Article Request Page ASABE Journal Article
NIR Reflectance and MIR Attenuated Total Reflectance Spectroscopy for Characterizing Algal Biomass Composition
Y. Ge, J. A. Thomasson
Published in Transactions of the ASABE 59(2): 435-442 (doi: 10.13031/trans.59.11396). Copyright 2016 American Society of Agricultural and Biological Engineers.
Submitted for review in June 2015 as manuscript number ITSC 11396; approved for publication by the Information, Technology, Sensors, & Control Systems Community of ASABE in November 2015.
The authors are Yufeng Ge, ASABE Member, Assistant Professor, Department of Biological Systems Engineering, University of Nebraska-Lincoln, Lincoln, Nebraska; J. Alex Thomasson, ASABE Member, Professor, Department of Biological and Agricultural Engineering, Texas A&M University, College Station, Texas. Corresponding author: Yufeng Ge, 209 Chase Hall, University of Nebraska-Lincoln, Lincoln, NE 68583; phone: 402-472-3435; e-mail: email@example.com.
Abstract. Algae have long been investigated as a potential feedstock for renewable energy production. There is growing interest in rapid and cost-effective techniques for characterizing algal biomass composition relevant to biofuel production. The objective of this study is to investigate the usefulness of near-infrared (NIR) and mid-infrared (MIR) spectroscopy in determining neutral lipids, crude protein, gross calorific value (GCV), and ash content in lyophilized green algae samples ( and ). NIR spectra were acquired in diffuse reflectance mode, and MIR spectra were acquired in attenuated total reflectance mode. Partial least squares regression models were calibrated and validated to correlate laboratory chemical data with absorption spectra in the NIR and MIR regions. The results show that, for both spectral regions, crude protein can be predicted most accurately, with validation R2 higher than 0.85, followed by neutral lipids (R2 > 0.70). Validation accuracy for GCV and ash is somewhat lower (R2 from 0.55 to 0.70). Large ash content, with its diverse chemical composition, was determined to negatively impact the prediction accuracy of the NIR and MIR models. It is concluded that both NIR and MIR have the potential to characterize algal biomass composition and to support the future algae-based biofuel and bioproducts industry.
Keywords.Algae, MIR, NIR, Partial least squares regression, Renewable energy, Vibrational spectroscopy.
Algae have long been investigated as a potential feedstock for renewable energy production. The advantages of algae for energy production are widely recognized: (1) they exhibit higher productivity than most terrestrial oil plants, (2) they do not compete for arable land with food and feed crops, (3) they can grow in brackish water and potentially recover waste nutrients, and (4) they would reduce greenhouse gas emission and mitigate the concern about global climate change associated with fossil fuel burning (Bryant et al., 2012; Chisti, 2007).
Algae could be used in versatile ways for renewable energy production. Algal neutral lipids (triglycerides or TAG) can be converted to biodiesels (fatty acid methyl esters or FAME) through transesterification. Alternatively, whole algal biomass or lipid-extracted algae (LEA) can undergo chemical or thermal conversions to produce biogas and bio-oil. The use of algae for non-fuel bioproducts has also been widely discussed (Guedes et al., 2011; Harun et al., 2010). Algae produce a variety of high-value bioproducts such as pigments, ß-carotene, protein, and long-chain polyunsaturated fatty acids that are widely used in dietetics, therapeutics, cosmetics, and aquaculture (Mata et al., 2010). Of particular economic importance are omega-3 oils (eicosapentaenoic acid or EPA, and docosahexaenoic acid or DHA), which are associated with general health and were the major motivation of algae production in the past. In addition, LEA could be used as a high-nutritive primary or supplemental feed for livestock due to its high protein and mineral contents (Harun et al., 2010; Bryant et al., 2012). Overall, algal biofuel could play a significant role in meeting the national goal of 21 billion gal (79 billion L) of advanced biofuels by 2022 (U.S. Congress, 2007), and many value-added bioproducts would increase the economic feasibility of algal biofuel.
As research efforts on large-scale algae-based biofuel production are likely to expand in the future, there is growing interest in rapid and cost-effective techniques to characterize algal biomass composition (such as neutral lipid and protein contents). Throughout the algae-to-biofuel research and production chain, there are multiple points where these techniques would play significant roles. They can serve as high-throughput in vivo screening tools to identify high lipid-producing strains. They could also be flexibly integrated into cultivation processes for pond monitoring, automation, and management (Gitelson et al., 2000; Richardson et al., 2012). At the inlets of biorefineries, these techniques would be useful to characterize the quality of incoming feedstock and determine the most suitable conversion processes. Finally, they could also be applied to dry algal biomass in applications such as determining protein and mineral contents for blending animal feed rations.
Nile red staining and fluorescence (NRf), initially put forward by Greenspan and Fowler (1985) and further investigated and developed by many researchers (Lee et al., 1998; Elsey et al., 2007; Chen et al., 2011; Doan and Obbard, 2011), is a promising high-throughput sensing technique for neutral lipid determination. However, concerns have been raised about the unevenness of dye uptake across algal strains that have differences in cell wall thickness and rigidity. Other new and emerging techniques for rapid lipid quantification include pigment-related optical measurement (Solovchenko et al., 2011), liquid-state 1H nuclear magnetic resonance (Davey et al., 2012), and colorimetric methods based on copper soaps of fatty acids (Wawrik and Harriman, 2010). More recently, digital image analysis has been used for algal biomass quantification (Murphy et al., 2013; Sarrafzadeh et al., 2015).
Near-infrared (NIR, 1000 to 2500 nm) spectroscopy and mid-infrared (MIR, 4000 to 400 cm-1) spectroscopy are two techniques that may have great potential for characterizing algal biomass. Like other biological samples, algal cells primarily consist of oil, protein, and carbohydrates. These molecules are abundant with C–H, N–H, O–H, C=O, C–C, and C=C bonds that cause characteristic vibrational absorption features in the MIR region, and overtones and combinations of these fundamental bands show up in the NIR region. Both techniques have been used to analyze a wide variety of agricultural and biological samples, including food, feed, and plant materials (Shenk et al., 2008). Good results have been achieved for both oil and protein (Orman and Schumann, 1991; Pérez-Vich et al., 1998; Delwiche and Hruschka, 2000; Kovalenko et al., 2006). Several studies have applied the technique to algae compositional analysis, mainly focusing on sensing physiological changes of algae in natural environments (e.g., lakes) in response to nutrient deprivation (Stehfest et al., 2005; Wagner et al., 2010; Dean et al., 2012).
Laurens and Wolfrum (2011) appear to be the first to investigate NIR and MIR spectroscopy for algal biomass characterization in the context of biofuel production. Their study focused on lipids and used exogenously spiked lipids as the variable of interest. Therefore, the implications of their study may be limited when naturally occurring intracellular lipids are considered. In another study, Mulbry et al. (2012) analyzed a set of algal turf scrubber samples with both NIR and MIR reflectance spectroscopy for ash, total and mono sugars, nitrogen (N), phosphorus (P), and total lipids and fatty acids.
The objective of this study is to investigate the usefulness of NIR and MIR spectroscopy as rapid and cost-effective techniques to characterize the biomass composition of algae samples. In particular, we are interested in the quantification of neutral lipids, crude protein, gross calorific value, and ash content.
Materials and Methods
A total of 99 algae samples were obtained from the algae production facility of Texas A&M AgriLife Research in Pecos, Texas. All samples belong to the green algae genera Nannochloropsis and Nannochloris. These samples were obtained in large-volume open-pond raceways (several thousand liter capacity) from 2010 to 2012, with the goal of evaluating various procedures for cultivating algae in outdoor environments. As a result, they were subjected to different growing conditions and dewatering and harvesting methods (centrifugation and coagulation). After harvest, the algae were lyophilized with an average moisture content ranging from 1% to 4% for long-term storage. For analysis, the samples were extracted from storage and split into two parts: one for laboratory chemical analysis and the other for NIR and MIR spectral measurement.
The biomass composition of interest in this study included neutral lipids, crude protein (CP), gross calorific value (GCV), and ash content. The samples were first analyzed for ash-free dry weight (AFDW) by combustion. Gas chromatography (Agilent 6890N) mass spectrometry (Agilent 5975B) with a flame ionization detector (GCMS-FID) was used to resolve and measure the FAME components from C14 to C24. Neutral lipid was then calculated as the sum of all FAME components. GCV was measured with an isoperibol bomb calorimeter (model 6200, Parr Instrument Co., Moline, Ill.). CP was measured with the combustion method for total N and then scaled by 6.25. Ash content was derived as 100% less AFDW.
An AgriSpec spectroradiometer (Analytical Spectral Devices, Boulder, Colo.) was used to acquire reflectance spectra of the algae samples in the NIR range. Many of the lyophilized algae samples were flaky and were therefore ground in an agate mortar to reduce the particle size and minimize particle size effects on light scattering before analysis. However, because sample particle size was not measured or controlled, its effect on light scattering could be a source of variability in the NIR spectra (Chen and Thennadil, 2012). The spectral range used in this study was 1000 to 2500 nm, and the spectral sampling interval was 1 nm. About 1.0 g of dry mass was placed in a borosilicate glass petri dish and scanned using the AgriSpec’s high-intensity mug lamp. A Spectralon panel was used for white referencing, which was performed after every four or five samples. Each reflectance spectrum collected was ten co-added instantaneous internal scans. Two spectra were collected and then averaged for each sample, with the petri dish turned 90° to compensate for the directional effect of diffuse reflectance. Reflectance (R) was converted to absorbance (A) by A = log10(1/R).
An FT-IR spectrometer (model 6700, Thermo-Nicolet, Madison, Wisc.) was used to acquire attenuated total reflectance (ATR) spectra in the MIR region (4000 to 525 cm-1). ATR has the advantage of eliminating the need for sample preparation (i.e., diluting samples in an IR-transparent substrate such as KBr). About 10 mg of biomass was placed on the spectrometer’s Smart Orbit accessory and pressed against the ATR crystal with a constant force of 200 N. The ATR crystal is a single reflection diamond crystal with an incidence angle of 45°. Other important configuration parameters of the spectrometer are as follows: beam splitter = CsI (cesium iodide), detector = thermo-electrically cooled DTGS (deuterated triglycerine sulfide) CsI, spectral resolution = 4 cm-1, sampling resolution = 1.92 cm-1, optical (moving error) velocity = 0.632 cm s-1, purge gas = none, co-added interferograms per sample = 32, and signal processing = automatic atmospheric suppression. The background ATR spectrum was collected after every sample measurement using atmosphere as the background. Figure 1 shows how the algae samples were scanned with the NIR and FT-IR spectrometers.
Figure 1. Near-infrared (NIR) and mid-infrared (MIR) spectral measurement of lyophilized algae samples. NIR (top) was measured in diffuse reflectance mode, and MIR (bottom) was measured in attenuated total reflectance mode.
The presence of moisture in samples usually has a significant effect on the samples’ NIR and MIR spectra. However, in our study, all samples were lyophilized to very low moisture content (1% to 4%, as previously mentioned) and carefully stored to avoid hygroscopic moisture. Therefore, the effects of varying moisture on the samples’ NIR and MIR spectra were negligible.
The NIR and MIR-ATR spectra of the algae samples were first assessed by relating the major absorption bands to specific chemical constituents. Partial least squares regression (PLSR) was performed to model the four biomass properties with NIR absorbance and MIR-ATR spectral data. Several spectral preprocessing techniques, including standard normal variate, multiplicative signal correction, and first and second derivative using Savitzky-Golay smoothing, were attempted. However, because no substantial improvement was achieved with these preprocessing techniques, results are reported using the original spectral data. Two-thirds of the samples were used for model calibration, and the remaining 1/3 was used for model validation. Model performance was assessed using the coefficient of determination (R2) and root mean squared error (RMSE) between the predicted and laboratory-measured values. In addition, the relative performance difference (RPD, the ratio of the standard deviation of the variables to RMSE) is reported. Multivariate modeling was implemented with the chemometrics package (Filzmoser and Varmuza, 2011) in R statistical software (R Development Core Team, 2013).
Results and Discussion
Biomass Compositions of Algae Samples
Table 1 summarizes the algal biomass properties of interest. Neutral lipid content is usually the primary focus because it determines biodiesel yields. It ranged from 4.1 to 168.2 g kg-1 with a mean of 80.7 g kg-1. This is significantly lower than the published data for Nannochloropsis and Nannochloris (Chisti, 2007). There are two reasons for this. First, the published data include not only TAGs but also other complex oils. Second, the published data are reported on an algal biomass basis, while table 1 is reported on a total mass basis, lowering the percentage of neutral lipids and other variables due to the presence of ash.
Our sample set had high ash content, ranging from 10% to over 83% with a mean of 29%. In contrast to samples described in the literature, our samples came from large-scale production in open-pond raceways with capacities of several thousand liters. The high ash content was attributable to different cultivation and processing conditions. For instance, some batches were cultivated with elevated bicarbonate concentrations in growth media, and the salt was co-precipitated during centrifugation. Excessive aluminum and iron were found in samples that were electro-coagulated with aluminum or steel rods. Sand particles were also common in these samples due to the open-pond raceway conditions, in which soil contaminants were often blown into the ponds by strong winds. In summary, this sample set was different from the clean, pure algae usually obtained in laboratory benchtop processes. It had higher ash content, and the ash had quite different chemical profiles. It should be noted that high ash content of dried algal biomass as a result of open-pond environments has been discussed in the literature (Bryant et al., 2012).
Table 1. Summary statistics of neutral lipid content, crude protein, gross calorific value, and ash content of dried algae sample set (n = 99). Algae Property Max. Min. Mean SD Neutral lipids (g kg-1) 168.2 4.1 80.7 41.2 Crude protein (%) 57.40 6.85 30.08 11.35 Gross calorific value (kJ g-1) 20.45 3.86 16.32 3.80 Ash content (%) 83.8 10.2 29.1 15.4
Results of PLSR Modeling
Figure 2a shows the absorbance spectra of two samples in the NIR region (1000 to 2500 nm): one with low ash content and the other with high ash content. Unlike MIR spectra, in which distinctive bands of different functional groups can be assigned, NIR spectra have mainly weak overtones and combination bands of C–H, N–H, and O–H. Bands related to C–H are observed at around 1210, 1750, and 2310 nm, which are in the second overtone, first overtone, and combination band regions, respectively. Bands associated with O–H are observed at 1420 and 1920 nm. The first overtone of N–H is at around 1500 nm, which is overlapped with the O–H band at 1420 nm. These bands might be useful in predicting neutral lipids and protein. Compared to the low ash sample, the high ash sample exhibited lower absorption intensities in all bands. In addition, the high ash sample exhibited lower overall absorbance at all wavelengths. This is because a large fraction of ash is inorganic salts, which are non-absorbing in the NIR region but strongly scattering. The algae NIR spectra are similar to the spectra reported by Lauren and Wolfrum (2011) and to the spectra of many other biological samples, including food, feed, and plant materials.
Figure 2. Absorbance spectra of two algae samples in the (a) NIR region (1000 to 2500 nm) and (b) MIR region (4000 to 525 cm-1). The solid line is a low-ash sample, and the dashed line is a high-ash sample. Major absorption peaks in MIR, their overtones and combination bands in NIR, as well as their major contributors are identified and labeled.
Table 2. Summary of partial least squares regression model calibration for four algal biomass properties using NIR and MIR spectra. Algae Property NIR Spectra MIR Spectra RMSE R2 RPD Factor RMSE R2 RPD Factor Neutral lipids (g kg-1) 20.1 0.77 2.10 5 25.4 0.63 1.67 5 Crude protein (%) 5.49 0.77 2.10 7 4.58 0.84 2.52 5 Gross calorific value (kJ g-1) 2.08 0.77 2.12 7 1.75 0.84 2.53 8 Ash content (%) 11.1 0.53 1.48 5 10.9 0.55 1.51 3
Figure 2b shows the MIR-ATR spectra of the same two algae samples. Several absorption features resulting from the molecular structure of the organisms can be identified. These include a broad band attributed to O–H and N–H stretching at around 3300 cm-1, two peaks attributed to C–H symmetric and asymmetric stretching at around 2920 and 2850 cm-1, a carbonyl band due to C=O stretching of ester at around 1735 cm-1, amide I (C=O stretching of amide) and amide II (N–H bending of amide) at around 1635 and 1540 cm-1, and a group of overlapped bands from 1200 to 1000 cm-1 attributed to C–O and C–O–C in carbohydrates. The C–H and carbonyl C=O bands are commonly used for lipid quantification, whereas the amide I and II bands are used for protein quantification.
These spectra are in good agreement with the spectra reported by Pistorius et al. (2009) and Laurens and Wolfrum (2011). The low ash sample exhibited lower intensity for all the major absorption peaks except for the O–H band at 3300 cm-1. The low ash sample also exhibited an additional peak at around 870 cm-1, which corresponds to CO32-. This is obviously due to the bicarbonate salts co-precipitated with the algal biomass during harvest. While not shown in figure 2, some samples also showed Si–O bands in their MIR spectra due to sand or soil contamination. In addition to these major peaks, there is an almost undetectable band at 3015 cm-1. This is attributed to cis C=C–H vibration. Its small amplitude indicates that the degree of unsaturation in neutral lipids is quite low. Potentially, this can be a useful band for identifying algae with higher content of long-chain polyunsaturated fatty acids. Finally, the noise from 2300 to 1900 cm-1 was caused by the diamond crystal used in the ATR experiment. This region was excluded from the following quantitative modeling.
The results of PLSR model calibration using the NIR and MIR spectra are given in table 2. For NIR, neutral lipids, crude protein, and GCV showed almost the same prediction accuracy, with R2 of 0.77 and RPD around 2.1. Ash content showed lower accuracy, with R2 of 0.53 and RPD close to 1.5. For MIR, crude protein and GCV had the highest prediction accuracy, with R2 of 0.84 and RPD around 2.5. Neutral lipids showed an intermediate accuracy, with R2 of 0.63 and RPD about 1.7. Ash content again had the lowest accuracy.
Figures 3 and 4 summarize the validation results of the NIR and MIR models, respectively. For NIR, crude protein can be predicted best (R2 = 0.90), followed by neutral lipids (R2 = 0.73) and ash content (R2 = 0.70). GCV was validated with the lowest accuracy, with R2 = 0.57. For MIR, crude protein can still be predicted with the highest accuracy (R2 = 0.85), followed by neutral lipids (R2 = 0.70) and ash content (R2 = 0.68). GCV was again validated with the lowest accuracy. Overall, the results of model validation were similar to those of calibration in table 2, meaning that these models are stable and consistent and can potentially be applied to other algae samples of the same species. Chang et al. (2001) suggested an RPD-based criterion to assess the accuracy and stability of multivariate regression models: RPD > 2.0 indicates a stable and accurate model, 1.4 < RPD < 2.0 indicates a fair model with potential for improvement, and RPD < 1.4 indicates a poor model. By this criterion, the models for crude protein are stable and accurate, whereas the models for all other variables fall into the second category, fair but with potential for improvement.
Figure 3. Validation result of partial least squares modeling of four biomass properties of algae samples with NIR spectra.
Of the four algal properties, NIR performed slightly better than MIR for neutral lipids, crude protein, and ash content. In spectroscopy, MIR has the advantage of working with fundamental vibrational bands that are distinctive and strong, while NIR has the advantage of higher spectral fidelity due to more robust instrumentation optics and photonics. Traditionally, MIR measurement of solid samples requires mixing the samples with an IR-transparent substrate at accurate weight ratios. This substantially lowers the throughput of the analysis. With the ATR technique, neat samples can be analyzed, eliminating the need for sample preparation and increasing the throughput. In our experiment, obtaining an MIR-ATR spectrum required about 30 s per sample.
Compared to the results of our study, Laurens and Wolfrum (2011) obtained a prediction accuracy of R2 = 0.97 and RMSE of 0.18% using NIR spectra and R2 of 0.91 and RMSE of 0.30% for MIR with exogenously spiked neutral lipids in algal biomass. The spike levels in their study were fairly low (1% to 3% (w/w)) compared to the common levels of naturally occurring neutral lipids in algal biomass. This might explain the low RMSE and high R2 for their spectroscopic models. In working with algae turf scrubber samples (e.g., high levels of minerals and other inorganics), Mulbry et al. (2012) reported that lipids and total Kjeldahl N (or TKN, similar to our crude protein) had validation R2 values of 0.643 and 0.737, respectively, for NIR; and 0.554 and 0.709, respectively, for MIR. These results are slightly inferior to our results. In contrast, Mulbry et al. (2012) reported excellent modeling accuracy for ash content (R2 > 0.95) for both NIR and MIR. This was probably due to the relative uniformity of the chemical composition of ash in their samples as compared to our samples.
Compared to the prediction of protein and lipid content in other food and feed samples, our NIR and MIR models are somewhat inferior (Orman and Schumann, 1991; Pérez-Vich et al., 1998; Delwiche and Hruschka, 2000; Kovalenko et al., 2006; Shenk, 2008). This is mainly attributable to the high ash content as well as its diverse chemical composition in the algae samples. Some mineral salts are largely NIR and MIR transparent, but their presence changes the scattering behavior of IR energy. Other salts, such as bicarbonates and silicates, have spectral absorption in both the NIR and MIR regions. It is speculated that, as a result of scaled-up production, algal biomass will exhibit large variations in ash content due to different cultivation and harvesting methods. Therefore, a rapid and cost-effective method to quantify ash content would be very useful to determine the best use of the algal biomass.
Because water is a very strong absorber in both the NIR and MIR regions, direct application of these techniques on algae culture (99.9% water) is challenging. However, NIR and MIR based rapid characterization techniques would be useful in the algae-based biofuels and bioproducts industry. For example, it is proposed that algae be dewatered to contain 15% to 20% biomass before transport, processing, or conversion. At this moisture level (similar to that of fresh biological samples such as plant leaves), NIR and MIR might be useful for characterizing algal biomass for downstream process decision-making. Lipid-extracted algae could be used for animal feed or soil amendments (similar to distillers’ grain in ethanol production). In such a scenario, rapid characterization of biomass properties such as protein, fiber, moisture, and ash content would be valuable for determining the best use of the LEA biomass. Finally, Raman spectroscopy could be complementary to NIR and MIR for addressing the challenge presented by the water content of algae for rapid and non-destructive compositional analysis (Huang et al., 2009; Samek et al., 2011). Weiss et al. (2010) demonstrated the in vivo characterization of the structure of botrycoccenes from Raman signals of Botryococcus braunii samples (also a green algae). Raman signals are not affected by the presence of water and therefore would be a powerful tool for characterizing aqueous algae samples without time-consuming sample harvesting, dewatering, and drying.
In this study, a total of 99 lyophilized green algae samples were obtained to investigate the usefulness of NIR and MIR spectroscopy as a rapid and nondestructive method to quantify algal biomass composition. The algae samples were harvested from large-scale open-pond raceways over three years, with a wide range of cultivation and harvest practices. The biomass properties of interest included neutral lipids, crude protein, gross calorific value, and ash content. PLSR was employed for spectroscopic model calibration and validation. It was found that, due to the wide range of cultivation and harvest methods, the ash content of the sample set was high (10.2% to 83.8%). High ash content made the reported neutral lipid value significantly lower compared to literature values and impacted the spectral measurement of samples in both the NIR and MIR regions. NIR and MIR spectroscopy both predicted algal biomass composition reasonably well, with calibration R2 ranging from 0.55 to 0.84 and validation R2 ranging from 0.57 to 0.90. In terms of prediction accuracy, the NIR and MIR models performed similarly. It is concluded that NIR and MIR are both potentially useful for characterizing algal biomass rapidly and cost-effectively to support the algae-based biofuels and bioproducts industry. Future work should expand NIR and MIR spectroscopy to a wider variety of algae species so that the wider applicability of these spectroscopic models can be evaluated and tested.
We would like to acknowledge funding of this work by the U.S. Department of Energy under Contract No. DE-EE0003046 awarded to the National Alliance for Advanced Biofuels and Bioproducts. We would also like to acknowledge Mr. Lou Brown and Ms. Yola Brown at Texas A&M AgriLife Research Center for providing the lyophilized algae samples and associated lipid profile data.
Bryant, H. L., Gogichaishvili, I., Anderson, D., Richardson, J. W., Sawyer, J., Wickersham, T., & Drewery, M. L. (2012). The value of post-extracted algae residue. Algal Res., 1(2), 185-193. http://dx.doi.org/10.1016/j.algal.2012.06.001
Chang, C. D., Laird, D. A., Mausbach, M. J., & Hurburgh Jr., C. R. (2001). Near-infrared reflectance spectroscopy: Principal component regression analysis of soil properties. SSSA J., 65(2), 480-490. http://dx.doi.org/10.2136/sssaj2001.652480x
Chen, W., Sommerfeld, M., & Hu, Q. (2011). Microwave-assisted Nile red method for in vivo quantification of neutral lipids in microalgae. Bioresour. Tech., 102(1), 135-141. http://dx.doi.org/10.1016/j.biortech.2010.06.076
Chen, Y.-C., & Thennadil, S. N. (2012). Insight into information contained in multiplicative scatter correction parameters and the potential for estimating particle size from these parameters. Analytica Chimica Acta, 746(9), 37-46. http://dx.doi.org/10.1016/j.aca.2012.08.006
Chisti, Y. (2007). Biodiesel from microalgae. Biotech. Adv., 25(3), 294-306. http://dx.doi.org/10.1016/j.biotechadv.2007.02.001
Davey, P. T., Hiscox, W. C., Lucker, B. F., O’Fallon, J. V., Chen, S., & Helms, G. L. (2012). Rapid triacylglyceride detection and quantification in live micro-algal cultures via liquid state 1H NMR. Algal Res., 1(2), 166-175. http://dx.doi.org/10.1016/j.algal.2012.07.003
Dean, A. P., Nicholson, J. M., & Sigee, D. C. (2012). Changing patterns of carbon allocation in lake phytoplankton: An FTIR analysis. Hydrobiologia, 684(1), 109-127. http://dx.doi.org/10.1007/s10750-011-0973-0
Delwiche, S. R., & Hruschka, W. R. (2000). Protein content of bulk wheat from near-infrared reflectance of individual kernels. Cereal Chem., 77(1), 86-88. http://dx.doi.org/10.1094/CCHEM.2000.77.1.86
Doan, T. Y., & Obbard, J. P. (2011). Improved Nile red staining of Nannochloropsis sp. J. Appl. Phycol., 23(5), 895-901. http://dx.doi.org/10.1007/s10811-010-9608-5
Elsey, D., Jameson, D., Raleigh, B., & Cooney, M. J. (2007). Fluorescent measurement of microalgal neutral lipids. J. Microbiol. Methods, 68(3), 639-642. http://dx.doi.org/10.1016/j.mimet.2006.11.008
Filzmoser, P., & Varmuza, K. (2011). Chemometrics: Multivariate statistical analysis in chemometrics. R package version 1.3.7. Retrieved from http://CRAN.R-project.org/package=chemometrics
Gitelson, A. A., Grits, Y. A., Etzion, D., Ning, Z., & Richmond, A. (2000). Optical properties of Nannochloropsis sp. and their application to remote estimation of cell mass. Biotech. Bioeng., 69(5), 516-525. http://dx.doi.org/10.1002/1097-0290(20000905)69:5<516::AID-BIT6>3.0.CO;2-I
Greenspan, P., & Fowler, S. D. (1985). Spectrofluorometric studies of the lipid probe, Nile red. J. Lipid Res., 26(7), 781-788.
Guedes, A. C., Amaro, H. M., & Malcata, F. X. (2011). Microalgae as sources of high added-value compounds: A brief review of recent work. Biotech. Prog., 27(3), 597-613. http://dx.doi.org/10.1002/btpr.575
Harun, R., Singh, M., Forde, G. M., & Danquah, M. K. (2010). Bioprocess engineering of microalgae to produce a variety of consumer products. Renew. Sustain. Energy Rev., 14(3), 1037-1047. http://dx.doi.org/10.1016/j.rser.2009.11.004
Huang, Y. Y., Beal, C. M., Cai, W. W., Ruoff, R. S., & Terentjev, E. M. (2009). Micro-Raman spectroscopy of algae: Composition analysis and fluorescence background behavior. Biotech. Bioeng., 105(5), 889-898. http://dx.doi.org/10.1002/bit.22617
Kovalenko, I. V., Rippke, G. R., & Hurburgh, C. R. (2006). Measurement of soybean fatty acids by near-infrared spectroscopy: Linear and nonlinear calibration methods. JAOCS, 83(5), 421-427. http://dx.doi.org/10.1007/s11746-006-1221-z
Laurens, L. M., & Wolfrum, E. J. (2011). Feasibility of spectroscopic characterization of algal lipids: Chemometric correlation of NIR and FTIR spectra with exogenous lipids in algal biomass. Bioenergy Res., 4(1), 22-35. http://dx.doi.org/10.1007/s12155-010-9098-y
Lee, S. J., Yoon, B. D., & Oh, H. M. (1998). Rapid method for the determination of lipid from the green alga Botryococcus braunii. Biotech. Techniques, 12(7), 553-556. http://dx.doi.org/10.1023/A:1008811716448
Mata, T., Martins, A., & Caetano, N. (2010). Microalgae for biodiesel production and other applications. Renew. Sustain. Energy Rev., 14(1), 217-232. http://dx.doi.org/10.1016/j.rser.2009.07.020
Mulbry, W., Reeves, J., Liu, Y., Ruan, Z., & Liao, W. (2012). Near- and mid-infrared spectroscopic determination of algal composition. J. Appl. Phycol., 24(5), 1261-1267. http://dx.doi.org/10.1007/s10811-011-9774-0
Murphy, T. E., Macon, K., & Berberoglu, H. (2013). Multispectral image analysis for algal biomass quantification. Biotech. Prog., 29(3), 808-813. http://dx.doi.org/10.1002/btpr.1714
Orman, B. A., & Schumann Jr., R. A. (1991). Comparison of near-infrared spectroscopy calibration methods for the prediction of protein, oil, and starch in maize grain. J. Agric. Food Chem., 39(5), 883-886. http://dx.doi.org/10.1021/jf00005a015
Pérez-Vich, B., Velasco, L., & Fernández-Martínez, J. M. (1998). Determination of seed oil content and fatty acid composition in sunflower through the analysis of intact seeds, husked seeds, meal, and oil by near-infrared reflectance spectroscopy. JAOCS, 75(5), 547-555. http://dx.doi.org/10.1007/s11746-998-0064-1
Pistorius, A. M. A., DeGrip, W. J., & Egorova-Zachernyuk, T. A. (2009). Monitoring of biomass composition from microbiological sources by means of FTIR spectroscopy. Biotech. Bioeng., 103(1), 123-129. http://dx.doi.org/10.1002/bit.22220
R Development Core Team. (2013). R: A language and environment for statistical computing. Vienna, Austria: R Foundation Stat. Comput. Retrieved from www.R-project.org
Richardson, J. W., Johnson, M. D., & Outlaw, J. L. (2012).
Economic comparison of open pond raceways to photo bio-reactors for profitable production of algae for transportation fuels in the Southwest. Algal Res., 1(1), 93-100. http://dx.doi.org/10.1016/j.algal.2012.04.001
Samek, O., Zemánek, P., Jonáš, A., & Telle, H. H. (2011). Characterization of oil-producing microalgae using Raman spectroscopy. Laser Phys. Letters, 8(10), 701-709. http://dx.doi.org/10.1002/lapl.201110060
Sarrafzadeh, M. H., La, H.-J., Lee, J.-Y., Cho, D.-H., Shin, S.-Y., Kim, W.-J., & Oh, H.-M. (2015). Microalgae biomass quantification by digital image processing and RGB color analysis. J. Appl. Phycol., 27(1), 205-209. http://dx.doi.org/10.1007/s10811-014-0285-7
Shenk, J. S., Workman Jr., J. J., & Westerhaus, M. O. (2008). Chapter 17: Application of NIR spectroscopy to agricultural products. In D. A. Burns, & E. W. Ciruczak (Eds.), Handbook of Near-Infrared Analysis (pp. 347-386). Boca Raton, Fla.: CRC Press.
Solovchenko, A., Khozin-Goldberg, I., Recht, L., & Boussiba, S. (2011). Stress-induced changes in optical properties, pigment, and fatty acid content of Nannochloropsis sp.: Implications for non-destructive assay of total fatty acids. Marine Biotech., 13(3), 527-535. http://dx.doi.org/10.1007/s10126-010-9323-x
Stehfest, K., Toepel, J., & Wilhelm, C. (2005). The application of micro-FTIR spectroscopy to analyze nutrient stress-related changes in biomass composition of phytoplankton algae. Plant Physiol. Biochem., 43(7), 717-726. http://dx.doi.org/10.1016/j.plaphy.2005.07.001
U.S. Congress. (2007). H.R.6: Energy Independence and Security Act of 2007 (EISA). Washington, D.C.: U.S. Congress. Retrieved from www.congress.gov/bill/110th-congress/house-bill/6
Wagner, H., Liu, Z., Langner, U., Stehfest, K., & Wilhelm, C. (2010). The use of FTIR spectroscopy to assess quantitative changes in the biochemical composition of microalgae. J. Biophotonics, 3(8-9), 557-566. http://dx.doi.org/10.1002/jbio.201000019
Wawrik, B., & Harriman, B. H. (2010). Rapid, colorimetric quantification of lipid from algal cultures. J. Microbiol. Methods, 80(3), 262-266. http://dx.doi.org/10.1016/j.mimet.2010.01.016
Weiss, T. L., Chun, H. J., Okada, S., Vitha, S., Holzenburg, A., Laane, J., & Devarenne, T. P. (2010). Raman spectroscopy analysis of botryococcene hydrocarbons from the green microalga Botryococcus braunii. J. Biol. Chem., 285(42), 32458-32466. http://dx.doi.org/10.1074/jbc.M110.157230