Application of straight-chain lipids as
biomarkers in the reconstruction of
landscape development in a driftsand
landscape in the Netherlands
A.J. de Vreng
MSc. thesis
Earth Science
Supervisors:
Dr. J.M. van Mourik
Dr. B. Jansen
Prof. Dr. K. Kalbitz
March 2011
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Netherlands
Abstract
Historical land use systems as shifting cultivation and plaggen agriculture on nutrient poor
sandy soils caused serious degradation of soil and vegetation, resulting in Late Holocene
sand drifting. Locally the coversand landscape transformed into a driftsand landscape. On
sheltered deposition sites polycyclic driftsand profiles developed. They are valuable geoecological records of alternating instable (sand drifting) and stable (soil formation) phases in
landscape development.
Previous research has been done on the reconstruction of the landscape genesis, based on
pollen analysis and radiocarbon dating of polycyclic driftsand deposits. Recently, the
application of
Optically Stimulated
Luminescence
(OSL)
dating improved
the
geochronology. The principles of pollen transport and deposition does not allow to select
the plant species from pollen spectra, responsible for the on-site production of soil organic
matter during soil formation in stable periods.
A possible proxy for this, that could add information, is the application of straight-chain
lipids analysis. In this pilot study we introduced the biomarker method to separate local and
regional species, found in pollen spectra. Different species often have their own specific
patterns of alcohols and alkanes of different carbon chainlengths. Therefore, comparing the
straight-chain lipid patterns of both (expected) vegetation species and selected soil samples is
a means to assess the landscape history. In order to do this the recently developed VERHIB
model was used in combination with a cluster analysis.
The results obtained from VERHIB and the cluster analysis suggest a dynamic sequence of
dominant vegetation species, which in majority support the existing theories about the
landscape history of polycyclic deposits in the Netherlands. Some results (e.g. the
distribution of Pinus throughout the profile) are much harder to link to earlier research and
are likely to be incorrect.
Although a couple of problems arose during the analytical process, the application of
straight-chain lipids is found to be certainly a very promising technique. Nevertheless, a lot
more research has to be carried out to draw more solid conclusions on the strength of the
biomarker technique in sandy environments and its impact on the ideas about the culturalecological history of polycyclic soils in the Netherlands.
3
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Netherlands
Contents
Abstract ................................................................................................................................3
1. Introduction ......................................................................................................................7
2. Objectives ....................................................................................................................... 11
3. Research area................................................................................................................... 13
3.1. Summary of earlier work Defensiedijk ...................................................................... 14
3.1.1. Pollen analysis (profile 1984) .............................................................................. 14
3.1.2. 14C dating (profiles 1984, 1986 and 2002) ........................................................... 14
3.1.3. Micromorphology (profiles 1986 and 2002) ....................................................... 16
3.1.4. OSL dating (profile 2002) .................................................................................. 17
4. Methodology ................................................................................................................... 19
4.1. Field methodology .................................................................................................... 19
4.1.1. Soil sampling...................................................................................................... 19
4.1.2. Vegetation sampling........................................................................................... 21
4.2. Laboratory methodology .......................................................................................... 22
4.2.1. Preparation ........................................................................................................ 22
4.2.2. ASE extraction................................................................................................... 22
4.2.3. Clean-up ............................................................................................................ 23
4.2.4. Derivatisation..................................................................................................... 23
4.2.5. GC/MS analysis................................................................................................. 24
4.3. Data analysis............................................................................................................. 25
4.3.1. Lipid quantification............................................................................................ 25
4.3.2. Cluster analysis................................................................................................... 25
4.3.3. VERHIB model ................................................................................................. 26
5. Results and discussion ..................................................................................................... 29
5.1. Lipid quantification................................................................................................... 29
5.2. Cluster analysis ......................................................................................................... 34
5.3. VERHIB model........................................................................................................ 37
5.3.1. Output analysis .................................................................................................. 38
5.3.2. Alternative runs ................................................................................................. 40
5.4. Relation to data from previous studies ...................................................................... 42
6. Conclusions..................................................................................................................... 43
References........................................................................................................................... 45
APPENDICES ................................................................................................................... 49
Appendix A: Photos of Defensiedijk 2008 profile ........................................................... 49
Appendix B: Matlab code ................................................................................................ 53
Appendix C: GC/MS results ........................................................................................... 59
5
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Netherlands
1. Introduction
Late glacial aeolian coversand dominates the surface geology of large parts in the east and the
south of the Netherlands. During prehistoric and early historic time, forest grazing, wood
cutting and shifting cultivation gradually transformed natural forest into heath land. Part of
this heath land is later utilized in the plaggen agriculture.
In extensive parts of North-western Europe this plaggen agriculture was being carried out
starting about 3000 years ago, increasing from the early middle Ages on. This went on until
the introduction of chemical fertilizers in the beginning of the 20th century made the use of
plaggen management obsolete (e.g. Spek, 2004).
Heather, forest and grass sods were used for animal bedding in stables and afterwards
applied to the fields. This caused deep humus rich plaggen soils to be formed near the
villages (e.g. Blume & Leinweber, 2004), with significantly improved fertility. Large parts of
The Netherlands are influenced by plaggen management. More than 221,000 ha of land has
soils with a humic sand cover thicker than 50 cm, and more than 196,000 ha has soils with a
cover from 30 to 50 cm thick (Pape, 1970). These soil are generally classified as Plaggic
Podzols and Plaggic Anthrosols (IUSS Working Group WRB, 2007).
As an effect of the plaggen agriculture, serious degradation of soil and vegetation occurred in
the areas where the removal of plaggen took place. This resulted in the comeback of sand
drifting and locally the coversand landscape transformed into a driftsand landscape.
Characteristic for this landscape are soils consisting of polycyclic driftsand deposits, which
can be utilized as geo-ecological records of alternating instable and stable phases in
landscape development. Every cycle of a polycyclic profile reflects a period of landscape
instability (deposition) and landscape stability (soil development).
In the past quite some research has been done on the historic reconstruction of cultural
sandy soils in the Netherlands using radiocarbon dating, Optically Stimulated Luminescence
(OSL) dating and pollen records. Due to the complex composition of soil organic matter,
the interpretation of conventional radiocarbon dating results has never been satisfactory (van
7
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
Mourik et al., 1995). This is because some of the components of soil organic material
developed in situ, while other (often older) components were transported into the area with
the sods. To overcome this problem OSL dating was introduced (e.g. Bokhorst et al., 2005).
This technique dates the last time quartz grains were exposed to light.
Another source of uncertainty is in the origin of the pollen influx. Wind dispersal of the
pollen prior to deposition might have a significant effect (e.g. Okubo and Levin, 1989).
Therefore finding a proxy that is not influenced by this wind dispersal is essential to draw
more solid conclusions concerning past vegetation distributions. Lipid biomarkers is such a
proxy as it is based on root (no wind dispersal) and leaf (only very local wind dispersal)
biomass of the vegetation.
Lipids are important components in soil organic matter (SOM) due to their hydrophobic
nature (Quenea et al., 2006). As a result they are of major importance for various soil features
such as aggregate stability, water retention and fertility. These compounds derive from the
epicuticular waxes in leaves as well as waxes from roots and are usually present in distinctive
odd-over-even carbon chain length predominance for the n-alkanes and an even-over-odd
predominance for the n-alcohols (Jansen et al., 2006a). Solvent extractable lipids are usually
regarded as specific indicators of terrestrial higher plant input (Rieley et al., 1991; van Bergen
et al., 1997), as different plant species often have distinctive lipid compositions. If lipids are
well preserved within the soil, peat or sediment and the stratification is still intact, the lipid
composition within the soil profile can be used to determine past vegetation distributions
and related dominating biogeochemical processes. Until recently the focus of research on
lipids as biomarkers has been on lake sediments and peat bogs (e.g. Ficken et al., 1998;
Ishiwatari et al., 2005). The application of lipid biomarkers in terrestrial soils is a
comparatively unexplored terrain. In one of the few related studies Jansen et al. (2006a)
utilize ratios of straight-chain lipids of different chain-lengths to differentiate between
páramo and forest species in order to reconstruct past upper forest line positions in Ecuador
as part of the RUFLE project. It shows that in this particular case, n-alkanes and n-alcohols
compositions of the two different types of vegetation are the best suitable for differentiating
between them. N-fatty acids, wax esters and n-aldehydes appear to be less suitable as
8
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
vegetation biomarkers due to their apparent lower specificity (n-fatty acids) and absence
from a significant number of the species investigated (wax esters and n-aldehydes).
Bioturbation results in mixing of old and younger soil organic matter (SOM). When this
process is dominant the applicability of lipid biomarkers is very limited. However, soil
profiles of cultural sandy soils in the Netherlands show clear stratification, which together
with the results of both radiocarbon and OSL dating suggests that bioturbation is not a
dominant process in this system. As the n-alkanes and n-alcohols of interest (C20-C36, as these
are typical for higher plant input) also have a hydrophobic nature (Quenea et al., 2004), and
have shown to have a relatively high resistance to biodegradation, it is expected that they
remain in situ after deposition for a reasonably long time. However it should be noted that
roots grow downwards into the profile and their lipids might therefore be younger than the
surrounding soil material.
For this pilot study we selected a previously investigated polycyclic driftsand sequence,
profile Defensiedijk. The pollen diagram suggests the presence of plant species during stable
periods with soil formation. The separation of in situ species and regional species can be
based on the comparison of the pollen spectra with a carefully composed dataset of bio
markers.
9
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Netherlands
2. Objectives
As few is known about the effectiveness of the lipid biomarker approach applied on cultural
sandy landscapes, the primary goal of this research will be to determine its potential as an
added resource of information (proxy) regarding landscape development. To achieve this
goal, the following research questions have been addressed:

Are straight-chain lipids well preserved within the organic soil layers of interest in a
meaningful chronological order?

Are the straight-chain lipid compositions of the plant species responsible for the
dominant biomass input in the soil records in the area unique? Is it possible to
distinguish between groups of heather, grasses and trees? Is further identification on
species level achievable?

Can the straight-chain lipid composition within the soil samples be linked to the
different plant(group) specific compositions (both qualitative and quantitative)?

What is the relation between the results from the biomarker approach and the
existing palynological information derived from the analysis of fossil pollen?
The secondary objective of the research is to gain more insight into the landscape
development in cultural sandy landscapes. The following question has been formulated to
achieve this objective:

What conclusions can be drawn by combining the results of the biomarker approach
and the existing palynological information about the change in abiotic conditions
that were present during the transition of the coversand landscape into a driftsand
landscape?
11
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Netherlands
3. Research area
Landforms and soils around the city of Weert are representative for the cultural landscapes
that developed on chemically poor Late Weichselian aeolian coversands in NW Europe (van
Mourik, 1988). Historically, there was a close relation between the development of Plaggic
Anthrosols (IUSS Working Group WRB, 2007) and driftsand deposits in these cultural
landscapes (Bokhorst et al., 2005).
Figure 3.1: historical map of the research area ( adapted from van Mourik et al. (2010))
Profile Defensiedijk is situated in the Weerterbergen, SE of the city of Weert. Figure 3.1
shows a fragment of a historical map (1900 AD) with the characteristic land use of cultural
landscapes on chemically poor sandy soils. Around the city of Weert, arable fields (Plaggic
Anthrosols) are visible. More to the west we can see the extensive heath (Podzols) with
complexes of land dunes (Arenosols) and the first generation of pine plantations. Profile
Defensiedijk is considered as an important paleo-ecological record of stable and instable
periods in the development of the cultural landscape. In 1984, the profile was sampled for
pollen analysis. Also samples were taken for radiocarbon dating, applied on bulk samples. In
1986 the same profile was resampled for fractionated radiocarbon dating and soil
micromorphology. Finally, in 2002 samples were taken for OSL dating.
13
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
3.1. Summary of earlier work Defensiedijk
The main reason why Defensiedijk was chosen as the area of interest of this research is that
the abovementioned research was already carried out there, which gives the opportunity to
get a good idea of the applicability of the lipid biomarker approach. The following section
will therefore summarize the findings of the previous research done at the Defensiedijk
profile (van Mourik et al., 2010).
3.1.1. Pollen analysis (profile 1984)
Pollen extractions of samples from all horizons using the pollen determination key of
Moore et al. (1991) were carried out. This resulted in a pollen diagram (figure 3.2). This
diagram gives a glance at the landscape evolution of the profile Defensiedijk and its
surrounding. It shows the degradation from forest into heath as an effect of deforestation
(zone 1), the result of and the plaggen removal, with a replacement of heather by grasses
(zone 2) and the introduction of pine trees (plantation of Pinus started after 1550 AD) to
stabilize the driftsand landscape (zone 3). Zone 4 represents the most recent period of sand
drifting. Only since 1995 the area is stabilizing under a vegetation of grasses.
3.1.2. 14C dating (profiles 1984, 1986 and 2002)
Conventional radiocarbon ages of bulk samples of buried A horizons of the 1984 profile
were performed in order to interpret the chronology of pollen zones and sand deposits by
the CIO (Centrum voor Isotopen Onderzoek), Rijks Universiteit Groningen, The
Netherlands, following the methods of Mook and Streurman (1983). The 14C dates of bulk
samples (profile 1984) did not provide a clear geochronology (see table 3.1), as the top of the
profile is clearly dated too old. Therefore the profile was resampled in 1986 for fractionated
14
C dating (van Mourik et al., 1995) consisting of three fractions: fulvic acids (FUL), humic
acids (HAC) and humin (HUM). The biological decomposition rate of FUL is relatively high;
this fraction migrates easily downwards through the soil profile. This makes them unsuitable
for dating and is therefore discarded. The biological decomposition rate of HUM is lower
than of that of FUL. Compared to FUL, HUM is immobile and reliable for dating purposes.
The 14C age of HAC will be the closest to the moment of burying of the soil. HUM will
accumulate during an active period of soil development; therefore, 14C ages of this fraction
will overestimate the time of fossilization of the soil. It is assumed that the differences
14
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
Figure 3.2: pollen diagram of the Defensiedijk 1984 profile (from van Mourik et al. (2010))
The numbers under “Zone” refer to the zones decribed in 3.1.1. “S” stands for
stable period; “D” for a period in which deposition was important.
15
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
between the ages of HUM and HAC increase during an active period of soil formation. The
big difference in 14C ages between HUM and HAC in the 2A, 3A and 4A horizons confirms
this. Although the radiocarbon dating gives valuable insights in the development of the soil
profile, the dates are not very reliable. This is best shown by comparing the HAC ages of the
1986 and 2002 profiles of horizons 4A and 4B. The 2002 horizons are dated 1000-2000 years
younger, probably caused by influence of the above lying horizons (the distance between
horizons 3A and 4A is considerably smaller in the 2002 profile in comparison with the 1986
profile).
Table 3.1: 14C ages of different samples taken at profile Defensiedijk (from Van Mourik et al. (2010))
Depth
(cm)
25 - 27
Horizon
Bulk profile
HUM profile
HAC profile
Depth
HAC profile
1984 (a)
1986 (a)
1986 (a)
(cm)
2002 (a)
2A
1130 ±60
3230 ±110
410 ±45
69 - 71
410 ±35
127 - 129
3A (top)
1075 ±30
1350 ± 50
1365 ±25
128 - 130
1230 ±35
129 - 131
3A (bottom)
1900 ±110
1675 ±30
173 - 175
4A (top)
4110 ± 90
3615 ±35
154 - 156
2645 ±40
175 - 176
4A (bottom)
4430 ±165
3965 ±40
185 - 186
4B (top)
170 - 172
2080 ±35
187 - 188
4B (middle)
189 - 190
4B (bottom)
3920 ±40
3985 ±35
3535 ± 80
3730 ±35
3700 ±50
3.1.3. Micromorphology (profiles 1986 and 2002)
For micromorphological analysis, undisturbed samples were taken in Kubiena boxes for the
production of thin sections (7 cm/4 cm/20 μm) (Brewer, 1976). The main goal of this
analysis was to assess the reliability of the radiocarbon dating. Some of the main findings are
the presence of channels throughout the profile and sin-sedimentary charcoal particles in the
2C horizon. These signs of bioturbation explain the contamination of driftsand with SOM
from older soil horizons. Also pollen grains are found in sin-sedimentary transported
aggregates, which present another complication in the interpretation of pollen diagrams in
buried humic horizons.
16
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
3.1.4. OSL dating (profile 2002)
Unlike
14
C dating, luminescence dating uses the
constituent mineral grains of the sediment itself, and
it makes direct determination of sedimentation ages
possible (e.g. Aitken, 1998). The profile Defensiedijk
had
previously
been
sampled
for
thermo-
luminescence (TL) dating using feldspar, but the age
results lacked precision (Dijkmans et al., 1992). More
recently, optically stimulated luminescence (OSL)
proved to be better suited to date sediments, as it was
Table 3.2: OS L ages of soil horizons of
the 2002 profile
(adapted from Van Mourik et al. (2010))
Depth (cm)
45
55
70
105
105
125
137,5
145
152,5
162,5
Horizon
1C1
1C2
1C3
2Ah
2C1
2C1
2C2
3Ah
3E
4Bh
OSL age (a)
82 ± 8
100 ± 10
90 ± 10
350 ± 30
590 ± 50
670 ± 60
1300 ±100
5800 ±500
4700 ±400
9200 ±800
successfully applied to Holocene (Derese et al., 2010)
and late Pleistocene sediments (Vandenberghe et al., 2004) and soils in cultural landscapes
(Bokhorst et al., 2005). Therefore, in 2002, the profile Defensiedijk was resampled for OSL
dating. OSL dating on quartz grains was performed in the luminescence dating laboratory at
Ghent University. The methodology, luminescence characteristics of the samples and OSL
dating results have previously been presented by Vandenberghe et al. (2005).
The results of the OSL dating are shown in table 3.2. The ages for the samples from the 1C ,
2Ah and 2C are in chronological order and therefore stratigraphical consistent with their
position. Conversely, the samples of 3Ah and 3E show an age inversion. Field
observatations show a complicated sedimentary structure, which might be caused by short
distance re-sedimentation, explaining the age inversion. The sample of 4Bh was taken from
the coversand unit. Coversand deposition is generally believed to have stopped in the Late
Pleistocene, which implicates that the age of the sample of 4Bh of 9200 years must be
regarded as (at least 3000 years) too young. As discussed before, the micromorphological
analysis shows some signs of bioturbation; contamination through this bioturbation of
vertically transported younger mineral grains could have caused the OSL age of the 4Bh
sample to be underestimated. It cannot be ruled out completely that other parts of the
profile suffer from similar over- or underestimations of OSL ages.
Nevertheless, based on the OSL dating results, the chronology of the polycyclic sequence
became more clear (table 3.3). Multiple phases of sand drifting in the last approximately 1300
17
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
years can now be distinguished, and it shows signs of an extra phase of landscape instability
roughly 5000 years ago.
Table 3.3: OS L based geochronology of profile Defensiedijk (from van Mourik et al. (2010))
Cycle
4. Driftsand
3. Driftsand/micro-podzols
2. Driftsand/podzols
1. Coversand/podzols
Sedimentation
OSL ages (a)
100 - recent
1200 - 350
5000 - 4500
> 9200
Soil formation
OSL ages (a)
Initial
350 - 150
4500 - 1200
9200 - 5000
18
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
4. Methodology
In order to achieve the desired objectives the following methodology, consisting of
fieldwork, labwork and data analysis, was followed.
4.1. Field methodology
The field methodology consisted of collecting
samples of both humic soil horizons of the
Defensiedijk profile and plant tissue samples of
the species that were expected to have played a key
role in the landscape development. This fieldwork
was carried out in November 2008.
4.1.1. Soil sampling
At the Defensiedijk site the points of interest were
the different stages of vegetation development and
the related conditions in between the various
stadiums of sand drifting. Consequently, samples
were collected only of the soil horizons rich in soil
organic matter. When thickness of the organic
horizons was sufficient, separate samples were
taken from both upper and lower parts of the
horizon. For this, a soil pit of approximately 1 m2
surface and 1.5-2 meter deep was dug. The pit
(figure 4.1; additional pictures can be found in
appendix A) was dug near the pits for the
radiocarbon and OSL age analysis and pollen
analysis. It was considered that by doing this,
linking data of old and new pits would be rather Figure 4.1: Pit of Defensiedijk 2008 profile
straightforward (by using depths and boundaries of soil horizons). The stratification of the
soil profile was recorded (table 4.1 and figure 4.2), and 9 samples were taken. They were
collected into airtight glass containers to prevent contamination prior to analyses.
19
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
Table 4.1: Stratification of profile Defensiedijk 2008, with sample numbers
depth(cm)
horizon
sample nr. other characteristics
0-2
2-15
15-34
34-36
36-63
(Ah)
C
2C
3(Ah)
3C
1
63-64
64-77
4(Ah)
4C
3
77-78
78-80
80-88
5Ah
5AE
5Bw
4
88-100
100-115
115-115.5
115.5-130
130-134
5C
6C
7Ah
7C
8Ah
134-151
8E
151-154
154-168
168-172
172-180
9Ah
9E
9Bh
9Bs
9C
180 -
ochric Ah
laminated, light coloured sand
laminated driftsand, variation of humic and less humic lamineae
2
laminated driftsand, deposited on micro-dune
laminated driftsand
intermitting, in pockets
somewhat brownish, signs of bioactivity visible
laminated light coloured driftsand
laminated driftsand, colour suggest source is from A and B horizons
5
laminated driftsand, colour suggest source is from A and B horizons
6(130-132)
7(132-134)
traces of oxidation and reduction are visible,
which suggests water stagnation
8
9
Figure 4.2: Schematic
representati on of the
Defensiedijk 2008 profile
When comparing stratifications of the pit dug for this research with the ones done for
previous research it appeared that they were not as similar as anticipated beforehand.
Apparently, the driftsand landscape was highly dynamic and consequently local differences in
soil formation are big. Both depth and thickness of the various soil horizons in the 2008
profile are significantly different from the previously described profiles. Because of this, it
will not be possible to link data with great certainty. Table 4.2 shows the most likely
relationship between the sampled horizons of the 2008 Defensiedijk profile and the 1984
profile (see figure 3.2).
20
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
Table 4.2: Probable relationship between 2008 and 1984 Defensiedijk profiles
2008 horizon
(Ah)
3(Ah)
4(Ah)
5Ah
7Ah
8Ah
9Ah
9Bh
1984 horizon
1C
2Ah
2C1
2C2
2C6
3Ah
4Ah
4Bh
4.1.2. Vegetation sampling
The biomass input of the sampled soil profile is expected to be of a certain set of plant
species. In order to be able to determine which of these species or groups of species were
mainly responsible for the biomass input at the different parts of the profile, a reference set
was assembled. Combining both information from the collected palynological information of
the sample locations and literature (Spek, 2004) the species mentioned in table 4.3 were
selected to be sampled and incorporated in this study.
Table 4.3: Sampled vegetation species
Scientific name
English name
Dutch name
Calluna vulgaris
Molinia caerulea
Corynephorus canescens
Deschampsia flexuosa
Betula pendula
Pinus sylvestris
Quercus robur
Bryophyta (exact species unknown)
Heather
Moor Grass
Grey hair-grass
Wavy h air-grass
Silver Birch
Scots Pin e
Pedun culate Oak
moss
Struikheide
Pijpenstrooitje
Buntgras
Bo chtige smele
Ruwe berk
Grove den
Zomereik
mos
Ascomycota (exact species unknown)
lichen
korstmos
It has to be said that at the time of the fieldwork no complete knowledge was obtained about
these key species, and therefore the selection of species later appeared not to be ideal.
Sampled species Quercus and Betula were not found in the pollen record of the Defensiedijk
profile. It would have been better if they were replaced by Corylus and Alnus. However, all
four of these species belong to the group of deciduous trees and the lipid patterns can
21
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
therefore expected to be similar, as lipid patterns of related species are often to some extent
similar.
4.2. Laboratory methodology
In order to prepare for the GC/MS analysis, a complex multi-step laboratory procedure was
followed. It consisted of the following stages: sample preparation, the extraction of the
straight-chain lipids from the sample, the purification of the extract from polar components
and possible traces of moisture, and the derivatisation of thus obtained extract. Each of
these steps is explained in detail below. The description of the GC/MS analysis follows.
4.2.1. Preparation
The preparation of the soil and vegetation samples consisted in essence of the same stages.
The major difference in the procedure was that the vegetation samples, before putting into
the containers which would later be place into the freezer, had to be cut into small pieces
and divided into roots and leaves. For obvious reasons this was impossible to perform on
the moss and lichen samples. Both soil and vegetation samples were put into the freezer at a
temperature of approximately -10°C and stayed there overnight. Afterwards, the sample was
dehydrated using a freeze-dryer. The next step was grinding the samples with the use of a
grinding machine relevant for the sample kind. The soil samples were ground into particles
of a size smaller than 2μm using a planetary ball mill, while the vegetation samples were
ground into particles of a size smaller than 0.2mm using a rotor mill. Finally, the ground
material was homogenised and placed into clean 20ml plastic containers.
4.2.2. ASE extraction
In order to analyse the lipid composition of every sample, the first step is to extract the lipids
from the soil. The most established technique to do this is Soxhlet extraction (eg. Naafs et al.,
2004). However, Soxhlet extraction is very time consuming (16 hours per sample) and takes
relatively large volumes of solvent (250 mL or more). A relatively new alternative is the
technique of accelerated solvent extraction (ASE) (Richter et al., 1996). Basically it uses raised
temperature and pressure to shorten extraction time to less than one hour. Jansen et al.
(2006a) show that extraction performance of ASE is on the same level with Soxhlet
22
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
extraction. Therefore ASE extraction was chosen as the method to extract the lipids from
the soil in this study.
Roughly 100mg of every sample (the exact weight was registered) was extracted by the
method of Accelerated Solvent Extraction (ASE) employing a Dionex 200 ASE extractor
with 11 ml extraction cells and dichloromethane/methanol (DCM/MeOH) (93:7 v/v) as the
extractant. SiO 2 was used to fill the extraction cells. The extractions were performed at a
temperature of 75°C and a pressure of 17x10 6 Pa with a heating phase of 5 minutes and a
static extraction time of 20 minutes; these settings gave good results according to Jansen et al.
(2006b). The product of the extraction was approximately 20ml of each sample extract
collected in a glass vial.
4.2.3. Clean-up
The DCM/MeOH extractant was removed from the extract using a rotary evaporator.
Subsequently, the residue was redissolved in approximately 2-5ml DCM/2-propanol (2:1
v/v). Afterwards, this solution was filtered through a Pasteur pipette filled with defatted
cotton wool, 0.5cm Na 2SO 4(s) and 2cm SiO 2(s). This filtering is performed to remove
potential remnants of moisture (with Na 2SO) and very polar components (with SiO 2). The
filtered extract was then collected in 5ml glass vials which were weighed in advance. Then
the DCM/2-propanol was removed by evaporation at 30°C, accelerated by applying a mild
stream of N2(g) over the opened vials. The next step of the clean-up procedure consisted in
weighing the vials again in order to determine the amount of residue of the different
samples. After this, the residues were redissolved in 1ml DCM/2-propanol.
4.2.4. Derivatisation
In order to get a distinct signal while preventing pollution of the GC/MS a final
concentration of roughly 300mg residue per litre solvent was selected. After GC/MS analysis
it showed that this concentration was too low to get distinct signals of the lipids. Therefore
the concentrations were raised to approximately 3g residue per litre solvent. To get to these
concentrations it was calculated for every sample how much extract was needed and in how
much cyclo-hexane it had to be dissolved. The calculated amounts of cleaned-up extracts
were then dispensed into a new 5ml vial. Then, known amounts of an internal standard,
23
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
which consisted of a mix of deuterated lipids (d42-n-C20 alkane, d41-n-C20 alcohol and d39n-C20 fatty acid) in a solution of 320mg/l were added to the vials, depending on the final
volume (20µl internal standard per 1ml final volume). After this the DCM/2 -propanol was
again removed by evaporation at 30°C, supported by a stream of N2(g). To the obtained
residue 50µl BSTFA (N,O-bis(trimethylsilyl) trifluoroacetamide) containing 1% TMCS
(trimethylchlorosilane) was added. The closed vials were then heated at 70°C for 1 hour to
derivatise all free hydroxyl and carboxylic-acid groups to their corresponding trimethylsilyl
(TMS) ethers and esters. After cooling down they were dried another time by applying the
N2(g) stream at 30°C to remove the surplus of BSFTA. Finally, the residues were redissolved
in the calculated amounts of cyclo-hexane.
4.2.5. GC/MS analysis
Gas chromatography-mass spectrometry (GC/MS) analyses were performed on a
ThermoQuest Trace GC 2000 gas chromatograph connected to a Finnigan Trace MS
quadrupole mass spectrometer. Separation took place by on-column injection of 1.0µl of the
derivatised extracts on a 30-m Restek Rtx®-5Sil MS column with an internal diameter of
0.25mm and film thickness of 0.1µm, using He as a carrier gas. Temperature programming
consisted of an initial temperature of 50°C for 2 minutes, heating at 40°C/min to 80°C,
holding at 80°C for 2 minutes, heating at 20°C/min to 130°C, immediately followed by
heating at 4°C/min to 350°C and finally holding at 350°C for 10 minutes. The subsequent
MS detection in full scan mode used a mass-to-charge ratio (m/z) of 50-650 with a cycle time
of 0.65s and followed electron impact ionisation with an ionisation energy of 70eV (Jansen et
al., 2006b)
24
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
4.3. Data analysis
After chromatograms were obtained from the GC/MS analysis, a quantification of the
different lipid compounds was made. To assess how the lipid patterns differentiate between
the vegetation samples, a cluster analysis was performed. Finally a model was used to
reconstruct the vegetation history based on the lipid patterns.
4.3.1. Lipid quantification
Resulting GC/MS peaks in the chromatograms for n-alkanes and n-alcohols were interpreted
and peak areas calculated.
The selected compounds were identified from the chromatograms based on their mass
spectra and their retention times, using the peak area from selected dominant fragment ions
for each class of components. The fragment ions were represented by: m/z 57 for the nalkanes and m/z 75 for the n-alcohols. Absolute concentrations of the different compounds
were calculated in µg/g soil sample by assuming that the response of the GC/MS is the same
for the added deuterated standards and the natural compounds in the samples. Haussmann
(2006) showed that this assumption is valid for n-alkanes and n-alcohols with carbon chainlengths in the range of interest for this study (C16 to C36). The selected fragment ions for
the internal standards were represented by: m/z 66 for d42-n-C20 alkane and m/z 76 for
d41-n-C20 alcohol.
4.3.2. Cluster analysis
In order to assess the distinctiveness of the patterns of the lipids found in the vegetation
samples a cluster analysis was carried out. This also gives an impression whether the lipid
data of the different species can be upscaled to the vegetation types mosses, grasses, heather
and (deciduous and coniferous) trees. Ideally, a cluster analysis would result in clusters
containing the different vegetation species belonging to a single vegetation type.
Clustering was done by doing a hierarchical cluster analysis in PASW Statistics 18 (SPSS Inc.,
2009) using Ward’s method (Ward, 1963). The data to be clustered was the relevant lipid
concentrations (alcohols with even-numbered carbon chain lengths and alkanes with unevennumbered carbon chain lengths) of both leaf and root samples. The lipid concentrations
25
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
were standardized to ranges from 0 to 1 prior to clustering. The squared Euclidean distance
was used as a measure of similarity as this measure is the recommended distance measure for
Ward’s method (Everitt et al., 2001).
In some occasions the outcome of clustering is depended on the order of the clustered data.
To check for this so called robustness, the cluster analysis was done twice with differently
ordered data.
4.3.3. VERHIB model
Vegetation types are reconstructed by linking the signals of both the vegetation species and
the soil samples. Linking the signals is not very straightforward and difficult to do by hand.
Therefore, Jansen et al. (2010) developed the statistical model VERHIB (VEgetation
Reconstruction with the Help of Inverse modelling and Biomarkers) in MATLAB
(Mathworks, 2007). It was designed and successfully applied to determine whether a shift in
the upper forest line in Ecuador took place. However, its universal design makes application
on different datasets as the one obtained in this research possible.
The VERHIB model consists of a linear regression model to describe the way in which a
certain vegetation development over time at a certain location results in accumulation of
biomarkers (n-alkanes and n-alcohols) in a suitable archive (the soil samples at various
depths). An inversion of the forward model is used to reconstruct paleo-vegetation on the
basis of the observed accumulated biomarker signal (Jansen et al., 2010).
Furthermore, there are a number of model parameters that have to be selected in order to
run the model. They have been selected in the following ways:

Incorporation of leaf and root data. For every single species it can be selected
whether to use the leaf and/or root data. It was chosen to incorporate all the data
into the model. For both moss and lichen there were no separate samples taken for
leaf and root material. Here, the obtained biomarker data was introduced in
VERHIB as being both leaf and root material.

Leaf-root ratio. An expected leaf-root ratio can be chosen for every single species.
As this ratio is not determined in this study, it is set as the standard value of 3 for all
26
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
species. This implicates that the contribution of leaf-lipids to the soil is considererd
to be three times as large as the contribution of root-lipids.

Depth duration. For every sampled horizon it can be set how long it took to
develop its thickness. As the sampled horizons in this study are not continuous,
setting this parameter did not seem very useful. On top of this, testing VERHIB with
different values for this did not affect the modelled output at all. Therefore, for every
soil sample the depth duration was set to 200 years, to have some kind of realistic
value.

Leaf depth. For every species it can be set how much the leaves mix throughout
the soil horizon. As the horizons are mostly more than 10 cm apart from each other,
it was assumed that no mixing takes place.

Root depth. A similar parameter is incorporated to take into account the mixing of
roots with older layers. Grasses, mosses and heather have only very shallow roots
and mixing is therefore set as zero. Pinus is known to root deeply, and therefore the
parameter was set so that 20% of a horizons signal is modelled one soil horizon
below it and 10% was modelled two soil horizon below it. Deciduous trees root less
deeply than pine trees, hence these values were chosen as 10% and 5%, respectively.

Include chem. It can be chosen which specific compounds to base the model on.
As the typical vegetation lipid compounds are considered to be alcohols with even
numbered carbon chain-lengths and alkanes with uneven carbon chain-lengths,
alcohols C16,18,20,22,24,26,28,30,32,34,36 and alkanes C17,19,21,23,25,27,29,30,
31,33,35 were used as model input. Additionally, the model was also run only based
on alcohols and alkanes separately, to assess the importance of combining both with
vegetation reconstruction in mind.
The VERHIB model appeared to be very sensitive of differences in concentrations between
samples (both vegetation and soil samples). As it is desired to base the modelled results on
the difference in patterns, rather than on the differences in concentrations, first a
normalization of the data was carried out. Per sample the total concentration of biomarkers
was determined. Every compound concentration was then divided by this total
concentration.
27
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
The Matlab code of VERHIB used for this analysis is appended as appendix B. This was
thought to be useful as the model might be altered and extended in the future, and possible
confusion about the used version of VERHIB is hereby avoided.
28
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
5. Results and discussion
5.1. Lipid quantification
Figures 5.1a and 5.1b show the results for the vegetation samples of the GC/MS analysis.
These only show the results of the third analysis of the vegetation samples. The first tw o
Lichen
Moss
14
18
alcohols
alkanes
10
8
6
4
alcohols
16
lipid concentration (μg/g plant)
lipid concentration (μg/g plant)
12
2
alkanes
14
12
10
8
6
4
2
0
0
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
16
17
18
19
20
21
22
23
Carbon chainlength
24
25
26
27
28
29
30
31
32
33
34
Calluna: leaves
7
alcohols
alkanes
100
80
60
40
20
alcohols
6
lipid concentration (μg/g plant)
lipid concentration (μg/g plant)
36
Calluna: roots
120
alkanes
5
4
3
2
1
0
0
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
16
17
18
19
20
21
22
23
Carbon chainlength
24
25
26
27
28
29
30
31
32
33
34
35
36
Carbon chainlength
Molinia: leaves
Molinia: roots
45
3
alkanes
35
30
25
20
15
10
alcohols
lipid concentration (μg/g plant)
alcohols
40
lipid concentration (μg/g plant)
35
Carbon chainlength
alkanes
2,5
2
1,5
1
0,5
5
0
0
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
16
17
18
19
20
21
22
23
Carbon chainlength
24
25
26
27
28
29
30
31
32
33
34
36
Corynephorus: roots
Corynephorus: leaves
35
250
alcohols
alcohols
alkanes
200
150
100
50
30
lipid concentration (μg/g plant)
lipid concentration (μg/g plant)
35
Carbon chainlength
alkanes
25
20
15
10
5
0
0
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
16
36
17
18
19
20
21
22
23
24
25
26
27
28
Carbon chainlength
Carbon chainlength
Figure 5.1a: Results of the third GC/ MS analysis of the vegetation samples
29
29
30
31
32
33
34
35
36
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
analyses did not lead to satisfactory results, but have been included as appendix C. The first
analysis resulted in unexpected high concentrations of alkanes with even-numbered carbon
chainlengths in most of the vegetation samples. Therefore it was decided to carry out a
second identical analysis. This second analysis resulted in significant lower concentrations of
alkanes with even-numbered carbon chainlengths. This shows that somewhere during the
Deschampsia: leaves
Deschampsia: roots
60
35
alkanes
40
30
20
10
alcohols
30
lipid concentration (μg/g plant)
lipid concentration (μg/g plant)
alcohols
50
alkanes
25
20
15
10
5
0
0
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
16
17
18
19
20
21
22
23
Carbon chainlength
24
25
26
27
28
29
30
31
32
33
34
Pinus: leaves
0,9
alcohols
alkanes
10
8
6
4
alcohols
0,8
lipid concentration (μg/g plant)
12
lipid concentration (μg/g plant)
36
Pinus: roots
14
2
alkanes
0,7
0,6
0,5
0,4
0,3
0,2
0,1
0
0
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
16
17
18
19
20
21
22
23
Carbon chainlength
24
25
26
27
28
29
30
31
32
33
34
35
36
Carbon chainlength
Betula: leaves
Betula: roots
120
9
alcohols
alcohols
8
alkanes
100
lipid concentration (μg/g plant)
lipid concentration (μg/g plant)
35
Carbon chainlength
80
60
40
20
alkanes
7
6
5
4
3
2
1
0
0
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
16
17
18
19
20
21
22
23
Carbon chainlength
24
25
26
27
28
29
30
31
32
33
34
Quercus: roots
36
Quercus: leaves
4
25
alkanes
3
2,5
2
1,5
1
alcohols
lipid concentration (μg/g plant)
alcohols
3,5
lipid concentration (μg/g plant)
35
Carbon chainlength
alkanes
20
15
10
5
0,5
0
0
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
16
Carbon chainlength
17
18
19
20
21
22
23
24
25
26
27
28
Carbon chainlength
Figure 5.1 b: Results of the third GC/ MS anal ysis of the vegetation samples
30
29
30
31
32
33
34
35
36
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
first analysis, the samples must have been contaminated with alkanes not part of the
vegetation samples. Unfortunately, it was not possible to determine when exactly the
contamination took place. Although the results of the second analysis seemed to be better
than those of the first analysis, they were still unsatisfactory. Many of the peaks within the
chromatograms were difficult or even impossible to quantify as they were very low
compared to the noise within the chromatograms. Therefore a third analysis was carried out.
This time the concentration of the vegetation samples was raised from roughly 300mg to 3g
vegetation residue per litre solvent. This resulted in more distinguishable peaks, which seem
to be relatively accurate. However, due to having to repeat the procedure two times, no time
was left to assess the repeatability of the analysis with the increased concentrations. But as
the results of the third analysis appeared to be significantly better than those of the first two
analyses, modelling and statistical analysis was only based on the results of the third analysis.
Similarly, two GC/MS analyses were carried out of the soil samples. As the concentrations
of lipids in soil samples were generally lower than those in vegetation samples, results of the
first (300 mg soil residue per litre solvent) analysis showed few peaks, which were often
inaccurate because of the high noise to signal ratio. The second analysis (with roughly 3g soil
residue per litre solvent), showed much better results. Figure 5.2 shows these results.
Appendix C shows the results of the first analysis of the soil samples, additionally.
The concentrations of the lipids were based on the assumption that the response of the
GC/MS is the same for the added deuterated standards and the natural compounds in the
samples as is explained in the methodology description (4.3.1). However, from the results of
the GC/MS analysis it seems that this assumption might not be correct. Figure 5.3 shows the
ratio between the measured concentrations of the added deuterated standards d-20 alcohol
and d-20 alkane in the different. In theory (with a perfectly clean GC/MS), this ratio should
be the same for all samples, as the standards were added from a single solution with
deuterated standards. But in practice the ratio had a mean of 1.9, a standard deviation of 1.2
and minimum and maximum values of 0.8 and 5.6, respectively. An explanation for a
difference in these ratios could be the gradual pollution of the GC/MS throughout a
measuring sequence. However, when looking at the continuous sequence of the soil samples,
31
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
Soil sample 1 (0-2 cm)
Soil sample 2 (34-36 cm)
7
1,4
alcohols
alcohols
1,2
alkanes
lipid concentration (μg/g soil)
lipid concentration (μg/g soil)
6
5
4
3
2
1
alkanes
1
0,8
0,6
0,4
0,2
0
0
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
16
17
18
19
20
21
22
23
Carbon chainlength
24
25
26
27
28
29
30
31
32
33
34
Soil sample 3 (63-64 cm)
1,6
alcohols
0,6
alcohols
1,4
alkanes
lipid concentration (μg/g soil)
lipid concentration (μg/g soil)
36
Soil sample 4 (77-78 cm)
0,7
0,5
0,4
0,3
0,2
0,1
alkanes
1,2
1
0,8
0,6
0,4
0,2
0
0
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
16
17
18
19
20
21
22
23
Carbon chainlength
24
25
26
27
28
29
30
31
32
33
34
35
36
Carbon chainlength
Soil sample 5 (115-115.5 cm)
Soil sample 6 (130-132 cm)
0,5
3
alcohols
alcohols
0,45
alkanes
0,4
lipid concentration (μg/g soil)
lipid concentration (μg/g soil)
35
Carbon chainlength
0,35
0,3
0,25
0,2
0,15
0,1
alkanes
2,5
2
1,5
1
0,5
0,05
0
0
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
16
17
18
19
20
21
22
23
Carbon chainlength
25
26
27
28
29
30
31
32
33
34
35
36
Carbon chainlength
Soil sample 7 (132-134 cm)
Soil sample 8 (151-154 cm)
0,7
5
alcohols
0,6
alcohols
4,5
alkanes
lipid concentration (μg/g soil)
lipid concentration (μg/g soil)
24
0,5
0,4
0,3
0,2
0,1
alkanes
4
3,5
3
2,5
2
1,5
1
0,5
0
0
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
16
17
18
19
20
21
22
23
Carbon chainlength
24
Soil sample 9 (168-172 cm)
10
alcohols
lipid concentration (μg/g soil)
9
alkanes
8
7
6
5
4
3
2
1
0
16
25
26
27
28
Carbon chainlength
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
Carbon chainlength
Figure 5.2: Results of the second GC/ MS analysis of the soil samples
32
34
35
36
29
30
31
32
33
34
35
36
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
no trend of a gradual change of ratio could be observed. Therefore, basing the
concentrations of the lipids on the measured deuterated concetrations, seemed not to be
very accurate. However, there was no alternative. For the separate patterns of both alcohols
and alkanes of different carbon chainlengths, the absolute concentrations are not important
though.
ratio d41-n-C20 alcohol/d42-n-C20 alkane
6
5
4
3
2
1
0
Figure 5.3: Rati o between both added deuterised compounds
33
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
5.2. Cluster analysis
Hierarchical clustering of both leaf and root lipid (alcohols and alkane) concentrations
resulted in the clusters shown in the dendrogram shown below (figure 5.4). The proximity
matrix, resulting from calculating the squared Euclidean distances,which was used to carry
out the clustering is shown as table 5.1. Sample order did not influence the cluster analysis,
thus the clustering can be considered as robust.
Although the obtained dendrogram gives a quick insight on the similarities of the lipid
patterns of the different species, looking in detail at the squared Euclidean distances reveals
some additional information.
Of all the sampled species, Betula and Quercus were by far the most similar species. Based
on the lipids patterns they could not be distinguished at all. This must be taken into account
when analysing the VERHIB modelling results. Assuming that other deciduous trees also
have very similar patterns, it seems not to be such a serious omission that no samples of
Alnus and Corylus were analyzed. But there is no absolute certainty that this assumption can
be made.
Calluna’s lipid patterns were most closely related to Betula and Quercus. Squared Euclidian
distances to the other species are relatively big.
Lichen’s patterns were most closely related to those of Quercus, closely followed by Moss
and Betula. Moss was most closely related to Deschampsia, Molinia and Lichen. So, although
Lichen did not form a cluster with Moss, they are definitely related.
According to squared Euclidian distance, Molina was most closely related to Deschampsia,
Betula, Quercus and Corynephorus. This suggests that lipid patterns of grasses are quite
similar, but also that deciduous trees have such similar patterns, that makes a clear
distinction between grasses and deciduous trees difficult.
This can also be seen when focusing on Corynephorus and Molinia. Corynephorus had
34
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
relatively small squared Euclidian distances with Molina, Betula and Quercus. Deschampisa
had the smallest squared Euclidian distances with Molinia, Quercus, Moss and Betula.
Figure 5.4: Dendrogram, resulting from the cluster analysis of all the data
Table 5.1: Proximity matrix of the vegetation samples, based on leaves and roots signals
Case
Squared Euclidean Distance
Lichen
Moss
Calluna
Molinia
Corynephorus
Deschampsia
Pinus
Betula
Quercus
Lichen
,000
2,467
3,515
4,163
4,016
3,702
5,049
2,705
2,229
Moss
2,467
,000
3,681
2,014
2,659
1,960
3,901
2,522
2,528
Calluna
3,515
3,681
,000
2,279
2,663
3,126
3,472
,946
,981
Molinia
4,163
2,014
2,279
,000
1,836
1,351
2,459
1,450
1,489
Corynephorus
4,016
2,659
2,663
1,836
,000
2,467
2,716
1,939
1,995
Deschampsia
3,702
1,960
3,126
1,351
2,467
,000
3,336
2,096
1,959
Pinus
5,049
3,901
3,472
2,459
2,716
3,336
,000
2,492
2,563
Betula
2,705
2,522
,946
1,450
1,939
2,096
2,492
,000
,084
Quercus
2,229
2,528
,981
1,489
1,995
1,959
2,563
,084
,000
35
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
Pinus had the biggest average squared Euclidian distance to the other species. It had the
closest relation with Molinia, Betula and Quercus.
In general, the groups of vegetation that the different species belong to, were reflected in the
cluster analysis. However the relationships between the species belonging to the different
vegetation groups were not very strong. The uniqueness of the lipid patterns of the
researched species seems not to be ideal. Distinctions will be hard to make with great
certainty.
36
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
5.3. VERHIB model
The output of the VERHIB model, based on alcohols with even-numbered carbon
chainlengths and alkanes with uneven-numbered carbon chainlengths of both leaves and
root data is shown in figure 5.5. It should be noted that VERHIB will always produce the
best fit of the lipid patterns of the included vegetation species to the lipid patterns of the soil
samples, even if the input data is flawed or important vegetation species are missing.
VERHIB output
alkanes+alcohols, leaves+roots
0
20
modelled vegetation coverage (%)
40
60
80
100
soil sample depth (cm)
0-2
34-36
63-64
77-78
115-115.5
130-132
132-134
151-154
168-172
Lichen
Moss
Calluna
Molinia
Corynephorus
Deschampsia
Pinus
Betula
Quercus
Figure 5.5: visualization of VERHIB model output, based on all data
In the following section an attempt has been made to explain the modelled output by the
lipid patterns of both vegetation and soil samples. Values of less than 10% of modelled
vegetation coverage are not discussed here, as they are difficult to interpret and statistically
weak.
37
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
5.3.1. Output analysis
Soil sample 1 (0-2 cm) was modelled to consist almost completely (95%) of lipids produced
by Lichen. When the patterns of soil sample 1 are compared with the patterns of the Lichen
sample it can be seen why this is. These patterns look very similar, both for alkanes and
alcohols. Seemingly, lichens consist of quite some extractable lipids, which made up the
majority of the lipids that were found in the upmost soil horizon. The photos of the top of
the Defensiedijk profile (appendix A, e.g. photo 1) show that the soil profile is mostly
covered by lichens (or mosses, they are hard to distinguish on photos), and thus confirms
the outcome of the model.
The second soil sample (34-36 cm) consisted according to the model output of 49% Molinia,
21% Pinus and 20% Quercus. The high peak of alcohol C28 is the main reason why this
sample was modelled to consist of high amounts of Molinia. Quercus can be explained by
additional lower peaks of alcohols C22, C24 and C26. The 20% modelled Pinus is hard to
interpret; most characteristic peaks of Pinus were also present in the soil sample, but its very
characteristic high peaks of alcohols C22 and C24 were only present in low amounts in the
soil sample.
Soil sample 3 (63-64 cm) was build up by 31% Corynephorus, 25% Quercus, 22% Lichen
and 19% Pinus according to VERHIB. The high peak of C26 was responsible for the 31%
of modelled Corynephorus. All major lipid compounds of Quercus, Lichen and Pinus were
also present in the sample, which explains their presence in the output model. Although, the
low alcohol peaks of C22 and C24 makes the presence of Pinus in this sample also
questionable.
VERHIB’s output modelled soil sample 4 (77-78 cm) to exist mainly (78%) of Calluna and
some Pinus and Lichen (both 11%). Calluna is characterised by its large peaks of alkanes C31
and C33. These peaks were also present in this sample, which explains its high modelled
presence. The main lipid compounds of Lichen and Pinus were also present in the soil
sample, but it is impossible to further interpret these low amounts of modelled coverage.
38
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
Soil sample 5 (115-115.5 cm) consisted according to the model output of 44% Pinus and
32% Quercus. Pinus can be explained by the relative high peaks of alcohols C22 and C24
that are not present in the other species. The peaks of alkanes C25, C27, C29 and C31 were
responsible for the modelled Quercus.
The modelled output suggests that soil sample 6 (130-132 cm) consisted of Calluna (56%)
and Pinus (44%). This can be explained by the presence of their characteristic peaks (alkanes
C31 and C33 for Calluna, and alcohols C22 and C24 for Pinus).
Soil sample 7 (132-134 cm), which was sampled directly below soil sample 6 also had a
similar composition according to the modelled output; 42% of Pinus and 35% of Calluna.
Soil sample 8 (151-154 cm) had a high peak of alcohol C26. As this is characteristic for
Corynephorus, its modelled coverage of 67% can be explained by this. The 28% of Pinus
must have been modelled because the alcohols C22 en C24 were present. However, these
peaks were again low and might not be able to explain for these amounts of Pinus.
Finally, soil sample 9 (168-172 cm) consisted mainly of Molinia (70%) and some Pinus
(21%). The single high peak of alcohol C28 explains Molinia. The Pinus signal could also be
seen here, but similar problems as with soil samples 2, 3 and 8 make this uncertain.
The VERHIB model suggests that Pinus played a big role in the landscape development.
However, pollen diagrams show that Pinus was only present in recent history. The first
explanation that came to mind is that through the deep rooting of Pinus, more recent
biomass of Pinus roots contaminates the biomarker signal of the older soil horizons. But
when looking at the biomarker patterns of root and leaf biomass, it can be seen that both are
very different and that the root signal is very weak. It also showed that its leaf pattern is very
broad with presence of alcohols with chain lengths of 22, 24, 26, 28, 30 and 32. Most soil
horizons also contained these alcohols, which caused VERHIB to model Pinus as the main
vegetation in these horizons. Most likely there was another source besides Pinus that was not
taken into account in this study, which was responsible for the presence of alcohols with
chain lengths of 22 to 32.
39
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
5.3.2. Alternative runs
Of main interest for the research was to run the model with all available data. However,
running the model with only part of the data gave a lot of insight in the distinguishness of
the lipid patterns and the related strength of the previously discussed model output.
Figure 5.6 shows the output of VERHIB when only alcohols (left) or only alkanes (right)
alkanes only, leaves+roots
were used as model input. The results were very different from the combined model output
modelled vegetation coverage (%)
0
20 was completely
40
80 layer in100
(figure 5.4). For example
Lichen
absent60in the surface
the model
0-2
alkanes only, leaves+roots
alcohols only, leaves+roots
soil sample depth (cm)
0
modelled vegetation coverage (%)
34-36
20
40
60
100 0
80
soil sample depth (cm)
soil sample depth (cm)
63-64
77-78
63-64
77-78
115-115.5
115-115.5
130-132
132-134
130-132
132-134
151-154
151-154
168-172
modelled vegetation coverage (%)
40
60
80
100
0-2
0-2
34-36
20
168-172
Lichen
Deschampsia
34-36
63-64
77-78
115-115.5
130-132
132-134
151-154
168-172
Lichen
Moss
Calluna
Moss
Molinia
CorynephorusCalluna
Deschampsia
PinusPinus
Betula
Quercus
Betula
Lichen
Molinia
Molinia
Pinus
Quercus
Moss
Calluna
Corynephorus
Corynephorus
Deschampsia
Betula
Quercus
Figure 5.6: VERHIB model output, based on alcohol data (left) and alkane data (right)
output based solely on the alcohol patterns, while this was the main constituent in the
combined model. The most likely explanation for the difference in model output is that the
patterns of both alcohols and alkanes of the tested vegetation species on their own are not
distinct enough to reconstruct the paleo-vegetation based on extractable lipids within the
sampled soil profile, although the simple fact that only half of the data was used when
running the model for only one component class, is also likely to impair the quality of the
output. It also cannot be ruled out completely that the lipid signals of one component class
(either alcohols or alkanes) explained the species distribution very well, while the other
component class performed very poorly. In this way the combined model would give less
accurate output as the one for the well performing component class.
Another noticeable thing is that a lot more Pinus was modelled when only taking into
account alcohols or alkanes. It seems that Pinus has the most general, broad pattern for both
alcohols and alkanes, which caused VERHIB to model the remaining part of the pattern (the
40
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
part that cannot be explained by the lipid patterns of other species) as Pinus. This explains
the presence of Pinus in the combined model output, when its pattern cannot easily be
distinguished. It would be very interesting to see if adding an extra component class (e.g.
fatty acids) results in smaller amounts of Pinus in the output model, as it seems to be
important to use as much (relevant) data as possible as model input in order to obtain the
most accurate model output.
Figure 5.7 shows VERHIB’s output when only using either leaf lipids (left) or root lipids
(right) are used as model input. Again the results here are quite different from the results
obtained with the combined model. Especially when only using only root data as input, the
results are almost incomparable
withleaves+roots
the combined model. This is partly due to the higher
alkanes only,
importance that is given to the leaf patterns compared to the root patterns (3:1), but also
modelled vegetation coverage (%)
20
40 distinctive
60 than the root
80 lipid patterns.
100
because the 0leaf lipid patterns
are more
0
alkanes+alcohols, roots only
alkanes+alcohols, leaves only
34-36
20
modelled vegetation coverage (%)
40
60
100 0
80
63-64
34-36
77-78
63-64
77-78
115-115.5
130-132
115-115.5
132-134
130-132
132-134
151-154
modelled vegetation coverage (%)
40
60
80
34-36
63-64
77-78
115-115.5
130-132
132-134
151-154
151-154
168-172
168-172
Lichen
Deschampsia
20
0-2
0-2
soil sample depth (cm)
soil sample depth (cm)
soil sample depth (cm)
0-2
168-172
Lichen
Moss
Molinia
Pinus
Pinus
Moss
Calluna
Calluna
Corynephorus
Deschampsia
Betula
Betula
Quercus
Molinia
Quercus
Lichen Corynephorus
Moss
Calluna
Molinia
Corynephorus
Deschampsia
Pinus
Betula
Quercus
Figure 5.7: VERHIB model output, based on leaf data (left) and root data (right)
41
100
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
5.4. Relation to data from previous studies
When the results from the VERHIB model are compared to the results from the previously
carried out pollen analysis (see figure 3.2), the following observations can be made:
The relative abundance of deciduous trees (Alnus and Corylus) pollen found in the lower
part of the profile was not seen in the output of the VERHIB model. Instead, the VERHIB
model output suggests that the in situ vegetation consisted mainly of grasses. One reason for
this could be the absence of both Alnus and Corylus as vegetation samples. However, it is
expected that Betula and Quercus will have more similar lipid patterns to Alnus and Corylus
in comparison to grasses (Molinia, Corylus and Deschampsia). But the cluster analysis also
showed that grasses and deciduous trees have relatively similar lipid patterns and are
therefore hard to distinguish. This shows that that the use of lipid patterns (of alkanes and
alcohols) as a single proxy does not always explain the vegetation history correctly.
The two periods of high abundance of Ericaceae (Calluna) pollen in the (lower) middle of
the profile were very well reflected in the output of VERHIB; it models mainly Calluna in
the soil samples at depths 77-78 cm, 130-132 cm and 132-34 cm, which correspond with the
peaks in heather in the pollen diagrams.
The replacement of pollen of heather by pollen of grasses after the removal of heather in the
upper (middle) part of the Defensiedijk profile is also visible in the output of VERHIB (3436 cm, 63-64 cm).
The big peak of Pinus pollen in the top of the pollen diagram could not be seen in the model
output. Possibly this is caused by the very strong signal of Lichen that might blend the signal
of Pinus.
42
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
6. Conclusions
Application of lipid biomarkers as an added proxy in the reconstruction of landscape
development in cultural sandy landscapes appeared to be a promising methodology. This
research showed that the two most important conditions on which the lipid biomarker
approach depends are met. Firstly, both the signal of the n-alkanes and the signal of the nalcohols were found throughout the whole soil profile in somewhat similar concentrations.
This implicates that the decomposition of these (lipid) plant compounds is slow enough to
draw conclusions on the plant composition of organic rich soil layers up to at least a few
thousand years. Secondly, combining the ratios between the carbon chain length of both the
n-alkanes and the n-alcohols of the different investigated species, it is found to be possible to
distinguish the group of vegetation (grasses, heather and trees). Differentiation on a species
level appeared much more difficult and is only feasible for a few of the studied vegetation
species.
Generally, the results of the biomarker analysis support the alleged history of landscape
development, based on previous (palynological) studies, of the surroundings of the studied
polycyclic soil profile. In the cases where the results of the biomarker analysis point in
different directions than based on the results of previous research it is most likely not
because the landscape history is different than thought before, but because of the lack of
quantity and quality of obtained data used in the biomarker analysis.
To make sure the results of this study are valid and reliable, additional research has to be
carried out. The trustworthiness of the found lipid patterns of both soil and vegetation
samples must be improved to draw more solid conclusions.
43
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Netherlands
References
Aitken, M.J. (1998) An Introduction to Optical Dating. Oxford University Press, Oxford, p.267
Bokhorst, M.P., Duller, G.A.T. & van Mourik, J.M. (2005) Optical dating of a Fimic
anthrosol in the southern Netherlands. Journal of Archaeological Science, 32, pp. 547-553.
Blume, H.P. & Leinweber, P. (2004) Plaggen soils: landscape history, properties, and
classification. Journal of Plant Nutrition and Soil Science-Zeitschrift Fur Pflanzenernahrung
Und Bodenkunde, 167, pp. 319-327.
Brewer, R. (1976) Fabric and Mineral Analysis of Soils. Robert E. Krieger Publishing Company,
Huntingon, New York, p. 482.
Derese, C., Vandenberghe, D., Eggermont, N., Bastiaens, J., Annaert, R. & Van den haute,
P. (2010) A medieval settlement caught in the sand: Optical dating of sand-drifting at
Pulle (N Belgium). Quaternary Geochronology, 5, pp. 336-341.
Dijkmans, J.W.A., van Mourik, J.M. & Wintle, A.G. (1992) Thermoluminescence dating of
aeolian sands from polycyclic soil profiles in the Southern Netherlands. Quarternary
Science Reviews II, pp. 85-92.
Everitt, B.S., Landau S. & Leese, M. (2001) Cluster Analysis. Arnold.
Ficken, K.J., Barber, K.E., & Eglinton, G. (1998) Lipid biomarker, delta C-13 and plant
macrofossil stratigraphy of a Scottish montane peat bog over the last two millennia.
Organic Geochemistry, 28, pp. 217-237.
Haussmann, N.S. (2006) Application of the lipid biomarker approach to paleo-research in
northern Ecuador, MSc. Thesis, Universiteit van Amsterdam.
Ishiwatari, R., Yamamoto, S., & Uemura, H. (2005) Lipid and lignin/cutin compounds in
Lake Baikal sediments over the last 37 kyr: implications for glacial-interglacial
palaeoenvironmental change. Organic Geochemistry, 36, pp. 327-347.
IUSS Working Group WRB (2007) World Reference Base for Soil Resources 2006, first
update 2007. World Soil Resources Reports No. 103. FAO, Rome.
Jansen, B., Nierop, K.G.J., Kotte, M.C., de Voogt, P., Verstraten, J.M. (2006a) The
applicabilty of accelerated solvent extraction (ASE) to extract lipid biomarkers from
soils. Applied Geochemistry, 21, pp. 1006-1015
45
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
Jansen, B., Nierop, K.G.J., Hageman, J.A., Cleef, A. M. & Verstraten, J.M. (2006b) The
straight-chain lipid biomarker composition of plant species responsible for the
dominant biomass production along two altitudinal transects in the Ecuadorian
Andes. Organic Geochemistry, 37, pp. 1514-1536.
Jansen, B., van Loon, E.E., Hooghiemstra, H. & Verstraten, J.M. (2010) Improved
reconstruction of paleo-environments through unravelling of preserved vegetation
patterns. Paleogeography, Paleoclimatology, Paleoecology, 285, pp. 119-130
Mathworks (2007) MATLAB (software).
Mook, W.G. & Streurman, H.J. (1983) Physical and chemical aspects of radiocarbon dating.
In: Mook, W.G., Waterbolk, H.T. (Eds.), Proceedings of the First International
Symposium on 14C and Archaeology, PACT 8. Council of Europe, pp. 31-55.
Moore, P.D., Webb, J.A. & Collison, M.E. (1991) Pollen analysis. Blackwell Scientific
Publications, Oxford.
Naafs, D.F.W., van Bergen, P.F., Boogert, S.J. & de Leeuw, J.W. (2004) Solvent-extractable
lipids in an acid andic forest soil; variations with depth and season. Soil Biol. Biochem.
36, pp. 297-308.
Okubo, A & Levin, S.A. (1989) A Theoretical Framework for Data Analysis of Wind
Dispersal of Seeds and Pollen. Ecology, 70, pp. 329-338
Quenea, K., Derenne, S., Largeau, C., Rumpel, C., & Marlotti, A. (2004) Variation in lipid
relative abundance and composition among different particle size fractions of a
forest soil. Organic Geochemistry, 35, pp. 1355-1370.
Quenea, K., Largeau, C., Derenne, S., Spaccini, R., Bardoux, G., & Mariotti, A. (2006)
Molecular and isotopic study of lipids in particle size fractions of a sandy cultivated
soil (Cestas cultivation sequence, southwest France): Sources, degradation, and
comparison with Cestas forest soil. Organic Geochemistry, 37, pp. 20-44.
Pape, J.C. (1970) Plaggen Soils in the Netherlands. Geoderma, 4, pp. 229-255.
Rieley, G., Collier, R.J., Jones, D.M., & Eglinton, G. (1991) The Biogeochemistry of
Ellesmere Lake, Uk .1. Source Correlation of Leaf Wax Inputs to the Sedimentary
Lipid Record. Organic Geochemistry, 17, pp. 901-912.
Spek, T. (2004) Het Drentse esdorpenlandschap: een historisch-geografische studie.
Matrijs.
SPSS Inc. (2009) PASW Statistics 17 (software).
46
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
van Bergen, P.F., Bull, I.D., Poulton, P.R., & Evershed, R.P. (1997) Organic geochemical
studies of soils from the Rothamsted classical experiments .I. Total lipid extracts,
solvent insoluble residues and humic acids from Broadbalk wilderness. Organic
Geochemistry, 26, pp. 117-135.
van Mourik, J.M. (1988). De ontwikkeling van een stuifzandgebied. Netherlands Geographical
Studies, 74, pp. 5-42.
van Mourik, J.M., Nierop, K.G.J., Vandenberghe, D.A.G. (2010) Radiocarbon and optically
stimulated luminescence dating based chronology of a polycyclic driftsand sequence
at Weerterbergen (SE Netherlands). Catena, 80, pp. 170-181.
van Mourik, J.M., Wartenbergh, P.E., Mook, W.J., & Streurman, H.J. (1995) Radiocarbon
dating of paleosols in Eolian sands. Mededelingen Rijks Geologische Dienst, 52, pp. 425439.
Vandenberghe, D., Kasse, C., Hossain, M., De Corte, F., Van den haute, P., Fuchs, M. &
Murray, A.S. (2004). Exploring the method of optical dating and comparison of
optical dating and
14
C ages of Late Weichselian coversands in the southern
Netherlands. Journal of Quaternary Science, 4, pp. 73-86.
Vandenberghe, D., van Mourik, J.M., Buylaert, J.P., De Corte, F., Van den haute, P., (2005)
Optical dating of Late Holocene drift sands from southern Netherlands. Book of
Abstracts 11th' International Conference on Luminescence and Electron Spin Resonance Dating,
July 24-29, 2005. University of Cologne, Germany, p. 196.
Ward, J.H. (1963) Hierarchical groupings to optimize an objective function. Journal of the
American Statistical Association, 58, pp. 234-244.
47
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Netherlands
APPENDICES
Appendix A: Photos of Defensiedijk 2008 profile
Figure A.1: Photo of the top (0-39 cm) of the studied Defensiedijk 2008 profile.
49
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
Figure A.2: Photo of the part of the Defensiesijk 2008 profile between a depth of 29 and 82 cm. The
irregularity of the soil horizons i s clearly visi ble.
50
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
Figure A.3: Between a depth of 67 and 99 cm
Figure A.4: Between a depth of 124 and 151 cm
51
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Netherlands
Appendix B: Matlab code
B.1. verhib.m
function [plantmass,soilchem_pred,residchem,residdepth,residgroup] = verhib(varargin)
%
% [plantmass,soilchem_pred,residchem,residdepth,residgroup] = ...
% verhib(inputxls,regdepth,reggroup)
%
% or
%
% [plantmass,soilchem_pred,residchem,residdepth,residgroup] = ...
% verhib(soilchem,leafchem,rootchem,leafdepth,rootdepth, ...
%
plantgroups,leafrootratio,yearperlayer,regdepth,reggroup);
%
% If lsqlin is not found: add path to optimization toolbox + subdirs.:
% addpath( genpath('D:\RUFLE Project\modelleren met Emiel\verhib1\optim') )
%
%
% e.g.
%
% verhib('verhib_g15', 0.1,0.1)
%
if isstr( varargin{1} )
xlsInOut = true;
inputxls = [ 'data\' varargin{1} ];
regdepth = varargin{2};
reggroup = varargin{3};
[soilchem,leafchem,rootchem,leafdepth,rootdepth,plantgroups,useleaf,useroot,leafrootratio,yearperlayer,nrcomp,nrlayers,nrplan ts,nrgroups,depthlbl_num,species,chemlbl] = read_verhib_xls(inputxls);
else
soilchem = varargin{1};
leafchem = varargin{2};
rootchem = varargin{3};
leafdepth = varargin{4};
rootdepth = varargin{5};
plantgroups = varargin{6};
leafrootratio = varargin{7};
yearperlayer = varargin{8};
regdepth = varargin{9};
reggroup = varargin{10};
end
%% build matrices for regression function
[E,f,G,h,C,d,minmass,maxmass] = mk_verhib_matrices(soilchem,leafchem,rootchem,leafdepth,rootdepth,plantgroups,useleaf,useroot,leafrootratio,yearperlayer,regdepth,reggroup);
%% the actual regression
% set tolerance and maxnr of iterations to reasonable levels for this
% problem
options = optimset('TolFun',1e-1,'MaxIter',1000);
[par,resnorm,residual,exitflag] = lsqlin( E, f, G, h, C, d, minmass, maxmass, maxmass/nrplants, options );
%exitflag
% [X,RESNORM,RESIDUAL,EXITFLAG] = LSQLIN(C,d,A,b,Aeq,beq,LB,UB)
%
% Solves the least-squares, with equality constraints problem:
%
%
min 0.5*(NORM(C*x-d)).^2
%
x
%
%
subject to: A*x <= b and Aeq*x = beq
%
% so that the solution is in the range LB <= X <= UB. Use empty matrices for
% LB and UB if no bounds exist. Set LB(i) = -Inf if X(i) is unbounded
% below; set UB(i) = Inf if X(i) is unbounded above.
% par = lsie(E,f,G,h,C,d);
% this minimizes ||Ex - f|| subject to Gx >= h and Cx = d.
%% reformat output
[plantmass,soilchem_pred,residchem,residdepth,residgro up] = mk_verhib_output(nrcomp,nrlayers,nrplants,nrgroups,E,f,par,leafrootratio);
%% write output
if xlsInOut
write_verhib_xls(inputxls,regdepth,reggroup,depthlbl_num,species,chemlbl,plantmass,residchem)
else
disp('Output not written to xls-file')
end
B.2. read_verhib_xls.m
function [soilchem,leafchem,rootchem,leafdepth,rootdepth,plantgroups,useleaf,useroot,leafrootratio,yearperlayer,nrcomp,nrlayers,nrplan ts,nrgroups,depthlbl_num,species,chemlbl] = read_verhib_xls(inputxls)
%% load data from 'inputxls'
[dat,lbl] = xlsread( inputxls, 'includechem' );
% endcol = size(dat,2);
inchem = logical( [0 dat] );
chemlbl = lbl(1,inchem);
[dat,lbl] = xlsread( inputxls, 'plantgroups' );
useleaf = logical( dat(:,6) );
useroot = logical( dat(:,7) );
53
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
inplant = useleaf | useroot;
useleaf = logical( dat(inplant,6) )';
useroot = logical( dat(inplant,7) )';
leafrootratio = dat( inplant, end )';
plantgroups = dat( inplant, 2:end-1 )';
inplantlbl = [false; inplant];
family = lbl( inplantlbl, 1 );
genus = lbl( inplantlbl, 2);
species = lbl( inplantlbl, 3 );
[dat,lbl] = xlsread( inputxls, 'depthdur' );
yearperlayer = dat(:,2);
[dat,lbl] = xlsread( inputxls, 'soilchem' );
soilchem = dat( :, inchem )';
depthlbl_num = dat(:,1);
[dat,lbl] = xlsread( inputxls, 'leafchem' );
leafchem = dat( inplant, inchem )';
[dat,lbl] = xlsread( inputxls, 'rootchem' );
rootchem = dat( inplant, inchem )';
[dat,lbl] = xlsread( inputxls, 'leafdepth' );
leafdepth = dat( inplant, 2:end )';
depthlbl_str = lbl( 1, 5:end );
[dat,lbl] = xlsread( inputxls, 'rootdepth' );
rootdepth = dat( inplant, 2:end )';
% rootchem = [];
% rootdepth =[];
% determine the dimension of the problem
[nrcomp,nrlayers] = size(soilchem);
nrplants = size(leafchem,2);
% remove plantgroups without plants
plantsingroup = sum(plantgroups,2)>0;
plantgroups = plantgroups(plantsingroup,:);
nrgroups = size( plantgroups, 1 );
B.3. mk_verhib_matrices.m
function [E,f,G,h,C,d,minmass,maxmass] = mk_verhib_matrices(soilchem,leafchem,rootchem,leafdepth,rootdepth,plantgroups,useleaf,useroot,leafrootratio,yearperlayer,regdepth,reggroups,depthlimit)
% where
% soilchem is the matrix with soil chemical data (nrcomp,nrlayers)
% leafchem is the matrix with leaf chemical data (nrcomp,nrplants)
% rootchem is the matrix with root chemical data (nrcomp,nrplants)
% leafdepth is the matrix with leaf depth relation (nrlayers,nrplants)
% rootdepth is the matrix with leaf depth relation (nrlayers,nrplants)
% plantgroups is the matrix with plant groups (nrgroups,nrplants) can be left empty
% leafrootratio is the (1,nrplants) vector with the ratio leafmass:rootmass per plant
% yearperlayer is the (nrlayer,1) vector with the nr of years per depth layer
% regdepth is the regularization to weight the importance of stiffness over time
%
parameter (usually a value between 0
%
and 0.1, smaller means less regularization)
% reggroups is the regularization parameter for plant groups
% depthlimit specifies the number of layers to consider for spatial partitioning
%
- to make the problem easier to sovle
%
% output-matrices and vectors are for the least squares problem:
% ||Ex - f|| subject to Gx >= h and Cx = d
%
% the plantmass is accounted for in Gx >= h
% and the leaf-root ratio in Cx = d
%
%
%
%
%
%
%
%
%
%
%
%
%
%
%
%
%
%
%
%
%
%
%
%
%
for a case with 2 plants, and three layers,
and including the root info the resulting X
matrix should look like:
leafs
roots
<- nrplants*nrlayers -> <- nrplants*nrlayers ->
X = [ a1 a2 0 0 0 0 c1 c2 0 0 0 0
^
b1 b2 0 0 0 0 d1 d2 0 0 0 0
|
0 0 a1 a2 0 0 0 0 c1 c2 0 0
| nrcomp * nrlayers
0 0 b1 b2 0 0 0 0 d1 d2 0 0
|
0 0 0 0 a1 a2 0 0 0 0 c1 c2 |
0 0 0 0 b1 b2 0 0 0 0 d1 d2 V
.9 0 0 0 0 0 .7 0 0 0 0 0
^
0 .8 0 0 0 0 0 .6 0 0 0 0
|
.1 0 .9 0 0 0 .2 0 .7 0 0 0
|
0 .2 0 .8 0 0 0 .3 0 .6 0 0
| nrplants * nrlayers
0 0 .1 0 .9 0 .1 0 .2 0 .7 0
|
0 0 0 .2 0 .8 0 .1 0 .3 0 .6 |
.1 .1 0 0 0 0 .1 .1 0 0 0 0
^
0 0 .1 .1 0 0 0 0 .1 .1 0 0
| nrgroup * nrlayers
0 0 0 0 .1 .1 0 0 0 0 .1 .1 V
if ~exist('depthlimit','var')
depthlimit = size(leafdepth,1);
end
if ~exist('reggroups','var') || isempty(reggroups)
reggroups = 0.1;
end
if ~exist('regdepth', 'var') || isempty(regdepth)
regdepth = 0.1;
end
maxleafperyear = 1000; % (g per m2)
maxrootperyear = 1000; % (g per m2)
leafrootratio = leafrootratio(:)';
54
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
% determine the dimension of the problem
[nrcomp,nrlayers] = size(soilchem);
nrplants = size(leafchem,2);
% remove plantgroups without plants
plantsingroup = sum(plantgroups,2)>0;
plantgroups = plantgroups(plantsingroup,:);
nrgroups = size( plantgroups, 1 );
% leafpart = leafrootratio./(leafrootratio+1);
% rootpart = 1-leafpart;
% set values leafchem and rootchem to zero, if the respective plant
% component is not included
leafchem = leafchem .* repmat(useleaf,nrcomp,1);
rootchem = rootchem .* repmat(useroot,nrcomp,1);
if sum(rootchem(:))==0, rootchem=[]; disp('Roots not included'); end
if isempty(rootdepth) || isempty(rootchem)
% build the y-vector
ytime = zeros(nrlayers*nrplants-nrplants,1);
if isempty(plantgroups)
ygroups = [];
else
ygroups = reggroups * repmat(sum(plantgroups,2),nrlayers,1);
end
f = [ soilchem(:); ytime; ygroups];
% build the x-matrix
leafblocks = repmat( leafchem, nrlayers,nrlayers ) .* ...
spacetime_knitting( leafdepth./100, nrcomp, depthlimit );
% the blocks to control time-stiffness distribution
timebands = - eye(nrplants*nrlayers) + diag( ones( (nrplants*nrlayers - nrplants) ,1),-nrplants );
timebands = timebands( nrplants+1:end , : );
if isempty(plantgroups)
pgroupblocks = [];
else
pgroupblocks = blockdiag_lim(plantgroups,nrlayers);
end
E = [ leafblocks; regdepth*timebands; reggroups*pgroupblocks ];
%
%
totperlayer = blockdiag_lim( ones(1,nrplants), nrlayers );
totleafperlayer = yearperlayer * maxleafperyear;
G = [ eye(nrplants*nrlayers); -totperlayer ];
h = [ zeros(nrplants*nrlayers,1); -totleafperlayer ];
G = totperlayer ;
h = totleafperlayer ;
C = [];
d = [];
else % leaves and roots
% build the y-vector
ytime = zeros(2*nrlayers*nrplants - 2*nrplants,1);
if isempty(plantgroups)
ygroups = [];
else
ygroups = reggroups * repmat(sum(plantgroups,2),2*nrlayers,1);
end
%
yleafroot = zeros(nrplants*nrlayers,1);
f = [ soilchem(:); ytime; ygroups];
% the blocks for the biomarker mass in the plant parts
leafblocks = repmat(leafchem,nrlayers,nrlayers) .* ...
spacetime_knitting(leafdepth./100, nrcomp, depthlimit );
rootblocks = repmat(rootchem,nrlayers,nrlayers) .* ...
spacetime_knitting(rootdepth./100, nrcomp, depthlimit );
totalmass = [leafblocks rootblocks];
% the blocks to control time-stiffness distribution
timebands = - eye(nrplants*nrlayers) + diag( ones( (nrplants*nrlayers - nrplants) ,1),-nrplants );
timebands = timebands( nrplants+1:end , : );
zeroblock = zeros(nrplants*nrlayers - nrplants,nrplants*nrlayers);
totaltimebands = [timebands zeroblock; zeroblock timebands];
% the blocks for the plant groups
if isempty(plantgroups)
pgroupblocks = [];
else
pgroupblocks = blockdiag_lim(plantgroups,nrlayers);
end
zeroblock = zeros(nrgroups*nrlayers,nrplants*nrlayers);
totalpgroupblocks = [pgroupblocks zeroblock; zeroblock pgroupblocks];
%
%
% the blocks for the leafroot-ratio
leafrootblock = [ eye(nrplants*nrlayers), - diag( repmat(leafrootratio',nrlayers,1) ) ];
E = [ totalmass
regdepth * totaltimebands
reggroups * totalpgroupblocks ];
totperlayer = blockdiag_lim( ones(1,nrplants), nrlayers );
totplantperlayer = yearperlayer * ( maxleafperyear + maxrootperyear );
% de pijn zit hem er misschien in dat wortels en blad achter
% elkaar staan (en gezamenlijk > dan 0 moeten zijn)
% terwijl ze dat afzonderlijk ook zijn
%
% G = [ eye(nrplants*nrlayers) eye(nrplants*nrlayers);
%
-totperlayer
-totperlayer ];
55
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
%
h = [ zeros(nrplants*nrlayers,1); -totplantperlayer ];
G = [ totperlayer, totperlayer ];
h = totplantperlayer ;
C = [ eye(nrplants*nrlayers) , diag( repmat( -leafrootratio(:), nrlayers, 1) ) ];
d = zeros(nrplants*nrlayers,1);
end
nrpar = size(E,2);
minmass = zeros(nrpar,1);
maxmass = ones( nrpar/nrlayers,1 ) * h' ; % note: h is the maximum biomass production per layer
maxmass = maxmass(:);
%%
function [partitionmatrix] = spacetime_knitting(partition_in,nrcomp,nrlag)
[nrlayer, nrplant] = size(partition_in);
mindim = min(nrlayer,nrlag);
partitionmatrix = zeros(nrcomp*nrlayer,nrplant*mindim);
for diagonal = 1:mindim
i=0;
j=0;
for n = 1:nrlayer + 1 - diagonal
partition_submatrix = repmat( partition_in(diagonal,:),nrcomp,1 ) ;
rl = (diagonal-1)*nrcomp + i + 1 : (diagonal-1)*nrcomp + i + nrcomp;
cl = j+1:j+nrplant;
partitionmatrix(rl,cl) = partition_submatrix ;
i = i+nrcomp;
j = j+nrplant;
end
end
%%
% [output] = blockdiag_lim(input,nrblk)
%
% create a block diagonal matrix
% using nrblk blocks
%
% Emiel van Loon
function [output] = blockdiag_lim(input,nrblk)
[nrow,ncol]=size(input);
output = zeros(nrow*nrblk,ncol*nrblk);
i=0;
j=0;
for k = 1:nrblk
output(i+1:i+nrow,j+1:j+ncol)=input;
i = i+nrow;
j = j+ncol;
end
B.4. mk_verhib_output.m
function [plantmass,soilchem_pred,residchem,residdepth,residgroup] = mk_verhib_output(nrcomp,nrlayers,nrplan ts,nrgroups,X,y,par,leafrootratio)
% totale lengte van y is:
% nrcomp*nrlayers + nrplants*(nrlayers-1) + nrgroups*nrlayers
%
residual = y - X*par;
if length(par)==nrplants*nrlayers % only leaves were considered - no roots
totmass = par + par./repmat(leafrootratio(:),nrlayers,1);
plantmass = reshape( totmass, nrplants, nrlayers );
soilchem_pred = reshape(y(1:nrcomp*nrlayers),nrcomp,nrlayers);
residchem = reshape(residual(1:nrcomp*nrlayers),nrcomp,nrlayers);
brk = nrcomp*nrlayers;
residdepth = reshape( residual( brk + 1 : brk + nrplants*(nrlayers -1) ), nrplants, nrlayers-1);
brk = nrcomp*nrlayers + nrplants*(nrlayers -1);
residgroup = reshape( residual( brk + 1 : brk + nrgroups*nrlayers ), nrgroups, nrlayers);
else % both leaves and roots were considered.
% reshape the parameter output and calculate the relative contribution
% of each plant to the total
plantmass = reshape( par( 1 : nrplants*nrlayers ), nrplants, nrlayers ) + ...
reshape( par( nrplants*nrlayers+1 : 2*nrplants*nrlayers ), nrplants, nrlayers ) ;
% plantmass_relative = plantmass ./ repmat( sum(plantmass) ,nrplants,1 );
% should be the same
% rootpar = reshape( par(nrplants*nrlayers+1:end), nrplants, nrlayers );
% rootpar_relative = rootpar ./ repmat( sum(rootpar) ,nrplants,1 );
% reshape the predicted biomarker concentration to the right format:
% after 'reshape' each row is one component, each column is one layer,
% this is transposed.
% the same for the residuals and regularized values
soilchem_pred = reshape(y(1:nrcomp*nrlayers),nrcomp,nrlayers);
residchem = reshape(residual(1:nrcomp*nrlayers),nrcomp,nrlayers);
brk = nrcomp*nrlayers;
56
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
residdepth = reshape( residual( brk + 1 : brk + nrplants*nrlayers ), nrplants, nrlayers);
brk = nrcomp*nrlayers + nrplants*nrlayers;
residgroup = reshape( residual( brk + 1 : brk + nrgroups*nrlayers ), nrgroups, nrlayers);
end
B.5. write_verhib_xls
function write_verhib_xls(inputxls,regdepth,reggroup,depthlbl_num,species,chemlbl,plantmass,residchem)
warning('o ff','MATLAB:xlswrite:AddSheet');
% write the output (vegetation composition) to the source spreadsheet
% the regularisation parameters are used in the sheetn ame
outputsheet = [ 'veget_rd' num2str(regdepth) '_rg' num2str(reggroup)];
xlswrite(inputxls,{'Depth (cm):'},outputsheet, 'B1');
xlswrite(inputxls,depthlbl_num',outputsheet, 'B2');
xlswrite(inputxls,{'Species name:'},outputsheet, 'A2');
xlswrite(inputxls,species,outputsheet, 'A3');
xlswrite(inputxls,plantmass,outputsheet,'B3');
% write the residual to the source spreadsheet
% the regularisation parameters are used in the sheetname
outputsheet = [ 'resid_rd' num2str(regdepth) '_rg' num2str(reggroup)];
xlswrite(inputxls,chemlbl,outputsheet, 'B1');
xlswrite(inputxls,{'Depth (cm):'},outputsheet, 'A1');
xlswrite(inputxls,depthlbl_num,outputsheet, 'A 2');
xlswrite(inputxls,residchem',outputsheet, 'B2');
warning('o n', 'MATLAB:xlswrite:AddSheet');
57
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Netherlands
Appendix C: GC/MS results
Soil sample 1 (0-2 cm)
First analysis
2,5
alcohols
alkanes
2
1,5
lipid concentration (μg/g soil)
1
0,5
0
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
28
29
30
31
32
33
34
35
36
Soil sample 1 (0-2 cm)
Second analysis
7
6
5
4
3
2
1
0
16
17
18
19
20
21
22
23
24
25
26
27
Carbon chainlength
59
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
Soil sample 2 (34-36 cm)
First analysis
0,8
alcohols
0,7
alkanes
0,6
0,5
0,4
0,3
lipid concentration (μg/g soil)
0,2
0,1
0
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
28
29
30
31
32
33
34
35
36
Soil sample 2 (34-36 cm)
Second analysis
1,4
1,2
1
0,8
0,6
0,4
0,2
0
16
17
18
19
20
21
22
23
24
25
26
27
Carbon chainlength
60
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
Soil sample 3 (63-64 cm)
First analysis
0,4
alcohols
0,35
alkanes
0,3
0,25
0,2
0,15
lipid concentration (μg/g soil)
0,1
0,05
0
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
28
29
30
31
32
33
34
35
36
Soil sample 3 (63-64 cm)
Second analysis
0,7
0,6
0,5
0,4
0,3
0,2
0,1
0
16
17
18
19
20
21
22
23
24
25
26
27
Carbon chainlength
61
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
Soil sample 4 (77-78 cm)
First analysis
1,2
alcohols
alkanes
1
0,8
0,6
lipid concentration (μg/g soil)
0,4
0,2
0
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
30
31
32
33
34
35
36
Soil sample 4 (77-78 cm)
Second analysis
1,6
1,4
1,2
1
0,8
0,6
0,4
0,2
0
16
17
18
19
20
21
22
23
24
25
26
27
28
Carbon chainlength
62
29
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
Soil sample 5 (115-115.5 cm)
First analysis
0,45
alcohols
0,4
alkanes
0,35
0,3
0,25
0,2
0,15
lipid concentration (μg/g soil)
0,1
0,05
0
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
29
30
31
32
33
34
35
36
Soil sample 5 (115-115.5 cm)
Second analysis
0,5
0,45
0,4
0,35
0,3
0,25
0,2
0,15
0,1
0,05
0
16
17
18
19
20
21
22
23
24
25
26
27
28
Carbon chainlength
63
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
Soil sample 6 (130-132 cm)
First analysis
2
alcohols
1,8
alkanes
1,6
1,4
1,2
1
0,8
lipid concentration (μg/g soil)
0,6
0,4
0,2
0
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
28
29
30
31
32
33
34
35
36
Soil sample 6 (130-132 cm)
Second analysis
3
2,5
2
1,5
1
0,5
0
16
17
18
19
20
21
22
23
24
25
26
27
Carbon chainlength
64
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
Soil sample 7 (132-134 cm)
First analysis
14
alcohols
alkanes
12
10
8
6
lipid concentration (μg/g soil)
4
2
0
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
28
29
30
31
32
33
34
35
36
Soil sample 7 (132-134 cm)
Second analysis
0,7
0,6
0,5
0,4
0,3
0,2
0,1
0
16
17
18
19
20
21
22
23
24
25
26
27
Carbon chainlength
65
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
Soil sample 8 (151-154 cm)
First analysis
4
alcohols
3,5
alkanes
3
2,5
2
1,5
lipid concentration (μg/g soil)
1
0,5
0
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
28
29
30
31
32
33
34
35
36
Soil sample 8 (151-154 cm)
Second analysis
5
4,5
4
3,5
3
2,5
2
1,5
1
0,5
0
16
17
18
19
20
21
22
23
24
25
26
27
Carbon chainlength
66
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
Soil sample 9 (168-172 cm)
First analysis
4
alcohols
3,5
alkanes
3
2,5
2
1,5
lipid concentration (μg/g soil)
1
0,5
0
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
28
29
30
31
32
33
34
35
36
Second analysis
10
9
8
7
6
5
4
3
2
1
0
16
17
18
19
20
21
22
23
24
25
26
27
Carbon chainlength
67
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
Lichen
First analysis
160
alcohols
140
alkanes
120
100
80
60
40
20
0
16
17
18
19
20
21
22
23
24
26
27
Lichen
28
29
30
31
32
33
34
35
36
29
30
31
32
33
34
35
36
29
30
31
32
33
34
35
36
Second analysis
6
lipid concentration (μg/g plant)
25
5
4
3
2
1
0
16
17
18
19
20
21
22
23
24
25
26
27
Lichen
28
Third analysis
14
12
10
8
6
4
2
0
16
17
18
19
20
21
22
23
24
25
26
27
28
Carbon chainlength
68
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
Moss
First analysis
30
alcohols
alkanes
25
20
15
10
5
0
16
17
18
19
20
21
22
23
24
25
26
27
Moss
28
29
30
31
32
33
34
35
36
29
30
31
32
33
34
35
36
29
30
31
32
33
34
35
36
Second analysis
35
lipid concentration (μg/g plant)
30
25
20
15
10
5
0
16
17
18
19
20
21
22
23
24
25
26
27
28
Moss
Third analysis
18
16
14
12
10
8
6
4
2
0
16
17
18
19
20
21
22
23
24
25
26
27
28
Carbon chainlength
69
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
Calluna: leaves
First analysis
400
alcohols
350
alkanes
300
250
200
150
100
50
0
16
17
18
19
20
21
22
23
25
26
27
28
29
30
31
32
33
34
35
36
28
29
30
31
32
33
34
35
36
28
29
30
31
32
33
34
35
36
Second analysis
1,2
lipid concentration (μg/g plant)
24
Calluna: leaves
1
0,8
0,6
0,4
0,2
0
16
17
18
19
20
21
22
23
24
25
26
27
Calluna: leaves
Third analysis
120
100
80
60
40
20
0
16
17
18
19
20
21
22
23
24
25
26
27
Carbon chainlength
70
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
Calluna: roots
First analysis
350
alcohols
300
alkanes
250
200
150
100
50
0
16
17
18
19
20
21
22
23
25
26
27
28
29
30
31
32
33
34
35
36
28
29
30
31
32
33
34
35
36
28
29
30
31
32
33
34
35
36
Second analysis
3,5
lipid concentration (μg/g plant)
24
Calluna: roots
3
2,5
2
1,5
1
0,5
0
16
17
18
19
20
21
22
23
24
25
26
27
Calluna: roots
Third analysis
7
6
5
4
3
2
1
0
16
17
18
19
20
21
22
23
24
25
26
27
Carbon chainlength
71
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
Molinia: leaves
First analysis
50
alcohols
45
alkanes
40
35
30
25
20
15
10
5
0
16
17
18
19
20
21
22
23
25
26
27
28
29
30
31
32
33
34
35
36
29
30
31
32
33
34
35
36
29
30
31
32
33
34
35
36
Second analysis
25
lipid concentration (μg/g plant)
24
Molinia: leaves
20
15
10
5
0
16
17
18
19
20
21
22
23
24
25
26
27
28
Molinia: leaves
Third analyis
45
40
35
30
25
20
15
10
5
0
16
17
18
19
20
21
22
23
24
25
26
27
28
Carbon chainlength
72
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
Molinia: roots
First analysis
14
alcohols
12
alkanes
10
8
6
4
2
0
16
17
18
19
20
21
22
23
25
26
27
28
29
30
31
32
33
34
35
36
28
29
30
31
32
33
34
35
36
28
29
30
31
32
33
34
35
36
Second analysis
35
lipid concentration (μg/g plant)
24
Molinia: roots
30
25
20
15
10
5
0
16
17
18
19
20
21
22
23
24
25
26
27
Molinia: roots
Third analysis
3
2,5
2
1,5
1
0,5
0
16
17
18
19
20
21
22
23
24
25
26
27
Carbon chainlength
73
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
Corynephorus: leaves
First analysis
45
alcohols
40
alkanes
35
30
25
20
15
10
5
0
16
17
18
19
20
21
22
23
25
26
27
28
29
30
31
32
33
34
35
36
28
29
30
31
32
33
34
35
36
28
29
30
31
32
33
34
35
36
Second analysis
250
lipid concentration (μg/g plant)
24
Corynephorus: leaves
200
150
100
50
0
16
17
18
19
20
21
22
23
24
25
26
27
Corynephorus: leaves
Third analysis
250
200
150
100
50
0
16
17
18
19
20
21
22
23
24
25
26
27
Carbon chainlength
74
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
Corynephorus: roots
First analysis
90
alcohols
80
alkanes
70
60
50
40
30
20
10
0
16
17
18
19
20
21
22
23
25
26
27
28
29
30
31
32
33
34
35
36
28
29
30
31
32
33
34
35
36
28
29
30
31
32
33
34
35
36
Second analysis
25
lipid concentration (μg/g plant)
24
Corynephorus: roots
20
15
10
5
0
16
17
18
19
20
21
22
23
24
25
26
27
Corynephorus: roots
Third analysis
35
30
25
20
15
10
5
0
16
17
18
19
20
21
22
23
24
25
26
27
Carbon chainlength
75
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
Deschampsia: leaves
First analysis
300
alcohols
alkanes
250
200
150
100
50
0
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
Deschampsia: leaves
Second analysis
lipid concentration (μg/g plant)
60
50
40
30
20
10
0
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
28
29
30
31
32
33
34
35
36
Deschampsia: leaves
Third analysis
60
50
40
30
20
10
0
16
17
18
19
20
21
22
23
24
25
26
27
Carbon chainlength
76
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
Deschampsia: roots
First analysis
30
alcohols
alkanes
25
20
15
10
5
0
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
28
29
30
31
32
33
34
35
36
28
29
30
31
32
33
34
35
36
Deschampsia: roots
Second analysis
8
lipid concentration (μg/g plant)
7
6
5
4
3
2
1
0
16
17
18
19
20
21
22
23
24
25
26
27
Deschampsia: roots
Third analysis
35
30
25
20
15
10
5
0
16
17
18
19
20
21
22
23
24
25
26
27
Carbon chainlength
77
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
Pinus: leaves
First anlysis
60
alcohols
alkanes
50
40
30
20
10
0
16
17
18
19
20
21
22
23
24
25
26
27
28
Pinus: leaves
29
30
31
32
33
34
35
36
29
30
31
32
33
34
35
36
29
30
31
32
33
34
35
36
Second analysis
8
lipid concentration (μg/g plant)
7
6
5
4
3
2
1
0
16
17
18
19
20
21
22
23
24
25
26
27
28
Pinus: leaves
Third analysis
14
12
10
8
6
4
2
0
16
17
18
19
20
21
22
23
24
25
26
27
28
Carbon chainlength
78
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
Pinus: roots
First analysis
70
alcohols
60
alkanes
50
40
30
20
10
0
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
28
29
30
31
32
33
34
35
36
28
29
30
31
32
33
34
35
36
Pinus: roots
Second analysis
1,6
lipid concentration (μg/g plant)
1,4
1,2
1
0,8
0,6
0,4
0,2
0
16
17
18
19
20
21
22
23
24
25
26
27
Pinus: roots
Third analysis
0,9
0,8
0,7
0,6
0,5
0,4
0,3
0,2
0,1
0
16
17
18
19
20
21
22
23
24
25
26
27
Carbon chainlength
79
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
Betula: leaves
First analysis
900
alcohols
800
alkanes
700
600
500
400
300
200
100
0
16
17
18
19
20
21
22
23
24
25
26
27
Betula: leaves
28
29
30
31
32
33
34
35
36
28
29
30
31
32
33
34
35
36
28
29
30
31
32
33
34
35
36
Second analysis
80
lipid concentration (μg/g plant)
70
60
50
40
30
20
10
0
16
17
18
19
20
21
22
23
24
25
26
27
Betula: leaves
Third analysis
120
100
80
60
40
20
0
16
17
18
19
20
21
22
23
24
25
26
27
Carbon chainlength
80
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
Betula: roots
First analysis
30
alcohols
alkanes
25
20
15
10
5
0
16
17
18
19
20
21
22
23
24
25
26
27
28
Betula: roots
29
30
31
32
33
34
35
36
29
30
31
32
33
34
35
36
29
30
31
32
33
34
35
36
Second analysis
10
lipid concentration (μg/g plant)
9
8
7
6
5
4
3
2
1
0
16
17
18
19
20
21
22
23
24
25
26
27
28
Betula: roots
Third analysis
9
8
7
6
5
4
3
2
1
0
16
17
18
19
20
21
22
23
24
25
26
27
28
Carbon chainlength
81
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
Quercus: leaves
First analysis
300
alcohols
alkanes
250
200
150
100
50
0
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
28
29
30
31
32
33
34
35
36
28
29
30
31
32
33
34
35
36
Quercus: leaves
Second analysis
35
lipid concentration (μg/g plant)
30
25
20
15
10
5
0
16
17
18
19
20
21
22
23
24
25
26
27
Quercus: leaves
Third analysis
25
20
15
10
5
0
16
17
18
19
20
21
22
23
24
25
26
27
Carbon chainlength
82
MSc.thesis Earth Science A.J. de Vreng
Application of straight-chain lipids as biomarkers in the reconstruction of landscape development in a driftsand landscape in the Net herlands
Quercus: roots
First analysis
9
alcohols
8
alkanes
7
6
5
4
3
2
1
0
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
28
29
30
31
32
33
34
35
36
28
29
30
31
32
33
34
35
36
Quercus: roots
Second analysis
1,8
lipid concentration (μg/g plant)
1,6
1,4
1,2
1
0,8
0,6
0,4
0,2
0
16
17
18
19
20
21
22
23
24
25
26
27
Quercus: roots
Third analysis
4
3,5
3
2,5
2
1,5
1
0,5
0
16
17
18
19
20
21
22
23
24
25
26
27
Carbon chainlength
83