When herbaceous species are pulped for paper- making or used for bioenergy, most frequently,
the whole plant with all plant parts is used as raw material. Except for fibrous material from cell
walls, this biomass contains inorganic elements, which are essential or useful for plant growth and
development. However, these mineral substances have a negative effect on the pulping and combus-
tion processes, so their quantities should be as low as possible Keitaanniemi and Virkola, 1982; Il-
vessalo-Pfa¨ffli, 1995; Obernberger et al., 1997. Silicon Si, potassium, manganese, copper and
iron are harmful for the pulping process Keitaan- niemi and Virkola, 1982 and undesirable in fuel
use Obernberger et al., 1997. In papermaking, silicon wears out the installations of a factory
Watson and Gartside, 1976, lowers the paper quality Jeyasingam, 1988 and complicates the
recovery of chemicals and energy Ranua, 1977; Keitaanniemi and Virkola, 1982; Ulmgren et al.,
1990. In combustion, high alkali metal concen- trations decrease the melting point of ash and
cause deposits and damage in boilers Burvall, 1993; Obernberger et al., 1997.
The concentration of each particular mineral substance varies depending on the plant species
and the plant part Rexen and Munck, 1984; Petersen, 1988; Theander, 1991. The plant age or
stage of development when harvested and the concentration of other minerals have also a sig-
nificant influence Tyler, 1971; Gill et al., 1989; Marschner, 1995; Landstro¨m et al., 1996. The
main botanical components in a grass plant are stem nodes, internodes, leaf sheaths, leaf blades
and panicles, in legumes stem, leaves and pods or flowers, respectively. The weight distribution of
these components varies within and between plant species Salo et al., 1975; Petersen, 1988. Harvest-
ing at different stages of development also affects the stem to leaf ratio in the biomass. There are
differences in the chemical composition of stem and leaves Muller, 1960; Salo et al., 1975; Buxton
and Hornstein, 1986; Albrecht et al., 1987; Pe- tersen, 1988 which also cause variations in the
mineral content of the harvested biomass.
In the early 1990s, the Agricultural Research Centre and the University of Helsinki together
with the Finnish Pulp and Paper Institute set out to search for the most promising crop species for
raw material of papermaking Pahkala et al., 1995. In those studies, reed canary grass, tall
fescue, meadow fescue and goat’s rue were chosen for further studies, including field experiments to
determine the proper harvesting system and fertil- isation level for biomass production. The fibre
and mineral compositions of the total yields have been reported in an earlier study Pahkala et al.,
1994. In the present study, we wanted to find out whether the quality of the raw material could be
improved by screening for the plant fraction, which is most appropriate for the pulping and
combustion processes.
2. Material and methods
2
.
1
. Field experiments In field experiments, three grass and one legu-
minous species, reed canary grass P. arundinacea L., meadow fescue Festuca pratensis Hudson,
tall fescue F. arundinacea Schreber and goat’s rue Galega orientalis L., were studied. The data
presented in this paper were collected from previ- ous field experiments designed to determine the
proper harvesting system and fertilisation level for biomass production. The field experiment for each
grass species comprised two fertilisation levels combined with four harvest times in split-plot
design with 3 – 5 replicates. The fertilisation treat- ments were completely randomised into blocks.
For goat’s rue, four harvest times and one fertili- sation level were used. Because Pahkala et al.
1994 showed earlier that harvesting to produce biomass for industrial purposes is best at seed
ripening stage and in the following spring as senescent dry crop, we focused only on these two
harvest times in this study. The results presented here are from the lower fertilisation level 100 kg
N ha
− 1
, because the higher fertilisation level 200 kg N ha
− 1
was shown to be too high for economic reasons.
The experiments were performed on farm-scale fields of sandy loam with pH values between 5.8
and 6.3. The plot size was 15 m
2
. More details of cultivation practices are given in Table 1 and in
an earlier report Pahkala et al., 1994.
2
.
2
. Har6esting and sampling Harvest at seed ripening stage felt usually be-
tween the last week of July and the second week of August, depending on the year Table 1. Har-
vest in spring was at the end of April or in the beginning of May. Delayed harvest was done in
early spring after the snow and ice had melted and the soil had dried enough to carry the harvest
machine Haldrup forage plot harvester. For analysing the DM content in harvested biomass,
two samples of 200 g were dried first for 2 h at 105°C and then 17 h at 60°C. For plant part
analyses, a sample of 25 × 50 cm consisting about 100 – 120 plants was taken from each plot,
cutting the plants near the soil surface. The dried grass samples were separated into stems, leaf
blades, leaf sheaths and panicles. The goat’s rue samples were separated into stems, leaf blades and
pods. The weight distribution of the plant parts was determined after drying the samples for
17 h at 60°C.
2
.
3
. Mineral composition and crude fibre For grass species, chemical analyses were per-
formed on samples of stem, leaf sheaths and leaf blades separately, for goat’s rue on samples of
stem and leaf blades. After drying, the samples were milled to less than 1 mm. The concentrations
of K, Fe, Mn and Cu were measured by flame AAS flame atomic absorption spectrophotome-
ter, the concentration of silica SiO
2
and ash by gravimetry, in both cases after dry ashing at
500°C. Crude fibre is the fibre fraction, which is not soluble in the acid – alkali treatment. Crude
fiber was determined by using the Fibertec system M, which consists of hot and cold extraction
units. The sample was boiled first in dilute acid H
2
SO
4
and then in dilute alkali KOH. The residue, which contains cellulose, some hemicellu-
lose and lignin, was measured gravimetrically af- ter ashing it at 500°C. Analyses were performed at
the Chemistry Laboratory of the Agricultural Re- search Centre of Finland.
2
.
4
. Statistical analyses An analysis of variance was done for each
measured variable.
Each plant
species was
analysed separately. Statistical analysis was done for testing the harvest time effect. When testing
the differences between plant parts, the plant part was handled as a repeated factor. For reed canary
grass, the year was analysed as a repeated factor when analysing the total DM yield and stem
proportion. The repeated measurements were cor- related. This correlation was taken into account
Table 1 Cultivation practices, varieties, localities, sowing years and harvesting dates of the experiments
Sown Variety
Harvested Locality
Spring Autumn
DM yield and plant parts Reed canary grass
24th July 92
a
1990 27th April 93
a
Jokioinen Venture
Venture Jokioinen
Reed canary grass 1990
29th July 93 25th April 94
Reed canary grass Venture
Jokioinen 1990
3rd August 94 11th May 95
8th May 96 Jokioinen
1990 Reed canary grass
25th July 95 Venture
12th May 97 Venture
Reed canary grass Jokioinen
1990 16th August 96
11th May 98 Venture
Reed canary grass Jokioinen
1990 16th August 97
21st April 93
a
28th July 92
a
1988 Tall fescue
Helsinki Retu
21st April 93
a
Helsinki 1988
Meadow fescue 28th July 92
a
Kalevi Gale
Jokioinen 1990
Goat’s rue 5th August 92
a
6th May 93
a
Fibre Venture
Jokioinen 1993
19th August 96 16th May 97
Reed canary grass
a
Mineral composition of the plant parts.
Fig. 1. Dry matter yield kg ha
− 1
of reed canary grass in autumn 1992 – spring 1998. Relative proportion of different plant parts in total DM are given in columns.
on a selected model 1. The covariance structure of the repeated measurements was chosen by com-
paring potential structures using Akaike’s infor- mation criterion Wolfinger, 1996.
Y
ij
= m + T
i
+ B
j
+ o
ij
+ P
k
+ T × P
jk
+ B × P
ik
+ d
ijk
1 where m, T
i
, B
j
and o
ij
are equivalent in the analysis of variance for the classical randomised
complete-block design. P
k
and T × P
jk
represent the fixed effect of plant part or year in reed
canary grass and plant part or year × harvest time interaction. B × P
ik
represents the random effect of plant part or year × block interaction.
d
ijk
are the experimental error terms. Assumptions of the model were checked by
graphical methods; box-plot for normality of er- rors and plots of residuals for constancy of error
variance Neter et al., 1996. The parameters of the models were estimated by the restricted maxi-
mum likelihood REML estimation method us- ing the SAS system for Windows 6.12.
3. Results
