Visual analysis of olive oil and food colors in general are commonly carried out. In this regard, large panels of untrained panelists are normally considered to determine preferences, whereas trained panelists are used for descriptive assessments. Indeed, trained judges are used for quantitative descriptive analysis (QDA), which is useful for generating color descriptive terms, which can be correlated with instrumental or other type of data at a later stage.
The simplest way of assessing the color of an olive oil visually is to compare it with color scales. These scales consist of a series of colored solutions, which are rather stable and can be made from easily available colorants. Furthermore, the scale of reference solutions must encompass all the possible colors that can be found in the type of olive oils studied. It is very important to place the samples to be assessed in appropriate cuvettes, and that these match those used to contain the reference solutions, since both (the sample and the references) are eventually compared using a standardized blank of diffuse light (Naudet and Sambuc 1955; Belbin 1993).
In the case of olive oil, the bromothymol blue method (BTM) is widely used. This method is based on the establishment of a scale of indexes that do not contain reddish hues for the definition of the color of oils from olives and seeds. That is, the scale has to encompass hues ranging from yellow to green (AENOR 1963). A modification of this method for the visual color definition of virgin olive oils has also been proposed (Gutiérrez and Gutiérrez 1986).
Colored glasses can also be used as reference standards as an alternative to colored solutions. These types of glasses are used in the Lovibond method (AOCS 1992). This method consists in the visual color matching of the light transmitted through a specific surface of oil with the color of the light originated by the same source transmitted through the colored glass standards. The apparatus enables to observe the oil under controlled conditions, since the illumination is standard and the vision angle is set. The optic system is designed to help the observer see simultaneously the field of the reference blank and that of the olive oil sample. The series of reference colored glasses is numbered and comprises colors ranging from unsaturated water-white to completely saturated red, yellow, and blue colors. These colors constitute the so-called Lovibond color scale. The color matching between the samples and the references is expressed in terms of Lovibond units of red, yellow, and/or blue (Belbin 1993).
Other color scales used to define the color of oils are the Gardner scale (mainly used to classify natural and synthetic oils, lecithins, fatty acids, and some oil derivatives) (AOCS 1964), the Wesson method (applicable to all the normal fats and oils) (AOCS 1992), or the fatty acid committee (FAC) scale, used to classify nonedible grease and dark oils (AOCS 1943).
The determination of the color of olive oils presents other drawbacks, such as the obtaining of appropriate reference material itself, the fact that these are not rigorously standardized, and the possibility that they undergo color changes due to oxidation processes catalyzed by heat, light, and oxygen as time goes by. Besides, the color matching by these means is rather imprecise and is very dependent on the observer. It is commonplace that different observers match the same sample differently and that the same observer matches the same sample differently when the test is repeated (Presnell 1949; Naudet and Sambuc 1955; Pohle and Tierneh 1957). On the other hand, in relation to the impact of color in the sensory analysis of olive oils, in general, a recent study concludes that the objective intended when using blue glasses for the oil tasting, which is to conceal the color so it does not influence the sensory assessment of aroma and taste, is not fulfilled. In this sense, it was reported that such glasses fail to effectively conceal the color such that the panelist can appreciate it visually. From this it can also be inferred that the color of olive oils is paid little importance in relation to the sensory analysis to the extent that it is regarded as an “annoying attribute.” This contrasts with the sensory analysis of other products for human consumption. For instance, in the case of wines, color is an important part of the sensory analysis and the panelists are expected to be sufficiently trained so as not to let themselves be influenced by it. In this study, it is concluded that the color of olive oils should also be carried out in a transparent glass so the sensory analysis is complete. In this sense, it is pointed out that the color intensity and the hue are much related to the degree of ripeness of the drupes. Additionally, the fact that unfiltered oils are acquiring certain relevance gives support to this recommendation, inasmuch as the consumer can regard olive oils having natural turbidity as more natural. On the other hand, the authors state that in order to effectively conceal the color of the oils, the chromatic characteristics of the tinted glass used for the sensory analysis should be established more carefully (Melgosa and others 2009).
To conclude this section, it can be stated that the need to reduce the subjectivity linked to the visual assessments led to the progressive development of instrumental methods. These methods also offer a series of advantages (simplicity, nondestructiveness, affordable and versatile equipment, portability, ultra-rapid measurements, possibility of automation, and so on) that can be harnessed for the quality control of foods in the field by the industry, or even in the marketplace.
The application of tristimulus colorimetry to olive oils
The physiological sensation to examine the appearance of a transparent liquid is dependent on 3 factors: the amount of radiant energy emitted by the source at each wavelength (emission spectrum of the source), the way in which this energy is transmitted by the sample observed (emission spectrum), and the response of the eye of the observer to the radiations of different frequencies (curve of sensitivity of the eye). The 2nd factor is the most important on which to base the proposal of the appreciation of the oil color. In this sense, what it is sought is to define a color from the sensations perceived by the human eye, copying in a “natural way” the language of the colors and the paintings. Thus, a color is characterized by its hue (yellowish, orange, greenish), its purity (the basal hue is more or less mixed with the white) and its brightness (the color is more dark or more bright).
The tristimulus colorimetry deals with the chromatic specification of color abiding by the basis of the trichromatic theory. The interest in the application of this theory to the color definition of oils started decades ago, above all, after it was demonstrated that their color assessment by visual comparisons or empirical equations yielded few reliable outcomes. Due to the variable amounts of the different pigments present in the oils, it is not possible to select only one wavelength at which the transmission is an exact function of the total or visual transmission of the oil. In this regard, it is considered that the expression of the results in terms of tristimulus values must be a more satisfactory method (Presnell 1949; Naudet and Sambuc 1955; Bigoni 1963).
The problem of the choice of the most adequate solvent to be used as blank reference has been tackled, as well as the validation of the Lambert–Beer Law and its application to fat solutions in different solvents and at different sample thicknesses. Additionally, the trichromatic theory has been applied in different cases like raw fatty compounds, decoloration of fats, and refined fats (Naudet and others 1956).
In the particular case of olive oils, the transmittance curves of a series of representative samples (3 refined oils, 3 blends of virgin and refined oils, 4 commercial olive oils, and 12 virgin olive oils obtained by the systems of pressing and partial extraction in the experimental olive mill of the Instituto de la Grasa (Sevilla, Spain) were obtained. From these samples, their coordinates in the International Commission on Illumination (CIE, from its name in French: Commission Internationale de l’Eclairage) system were calculated as a starting point to deduce the methodology to be applied to assess the color of those oils (Castro and others 1955). More specifically, the chromaticity coordinates (x,y) and the factor of luminance (Y) were calculated using the illuminant C and the tristimulus values X,Y,Z. In the chromatic diagram, it was observed that it was possible to trace a line that, with a very good approximation, represents the location of the chromaticity of olive oils. This fact would enable by establishing a scale along that line, to reduce the specification of the chromaticity of olive oils to only one parameter, and that of the “complete color” to 2 parameters, by including the luminance, which is not included in the diagram. On the other hand, the fact that this line was largely straight made it evident that it would be easy to design a simple colorimeter to determine the parameter to be used for the color definition without having to resort to standards. In relation to this, several simplified methods have been proposed in order to reduce the numerous calculations necessary to work out the trichromatic coordinates (above all when computing was not as advanced as it is nowadays) or to enable its application in the case that the necessary instrumentation to register the complete spectrum was unavailable. Likewise, the use of simple colorimeters to register only several wavelengths has also been proposed. In this regard, it seemed feasible to obtain the tristimulus values with a good approximation from functions of the transmission values at certain correctly chosen wavelengths. The procedure consisted basically of the application of Hardy's method of selected coordinates (Hardy 1936), although reducing the number of coordinates to a great extent. Presnell (1949) performed it from oils clarified by filtration and dehydration if necessary. For this purpose, 5-mm cuvettes were used, except for very dark oils, for which narrower cuvettes were employed and the results were finally corrected to 5-mm pathlengths. As blank reference, both water and CCl4 could be used (although the latter was recommended because its refractive index is closer to that of most of the oils), while the illuminant C was considered as reference for being the most widely used at that time. For the original calculation of the tristimulus values, oils from different origins exhibiting varied transmission spectra were used and 30 coordinates were selected. Subsequently, 3 coordinates were selected and the corresponding factors that could be applied to achieve a good approximation to the tristimulus values previously calculated were found. Eventually, the following expressions for X,Y,Z., obtained from the transmittance readings at 445, 555, and 600 nm, were arrived at:
Although the approximate values obtained applying these equations were in accordance with the values obtained by the method of the selected coordinates, Sambuc and Naudet (1956) made some observations to these results:
If the equations proposed by Presnell are applied to a perfectly transparent solution (T(λ) = 100), it can be observed that the trichromatic coordinates do not coincide with those of the point C (colorless).
X and Z are erroneous and the scattering of the results is considerable, especially in the case of Y.
The samples surveyed by Presnell were very bright oils with little variation in brightness.
The authors tried to enhance the results increasing to 4 the number of degrees of transmission used and choosing conveniently the wavelengths (444.4, 495.2, 551.8, and 624.2 nm), such that the following equations were eventually obtained:
As a result of these modifications, the deviation observed between the reference values and the practical values were markedly reduced, considering that the main source of error was due to raw and little refined oils that exhibited a marked variation of transparency in a very narrow region of the spectrum.
According to Bigoni (1963) the previous methods were based in the relative continuity of the absorption spectrum of the oils, normally without any characteristic band, which justifies partially the selection of only 4 wavelengths in correspondence to the 3 primary sources. Thus, some inconveniences were pointed out:
The longest wavelength for the expression of the tristimulus values was 625 nm, so the absorption band due to chlorophylls at 670 nm, which is characteristic of some oils like olive oils, was not considered.
The method restricts itself to give as results of the measurements the values XYZ. Such numerical values are not appropriate to convey an exact idea of the color, with the exception of the brightness (Y).
Consequently, the author developed a modification of the method, based on the following principles: to add to the equation of X (that considers the region 600 to 700 nm) another term that took into account the absorption band due to chlorophylls and the proposal of 2 simple equations that manage to express directly the values of saturation (S) and hue (λ), avoiding the graphical representation:
More recently, a methodology for the rapid determination of the color of virgin olive oils from the absorbance or transmittance readings at 480 and 670 nm of the pure samples has been proposed (Escolar and others 1997). These values correspond to the absorption maxima of the major pigments accounting for the color of the samples and, therefore, for the main absorption bands in their visible spectra. Indeed, the visible spectrum of a virgin olive oil can be simulated by means of an adequate combination of the spectra of carotenoids and chlorophylls. Moreover, simplified methods for the determination of the color of olive oils based on the application of the analysis of characteristic vectors have been proposed (Ayala and others 1994; Moyano and others 2001).
The use of filter photocolorimeters to assess the color of olive oils has also been described. When the filters are illuminated by means of a well-defined source and the emerging light is received on a photoelectric cell, the response curve of the cell as a function of the wavelength is identical to the contribution of each of the primary in that source. In other words, the light of each filter acts like a primary source. One of the most widely used filter photocolorimeters is the so-called Hunter photocolorimeter. With this apparatus the X,Y,Z values of the samples can be calculated from the degrees of transmission R,G,B read with each filter. Another similar apparatus has been devised from some of the simplified methods to determine the trichromatic coordinates (Cruz and others 1956; Sambuc and Naudet 1956, 1960; Bigoni 1963).
Up to this point, we have dealt with the development of instrumental methodologies that can be used to overcome the subjectivity linked to visual measurements. However, the comparison between different methods for the color assessment of oils has also been the objective of several studies. For instance, the correlation existing between different instrumental and visual methods to measure the color of rapeseed oils has been determined (Brát and others 1988, 1993).
In recent years, the BTM has been the subject of several investigations, in which some of the limitations of this methodology were pointed out. Thus, the trichromatic coordinates corresponding to 90 virgin olive oil samples and to these standard solutions were evaluated to examine the validity of the BTM scale. Interestingly, it was observed that the scale only matched part of the oils studied. Furthermore, considering the typical spectra of one oil and that of a standard solution, it was suggested that there may be problems of metamerism (the phenomenon by which 2 stimuli with different spectral composition are visually perceived as equal) that could invalidate even more the use of the BTM scale in the industry. As a result of these observations, it was concluded that certain corrections were necessary to take advantage of the well-known practical utility of this scale (Burón and others 1989). Moreover, it has been demonstrated that there is a clear chromatic degradation of the standards with time, which advised against the use of this system (Moyano and others 1999; Melgosa and others 2001). In another investigation, conclusions concerning the low precision, accuracy, and uniformity of the scale were drawn and additional suggestions for its enhancement were provided (Melgosa and others 2000).
In a comparative study, several types of oils subjected to thermal oxidation were used to calculate the chromatic parameters from the standard methods recommended by the CIE and several simplified methods. As a result, it was concluded that the latter methods could only be applied to certain chromatic parameters and certain oils and that, therefore the standard methods recommended by the CIE must be applied for a complete study of the color parameters of the samples (Guillén and others 1991).
In another interesting study, the mistakes made in the calculation of the tristimulus values of olive oil by methods based on the use of several selected coordinates and an increase in the number of such coordinates were analyzed. For this purpose, 10 methods were evaluated. Four were old methods developed for other kinds of oils but eventually used in olive oils. The other 6 were methods that used a large number of transmittance values and that were developed to determine the influence of the number of data considered to calculate the tristimulus values. The study demonstrated, as expected, that a large number of coordinates provide better results in the definition of olive oil color (Escolar and others 1994).
In recent years, new studies concerning the objective measurement of olive oil color have been published. In one of them, a uniform color scale for virgin olive oils was developed from the chromatic parameters corresponding to several hundreds of Spanish olive oils from different varieties and origins. More specifically, this scale proposed a new set of color standards that appear more appropriate than those proposed by the BTM (Melgosa and others 2004). Furthermore, the CIE 1976 (CIELAB) color space for virgin olive oil has been recently determined and used to classify the color of over 100 Spanish samples of diverse origin (Escolar and others 2007).