Agronomy Journal Grow Your Career With ASA
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Web of Science (9)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Sobrino, E.
Right arrow Articles by Díaz, M. C.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Sobrino, E.
Right arrow Articles by Díaz, M. C.
Agricola
Right arrow Articles by Sobrino, E.
Right arrow Articles by Díaz, M. C.
Related Collections
Right arrow Sunflower
Right arrow Other Models
Right arrow Plant and Environment Interactions
Agronomy Journal 95:329-334 (2003)
© 2003 American Society of Agronomy

MODELING

Modeling the Oleic Acid Content in Sunflower Oil

Eduardo Sobrino*,a, Ana M. Tarquisb and M. Cruz Díazc

a Departamento de Producción Vegetal, Botánica, Polytechnic University of Madrid, Ciudad Universitaria, 28040 Madrid, Spain
b Departamento de Matemática Aplicada, Polytechnic University of Madrid, Ciudad Universitaria, 28040 Madrid, Spain
c Departamento de Edafología, Escuela Técnica Superior Ingenieros Agrónomos, Polytechnic University of Madrid, Ciudad Universitaria, 28040 Madrid, Spain

* Corresponding author (esobrino{at}pvb.etsia.upm.es)

Received for publication January 5, 2002.

   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Knowledge of the effects of temperature and geographic variables on the oleic acid content of sunflower (Helianthus annuus L.) oil allows us to predict the type of oil that will be produced in a particular area. This study was designed to establish a simple empirical model, which uses available variables of previously established effects, to estimate the final oleic acid composition of sunflower oil. Over two growing seasons, sunflower seeds were collected from Spain's main producing areas, and the oleic acid concentration of oil extracted from these samples was analytically determined. The effects of two types of variables (geographical position and temperature) on oil oleic acid content were determined according to three models based on the input variables: latitude, longitude, and altitude (Model I); mean minimum and maximum temperatures during the phenological stages of sunflower seed development and maturation (Model II); and a combination of both types of data (Model III). Through stepwise regression, it was established that best results were obtained using the temperature model (Model II) and the variables' mean minimum development and mean minimum and maximum maturation temperatures (r2 = 0.99, P < 0.001, n = 88). Of the three variables included in this model, the mean minimum maturation temperature provided the closest estimate of percentage oleic acid content. This regression model was statistically validated and is proposed as a method for crop managers to estimate oleic acid content based on local temperatures.

Abbreviations: MAE, mean absolute error • MSE, mean square error • tmaxd, mean monthly maximum temperature corresponding to the time of achene development • tmaxmat, mean monthly maximum temperature corresponding to the time of physiological maturity before harvesting • tmind, mean monthly minimum temperature corresponding to the time of achene development • tminmat, mean monthly minimum temperature corresponding to the time of physiological maturity before harvesting


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
THE USE OF SUNFLOWER OIL is mainly determined by its fatty acid composition (Piva et al., 2000). Nutritional use requires a high oleic acid content and a low level of saturated fatty acids. In contrast, the paint industry uses oil with a high proportion of linoleic acid. Several studies have focused on the beneficial effects on health of a diet rich in unsaturated fatty acids, particularly oleic acid (Krajcova-Kudlackova et al., 1997; Jing et al., 1997; Baldini et al., 2000). Further, oils with a high oleic acid content have the advantage of a greater resistance to heat oxidation, which makes them particularly suitable for frying.

Sunflower oil from standard cultivars is characterized by its high linoleic acid, moderate oleic acid, and low linolenic acid concentration. Saturated palmitic and stearic fatty acids make up <15% of this vegetable oil (Dorrel, 1978). Although the biosynthesis of these last two saturated fatty acids is unaffected by environmental conditions, the balance between oleic and linoleic acids is conditioned by these factors, and a strong inverse relationship is shown between their concentrations. The environmental effects on fatty acid composition of sunflower oil are well documented. Kinman and Earle (1964) showed that achenes produced in the cold climates of North America normally contain 70% or more linoleic acid in oil while those produced in more southern latitudes show levels as low as 30%.

The genetic modification of sunflower oil fatty acids has also been a subject of research. Soldatov (1976) obtained a stable sunflower mutant by dimethyl sulfate–induced mutagenesis that gave rise to oil levels exceeding 80% oleic acid. This high-oleic mutant showed no substantial variation in fatty acid composition in response to changes in environmental conditions (Fick, 1984).

Sunflower cultivars differ in the sensitivity of their oil properties to the environment. This sensitivity has been exploited both from a theoretical perspective, such as by enzyme inactivation induced by temperature changes under natural conditions, as well as from an applied standpoint because it is possible to produce oil of different characteristics at different latitudes.

Robertson et al. (1978) showed that fatty acid composition was highly correlated with latitude. Although they found that saturated fatty acids showed scarce variation associated with the environment, oleic acid ranged from 14 to 50% between north and south while linoleic acid levels changed from 41 to 75% in the reverse direction. In Europe, the effect of altitude and temperature on the oleic–linoleic acid balance in sunflower oil has also been explored by Lajara et al. (1990). These authors analyzed the fatty acid content of sunflower oil corresponding to different Spanish regions and were able to confirm the close negative relationship between fatty acids with linoleic acid contents as high as 70% for the northern Meseta and 49% for Jaen in southern Spain.

Oil accumulates in the achene, commonly known as the sunflower seed, during the postflowering period (Anderson, 1975; Robertson et al., 1978; Maeda et al., 1987). This accumulation continues for 33 d after the initiation of anthesis; the greatest rate of production occurs 15 to 33 d after the initiation of anthesis.

According to Villalobos (2000), sunflower models have received relatively less attention than those of other crops. Full models developed since the 1980s include only a few specific to sunflower (Steer et al., 1993), e.g., OILCROP-SUN (Villalobos et al., 1996). Many of the described models can be applied to any species, e.g., CROSYST (Stockle et al., 1994), or to a small group of species, e.g., EPICPHASE (Cabelguenne et al., 1999). Recently, attempts have also been made to achieve highly specific goals such as the evaluation of genotypes (Agueda et al., 1997) or irrigation management (Chapman et al., 1993; Debaeke et al., 1998). Villalobos (2000) highlights the need for models that can predict the quality of seed/oil on which to base agricultural price policies including sunflower use because past attempts to develop empirical models of oil composition were based on simple correlations (Robertson et al., 1978) or related composition only to air temperature (Harris et al., 1978).

The present study was designed to develop models to explain and predict the percentage of oleic acid in sunflower oil, in terms of geographic or climatic conditions. The input data for the models proposed are commonly available at local meteorological stations.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Sunflower achenes were collected from the main producing areas in Spain over two consecutive years. According to the Anuario de Estadística Agroalimentaria (Agroalimentary Statistics Yearly Report, MAPA, 1998), these areas produce 90% of the country's entire sunflower crop. A standard-type sunflower defined in terms of oleic acid content was included in the study. The crops were cultivated under the standard conditions of no irrigation.

Sampling Protocol
Eight main sunflower-producing areas were identified (Fig. 1) . Each area was assigned a number of sampling points proportional to the cultivated area from 2 to 10 (Fig. 1) for a total of 96 different sampling points over the study period, which included two crop seasons (N = 48 for each crop season). Thirty-five 1-kg achene samples were taken from each of these points. The achene specimens collected from each point were mixed to obtain a representative 20-g sample for subsequent oil extraction and fatty acid analysis. The oil was extracted by elution of a small quantity of pulverized sample (15 mg) with 3.5 mL of petroleum ether using a disposable filter column.



View larger version (37K):
[in this window]
[in a new window]
  		Fig. 1. Main sunflower-producing areas selected: (1) West Andalucia (Córdoba, Sevilla, Málaga, and Cadiz), (2) East Andalucia (Jaén and Granada), (3) Albacete-Murcia, (4) Badajoz, (5) Toledo-Ciudad Real, (6) Cuenca-Guadalajara, (7) North plateau (Soria, Burgos, Valladolid, Palencia, Zamora, Salamanca, Avila, and Segovia), and (8) Ebro River basin (Logroño, Navarra, Zaragoza, and Lérida).

 
Oleic Acid Determination
Fatty acid composition was determined by gas chromatography according to the official analysis method (MAPA, 1993). First, methyl esters were prepared by applying sodium methylate followed by extraction in hexane. The sample was then filtered and injected into a Hewlett Packard 5890 fitted with a Chrompack CP-WAX-52 CB silica gel capillary column (25 m by 0.25 mm). A flame ionization detector was used under the following operating conditions: carrier gas, He; injector temperature, 240°C; detector temperature, 280°C; and oven temperature, 200°C. The methyl esters used as standards were those corresponding to fatty acids found in the sunflower oil: myristic, palmitic, stearic, oleic, linoleic, linolenic, arachidic, behenic, and lignoceric. Fatty acid peaks were identified by comparing their retention times with the peaks shown by the standard methyl esters. Quantitative determinations were performed using an automatic integrator, calculating the area under the peaks. Oleic acid and linoleic acid contents were calculated as the percentage of the area under the corresponding peak with respect to the total area.

Model Variables
Meteorological data (mean monthly minimum and maximum temperatures) corresponding to the study years (1999–2000) were obtained from the gauging station (Instituto Nacional de Meteorología) closest to the sampling point. Data missing from the temperature time series at each meteorological station were filled in by linear regression between the monthly values from neighboring stations (Guerra-Gomez, 1985), checking previously the homoscedasticity of the series (Thom, 1966; Buendía, 1985). The percentage of the missing data filled by this methodology was <2% of the total data used.

The monthly temperatures used in the model were those corresponding to the time of achene development (2 d after flowering) [maximum (tmaxd) and minimum (tmind)] and the time of physiological maturity before harvesting (usually one month after flowering) [maximum (tmaxmat) and minimum (tminmat)].

The oleic acid content of the oil was modeled according to: geographical variables (longitude, latitude, and altitude) and climatic variables (temperature recorded during achene development and at physiological maturity). Complete datasets corresponding to each sampling point were used to develop the models. Longitude ranged from 0°47' E to 7°0' W, latitude from 31°7' N to 42°49' N, and altitude from 10 to 1118 m. The proportion of oleic acid (with respect to total fatty acids) ranged from 15 to 38%, with a mean of 25% and standard deviation of 5%.

Data from eight randomly selected points (four per season) were set aside for final validation testing of the established models by calculating the determination coefficient. Thus, the model variables used were complete data sets corresponding to 88 sampling points.

Modeling was performed according to the stepwise multiple linear regression method. The software used was Statgraphics Plus for Windows (Manugistics, 1998), and forward and backward variable selections were applied.

Regression Models
To describe the percentage of oleic acid in sunflower oil (y), three linear regressions were fitted using different variables. The general regression model applied was

[1]
where xi is the input variable used for each particular model, ai is the coefficient to be determined, and p is the number of input variables used after the stepwise procedure. The input variables for each of the models tested were Model I, longitude, latitude, and altitude; Model II, mean monthly temperature recorded during achene development (tmaxd and tmind) and at achene maturity (tmaxmat and tminmat); and Model III, which included all seven variables.

To select the model that best described the oleic acid content of the sunflower oil, the following were calculated for each model:

  1. Variable coefficients (ai) and their standard error (SE).
  2. Value of the statistic test t (t value) and the F value for each variable. The value 4 is based on the number of degrees of freedom in the numerator and denominator of the F ratio.
  3. r2, r2 adjusted (r2adj), mean square error (MSE), and mean absolute error (MAE):

    [2]
    where n is the number of observations (in this case 88) and p the number of variables selected. The expression (n - p) represents the degrees of freedom that affect the r2 adjusted. The MAE is given by:

    [3]
    where ej is the difference between the estimated and the real percentage value of oleic acid.

  4. Analysis of model residuals and normality testing of the residuals. To be reliable, models should not have any systematic bias. Thus, the regression residues (measured acid content - calculated acid content) should show a normal distribution. Coefficients of kurtosis and skewness of the residuals were used to determine whether the residues of each of the models were normally distributed.
  5. Correlation matrix for the estimated coefficients. The Pearson correlation coefficient for the estimated coefficients in the model is calculated according to Draper and Smith (1981).
  6. Robust coefficient (B2) was calculated using all of the data. This coefficient is obtained following the relation (Peña, 2000):

    [4]
    where yj values are the real data, j is an estimate of the value j using the fitted model with all the data (n = 96), and (j) is an estimate of j value using the fitted model with all of the points except j (95 points). It is obvious that the B2 value will vary between 0 and 1. Whether each data point is used in the fitted model or not, the closer the two sets of predictions are, the closer B2 will be to 1. In the reverse case, B2 will approach 0.


   RESULTS AND DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
CONCLUSIONS
 REFERENCES
 
Model I, based on geographic variables, was the first tested. The coefficients and F-ratio value indicated a large influence of latitude on the percentage oleic acid, as reported by Lajara et al. (1990) (Table 1). Thus, the greater the latitude is, the higher the oleic acid content. In contrast, greater longitudes and altitudes give rise to a diminished oleic acid percentage.


View this table:
[in this window]
[in a new window]
  		Table 1. Variables selected by the forward method with their corresponding coefficient values (ai), standard error (SE), t value, F value, and significance level (P) according to the model used.

 
Model II, based only on temperature, shows that the tmind correlated negatively with oleic acid content (Table 1). This was unexpected because generally the lower the oleic acid content is, the lower the minimum temperature of achene formation (Kinman and Earle, 1964). A possible explanation might be the colinearity shown between tmind and tminmat (Table 2) not detected by the stepwise regression. This has been previously described as a possible limitation of using such a model (Devore, 1995). Alternatively, it could indicate the existence of a temperature range over which the efficiency of oleic acid formation is optimum, with diminished production below and above temperatures that could be classified as critical. Of the four variables initially introduced, tmaxmat was eliminated in subsequent regression steps. The tminmat variable showed most influence and is consequently most explicative of the model.


View this table:
[in this window]
[in a new window]
  		Table 2. Correlation matrices for coefficient estimates according to the model using 88 data points.

 
Hypotheses proposed to explain the accumulation of oleic and linoleic acid are based on the action of the enzymes stearoyl-ACP-desaturase ({Delta}9-desaturase), which catalyzes the initial desaturation of stearoyl-ACP to oleoyl-ACP (McKeon and Stumpf, 1982), and oleyl-phosphatidyl-choline-desaturase ({Delta}12-desaturase), which is responsible for the second desaturation of oleyl-PC to linoleoyl-PC (Slack et al., 1979; Stymne and Appelqvist, 1980). In standard sunflower cultivars, it is possible that the activity or synthesis of the enzyme mediating the transformation of oleic into linoleic acid ({Delta}12-desaturase) is diminished at high temperatures (Champolivier and Merrien, 1996). This effect was specifically explored in the sunflower by Garcés and Mancha (1991), who established that the activity of {Delta}12-desaturase is strongly inhibited at temperatures above 20°C.

In Model III, obtained by introducing both geographical and temperature variables, only the variables tmaxmat, tminmat, and altitude remained after a selection generated by the stepwise regression method itself. The higher the temperature is during the maturation stage, the greater the percentage oleic acid, and the greater the altitude is, the lower the percentage obtained (Table 1). This might indicate that changes in oleic acid level could be due more to the second stage (time of physiological maturity before harvesting) than the first stage (time of achene development) because at our latitudes, it is not as determining a factor as in other areas (Kinman and Earle, 1964). In this model and in Model II, tminmat is one of the most explicative variables, with an F value of 50.652 and 207.559, respectively (Table 1). In Model III, the correlation matrix shows a colinearity between tmaxmat and tminmat (Table 2) although when the backward-and-forward method is applied, the variables selected are the same. It is obvious that a relationship exists between the maximum and minimum temperatures coming from the same place; for this reason, the analysis of Model III has been continuing and not rejected at this step.

The fact that each model shows a high r2 (Table 3) does not imply good prediction (Weisberg, 1985). By plotting the residuals generated by each model (Fig. 2) , it becomes evident that in Model I (Fig. 2A), there is a tendency to underpredict at low acid levels and overpredict at high acid levels. This effect is not so obvious in the other two models (Fig. 2B and 2C). Model I shows a higher MSE and MAE (Table 3), and at the same time, the coefficient of skewness of the residuals is the highest of the three models tested (0.56), in accordance with the residuals observed in Fig. 2B.


View this table:
[in this window]
[in a new window]
  		Table 3. R2 adjusted (r2 adj), mean square error (MSE), mean absolute error (MAE), and coefficients of skewness (skewness) and kurtosis (kurtosis) of the residuals for each model using 88 observations with 85 degrees of freedom.

 


View larger version (16K):
[in this window]
[in a new window]
  		Fig. 2. Residual plot of observed values for each model using: (A) geographical variables (Model I), (B) temperature variables (Model II), and (C) temperature and altitude variables (Model III).

 
When the predicted values were compared to the real values (Fig. 3) , the same patterns described above were observed. However, Models II and III overpredicted oleic acid levels above 28% (see Fig. 3B and 3C), which points to the possibility that a new variable might improve the models in these cases.



View larger version (18K):
[in this window]
[in a new window]
  		Fig. 3. Predicted vs. observed values of the data points used in model fitting according to each model: (A) Model I based on geographical variables (r2 = 0.97), (B) Model II based on temperature variables (r2 = 0.99), and (C) Model III based on temperature and altitude variables (r2 = 0.99).

 
If we analyze the statistical tests applied and compare the mean errors of the predictions, Model II appears to be optimal for the data under study because a high r2 (r2 = 0.99) was accompanied by the lowest MAE and kurtosis and asymmetry coefficients were low, indicating a normal distribution of residuals (Table 3).

The complementary validation test using the eight points set aside (Fig. 4) indicated that the best prediction was made by Model II, showing an r2 of 0.86, compared with Model I (r2 = 0.39) and Model III (r2 = 0.82). Besides this, the B2 values confirm the determination coefficient results: 0.71 in Model II, 0.63 in Model III, and 0.47 in Model I. All these tests point to discarding Model III and show the effect that the colinearity has between two temperature variables (tmaxmat and tminmat).



View larger version (14K):
[in this window]
[in a new window]
  		Fig. 4. Predicted vs. observed values of the eight data points not used in the model-fitting procedure: (A) using geographical variables (Model I), (B) using temperature variables (Model II), and (C) using temperature and altitude variables (Model III). Dotted line corresponds to y = x. The regression equation and r2 for each model are shown in the graph.

 

   CONCLUSIONS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
REFERENCES
 
As also suggested in the literature cited in the introduction, the temperature of development and maturation of sunflower achenes is among the most determining factors for the production of oleic acid.

Of the models tested, Model II yielded the best prediction and fit the data analyzed the best in terms of tmaxd, tmind, and tminmat. This model also showed the most statistical power. These findings suggest that further work should include a search for new variables that might improve prediction in cases of high oleic acid contents (above 28%). This search should consider variables such as sowing date, nutrient level, soil water availability, etc. Basically, these factors affect the phenological stage of the crop, which in turn, interacts with the temperature regime.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 




This article has been cited by other articles:


Home page
 Agron. J.Home page
V. D. Zheljazkov, B. A. Vick, B. S. Baldwin, N. Buehring, T. Astatkie, and B. Johnson
Oil Content and Saturated Fatty Acids in Sunflower as a Function of Planting Date, Nitrogen Rate, and Hybrid
Agron. J., July 7, 2009; 101(4): 1003 - 1011.
[Abstract] [Full Text] [PDF]

Home page
 Agron. J.Home page
V. D. Zheljazkov, B. A. Vick, M. W. Ebelhar, N. Buehring, B. S. Baldwin, T. Astatkie, and J. F. Miller
Yield, Oil Content, and Composition of Sunflower Grown at Multiple Locations in Mississippi
Agron. J., May 7, 2008; 100(3): 635 - 642.
[Abstract] [Full Text] [PDF]

Home page
 Agron. J.Home page
N. G. Izquierdo, L. A.N. Aguirrezabal, F. H. Andrade, and M. G. Cantarero
Modeling the Response of Fatty Acid Composition to Temperature in a Traditional Sunflower Hybrid
Agron. J., April 11, 2006; 98(3): 451 - 461.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Web of Science (9)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Sobrino, E.
Right arrow Articles by Díaz, M. C.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Sobrino, E.
Right arrow Articles by Díaz, M. C.
Agricola
Right arrow Articles by Sobrino, E.
Right arrow Articles by Díaz, M. C.
Related Collections
Right arrow Sunflower
Right arrow Other Models
Right arrow Plant and Environment Interactions

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
The SCI Journals Crop Science Vadose Zone Journal
Journal of Natural Resources
and Life Sciences Education		 Soil Science Society of America Journal
Journal of Plant Registrations Journal of
Environmental Quality		 The Plant Genome