1 Introduction

1.1 Microalgae biofuels and life cycle assessment

Microalgae have long been considered a possible source for future low-carbon transport fuel (Laurens et al. 2017) and the most promising photoautotrophs for fixing the carbon in the atmosphere (Shahid et al. 2020). One of the primary reasons for this is the low use of prime agricultural land and high productivity per hectare compared with other crops. Microalgae are single-celled organisms containing carbon-rich lipids; these lipids can be converted into biodiesel through transesterification, with glycerol as a by-product (Skorupskaite et al. 2016). Biodiesel production from aquatic biomass (such as microalgae) is a promising option that could address most of the challenges with the previous generations of biofuels and thus has received considerable attention (Barr and Landis 2018; Sander and Murthy 2010). However, there is significant debate over the sustainability of all bioenergy sources (Zheng et al. 2020), including microalgae-derived biofuels, that requires further investigation using sustainability assessments in general and environmental impact assessment in particular (Rafiaani et al. 2020; Villarroel Walker et al. 2017). Moreover, the non-lipid parts of the microalgae can be used as biomass for energy generation or to produce high-value products such as β-carotene (Di Caprio et al. 2020), astaxanthin (García Prieto et al. 2017), docosahexaenoic acid (Yin et al. 2019) and extracts for use in cosmetics. Nevertheless, biomass conversion from microalgae to bioproducts encompasses complex processing stages and different process pathways that need to be examined.

Life cycle assessment (LCA) methodology can quantify the environmental aspects and potential environmental impacts associated with a product, process or service (ISO 140402006). Depending on the system boundary, LCA includes the initial resource mining and refinement, transportation, fabrication, use and end-of-life (EoL). The accuracy of an LCA is highly dependent on the quality and accuracy of the underlying data and assumptions used to accomplish the analysis (Velasquez-Orta et al. 2018; Torkayesh et al. 2022). LCA studies of the production of microalgae-derived biofuels have shown that results vary significantly (Garcia et al. 2020) due to alternative methodologies and assumptions used (Azadi et al. 2014; Collet et al. 2014; Quinn and Davis 2015).

1.2 Previous LCA studies on microalgae

Although there is a great interest in microalgae biofuel LCA, there is a distinct lack of LCA studies that use real-world primary data from microalgae production facilities. In fact, the majority of microalgae biofuel LCA research is based on hypothetical industrial facilities that is further based on scaled-up or theoretical calculations. Some of these extrapolated data from their own smaller lab tests, whilst others used secondary data from the scientific literature. There are a number of issues with these, as they are based on either dramatically extrapolated data from laboratory tests or on data from multiple sources that are not representative of a particular single microalgae biofuel facility. Within the IEA 2017 State of Technology Review for Microalgae Bioenergy (Laurens et al. 2017), there was significant criticism of two highly cited papers (Chisti 2007; Hu et al. 2008), which seem to form the foundation of many LCAs of hypothetical facilities. The IEA review considered that these two papers provide unrealistically high levels of operational time, oil content and aerial productivities.

Aside from the data gaps, there are numerous issues with the LCA of microalgae-derived biofuels, which can mostly be considered to be generic and applicable to the whole of LCA as a field. One significant issue, which can be argued to be general to LCA, is the lack of comparability within apparently similar LCAs. Although ISO 14040 and 14044 present the general framework and requirements for performing a standardised LCA, they do not propose specific methods for making the various choices within the LCA. The choice of functional unit, boundary conditions and impact categories are left up to the practitioners; thus, seemingly similar LCAs could be, in fact, quite different in assumptions, methodology and inconsistent in data quality (Sills et al. 2020; Valente et al. 2019) and thus incomparable. Additionally, elements of ISO 14040 and 14044, such as the range of impact categories to be studied, are often not followed. As a result, many LCAs provide widely different results for microalgae biodiesel for all impact categories. This research compares the environmental impacts of microalgae biodiesel production—under different energy regimes and—with petroleum-derived diesel across different impact categories using LCA approach.

Purely looking at GHG emissions, they range from −0.75 kg CO2eq/MJ (Batan et al. 2010) to 3.8 kg CO2eq/MJ. Various meta-analyses have been undertaken to deal with this issue; however, the scientific literature uses widely differing assumptions, meaning that the results cannot be compared. Dasan et al. (2019) compared different microalgae biofuel systems, i.e. raceway open pond (OP), bubble column photobioreactor (BC-PBR) and tubular photobioreactor (TB-PBR) cultivation systems using LCA and life cycle costing (LCC) considering a cradle-to-gate approach. The Life Cycle Inventory (LCI) for the cultivation and biofuel production was extracted from secondary sources. The results showed that the dehydration and lipid extraction of microalgae biomass required 21–30% and 39–57% of the total energy consumption in the life cycle (Dasan et al. 2019). Therefore, the source of energy for these stages could have a significant impact on the life cycle environmental impacts of microalgae biofuel production systems. Almost all studies are based on scientific literature data, estimated cultivation and scaled-up lab experiments, with few specific exemptions that offer empirical evidence using data from large-scale, real-world facilities.

In terms of real-world data, Passell et al. (2013) performed an LCA based on data from Seambiotic, Inc., in Israel and Solution Recovery Services (SRS) Inc. (now trading as Valicor), based in Dexter, MI, USA. The authors combined data from the two facilities which produce microalgae for commercial purposes. The Seambiotic facility is a kind of carbon capture and utilisation (CCU) facility, as it is co-located with a fossil fuel power station; thus, the CO2 was treated as burden-free. The microalgae strains produced in the Seambiotic facility are Nannochloris sp. and Nannochloropsis salina. The facility uses raceways ponds with paddle wheels, producing microalgae biodiesel with a GWP100 climate change impact of 2.88 kg CO2eq/MJ biodiesel, with a future case of 0.18 kg CO2eq/MJ biodiesel. Another work by the University of Texas measured the production of Chlorella protothecoides for biofuels (including data on transesterification). The data for this is discussed in Beal et al. (2012a) and Beal et al. (2012b); however, these are focused purely on the energy, not the environmental impacts.

The All-Gas project is another example that published results on a system fed by sewerage (Maga 2017), which followed the same methodology described in Bradley et al. (2015). As detailed in Maga (2017), the climate change (GWP100) impacts of 1 MJ of microalgae-derived biogas were 2.08 × 10−2 kg CO2eq, which is the most favourable result in terms of climate change for microalgae biofuels, and compared to the fossil fuel reference considered (i.e. compressed natural gas), it showed a 41% improvement. However, for particulate matter formation, photochemical oxidant formation and terrestrial acidification, the microalgae system performed worse.

Pérez-López et al. (2017) studied the AlgaePARC system at Wageningen UR, comparing different photobioreactor technologies. The technologies considered were Horizontal PBRs (0.56 m3), Vertical PBRs (1.06 m3) and raceways (4.73 m3). Unlike other scientific literature on real-world facilities, their study did not compare microalgae with alternative biofuels or petroleum-derived fuels. Branco-Vieira et al. studied the environmental impacts of a 2.5-hectare facility in Concepción, Chile (Branco-Vieira et al. 2020a, 2020b). The facility itself was detailed by Branco-Vieira et al. (2018), the microalgae species was Phaeodactylum tricornutum, and the boundary conditions were cradle-to-gate. The data were taken from a single 0.8 m3 bubble column photobioreactor on the site (Branco-Vieira et al. 2018), scaled up using the methodologies from the EnAlgae project (Spruijt et al. 2015). In terms of climate change, the impacts of microalgae biodiesel was reported as 5.74 kg CO2-eq/MJ biodiesel (Branco-Vieira et al. 2020b).

The EU-funded ABACUS Project, as described by Onorato and Rösch (2020), undertook a detailed LCA. Three types of photobioreactors were tested in their study: Flat Panel Airlift (FPA), Unilayer Horizontal Tubular PBR (UHT-PBR) and the Green Wall Panel (GWP). These were tested at real-world facilities, specifically Subitec Gmbh, A4F and Microphyt, respectively. The Subiyrc and A4f data were based on 93 m3 volume systems, whilst the Green Wall Panel data was based on a small 0.1m3 system. Dependent on the technology and electricity source, the paper showed climate change impacts varying from 21 to 265 kg CO2eq/kg microalgae. If we consider a value of approximately 38 MJ/kg of microalgae, this gives us a range of 0.55 to 9.6 kg CO2eq/MJ microalgae.

Considering the above research gaps on microalgae biofuel production, the current study assesses the environmental impacts of biofuel production (i.e. biodiesel) from microalgae using a set of primary data (i.e. real-world). It considers the cultivation stage, compares the microalgae production under different energy regimes and compares it to diesel fuel as the fossil counterpart. The novelties of our study are that (a) the work contributes to the slowly growing scientific literature in the field using real-world data for the LCA of microalgae-derived products and (b) the facility under investigation, which mixed autotrophic and heterotrophic growth methods, was unique and thus the performed LCA could be of importance. Considering the fact that biofuel production from algal resources is still in its infancy, studying the supply chain, resource management and process optimisation could significantly help the biofuel industry and specifically microalgal biofuel meet environmental targets. Therefore, the current study also aims at studying the hotspots in biofuel production from microalgae and providing some recommendations for improving the environmental profile of biofuel production from microalgae. Critically to date, whilst there is now a growing body of LCA studies of microalgae-derived fuels based on real-world data, this work is the only LCA of an integrated real-world system of fermenters and photobioreactors. In addition, with 60 m3 of PBRs and 1 m3 of fermenter, the data used within this work represents one of the largest industrial systems within the scientific literature from the past 5 years (with the exception of the A4f data within Onorato and Rösch (2020)). 

The importance of this work is threefold. First, it is amongst a small number of real-world work pieces of research into the LCA of microalgae, not just for biofuels but for a range of other products. Secondly, it adds to the continuing debate over biofuels, which with the current political climate over energy can be reasonably expected to increase in importance. Third, it considers a mixed system of heterotrophic and phototrophic systems, which is seldom considered within the literature.

2 Material and methods

2.1 Life cycle assessment (LCA): goal and scope definition

The current LCA study followed ISO standards for LCA (ISO 140402006; ISO 140442006) and aimed to compare the environmental impacts of microalgae biodiesel production under different energy regimes and with petroleum-derived diesel.

The system under study is from the InteSusAl (demonstration of integrated and sustainable microalgae cultivation with biodiesel validation) project, in which the consortium designed, built and operated a microalgae production facility at Necton S.A. in Olhão, Portugal. InteSusAl was a part of the Algae Cluster, which additionally included the All-Gas and BIOFAT projects; all three projects have built demonstration facilities that produced microalgae for fuel production. The facility in Olhão comprised of:

  • Inoculation: 10 green wall panels

  • Photobioreactors: 16 km of photobioreactors, with a combined volume of 60,000 L, divided into four separate systems, growing Phaeodactylum and Nannochloropsis

  • Fermenters: Three 1000 L low-cost steel fermenters, growing Chlorella

  • Harvesting: Evodos 25 Centrifuge

The system at the Centre for Process Innovation (CPI) comprised of:

  • Fermenters: 1000 L low-cost steel fermenter

The InteSusAl concept (Fig. 1) is based on integrating heterotrophic and autotrophic growth systems to produce microalgae biomass, thereby ensuring maximum recycling between the systems. The reason for mixing heterotrophic and autotrophic systems together was that at the time of the original project conceptualisation, this was considered a possible option to increase production efficiency. As shown in Fig. 1, microalgae is produced heterotrophically within low-cost steel fermenters, feeding the microalgae an optimised concentration of glycerol, yeast extract, di-potassium hydrogen phosphate and trace levels of other chemicals (Fig. 2). Tubular photobioreactors (TPBR) (Fig. 3) act as an autotrophic microalgae-based carbon capture and utilisation (CCU) system, utilising the waste CO2 from the fermenters’ off-gas stream to facilitate the growth of more microalgae. The O2 produced from photosynthesis within the TPBRs is fed into the fermenters to further enrich the system and promote a higher microalgae growth density. The fermenter and TPBR systems do not mix microalgae but only exchange gases. Finally, microalgae from both systems are harvested and subjected to dewatering and lipid extraction before undergoing a transesterification process, which produces fatty acid methyl esters (biodiesel) and glycerol. The glycerol is then recycled back into the fermenters within a process supplemented with additional glycerol from industrial sources. It should be noted that the water and nutrients in the harvesting stage are recycled back into the fermenters and TPBRs. There are four core elements to the system: (a) growth, (b) harvesting, (c) biodiesel production and (d) use.

Fig. 1
figure 1

The InteSusAl concept. Schematic diagram of integrated autotrophic and heterotrophic concepts within the project for biodiesel production

Full size image
Fig. 2
figure 2

The heterotrophic fermentation systems (3 lines). 1000 L stainless steel fermenter (centre) with corresponding stainless steel feed vessel and harvest vessel

Full size image
Fig. 3
figure 3

The photobioreactors. These make up the carbon capture and utilisation system, containing Nannochloropsis salina (left, green) and Phaeodactylum tricornutum (right, brown)

Full size image

In line with the goal of the study, three energy scenarios were defined as follows:

  • Scenario A: In this scenario, the average energy mix for the EU-27 countries (including the UK—the UK was part of the EU27 in 2012) in the year 2012 was considered, which was the year the InteSusAl project began and ensured that the work in this paper is comparable to (Maga 2017). (Note: Because this work and Maga’s work were both part of the Algae Cluster; thus, the same methodologies were used to make the two studies comparable).

  • Scenario B: In this scenario, the average energy mix for the EU-27 + UK for 2020 was considered.

  • Scenario C: In this scenario, solar photovoltaic (PV) electricity from a PV farm was modelled as the source of electricity for microalgae production.

These scenarios, along with the base case scenario (i.e. petroleum-derived diesel fuel), shape the four scenarios considered within this study. Within these scenarios, the models are split into infrastructure and operational impacts.

The current LCA study supports decisions on the micro decision-making level rather than on the macro level, and thus, it adopts an attributional modelling framework (Rajaeifar et al. 2017). The functional unit was considered as “combustion of 1 MJ (lower heating value) of algal biodiesel in an internal combustion engine (B100)”. Figure 4 shows the scope of the present study and its system boundaries. Accordingly, boundary conditions were set to embrace “well-to-wheel”. The top dashed box shows the system boundaries for microalgae biodiesel, which include cultivation, harvesting and lipid extraction, biodiesel production and its combustion. The bottom dashed box shows system boundaries for petroleum diesel which include extraction of crude oil, transportation to refineries, refinery operations and combustion. It should be noted that the transportation of fuels (either biodiesel or diesel) from the facility/refinery to the pump station was not considered in the calculations, as this was considered to be similar in both life cycles. Full descriptions of the logic behind this methodology are available within Bradley et al. (2015).

Fig. 4
figure 4

The scope of the present study and its system boundaries

Full size image

The LCA study is further divided into two main parts, i.e. the infrastructure model, based on considering the environmental and energy impacts from the construction of the facility, and the operational model, which is based on considering the environmental and energy impacts from the operation of the facility. Infrastructure must be considered, as microalgae biofuels require extensive new facilities to be constructed. In contrast, other biofuels can partially utilise existing supply chains and facilities within the agricultural industry, therefore not leading to an immediate additional environmental impact. To ensure the study was comparable with other European Commission funded microalgae projects, a harmonised methodology for LCA of algae biofuels presented by Bradley et al. (2015) was agreed upon between the InteSusAl, All-Gas and BIOFAT projects. The only area in which the InteSusAl project was unable to follow the principles was with the co-products. It was assumed that co-products would be allocated based on energy content; however, there are various other possibilities for co-products, such as high-value products, as described earlier.

In terms of co-products, the system produces glycerol, which is recycled back and thus accounted for within the analysis. Waste products are produced within the project, and the assumption used is that, as an energy-based facility, these other resources would also be used for energy production, most probably via anaerobic digestion. This moves into the issues of multifunctionality, where the system has created two useful products. Whilst it can be argued there is no correct approach to multifunctionality, we can consider the purpose of the system to guide our decisions. Within this article, an energy-based allocation has been chosen. It is important to note that the energy density of the waste materials and the lipids are similar, and hence, a mass-based allocation would provide similar results.

The software used was Thinkstep GaBi 6.2, and the primary database used was the Ecoinvent 3.2 Cut-Off System Model (Wernet et al. 2016). However, a small number of non-Ecoinvent processes were adopted when appropriate Ecoinvent data were not found (see further information in Sect. 2.2).

2.2 Life cycle inventory

The foreground data for this article was provided by the InteSusAl project, part-funded under the European Union Framework Programme 7. Additional data was provided within the project by CPI, who ran a fermenter system in Teesside, UK. As the fermenter is a closed, controlled environment, the actual location should not impact the operation. CPI produced data on lipid extraction. Since the facility itself did not convert the extracted lipids into biodiesel, the LCI data for the transesterification and use phases, data was converted from the Ecoinvent database and US Life Cycle Inventory (LCI) Database (National Renewable Energy Laboratory 2012). This data is listed in the supplementary materials Tables S1 and S2. It should be noted that the Ecoinvent models used were “Allocation, cut-off by classification”, based on the “recycled content” or “cut off” approach. This methodology provides no environmental benefits to the product producer for recycling but instead incentivises the use of recycled products, which come with zero burdens.

Within the InteSusAl project, the majority of the microalgae production side of the overall concept was constructed and operated in at the Necton S.A. facility in Olhão, Portugal. The system is comprised of four TPBR systems, each with a 15,000 L capacity (total capacity of 60,000 L). Three 1000 L fermenters were operated in Portugal, with a fourth 1000 L fermenter running in Wilton, UK. Growth productivity trials were conducted from October 2015 to June 2016 for Chlorella protothecoides (fermenters), Phaeodactylum tricornutum (TPBR) and Nannochloropsis salina (TPBR). During these trials, the required chemicals, energy use and microalgae volume produced were measured. The CO2/O2 gas exchange system, water recycling and glycerol recycling systems were not implemented. Thus, the LCA models do not include the processes of CO2/O2 gas exchange or water recycling. It is important to note that as the trials ran from 2015-2016, whilst the analysis is based on real data, productivities of facilities (including the Necton. S.A. facility in this work) have since improved. 

In some cases, appropriate Ecoinvent data did not exist for some background data, and therefore models from other databases were used. However, it is acknowledged that this is not ideal due to differences in the databases and underlying methodologies. The processes used and respected databases are demonstrated in Table 1. The whole life cycle inventory data for the article is provided in Tables S1–S10. Within the supplementary tables, it is made clear which data is primary data, and which data is secondary data, along with assumptions used. All the upstream data (i.e. before transesterification of algae) is primary data, whilst the sources for secondary data are also given in the supplementary materials.

Table 1 Processes used for the background system. A full life cycle inventory is available within the supplementary materials
Full size table

An issue highlighted post analysis was the lack of an esterification process within the models due to the use of a modified NREL model for soy biodiesel. This esterification process is required to eliminate the free fatty acid within the algae oil (Petchsoongsakul et al. 2017). An examination of esterification models within Ecoinvent for a number of feedstocks showed the impacts of this stage to be negligible compared with the growth and harvesting stages, and therefore the process was eliminated from the model. This is also indicated by the low impacts of the processing stage, as detailed in Sect. 3.2 (Fig. 6).

2.2.1 Proxy processes

In the case of potassium dihydrogen phosphate (CAS: 7778–77-0), due to a lack of access to industrial data, a proxy from Thinkstep was used within the models. The proxy model was tetrapotassium pyrophosphate (CAS: 7320–34-5) based on the Thinkstep trisodium phosphate process. The logic for the use of this proxy was based on comparing information from Bellussi et al. (2000) and the documentation for the GaBi tetrapotassium pyrophosphate model, which is based on a trisodium phosphate model. The Thinkstep LCI for this process was used to create a new process; please note, the databases were not mixed.

2.2.2 Transport impacts

Where possible, “Market” processes were used from Ecoinvent. These include data on the average transport impacts of each individual product. In some cases, such as the “Thinkstep data on-demand” processes, there was no “Market” data. In these cases, the transport used within Ecoinvent for fodder yeast was used. The motivation was that the yeast industry is primarily based in China, which provided an extreme case for the transport impacts of a microalgae production facility based within the EU.

2.2.3 Grid electricity model (scenario A and scenario B)

The grid electricity used was a European wide model which models average electricity mix for the European countries. The reason for focussing on this, rather than using a Portuguese electricity mix (which is more realistic for the facility) is that to use a country specific model would essentially make this article into a proxy for comparing electricity grid mixes and would also mean that the models could not be compared with other Algae Cluster work.

The electricity model for scenario A was the electricity mix for 2012. Whilst it may seem strange to use such an old electricity mix, there are two reasons for this. For All-Gas, InteSusAl and BioFAT, it was extremely important that the three projects were comparable, hence, the electricity mix available when all three projects commenced was used. The means that the work undertaken within the All-Gas and BIOFAT EU microalgae projects may be compared with this work if somebody is undertaking a meta-analysis of real facilities in the future. Additionally, the few papers which consider real-world data within algae facilities range from 2013 to 2020; therefore, by giving such a wide range of electricity values, these results can be considered in terms of the other scientific literature (although we accept that methodological differences would make such a comparison unwise and purely indicative). The 2012 electricity mix is based on the EU-27 grid mix from the GaBi Professional 2016 edition 3 database. Secondly, showing the impacts of electricity in 2012 and 2020 shows how dependent the emissions of the facility are due to electricity, and not purely due to the processes of the facility.

As the modelling work within InteSusAl and All-Gas was undertaken prior to 2020, a hypothetical electricity grid mix was created. The EU-27 + UK 2020 grid mix model for scenario (B) was based on the data from National Renewable Energy Action Plan (NREAP) (Beurskens 2013; Beurskens et al. 2011) for each EU country as well as their nuclear programmes. There were two sets of NREAP data used: the data for the EU-27 + UK in 2010 and their targets for 2020. The model began with the individual 2010 EU-27 country electricity mixes from the Thinkstep Professional Database. The grid mix within this database is divided into several categories of the generation source. These were aligned with NREAP 2010 categories for electricity mix. Most of these align well. For some, the NREAP categories were condensed as follows (numbers in bracket shows the specific NREAP category number):

  • “(07) [%] percentage power from biomass (solid)” covered both “Solid biomass” and “Bioliquids”

  • “(10) [%] percentage power from hydro power” covered “Hydropower < 1 MW”, “Hydropower 1 MW—10 MW”, and “Hydropower > 10 MW”

  • “(11) [%] percentage power from wind power” used “Onshore wind” and “Offshore wind”

There is no concentrated solar power model within the Thinkstep Professional database or the Ecoinvent database. Therefore, a worst-case assumption was taken, and “(12) [%] percentage power from photovoltaics” was used.

With these assumptions, the NREAP 2010 data was mapped to the Thinkstep Professional Database for individual EU-27 country grid mixes. These mostly agreed with the Thinkstep Database models. With the 2010 data mapped, the 2020 NREAP data was then mapped onto the modified Thinkstep grid mixes, taking into account the impacts of energy efficiency measures reported in the NREAP, as a percentage of total consumption. Following this, the nuclear plant capacities were also mapped. The percentage increase or decrease in renewable and nuclear generation was calculated and mapped onto the remaining fossil generation, reducing or increasing the fossil generation balance with the expected demand, to take account of the renewable/nuclear generation increase or decrease. Finally, countries that export power (France and Sweden) due to their renewable and nuclear targets being greater than their electricity demand were given credits according to the mean mix of the rest of the EU-27 + UK. Using these new grid mixes, each country’s total expected energy production was used to create a mean energy mix per kWh for the whole of the EU-27 + UK.

2.2.4 PV model (scenario C)

The PV LCA model used a PV Syst-based model for a 227 kWp rated solar farm based at the Necton site in Olhão, which would produce 409 MWh in the first year of operation. This model was based on a system with 880 “Q.PRO-G3 260” polycrystalline silicon photovoltaic modules (produced by Hanwha Q Cells) with 10 “Conext CL 20000E” inverters (produced by Schneider Electric). Using this information, an LCA model was created within GaBi by modifying existing Ecoinvent models for photovoltaic systems in Portugal. This model showed the GWP100 impact of electricity produced by photovoltaics in Olhão to be 0.05 kg CO2eq/kWh, as opposed to the EU-27 grid average electricity (2012), which emitted 0.5 kgCO2eq/kWh. This means electricity from photovoltaics had a climate change impact of only 10% of that of grid electricity.

2.3 Life cycle impact assessment (LCIA) and energy analysis

In this LCA study, ReCiPe (Goedkoop et al. 2009) and IPCC AR5 (GWP100 and GWP20) (Tredici et al. 2015) were implemented to assess the life cycle environmental impacts. The latter were used due to the importance of climate change impacts in biofuel life cycles and could give the opportunity for a detailed analysis of the climate change impact category under different characterisation factors (CFs) and time horizons. In addition to the Algae Cluster agreed impact categories (Bradley et al. 2015), the Global Temperature Potential (GTP) for 20, 50 and 100 years was also considered. The AR5 impact categories database within GaBi was compiled by the authors and Thinkstep (now part of Sphera Solutions). As described earlier, the model used data from some GaBi Professional-based data, which did not contain data on agricultural or urban land use. Therefore, this was not included within the impact categories.

Energy analysis, which is considered important for microalgae biofuels, is usually performed by calculating the net energy ratio (NER) (Passell et al. 2013). NER is defined as the sum of the direct energy consumed through the production of a fuel, divided by the energy content of the produced fuel (lower heating value, LHV) (Rajaeifar et al. 2013). In the case of this study, the direct energy used consists of the measured energy used for cultivation and harvesting and the energy for the hypothetical extraction and transesterification. The output was considered as microalgae biodiesel with an energy content of 38 MJ/kg (Delrue et al. 2012).

2.4 Sensitivity and uncertainty

A sensitivity analysis was undertaken for the operational model, varying input variables within the cultivation, harvesting and processing phases by 5%. This enabled the major sources of impacts to be clearly identified and also understood, for which input uncertainty would have the greatest impact on the end results. The uncertainty was undertaken using the existing sensitivity tools within GaBi. The full list of parameters is within Table S23, supplementary materials.

2.5 Validation

Validation of the approach and models was undertaken by replicating the results of Passell et al. (2013). The full LCI data for Passell et al. (2013) was provided by the authors of that specific article, which were then replicated within GaBi to understand if our modelling and approaches provided similar figures to the results of Passell et al. (2013). Comparing our results using the Passell et al. (2013) data and Passell et al. (2013) itself, there was a < 4% difference per impact category. The only exception was water depletion. Further investigations of the water data showed issues with the way with which this particular version of GaBi accounted for water. Thinkstep was alerted, and this issue is resolved in future versions of GaBi.

3 Results and discussions

The current LCA was performed to compare microalgae biodiesel production under different energy regimes and conventional petroleum-derived diesel fuel. To ensure the assessments between all the three Algae Cluster projects were comparable, a harmonised methodology was defined in Bradley et al. (2015) and was briefly presented in Sect. 2. Also, the LCA results fell into two main parts: the infrastructure model, based on the construction of the facility, and the operational model, which is based on the operation of the facility.

3.1 Infrastructure model

Detailed information was provided from the Necton and CPI sites on the construction of the facilities and equipment (see supplementary materials, Tables S11 and S12). This information was then converted into LCA models. Information on end-of-life (EoL), where appropriate Ecoinvent models did not exist, was gathered from a range of sources. Specifically, polyethylene terephthalate (PET) and high-density polyethylene (HDPE) recycling was based on Franklin Associates (2011); poly (methyl methacrylate) (PMMA) recycling was based on standard industry practices, and data within the Advanced THermal Analysis laboratory (ATHAS) (Wunderlich 1995); polypropylene (PP) data was from Hardwick (2015); and polyvinyl chloride (PVC) data was gathered from Stichnothe and Azapagic (2013).

The top five sources of AR5 GWP100 impacts were polyvinyl chloride (PVC), polymethyl methacrylate (PMMA), concrete, fibreglass-reinforced plastic (FRP), aluminium and stainless steel. In terms of climate change, the timescale considered is important; for example, over a 100-year period, concrete is the third largest impact (14.4% of the GWP100 impact), whereas over 20 years, the third-largest impact is that of FRP (13.7%) whilst concrete is 11.9% of the GWP20 impact. The impacts, including those from the infrastructure, are given in the supplementary materials (see Table 2 as well as Table S13).

Table 2 Environmental impacts for the three microalgae scenarios, with and without infrastructure*, compared with petroleum diesel burned in an engine (including infrastructure)
Full size table

3.2 Operational model

As with many LCA studies, the results of this part of the assessment proved sensitive to the source of electricity used by the facility; hence, three electricity source scenarios for microalgae production were considered: scenario A: EU-27 2012 grid mix; scenario B: EU-27 + UK 2020 grid mix; and scenario C: PV (photovoltaic) powered facility. Scenarios A and B cover the European grid average, whilst C is an LCA model of a hypothetical solar farm designed for Olhão using the solar farm design software PVSyst (PVSyst 2020). The underlying assumptions for these models are detailed in Sect. 2. The results of the three scenarios compared with petroleum-derived diesel fuel are given in Table 2, with the figures presented as percentages within the supplementary materials in Table S13.

The results proved that within scenarios A and B, the primary source of impacts for microalgae production was electricity generation. The use of photovoltaics -as also recommended in Taylor et al. (2013) and Tredici (2010) decreases the non-infrastructure GWP100 impacts by 58.1% compared with scenario A, so that they were 87.8% of those petroleum-derived diesel impacts; however, the photovoltaics still contributed to all midpoint and endpoint impacts, due to the necessary construction of the PV arrays. It should be noted that the construction of PV has been included in the non-infrastructure model, as infrastructure is included in the grid electricity models.

When considering climate change impacts, the impacts increased with the smaller timescale are considered (e.g., higher impacts for GWP20 compared with GWP100). This is due to the different global warming impact of the methane produced within a 100-year period. Most of the methane produced in the life cycle of microalgae biodiesel comes from the energy generation for both the operation of the facility in Olhão itself and for the production of the yeast. Due to its short lifetime in the atmosphere (12.4 years) (Myhre et al. 2013), biogenic methane has an impact of 84 times that of CO2 over 20 years, reducing to 28 times over 100 years. This highlights the question of whether short- or long-term timescales should be considered for climate change-based decision-making and their balance (Cooper et al. 2019; Pierrehumbert 2014; Shoemaker and Schrag 2013). A full breakdown of the contributors to GWP100 and GWP20 is given in the supplementary materials (Tables S14 and S15), as produced using GaBi. It is important to note that the AR5 values were used within this study, as this work was undertaken before the 2021 AR6 figures were published. 

Global Temperature Potential (GTP) was also considered in this study. Unlike GWP, this metric accounts for the impact of the temperature of the planet normalised against CO2 (Cherubini et al. 2016) rather than changes in levels of radiative forcing. For scenario C’s operational emission, the GTP varied from 7.02 × 10−2 (GTP 100-year) to 7.32 × 10−2 (GTP 50-year) and 8.91 × 10−2 (GTP 20 year), implying 26.9% reduction from GTP-100 to 20. Less detailed figures for the operational impacts, including GWP10, GWP50 and GTP 10, are presented in Fig. 5 (more information could be found in supplementary materials, Table S16).

Fig. 5
figure 5

GWP (a) and GTP (b) of scenario C. (Notes: a) this uses data from both Fig. 8 and SM.16, which are within the AR5 supplementary data (Myhre et al. 2013); b) the plots only use data for CO2, CH4 and N2O; c) compared with the full LCA models, these GHGs contributed to 99.46% (GWP20) and 99.32% (GWP100) of the GWP of scenario C)

Full size image

The above shows that to interpret results correctly, it is important for there to be a dialogue between LCA practitioners and climate science. When simplifying the complexity of climate change for policymakers, it is understandable that GWP100 is used; however, the reality is more complex; hence for decisions around strategic investment into new technologies, a range of methods should be considered. Also, there are further areas to consider, such as the interaction of chemicals (for example, methane and aerosols) (Drew et al. 2009) and the cumulative impacts of GHGs on the climatic system (Cherubini et al. 2016), which are not counted within LCA. Data for all GHGs considered are provided in Table S17 to allow the reader to employ their own climate change CFs.

Of concern for all scenarios are the areas which show increased impacts when compared with petroleum-derived diesel, which, when including infrastructure, is all of them. For example, even without infrastructure, water depletion [m3] for PV-powered microalgae biodiesel was 2350% greater than petroleum diesel. Using PV reduced 12 mid-point impact categories, whilst others such as ecotoxicity, toxicity, eutrophication, metal depletion and ozone depletion were increased. It is noted that scenario C doubles ozone depletion from the operation. Within the Ecoinvent data used, the ozone depletion impact from PV was found to be from tetrafluoroethylene use in cell manufacturing. However, the PV models within another LCA database, Thinkstep Professional, show no tetrafluoroethylene use. As this is a major impact on the models, further investigation is necessary, and further evidence from real industrial data is required.

The model itself was divided into four main stages: (a) growth, (b) harvesting, (c) biodiesel production and (d) Use. The LCI was divided into these parts, ensuring that the GaBi models would divide the impacts.

The cultivation phase created a significantly greater climate change contribution and Ecotoxicity impact, as shown in Fig. 6. A larger set of impacts are detailed in the supplementary materials, Fig. S1 and Table S18.

Fig. 6
figure 6

Comparison of impacts in the four main phases of microalgae biofuel production and use. (Notes: Left figure shows GWP100 and GWP20; right figure shows ecotoxicity)

Full size image

For scenarios A and B, electricity was the primary source of operational impacts. If PV-derived electricity is used (scenario C), this reduces this source to the second-largest (or joint first source of impacts), with the major or joint source the production of yeast extract for the fermenters (except in the case of terrestrial ecotoxicity and metal depletion where electricity has still the most significant impact). After these two impact sources, other contributors to impacts were freshwater and potassium pyrophosphate. Concerning AR5 GWP100 impacts, yeast was responsible for 68.2%, PV electricity 15.7%, potassium pyrophosphate 8.9% and freshwater 2.7%.

In all the impacts considered, these were the most relevant inputs, with methanol, phosphoric acid, sodium hypochlorite and hydrochloric acid each minor contributor to most impacts. Natural gas and organic solvents used in the transesterification process yielded minor contributions to ozone depletion (full data in supplementary materials, Tables S16 and S19 to S21 for scenario C). From this, it shows that engineers need to optimise the production systems and have a greater understanding of the impacts of feedstocks (particularly yeast extract if used). For example, could a microalgae biofuel facility use more sustainable yeast extract (potentially co-located on-site) and produced using electricity from renewable energy? In this case, the analysis performed during the course of this study showed that a further reduction of operational GWP100 could be achieved. By taking a critical eye and perhaps direct involvement in the production of the chemical feedstocks for microalgae biofuels, the industry can reduce impacts. For small systems, this is impractical, but for the 10–100 hectares of bio-refineries considered by industry in the near future, this could well be a viable option. Of course, yeast itself can produce a wide range of lipids (Parsons et al. 2018), so one could question the very logic of using yeast as a feedstock instead of using the yeast directly. This question requires a comparative analysis of fuel from algae and further work to answer that question.

It is important to note that the current LCA is based on the assumption that the glycerol fed to the fermenters (in addition to the internally recycled glycerol) is industrial waste. Using fresh glycerol will give very different results, which would not be favourable for microalgae biofuels.

3.3 Net energy ratio (NER)

If the energy content of the co-products is taken into account, the NER would reach 0.99, i.e. 0.99 MJ of electricity/gas was used for 1 MJ of biodiesel produced; however, if the energy through the whole value chain, including the feedstock chemical production, is considered, then the NER is calculated as 1.03. These are not ideal results but should be viewed in comparison with other technologies. Work by Brandt et al. (2015) shows that the NER can vary dramatically per oil field (in terms of crude oil, not diesel) from 0.5 to 0.01, dependent on location and technology. In terms of actual petroleum diesel, the average NER for US petroleum diesel is 1.20, although, with the increasing use of oil shale and tar sands, this has increased within the USA to 1.65 (Shirvani et al. 2011). This shows the InteSusAl system NER is comparable with other poorly performing fossil fuel extraction/production methods.

3.4 Sensitivity and uncertainty

The sensitivity analysis results are given for all three scenarios within the supplementary materials, Tables S22 to S27. This shows a similar pattern to the results previously given, with the system most sensitive to electricity and yeast extract. For scenario A, an increase of 5% in the electricity use would result in an increase of 3.2% of the GWP100. An increase of 5% of the yeast input would result in a 1.45% increase in AR5 GWP100. Within scenario C, the yeast input becomes the dominant factor; taking the example of AR5 GWP100 again, yeast will vary the final result by 3.42% if it is varied by 5%. In contrast, electricity will only cause a 0.77% change in the final result.

3.5 Implications of the system modelling

Clearly, both the operational and infrastructure impacts must be combined when compared with fossil fuels. In terms of InteSusAl, the infrastructure impacts will be larger than an established biorefinery which is optimally set out. To merge data, the impacts of the infrastructure were equally distributed, assuming 38 MJ energy content for microalgae biodiesel, 15.27 tonnes hectare−1 year−1 production of microalgae and 20 years lifetime. The 15.27 tonnes hectare−1 year−1 production was based on the extrapolated productivity from the photobioreactor and fermenter trials. This showed, for all scenarios, that microalgae biofuels do not compare well with fossil fuels in such a production quantity. If the productivity of > 25.6 tonnes hectare−1 year−1 was achieved, which is a reasonable level, then a PV-powered system (scenario C) would be on a par with fossil fuels in terms of climate change (GWP100). A productivity of 31.4 tonnes hectare−1 year−1 would lead to equivalence in terms of ozone depletion, but 313.2 tonnes hectare−1 year−1 would be needed for equivalence with petroleum diesel in terms of eutrophication. This shows that the InteSusAl system is within reach of petroleum diesel in some areas of environmental impacts, such as ozone depletion, terrestrial ecotoxicity, and climate change, but not the majority of others especially impacts such as freshwater ecotoxicity, freshwater eutrophication, various types of depletion and land use.

3.6 Comparisons with other works

A comparison with other work, as previously mentioned, is difficult without a meta-analysis due to different methodological choices. In terms of Maga (2017) (All-Gas project), they performed analyses mostly followed Bradley et al. (2015), as does this article, although there are some divergences. Comparing these two similar studies (Table 3), it shows that in terms of AR5 GWP100 climate change impacts, microalgae-derived biogas has much lower impacts than microalgae-derived biodiesel from the InteSusAl system. However, due to various emissions from the All-Gas water treatment works, in terms of other impact categories considered, the InteSusAl and All-Gas facilities are comparable in impacts per MJ of fuel. It is important to note that the sewerage system included clean water as a co-product.

Table 3 Comparison of impacts between this article (Scenario A) and Maga (2017) (All-Gas project)
Full size table

This shows the complexity in terms of comparing systems, as whilst climate change favours the system within Maga (2017), marine eutrophication and terrestrial acidification favour the InteSusAl system. One important additional consideration is that food, feed, pharmaceutical and nutraceutical co-products are difficult to sell from systems with human sewerage for regulatory reasons.

Other works (see Table 4) have not worked closely in the same way as Maga (2017) and this article; therefore, it is difficult to draw conclusive conclusions without a detailed meta-analysis. For example, Branco-Vieira et al. (2020b) and related papers include excellent and detailed analyses, but the electricity grids considered by them are different (EU vs Chilean), and more substantial analyses are necessary on the underlying data to draw a sensible comparison. However, the results within Branco-Vieira et al. (2020b) are at least one order of magnitude higher than this article for all impact categories where the same units are used (climate change, fossil depletion, freshwater ecotoxicity, freshwater eutrophication, marine ecotoxicity, ozone depletion, terrestrial acidification, terrestrial ecotoxicity, water). The reasons for these differences are not immediately clear, but further analysis is necessary.

Table 4 Impacts for three impact categories for those papers which used energy as a functional unit. Note that the data from Branco-Vieira et al. (2020b) could not be compared as it used different units for the impact categories. Liu only addressed climate change. Also, Liu et al. (2013) and Branco-Vieira et al. (2020b) were cradle-to-gate, but the CO2 emissions from burning would be biogenic, hence their inclusion within this table (although clearly photochemical oxidant formation and particulate matter would be impacted by tailpipe emissions)
Full size table

Two articles have not considered the energy production and purely the production of 1 kg of algae ((Pérez-López et al. 2017) and (Onorato and Rösch 2020)), whereas Pérez-López et al. (2017) show extremely high impacts per kg of algae dry weight (DW) produced, which when converted to a per MJ functional unit vary (depending on technology and season) to between 6 and 11 times that of scenario B (the most relevant scenario, as both use grid electricity close to 2020). Human toxicity, freshwater ecotoxicity, marine ecotoxicity and terrestrial ecotoxicity are all comparable with this article’s results, although it is important to note that with the cut off at 1 kg of algae, there are the energy requirements of drying and transesterification to also include, although as shown in, this was comparatively a small element. Interestingly, extremely low impacts for ozone depletion, terrestrial ecotoxicity and photochemical oxidant formation (Pérez-López et al. 2017) compared with this work. If we consider Onorato and Rösch (2020), then we find a range of different impacts based on technology. In terms of the UHT (the A4f system), we find climate change impacts for 1 MJ of algae of 230% of scenario B. Except for freshwater eutrophication and ionising radiation, all common impact categories are between 100 and 370% of this work. This is interesting, as the A4f system was similar to the photobioreactor elements of the InteSusAl system, as the two organisations were related. Fermenter systems are considered by Siqueira et al. (2018) using lab scale data and show results for climate change within the same order of magnitude as reported by Branco-Vieira et al. (2020b).

Ultimately, all systems within the scientific literature vary, and the system within this article is different yet again in terms of horizontal tubular photobioreactors and a fermenter working as single system. In terms of drying, the system considered within this work uses a centrifuge, as does Passell et al. (2013) and Pérez-López et al. (2017). However, Branco-Vieira et al. (2020b) use flocculation and centrifuge, Onorato and Rösch (2020) use centrifugation and spray dryer, Liu et al. (2013) use dissolved air floatation and decanter centrifugation, whilst Beal et al. (2015) investigate a range of options (centrifugation and solvent extraction (POS Biosciences), thermochemical conversion (Valicor), hydrothermal liquefaction (PNNL), catalytic hydrothermal gasification (Genifuel), combined heat and power, wet extraction (OpenAlgae) and fermentation). Maga (2017) does not dry the biomass, as it is moved towards anaerobic digestion for the production of biogas. In terms of the downstream technologies applied; the majority of current work which includes the fuel production uses transesterification, although Liu et al. (2013) use hydrothermal liquefaction and Maga (2017) uses anaerobic digestion. Ultimately, comparisons with previous work show that the differences in technology, assumptions and methodological choices require a complete in-depth analysis as a separate article.

3.7 Limitations

The major limitation of this work is the age of the data, which was from a project which ran from 2010 to 2015. Since then, there have been substantial changes in the industry, including the facilities which exist at Necton. One noticeable change is that Necton has constructed a photovoltaic system at their facility to supply energy and hence reduce the environmental impacts of their production. In terms of advantages, this paper does contain primary pilot-scale data, which facilitates a realistic view of microalgae production.

3.8 High value product

Algae contain many different types of high-value products (such as β-carotene, astaxanthin, fucoxanthin and β-glucans). The production of these can be considered to be “Conventionally Supplementary Systems”, as described by Kiatkittipong et al. (2009). The process of extracting the lipids will effectively waste some high-value products, whilst the waste can be expected to be a mixture of unextracted lipids, polysaccharides or fibre, some proteins and minerals. Changes could be made to the downstream processes to extract further high-value products at the cost of producing less biodiesel. If the extraction of high-value products leads to low enough impacts compared with current industrial processes, then this may reduce the impacts of the whole system to the point where environmentally microalgae-derived biodiesel is competitive with petroleum-derived diesel. However, this is an open question which requires further investigation. Furthermore, a more detailed analysis of high-value products could lead to the conclusion that microalgae should not be used as a feedstock for fuels. This current article does not contain enough data to draw a conclusion on this question.

4 Conclusions

This life cycle assessment study demonstrates the impacts of a functioning case study microalgae production facility, which used a mixture of heterotrophic and autotrophic growth systems to generate microalgae for biodiesel production.

In terms of the energy used by the system, a net energy ratio of ~ 1 is found. This shows the facility’s net energy ratio was similar to poorly performing oil fields. Nevertheless, since these results are for a small pilot site, they are encouraging.

The assessment shows that, when infrastructure is included, microalgae-derived biofuels are not yet favourable over petroleum-derived fuels in terms of climate change impacts over 100 years, but this becomes worse over shorter timescales. For Fifth Assessment Report based Global Warming Potentials over 100 years, the scenarios range from 2.56 × 10−1 kg CO2eq/MJ (scenario A, 2012 electricity grid mix) 2.38 × 10−1 kg CO2eq/MJ (scenario B, 2020 electricity grid mix) and 1.48 × 10−1 kg CO2eq/MJ (scenario C, photovoltaic electricity). These compare with petroleum-derived diesel, with an impact of 8.84 × 10−2 kg CO2eq/MJ. Additionally, when considering infrastructure, for all other environmental impacts, further improvements are necessary. For example, for scenario B, the impacts of freshwater ecotoxicity for microalgae-derived biodiesel were 1320% times that of petroleum-derived diesel. For freshwater eutrophication, the value was 1610% that is highly concerning. If these issues are shown to be common within algae facilities, then a serious rethink must be made on microalgae-derived fuels.

In terms of relevance to industries, this work shows how the specific systems studied within this work do not have a future as biofuel production systems. Instead, other high-value products should be considered, although to understand which one is appropriate, there will be a necessity of associated LCA and techno-economic analyses.

In terms of recommendations, aside from moving from biofuel to other products and ensuring that low carbon energy is used, the productivity of systems needs to be significantly improved to lower the impacts of microalgae production. As also stated previously, there needs to be careful consideration of the feedstocks used, to ensure that they also have as low impacts as possible.