Bioresource Technology 102 (2011) 226–234

Contents lists available at ScienceDirect

Bioresource Technology

journal homepage: www.elsevier .com/locate /bior tech

Seasonal variation in the chemical composition of the bioenergy feedstock Laminariadigitata for thermochemical conversion

J.M.M. Adams a, A.B. Ross b,*, K. Anastasakis b, E.M. Hodgson a, J.A. Gallagher a, J.M. Jones b, I.S. Donnison a

a Bioenergy and Biorenewables, Institute of Biological, Environmental and Rural Sciences (IBERS), Aberystwyth University, Gogerddan, Aberystwyth SY23 3EB, United Kingdomb Energy and Resources Research Institute, School of Process, Environmental and Materials Engineering, University of Leeds, Leeds LS2 9JT, United Kingdom

a r t i c l e i n f o

Article history:Received 31 March 2010Received in revised form 17 June 2010Accepted 25 June 2010Available online 3 August 2010

Keywords:AlgaeBio-oilBiorefineryPy–GC–MSTGA

0960-8524/$ - see front matter � 2010 Elsevier Ltd. Adoi:10.1016/j.biortech.2010.06.152

* Corresponding author. Tel.: +44 0 113 343 1017;E-mail address: a.b.ross@leeds.ac.uk (A.B. Ross).

a b s t r a c t

To avoid negative impacts on food production, novel non-food biofuel feedstocks need to be identifiedand utilised. One option is to utilise marine biomass, notably fast-growing, large marine ‘plants’ suchas the macroalgal kelps. This paper reports on the changing composition of Laminaria digitata throughoutit growth cycle as determined by new technologies. The potential of Laminaria sp. as a feedstock for bio-fuel production and future biorefining possibilities was assessed through proximate and ultimate analy-sis, initial pyrolysis rates using thermo-gravimetric analysis (TGA), metals content and pyrolysis gaschromatography–mass spectrometry.

Samples harvested in March contained the lowest proportion of carbohydrate and the highest ash andalkali metal content, whereas samples harvested in July contained the highest proportions of carbohy-drate, lowest alkali metals and ash content. July was therefore considered the most suitable month forharvesting kelp biomass for thermochemical conversion to biofuels.

� 2010 Elsevier Ltd. All rights reserved.

1. Introduction

Climate change and energy security are major drivers for a shiftfrom the use of fossil fuels to renewable energy. A range of renew-able energy options exist including wind, solar and tidal but theseare intermittent and only suitable for the production of heat andelectricity. Biomass is an important part of any renewable energymix because it is not only capable of providing a stored means ofgenerating heat and electricity, but also of being converted to arange of end-products including transport fuels and platformchemicals.

1.1. Marine biomass

The majority of biomass currently used for biofuel production isfrom terrestrial sources. Growing biomass on land for fuel can dis-place other agricultural activities including food production. An in-crease in the demand for land has caused deforestation andshortages of food result in increased food prices and civil unrest.Solutions to an increased demand for plant products for food, feed,fibre and fuel, include an increase in yields of all crops, and a great-er utilisation of marine biomass. For example marine biomass ac-counts for over 50% of the primary production of global biomass(Carlsson et al., 2007) yet, relative to terrestrial biomass, little is

ll rights reserved.

fax: +44 0 113 343 2549.

used. Marine biomass comprises of macro and microalgae and bothhave been associated as potential biofuel feedstocks. In general,microalgae are potential sources of bio-oils whilst macroalgaeare potential sources of carbohydrates for fermentation or ther-mo-chemical based conversions. This paper is concerned withmacroalgae as a biofuel feedstock.

Macroalgae are multicellular, macroscopic algae capable of gen-erating more kg of dry biomass m�2 year�1 than fast-growing ter-restrial crops such as sugar cane (Gao and McKinley, 1993). Thelargest growing macroalgae species are within the phaeophyceaeand are termed ‘kelps’. In the Atlantic waters surrounding the UKthe kelps are primarily members of the laminariales order, growingup to 4 m in length (Hayward et al., 1996). One of the main consid-erations for the production of biomass for biofuels is the weight ofdry feedstock produced. The focus of this research was therefore onLaminaria digitata, the most prevalent kelp species growing off themid-Welsh coastline where the samples for this research werecollected.

1.2. Macroalgae biofuels

Previous research on fuels and energy from kelps have primarilyfocused on the production of methane (Moen et al., 1997), metha-nol (Horn, 2000) and more recently ethanol (Adams et al., 2009;Horn et al., 2000). A further study investigated the use of a rangeof macroalgae species as combustion fuels (Ross et al., 2008). How-ever all these trials were conducted on macroalgae harvested at

J.M.M. Adams et al. / Bioresource Technology 102 (2011) 226–234 227

one period in the year, and did not consider the seasonal variationwhich can alter the composition dramatically. Changes in compo-sition is not a new observation, as a study on seasonal variationin Laminaria species by Black (1950) demonstrated. Black mea-sured the main carbohydrate compounds (laminarin and manni-tol), dry matter and ash content in whole plants and fronds andstipes separately in two environments (open sea and loch), overtwo years.

Laminarin is the main storage carbohydrate in Laminaria spe-cies, consisting of a b-(1,3) glucan chain of approximately 25 d.p.with occasional b-(1,6) linkages (Nelson and Lewis, 1974). Theother main carbohydrate present is mannitol, which is the alcoholform of the sugar mannose. The dry weight proportion of laminarinvaried between <1 and 25% dry weight in L. digitata, peaking inOctober and was absent, or present only at low concentrations,during the winter and spring months (Black, 1950). Mannitol con-centration also varied across the seasons, ranging from 3% to 21% ofdry weight (Black 1950). The highest concentrations occurred inJuly and October (open sea) and June and Sept (loch) and the low-est concentrations occurred during winter. The dry matter con-tents mimicked the laminarin proportion but the ash proportionof L. digitata fluctuated conversely, peaking around March and low-est in October (Black, 1950).

Following the study of Black (1950), there has been no subse-quent report of the seasonal variation in the composition of UKseaweeds. Seasonal composition has been noted in articles on UKmacroalgae e.g. on heavy metal accumulation (Fuge and James,1973) or radioactive compound accumulation (Nawakowskiet al., 2004) but has not been studied further beyond acknowledge-ment of its occurrence. Given the time that has elapsed and thedevelopment of new analytical techniques since the work of Black,it is timely to revisit the topic of seasonal variation in macroalgaecomposition. In addition as these compositional changes will havean impact on the potential of macroalgae as a biofuel feedstock, theaim of this paper is to investigate the composition and potential forconversion over the year. For example macroalgae can be biologi-cally converted to ethanol or methane through fermentation oranaerobic digestion. Thermochemical conversion methods suchas pyrolysis can also be used to produce bio-oil, fractions of whichmay be used as a direct replacement of fossil fuel-derived diesel.Other energy generation routes could include gasification, combus-tion and hydrothermal liquefaction.

2. Methods

2.1. Sample collection and preparation

Samples of L. digitata were harvested from wild stock at after-noon spring low tides on a rocky outcrop off Aberystwyth beach,Ceredigion, UK (ordnance survey reference SN 581823). Plantswere frozen within 1 h of harvesting and subsequently dried at70–80 �C in a Gallenkamp Hotbox oven (Gallenkamp, Loughbor-ough, UK). Dried material was milled using an A11 Basic IKA mill(IKA, Staufen, Germany) to produce a flour with >90% by weight<1 mm particle size. The flour was then further milled using a SPEX6770 SamplePrep freezer/mill (Stanmore, Middlesex, UK) for2 � 1 min pre-chilled with liquid nitrogen to a fine powder.

2.2. Laminarin and mannitol determination

Aliquots of 20 mg ground sample were incubated in duplicate in2 ml screwcap microfuge tubes, with or without, 1U laminarinase(Trichoderma sp., Sigma) in 50 mM succinic acid (pH 5.0) at a finalvolume of 1 ml. Prepared samples were incubated at 37 �C,150 rpm for 2 h and the glucose release determined using the glu-

cose determinant assay kit (Megazyme, Bray, Ireland) scaled downto a 1 in 10 volume. A 10 ll aliquot was incubated with 300 ll glu-cose oxidase–peroxidase (GOPOD) reagent in a flat-bottomed 96well plate and incubated at 50 �C for 20 min. The plate was readusing a lQuant plate reader (Bio-Tek Instruments, Winooski,USA) at 510 nm and the laminarin content of the sample deter-mined from the difference in glucose release between the samplesand controls.

A 2% w/w ground L. digitata solution was prepared and allowedto equilibrate. An aliquot was centrifuged at 2500g for 5 min and100 ll supernatant mixed with 900 ll of 5 mM H2SO4 containing10 mM crotonic acid (Sigma). This solution was filtered through a0.45 lm PVDF Durapore filter (Millex-HV, Millipore, Billerica,USA) into 0.2 ll glass-insert vials and analysed using an HPLC sys-tem through a Resex ROA-organic acid H+ column at 30 �C in 5 mMsulphuric acid mobile phase at 0.6 ml min�1 (Jasco, Great Dunmow,Essex, UK). A refractive index detector determined peak areaswhich were compared to calibration and internal standards usingthe software programme EZChrom Elite Version 3.2 (Scientific Soft-ware, Agilent Technologies, Santa Clara, USA) and the concentra-tion of mannitol determined.

2.3. Proximate and ultimate analysis

Moisture content of samples were determined by drying 5.0 g ofeach sample in a B&T Unitemp oven (LTE, Oldham, UK) calibratedto 105 ± 2 �C overnight. The ash contents were obtained by heating500 ± 1 mg of each sample to 550 �C for 12 h in an Elite high tem-perature furnace (Elite Thermal Systems Ltd, Leicestershire, UK)and calculating the proportion retained. The volatile proportionwas determined using a thermo-gravimetric analysis (TGA) meth-od representing the initial pyrolysis step. Samples of 5–6 mg wereheated from 40 to 900 �C at 25 �C min�1 under a nitrogen flow witha 2219 Multitemp II thermostatic circulator (LKB Bromma) and athermal analyser STA-780 series (Stanton Redcroft). Data was gen-erated by comparing the sample weight against a control cruciblewhich was monitored electronically every 1.67 s. This was elec-tronically recorded using PicoLog recorder software and analysedfurther using Excel (Microsoft).

The C, H, N and S contents of the biomass were calculated usinga CE instruments Flash EA 1112 series elemental analyser in dupli-cate. The relative percentage of each was determined and the oxy-gen content calculated by difference and corrected for ash.

Higher heating values (HHV) were determined from the ulti-mate analysis values using the equation of Channiwala and Parikh(2002). Lower (net) heating values (LHV) were calculated from theHHV using the equation used by the Energy Research Centre forThe Netherlands (ECN) (ECN, 2010).

2.4. Metal analysis

Samples of 200 mg were wet digested in HNO3 in a closedvessel. Metal concentrations were determined using an Optima5300 DV inductively coupled plasma spectrometer (ICP) withoptical emission spectrometry (Perkin Elmer, Cambridge, UK).

2.5. Pyrolysis–gas chromatography–mass spectrometry (Py–GC–MS)

Pyrolysis–gas chromatography–mass spectrometry analysiswas performed on a CDS 5000 series pyrolyser connected to aGC-2010 gas chromatograph and a GCMS-QP2010 chromatographmass spectrometer (all Shimadzu, Kyoto, Japan). Samples weighing2–4 mg were prepared in duplicate and pyrolysed at 500 �C at aramp rate of 20 �C per milli-second with a hold time of 20 s. Sepa-ration occurred on an Rtx 1701 60 m capillary column, 0.25 i.d.,

228 J.M.M. Adams et al. / Bioresource Technology 102 (2011) 226–234

0.25 lm film thickness, using a temperature programme of 40 �C,hold time 2 min, ramped to 250 �C and held for 30 min. Columnhead pressure at 40 �C of 30 psi.

Data processing was performed using National Institute of Sci-ence and Technology (NIST) Automated Mass spectral Deconvolu-tion and Identification System (AMDIS, (v2.65) and compoundidentification was performed using NIST Mass Spectral library incombination with referenced literature data. The weight of thesample tubes were recorded pre- and post-analysis and the peakareas recalculated to provide values for each key marker com-pound studied per mg sample.

2.6. Statistical analysis

Statistical analysis on Py–GC–MS key marker compounds wasconducted on the area of the component peak relative to the totalion count (TIC) for the entire chromatogram per mg sample. Thisassumes all peak areas are comparable and all compounds have asimilar density but does also allow semi-quantification of the com-pounds identified.

Data was analysed using a two-way Analysis of Variance (ANO-VA) to determine whether the group means (where a group wasdefined as the month of collection) were all equal or not. A multi-ple comparison of the data was conducted using the Student–New-man–Keuls test (P 6 0.05) to determine which months differedsignificantly from one another. The tests were conducted withinthe software programme GenStat Version 11 (VSN InternationalLtd).

3. Results and discussion

3.1. Laminarin and mannitol concentrations

The concentrations of laminarin and mannitol were deter-mined using biochemical methods to allow comparisons to bedrawn between the composition of the samples and subsequentanalysis. Mannitol concentration was approximately 5% of dryweight (d.w.) at the beginning of the year (Jan–Apr), peaked inJune (32.1% d.w.), with a second smaller mannitol peak in August,and fell to �15% d.w. for the remainder of the year (Fig. 1). Lami-narin concentration was very low until June, peaked at 24.6% d.w.in July and gradually decreased through the remainder of the year(Fig. 1).

0

5

10

15

20

25

30

35

Jan Feb Mar Apr May Jun

Com

poun

d pr

opor

tion

(% d

.w.)

Harvest m

Fig. 1. Biochemical determination of laminarin and mannitol proportions of dry L. digita(Sigma) to convert the laminarin to glucose followed by glucose determination by kit (Mreference to an internal standard (crotonic acid, Sigma) and a previously run set of stan

3.2. Proximate and ultimate analysis

Thermo-gravimetric analysis (TGA) represents the initial pyro-lysis step with differences in mass loss at set temperatures indicat-ing a difference in pyrolysis behaviour. Pyrolysis decompositionhas previously been noted as occurring at a lower temperaturefor macroalgae than for terrestrial biomass with high cellulose orlignin content (e.g. energy grasses and woody biomass) (Rosset al., 2008). A direct comparison between terrestrial biomassand macroalgae is not possible due to the differences in composi-tion and elemental influences.

The majority of the samples exhibited decomposition profileswhere the largest decrease in sample weight (Tmax) occurred at�250 �C, and a second period of decomposition occurred at�300 �C. For samples harvested from June to August, Tmax wasat �300 �C, with a smaller decrease in sample weight at �250 �C.In all samples, there was a continuous decrease in weight through-out the TGA process unlike in terrestrial biomass where weight lossis minimal at higher temperatures. Loss6 105 �C was considered tobe moisture loss and material lost at 105 6 x6 500 �C was definedas the volatile content. The material still present at 900 �C was de-fined as ash and the proportion lost 500 6 x 6 900 �C defined aschar. These values were plotted with the most rapid decrease insample weight (Tmax) (Fig. 2).

The moisture content (MC) of the samples remained relativelyconstant, ranging from 3.5% to 6.6%, similar to the MC values gen-erated in oxygen (Section 2.3) of (3.4–6.1%) (Table 1). This demon-strated that little difference had occurred during the preparationand storage of samples prior to analysis. A greater variation wasobserved in the volatile and ash fractions, with the volatile contentdecreasing from January to March (48.9% d.w.) followed by an in-crease peaking in volatile content in June (69.1% d.w.). The volatileconcentration then remained relatively constant at �65% d.w. untilOctober when it decreased. The changes in the volatile proportionof the biomass mimicked the seasonal variation in carbohydratecontent, with maximum weight loss at higher temperatures forplants harvested in the summer months when the maximum car-bohydrate concentration was highest, and minimum volatile re-lease in early spring when the carbohydrate concentration waslowest (Fig. 1). The variation in ash content exhibited the oppositetrend to the volatile content, being highest in early spring and win-ter when there was little carbohydrate and lowest in the summerwhen the carbohydrate concentrations peaked.

Jul Aug Sept Oct Nov Dec

onth (2008)

ta material harvested throughout 2008. Laminarin determined using a laminarinaseegazyme). Mannitol concentration was determined by HPLC, using peak area with

dards.

100

150

200

250

300

350

400

0

10

20

30

40

50

60

70

80

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Tmax

tem

pera

ture

(ºC

)

Mas

s pr

opor

tion

(%)

Harvest month (2008)

Fig. 2. Laminaria digitata composition through the year as determined by thermo-gravimetric analysis (TGA) and Tmax – the temperature when maximum volatile material islost. Filled diamond = volatile proportion. Filled triangle = char proportion. Filled square = ash proportion. Filled circle = TGA moisture content. Open triangle = Tmax (usingsecond y-axis).

Table 1Laminaria digitata proximate and ultimate analysis. HHV and LHV determined using equations by Channiwala and Parikh (2002) and the Energy Research Centre forThe Netherlands (ECN), accessed 2010.

Month (2008) Proximate moisture content and ash analysis (%) Ultimate (% by weight) Heating values

M.C. Ash C H N O S H/C O/C HHVdry (MJ kg�1) LHVdry (MJ kg�1)

Jan 5.2 27.7 29.4 4.5 2.7 34.9 0.8 1.83 0.89 11.9 11.0Feb 5.2 29.5 28.8 4.4 3.0 33.4 0.8 1.83 0.87 11.8 10.8Mar 5.2 34.8 26.4 4.0 3.4 30.5 0.9 1.83 0.87 10.8 9.9Apr 5.2 32.4 27.3 4.1 3.5 32.0 0.9 1.77 0.88 11.0 10.1May 4.8 33.2 27.5 4.2 3.3 30.8 0.9 1.81 0.84 11.3 10.4Jun 3.4 22.5 32.2 5.3 1.7 37.6 0.8 1.97 0.88 13.6 12.5Jul 4.8 13.8 36.2 5.6 1.3 42.5 0.6 1.84 0.88 14.9 13.7Aug 4.9 16.5 35.1 5.5 1.1 41.2 0.7 1.86 0.88 14.5 13.3Sept 5.3 19.0 33.8 5.3 1.4 39.7 0.8 1.87 0.88 14.0 12.8Oct 4.6 21.1 33.3 5.1 1.7 37.8 0.9 1.84 0.85 13.8 12.6Nov 5.0 22.4 32.5 5.0 1.9 37.3 0.9 1.83 0.86 13.4 12.3Dec 6.1 23.7 31.7 4.9 2.1 36.8 0.9 1.83 0.87 13.0 12.0

M.C. = moisture content. HHV = higher heating value. LHV = lower heating value.

J.M.M. Adams et al. / Bioresource Technology 102 (2011) 226–234 229

At the beginning of the year, ash d.w. of 33.2%, 27.7% and 31.7%were measured in February, March and April, respectively. The fallin ash content during March may have been associated with a par-ticularly high elemental content for that month and this is dis-cussed in more detail in Section 3.3. The ash content declined inMay to �20% and remained at this level until October when it in-creased again as the carbohydrate concentration decreased. Ashcontent as determined by aerobic combustion (12 h at 550 �C,Table 1) was similar to ash content as determined by TGA(Fig. 2). The seasonal ash content determined using both methodsfollowed the same trends throughout the year with the exceptionof that in May, which was low following TGA and high in theNREL-determined ash fraction.

The char content was relatively constant at �14% throughoutthe year except in March, May and September when peaks in con-centration occurred. As the char content was calculated from theash fraction, these fluctuations can be primarily explained as achange in the ash content rather than a change exclusively in thechar fraction. It is hypothesised that the char defined in this study,the proportion of material lost between 500 and 900 �C, is differentto the char of lignocellulosic-rich terrestrial biomass. For examplethe TGA profiles of L. digitata exhibited a continuous decrease inweight loss over temperature throughout the temperature gradientwhich differs from a terrestrial biomass profile.

The carbon (C) fraction of the samples declined from January(29.4%) until March (26.4%), then increased until July (36.2%) whenthe laminarin concentration peaked (Fig. 1). Thereafter the C frac-tion decreased throughout the rest of the year. Hydrogen (H) fol-lowed the same trend as for C (min H = 4.0% in March, maxH = 5.6% July) as was also demonstrated by the comparatively lin-ear H/C ratio in Table 1. The same trend also occurred with the oxy-gen (O) concentration, with linear O/C ratios throughout the yearsuggesting that the majority of these elements are being utilisedin similar processes at similar ratios. By contrast, the nitrogen(N) proportion was almost the inverse of C, H and N, with the high-est O concentration during winter and spring and lowest concen-tration in the summer. The N concentrations mimicked therelative concentrations in the surrounding seawater (Sandersonet al., 2008), but may also be decreasing as a proportion as largeamounts of carbohydrate and C-based structures are produced asthe plant grows. The sulphur (S) concentration remained relativelyconstant throughout the year at 0.6–0.9%. This was possibly due tothe small concentration present but could also be explained in thatS-rich proteins and other S compounds could be sequestered in thestipe in winter months or selectively accumulated. Higher andlower heating values followed the C, H and O proportion through-out the year, with the lowest values in March (10.8 and 9.9MJ kg�1, respectively) and the highest in July (14.9 and 13.6

230 J.M.M. Adams et al. / Bioresource Technology 102 (2011) 226–234

MJ kg�1, respectively) (Table 1). L. digitata biomass for combustionshould therefore be harvested in or around July to obtain optimalenergy production per kg of feedstock.

3.3. Metal contents

Macroalgal samples were examined using inductively coupledplasma spectrometer (ICP) with optical emission spectrometryand the metal contents (ppm) of 15 elements detected (Table 2).Samples were also analysed to determine Li concentration but nodetectable amount was determined for any month. The main ele-ments present, on a ppm basis, in the L. digitata samples were K,Na, Ca and Mg, followed by Sr, Fe, Zn and Al, with a mean value ofthe other elements present at <10 ppm. Seaweed ash has previouslybeen reported to contain potassium-, sodium- and calcium-carbonate(Ruperez, 2002) and high concentrations of these compounds willlead to increased slagging, fouling and other ash-related prob-lems during thermochemical conversion. Generally, the highestconcentration of each element was detected at the beginning andend of the year with the lowest concentration detected in themiddle of the year. The sum of the alkali metals Ba, Ca, K, Mg, Naand Sr in ppm peaked in March and was lowest in July (Fig. 3).This trend was exhibited by many of the individual elements(Table 2), with two elements (Ca and Zn) exhibiting a minima inJune. A total of 10 elements (Al, Ba, Cd, Cr, Fe, K, Mg, Mn, Na andSr) exhibited minima in July and two elements (Cu and Ni) minimain August.

Table 2Laminaria digitata metals analysis. Determined using an inductively coupled plasma spect

Month Metal contents (ppm)

Al Ba Ca Cd Cr Cu Fe K

Jan 94 12.6 12,761 2.6 1.9 4.2 112 5Feb 73 12.8 12,692 2.2 2.9 5.8 150 5Mar 92 12.2 11,803 2.2 3.0 5.2 203 7Apr 71 12.7 12,980 1.5 2.0 5.5 157 7May 86 9.4 10,295 1.7 1.2 4.7 124 8Jun 53 7.0 8361 1.7 0.7 3.1 90 4Jul 13 6.5 8510 1.4 0.5 3.6 35 1Aug 64 7.0 8709 2.5 2.7 3.0 112 2Sept 31 7.2 9629 1.7 0.7 4.0 63 2Oct 35 10.3 11,371 2.2 0.5 4.4 79 3Nov 71 10.0 10,206 3.1 0.5 5.3 112 4Dec 44 11.6 11,213 2.8 0.9 5.5 97 4

0

20000

40000

60000

80000

100000

120000

140000

160000

Jan Feb Mar Apr May Ju

Con

cent

ratio

n (p

pm)

Harvest

Fig. 3. Sum of the alkali metals Ba, Ca, K, Mg, Na, Sr as parts per million showing seasonamass spectrometry detection.

The maximum concentration of the metals (in ppm) was lessseasonally specific, with maxima distributed through the year.The highest concentration of individual elements occurred in Janu-ary–May, August, September and November. Of the four elementsdetected at the highest concentrations, Mg and Na were at amaxima in March, Ca in April and K in May. These months corre-sponded with the detection of the highest elemental concentrations(Fig. 3). The concentration of two key elements – K and Na – can beused to calculate an alkali index which provides information on thelikelihood of slagging and fouling in a boiler if combusted or pyrol-ysed. This data is summarised in Table 3 and indicates that all har-vests are above the index value of 0.34 when slagging and foulingbecomes ‘virtually certain’ (Masia et al., 2007). However, there isvariation between months and March exhibited the highest alkaliindex (14.86) and July the lowest (3.65). This means that a July har-vest would provide the highest heating value and the lowest ashand alkali index values, making it the best month for harvestingfor thermochemical conversion in this study. Elements present inrelatively high concentrations in macroalgae can also be toxic orproblematic when released as volatiles. This includes bound As,which if combusted can be released to the inorganic form, whichis significantly more toxic than in the original arsenosugar form.Though clearly problematic, samples from July would be the mostsuitable for use in pyrolysis and hydrothermal liquefaction conver-sions to produce bio-oil or utilisation in co-firing.

The metal concentrations reported in this manuscript weregenerally lower than those reported by Ross et al. (2008) which

rometer with optical emission spectrometry (ICP–OES).

Mg Mn Na Ni Pb Sr Zn

0,913 8475 4.9 38,899 1.2 0.0 1079 858,723 9214 4.3 45,952 0.7 0.0 1105 764,292 9476 6.5 52,459 1.0 0.0 1121 742,153 8611 7.7 42,108 0.7 0.0 1172 751,442 8015 5.0 42,468 0.7 0.0 966 813,147 6832 3.9 33,977 0.5 0.0 670 308,993 6373 2.6 23,308 0.5 0.0 632 512,835 6373 4.2 29,767 0.2 3.5 643 559,343 6797 3.5 31,458 0.7 0.0 697 1156,805 7430 3.9 31,388 0.7 0.0 883 921,746 6990 4.1 30,813 0.5 0.0 855 573,791 7258 3.8 33,934 0.7 0.0 996 69

ne July Aug Sep Oct Nov Dec

month (2008)

l variation. Metals identified by inductively coupled plasma spectrometry (ICP) with

Table 3Alkali index values of Laminaria digitata showing seasonal variation. Index calculated from Masia et al. (2007) as (Na2O + K2O) kg GJ�1.

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Alkali index 9.53 11.25 14.86 13.12 13.72 7.18 3.65 4.67 5.57 6.30 6.86 7.55

J.M.M. Adams et al. / Bioresource Technology 102 (2011) 226–234 231

analysed L. digitata harvested in Cornwall (South West UK).Samples in the Ross et al. (2008) study were collected in February,but though concentrations for the most abundant elements (K, Na,Ca, Mg) were within the seasonal range for Aberystwyth-sourced L.digitata, the majority of other elements analysed in both samplesets (Zn, Cu, Al, Fe, Mn, Pb) were at a higher concentration in theRoss et al. (2008) study. Of the remaining metals measured in bothstudies, Cr concentration in the Cornwall samples were within theAberystwyth seasonal range, and Sr and Cd concentrations wereconsistently higher in the Aberystwyth samples. Some variationcould be due to weather differences such as rainfall prior to samplecollection. However the lower concentrations of the majority of themetals in the Aberystwyth (Cardigan bay) samples suggest that thewater was cleaner than that off the Cornish coastline. The low cur-rent elemental concentrations in the Cardigan bay contrasts with areport from 1973 that stated that ‘appreciable’ amounts of Pb, Zn,Cu and Cd flowed into the bay as contamination from mine effluentand tips drained into the rivers at this time (Fuge and James, 1973).Concentrations of trace-elements in macroalgae depend upon theconcentrations of metal compounds in the surrounding waterand the ability of macroalgae to bio-accumulate such compounds.Macroalgae will often therefore have a higher concentration ofmetals than that naturally occurring in the surrounding waters.Zn content in Fucus vesiculosis (no Laminaria were sampled) of358 ppm during June contrasts to the 30 ppm detected in June inL. digitata in the current study. Given the size of this difference, itseems unlikely that this represents only a difference between thespecies ability to accumulate Zn. Smaller differences existed be-tween the two studies in the concentration of other elements pres-ent at low amounts, but Mn and Ni were �10-fold lower in thecurrent Laminaria study than in the earlier Fucus report. A contin-ued decline of elemental contamination in macroalgae would in-crease the suitability of this feedstock for thermochemicalconversion pathways.

3.4. Pyrolysis–gas chromatography–mass spectrometry

Pyrolysis of biomass is a complex process, with variation inproducts caused by several factors including biomass composition,the presence of inorganic material and the heating rate (Nowakow-ski and Jones, 2008; Nowakowski et al., 2007, 2008). Using fastpyrolysis with an increase of >100 �C s�1, the technique is a simu-lation of flash pyrolysis on a small scale, where pyrolysis oil is pro-duced and analysed using the GC–MS. Although the decompositionof the material can be irregular, the origin of the pyrolysis productscan still be determined in the majority of cases. Py–GC–MS haspreviously been predominantly used to characterise the pyrolyticdecomposition of lignocellulosic materials, however this techniquehas also recently been successfully used to characterise macroalgae(Ross et al., 2009, 2008).

In the current study, Py–GC–MS was performed at 500 �C on theL. digitata samples in duplicate. A total of 29 peaks were identifiedas being consistently present in the sample spectra. Twelve ofthese compounds were selected as key marker compounds for L.digitata and used to screen for compositional changes over time.Ten of the 12 key marker compounds (toluene; pyrrole; furfural;ethanone, 1-(2-furanyl); furfural, 5-methyl; 1,2-cyclopentanedi-one, 3-methyl; phenol; dianhydromannitol; indole; 3,7,11,15-tet-ramethyl-2-hexadecen-1-ol) have been previously identified as

major pyrolysis products of L. digitata and other macroalgae spe-cies (Ross et al., 2008). In this study benzene acetonitrile and ben-zene propanenitrile were also identified as key pyrolysis productsof L. digitata under the pyrolysis conditions used.

All compounds were detected in all spectra except benzene ace-tonitrile (8), which was not detected in one of the replicates duringAugust, September and November. The identified compounds wereplotted throughout the year by peak area per mg of sample ana-lysed and separated into those of carbohydrate origin (Fig. 4A)and those of protein, lipid and phenolic origin (Fig. 4B). One ofthe May harvest replicates was contaminated, with peaks appear-ing in the first few minutes of the run and highly abnormal peaksappearing elsewhere in the spectrum so was not used in any statis-tical analyses but was included in Fig. 4 for comparative purposes.

Monthly replicates were highly comparable for all compoundsexcept dianhydromannitol. Dianhydromannitol appeared to stickslightly in the GC injection port and so peaks often appeared smal-ler than would be expected for the sample composition. Previouswork by the authors suggests that the higher peaks generated fordianhydromannitol more accurately reflects the actual quantityof the compound present, so n = 1 for June and July samples ofdianhydromannitol reported in Fig. 4A.

Those compounds of carbohydrate origin exhibited significantlylarger peak areas than those of other origins, reflecting the largecarbohydrate content of the macroalgae (Fig. 1). Seasonal trendsalso occurred in carbohydrate-derived compounds (Fig. 4) butthere were no clear trends for the protein, phenolic and lipid-de-rived compounds. A two-tailed ANOVA was performed to deter-mine whether there was a significant difference between theproportion of the key marker compounds, the time of year thatthey were harvested and the interaction between these two fac-tors. There were significant differences detected between the pro-portion of the compounds (P < 0.01), harvest date (P < 0.001) and asignificant interaction between the compounds and the harvestdate (P < 0.001) (Table 4).

The 12 compounds were also individually analysed by ANOVAto determine whether significant differences occurred in each com-pound across the year with the results summarised in Table 5. Themajority of the 12 compounds exhibited significant differences insamples from different months with a P 6 0.05 for all compoundsexcept toluene (1), dianhydromannitol (9) and 3,7,11,15-tetra-methyl-2-hexadecen-1-ol (12). With the exception of phenol(7), all other compounds had a significant difference of P < 0.01(Table 5). Two major patterns of changes in compound concentra-tion over the year were identified: (1) where concentrations duringJanuary–April where significantly different from June–December;(2) where concentrations during August where significantly differ-ent from all other months.

The peak area in May for more than half the compounds studiedand notably in all of the compounds where seasonal trends oc-curred, lay between those for April and June. The exception wasfurfural, where the proportion in May was 0.01% less than that ofApril. Dianhydromannitol was not significantly different betweenmonthly harvests of L. digitata as there was a large variation inpeak area between replicates as described above.

The proportionate release of ethanone 1-(2-furanyl) (4) for themonths Jan–Apr was significantly lower than that observed forsubsequent months. This trend was also exhibited by furfural, 5-methyl (5). Benzene propanenitrile (10) and benzene acetonitrile

0

5000000

10000000

15000000

20000000

25000000

Peak

are

a pe

r mg

sam

ple

Harvest month (2008)

2500000

0

500000

1000000

1500000

2000000

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Peak

are

a pe

r mg

sam

ple

Harvest month (2008)

A

B

Fig. 4. Compound peak area per mg L. digitata sample. A = Polysaccharide origin compounds: Filled circle = furfural. Open circle = ethanone, 1-(2-furanyl). Filled square = 5-methyl furfural. Open square = 1,2 cyclopentanedione. Filled triangle = dianhydromannitol (n = 1 for June and July reps). n = 1 for all May compounds. B = protein, lipid andpolyphenolic origin compounds: Filled triangle = toluene. Open triangle = pyrrole. Filled diamond = phenol. Open diamond = benzene acetonitrile. Filled square = benzenepropanenitrile. Open square = indole. Filled circle = 3,7,11,15-tetramethyl-2-hexadecen-1-ol. n = 1 for all May compounds.

Table 4Two way Analysis of Variance (ANOVA) considering the proportion of each compoundidentified (% whole sample) against harvest date.

Source ofvariation

Degrees offreedom(d.f.)

Sum ofsquares(s.s.)

Meansquared(m.s.)

Variance(v.r.)

F-testprobability(F. pr.)

Harvest 10 1.51 0.15 2.54 0.008Composition 11 67.16 6.11 103.01 <0.001Harvest 110Composition 15.46 0.14 2.37 <0.001Residual 132 7.82 0.06Total 263 91.96

232 J.M.M. Adams et al. / Bioresource Technology 102 (2011) 226–234

(8) exhibited the reverse trend with higher yields in Jan–Apr. Pyr-role was present at a significantly higher concentration for themonths Jan–Apr and June, with the single May value falling be-tween April and June. Indole was significantly more abundant inJan, Feb and April, compared to March and in later months.

The second group which exhibited a significant change in com-pound concentration during August included furfural (3) and furfu-ral, 5-methyl (5). In addition ethanone 1-(2-furanyl) (4) exhibitedsignificant changes during August and June.

It is proposed that furfural-based and nitro-benzene-basedcompounds generated through pyrolysis can be used as indicators

of feedstock composition. This was particularly the case with eth-anone 1-(2-furanyl) (4) where it was highly associated with themannitol fraction of macroalgae (Figs. 1 and 4 and Ross et al.,2009). Furfural concentration was associated, albeit to a much les-ser extent, to both laminarin and alginic acid concentrations (Figs.1 and 4 and Ross et al., 2009).

Another observation on the key marker compounds was thehigh presence of nitrogen-containing compounds pyrrole, benzeneacetonitrile and benzene propanenitrile in the early months of theyear. This occurrence was supported by the high concentration of Nin the macroalgae during January–May (see Table 1) and a knownhigh N concentration in marine water in the early months of theyear (Sanderson et al., 2008).

Py–GC–MS provided information on the potential use of themacroalgae as a feedstock for thermochemical conversion. Forexample a higher proportion of material was volatilised in thesummer months than in the rest of the year, identifying this asthe most suitable period to harvest macroalgae for pyrolysis. Thestatistical analysis also suggests compositional changes occurredin August which may be related to the increase in mannitol con-centration (Fig. 1). If macroalgae were to be used in a biorefinery,then the reduction in concentration of a number of the compoundsin August would make this month less suitable for a harvest.

Table 5Each Py-GC-MS key marker statistically compared through the year to determine significant differences per mg sample. Significant differences denoted by letters.

Compound Toluene Pyrrole Furfural Ethanone, 1-(2-furanyl)

Furfural, 5-methyl

1,2-Cyclopentanedione, 3-methyl

Peak No. 1 2 3 4 5 6

Mean 0.41 0.26** 0.52*** 1.42*** 0.35*** 0.61**

Std error 0.06 0.02 0.05 0.16 0.04 0.14

% Weightmg-1

% Weightmg-1

% Weightmg-1

% Weight mg-1 % Weightmg-1

% Weight mg-1

Jan a abc ab a ab abcFeb a a ab a a abcMar a ab ab a a aApr a a a a a abJun a abc ab b bc cJul a c b cd c cAug a bc c b d abcSep a bc b c c cOct a bc ab cd c bcNov a bc ab c c abcDec a bc ab d bc abc

Compound Phenol Benzeneacetonitrile

Dianhydromannitol

Benzenepropanenitrile

Indole 3,7,11,15-Tetramethyl-2-hexadecen-1-ol

Peak No. 7 8 9 10 11 12

Mean 0.25* 0.07*** 1.74 0.11*** 0.09** 0.37Std error 0.02 0.02 0.82 0.01 0.03 0.06

% Weightmg-1

% Weightmg-1

% Weightmg-1

% Weight mg-1 % Weightmg-1

% Weight mg-1

Jan ab a a ab ab aFeb ab b a a a aMar abc b a abc bc aApr ab c a a ab aJun c d a de c aJul ac d a d bc aAug abc d a d c aSep abc d a de bc aOct b d a bce bc aNov ab d a ce bc aDec ab d a bce bc a

* P < 0.05.** P < 0.01.

*** P < 0.001.

J.M.M. Adams et al. / Bioresource Technology 102 (2011) 226–234 233

4. Conclusions

The macroalgae used in this study contains a large proportion ofwater and a high concentration of metals, meaning that Laminariais unlikely to be an ideal feedstock for combustion or pyrolysis.However, by comparing the variation in biomass compositionacross the year, it was possible to determine the optimal harvesttime for L. digitata, for utilisation as a biofuel, was July when metalconcentrations were low, carbohydrate concentrations high, andHHV and LHV highest. Macroalgae, or macroalgal residues, couldbe pyrolysed to create a bio-oil or used in hydrothermal liquefac-tion to make bio-crude in a process which does not require an ini-tial drying of the feedstock.

Acknowledgements

This work was supported by the Engineering and Physical Sci-ences Research Council (EPSRC); grant number GR/S28204 to theSUPERGEN Bioenergy consortium. Work undertaken at the Univer-sity of Leeds was supported by the EPSRC through the SUPERGENtraining fund.

References

Adams, J.M., Gallagher, J.A., Donnison, I.S., 2009. Fermentation study on Saccharinalatissima for bioethanol production considering variable pre-treatments. Journalof Applied Phycology 21, 569–574.

Black, W., 1950. The seasonal variation in weight and chemical composition of thecommon British laminariaceae. Journal of the Marine Biological Association ofthe United Kingdom 29, 45–72.

Carlsson, A., van Beilen, J., Moller, R., Clayton, D., 2007. In: Bowles, D. (Ed.), Micro-and Macro-Algae: Utility for Industrial Applications Outputs from the EPOBIOProject. CNAP, University of York, York, UK, p. 82.

Channiwala, S.A., Parikh, P.P., 2002. A unified correlation for estimating HHV ofsolid, liquid and gaseous fuels. Fuel 81, 1051–1063.

ECN, 2010. Phyllis database: determination of lower heating value (LHV). EnergyResearch Centre of the Netherlands.

Fuge, R., James, K., 1973. Trace metal concentrations in brown seaweeds, CardiganBay, Wales. Marine Chemistry 1, 281–293.

Gao, K., McKinley, K., 1993. Use of macroalgae for marine biomass production andCO2 remediation: a review. Journal of Applied Phycology 6, 45–60.

Hayward, P., Nelson-Smith, T., Shields, C., 1996. Sea Shore of Britain and NorthernEurope. Collins, London.

Horn, S.J., 2000. Bioenergy from Brown Seaweeds. Department of Biotechnology,Norweigian University of Science and Technology, Trondheim, Norway. p. 148.

Horn, S.J., Aasen, I.M., Ostgaard, K., 2000. Ethanol production from seaweed extract.Journal of Industrial Microbiology and Biotechnology 25, 249–254.

Masia, A.A.T., Buhre, B.J.P., Gupta, R.P., Wall, T.F., 2007. Characterising ash ofbiomass and waste. Fuel Processing Technology 88, 1071–1081.

Moen, E., Horn, S., Ostgaard, K., 1997. Alginate degradation during anaerobicdigestion of Laminaria hyperborea stipes. Journal of Applied Phycology 9, 157–166.

Nawakowski, C., Nicholson, M.D., Kershaw, P.J., Leonard, K.S., 2004. Modelling Tc-99concentrations in Fucus vesiculosus from the North-East Irish sea. Journal ofEnvironmental Radioactivity 77, 159–173.

Nelson, T.E., Lewis, B.A., 1974. Separation and characterization of soluble andinsoluble components of insoluble laminarian. Carbohydrate Research 33, 63–74.

Nowakowski, D.J., Jones, J.M., 2008. Uncatalysed and potassium-catalysed pyrolysisof the cell-wall constituents of biomass and their model compounds. Journal ofAnalytical and Applied Pyrolysis 83, 12–25.

234 J.M.M. Adams et al. / Bioresource Technology 102 (2011) 226–234

Nowakowski, D.J., Jones, J.M., Brydson, R.M.D., Ross, A.B., 2007. Potassium catalysisin the pyrolysis behaviour of short rotation willow coppice. Fuel 86, 2389–2402.

Nowakowski, D.J., Woodbridge, C.R., Jones, J.M., 2008. Phosphorus catalysis in thepyrolysis behaviour of biomass. Journal of Analytical and Applied Pyrolysis 83,197–204.

Ross, A.B., Jones, J.M., Kubacki, M.L., Bridgeman, T., 2008. Classification of macroalgaeas fuel and its thermochemical behaviour. Bioresource Technology 99, 6494–6504.

Ross, A.B., Anastasakis, K., Kubacki, M., Jones, J.M., 2009. Investigation of thepyrolysis behaviour of brown algae before and after pre-treatment using PY–GC/MS and TGA. Journal of Analytical and Applied Pyrolysis 85, 3–10.

Ruperez, P., 2002. Mineral content of edible marine seaweeds. Food Chemistry 79,23–26.

Sanderson, J., Cromey, C., Dring, M., Kelly, M., 2008. Distribution of nutrients forseaweed cultivation around salmon cages at farm sites in North-West Scotland.Aquaculture 278, 60–68.