spectrometry. Wood Sci. Technol. 54: 1139–1150.
Richter, H.G. and Dallwitz, M.J. (2000). Commercial timbers: descriptions,
illustrations, identification, and information retrieval. In
English, French, German, Portuguese, and Spanish. Version: 9th April
2019. Available at: https://www.delta-intkey.com/wood/index.htm
(Accessed 19 April 2023).
Saake, B., Flaig, M.L., Berger, J., Helmling, S., and Olbrich, A. (2021). Nachweis
von Tropenholz in Papier - Chemotaxonomie und Anatomie zur
Identifizierung von mixed tropical hardwood: MTH III - Abschlussbericht
zum DBU-Projekt AZ 34295/01. Bericht, Hamburg. Available at: https://
www.dbu.de/projektdatenbank/34295-01/ (Accessed 20 February 2024).
Savitzky, A. and Golay, M.J.E. (1964). Smoothing and differentiation of data
by simplified least squares procedures. Anal. Chem. 36: 1627–1639.
Schmitz, N., Beeckman, H., Cabezas, J.A., Cervera, M.T., Espinoza, E., and
Fernandez-Golfin, J. (2019). The timber tracking tool infogram. Overview
of wood identification methods’capacity. Global Timber Tracking
Network. Available at: https://www.researchgate.net/publication/
332223451 (Accessed 15 May 2023).
Schmitz, N., Beeckman, H., Blanc-Jolivet, C., Boeschoten, L., Braga, J.,
Cabezas, J.A., Chaix, G., Crameri, S., Degen, B., Deklerck, V., et al.
(2020). Overview of current practices in data analysis for wood
identification. A guide for the different timber tracking methods.
Available at: https://www.researchgate.net/publication/342281916
(Accessed 15 May 2023).
Sieburg-Rockel, J. and Koch, G. (2020). Identification of wood species used in
particleboard production. IAWA J 41: 751–760.
Silva, D.C., Pastore, T.C., Soares, L.F., de Barros, F.A., Bergo, M.C., Coradin,
V.T., Gontijo, A.B., Sosa, M.H., Chacón, C.B., and Braga, J.W. (2018).
Determination of the country of origin of true mahogany (Swietenia
macrophylla King) wood in five Latin American countries using
handheld NIR devices and multivariate data analysis. Holzforschung
72: 521–530.
Stein, S.E. and Scott, D.R. (1994). Optimization and testing of mass spectral
library search algorithms for compound identification. J Am Soc Mass
Spectrom 5: 859–866.
Wheeler, E.A. (2011). InsideWood –a web resource for hardwood anatomy.
IAWA J 32: 199–211.
Windhagen, K., Moldenhauer, T., Burkard, A., and Geiger, G.A. (2019). Paper
- annual report: ein Leistungsbericht (annual report). Verband Deutscher
Papierfabriken e.V., Bonn, Germany.
Zheng, C., Liu, S., Wang, J., Lu, Y., Ma, L., Jiao, L., Guo, J., Yin, Y., and He, T.
(2024). Opening the black box: explainable deep-learning
classification of wood microscopic image of endangered tree species.
Plant Methods 20: 1–11.
Zwanziger, H.W. and Sorkau, E. (2020). Kalibration analytischer Methoden.
Theorie & Techniken. Eine Orientierungshilfe. Hochschulverlag
Merseburg, Merseburg.
Supplementary Material: This article contains supplementary material
(https://doi.org/10.1515/hf-2024-0025).
502 M.L. Flaig et al.: Blind test of two wood taxa identification methods
Available via license: CC BY 4.0
Content may be subject to copyright.
