Agronomy Journal 92:761-765 (2000)
© 2000 American Society of Agronomy
SUNFLOWER
Improving Estimates of Individual Leaf Area of Sunflower
Michael P. Bangea,
Graeme L. Hammerb,
Stephen P. Milroya and
Kenneth G. Rickertc
a CSIRO Div. of Plant Industry, Cotton Res. Unit, Locked Bag 59, Narrabri, NSW 2390, Australia
b Queensland Dep. of Primary Industries, Agric. Production Systems Res. Unit (APSRU), 203 Tor St., P.O. Box 102, Toowoomba, QLD 4350, Australia
c School of Natural and Rural Systems Management, The Univ. of Queensland, Lawes, QLD 4343, Australia
michael.bange{at}pi.csiro.au
 |
ABSTRACT
|
|---|
Simple, accurate, and nondestructive methods of determining leaf area of plants are important for many experimental comparisons. Determining individual leaf area (LA) of sunflower (Helianthus annuus L.) can involve measuring leaf length (L) and breadth (B). The objectives of this field study were to compare published and new models to determine the most precise model to predict the area of individual leaves of sunflower plants and to test the applicability of these models between sowing times, with three genotypes differing in maturity and stature, and at different times during crop growth. The best model, which included both the L and B factors (LA = aL + dL + eLB + c), accommodated changing leaf shape during crop development. The relationship between leaf dimensions and LA was significantly improved when sowing time and the times of leaf sampling were included in the relationship; however, no significant differences were found among genotypes. The relative increase in precision with more complex models was considered. Both length and breadth measurements were needed to attain precision for all leaf sizes, and given the availability of computer-based statistical packages, the use of relatively complex models in the assessment of individual leaf areas would appear to be a practical option.
Abbreviations: B, breadth of leaf lamina • L, length of leaf lamina • LA, individual leaf area • RMS, residual mean squares • R1, bud visible • R5.1, first anthesis • S1, sowing 13 Sept. 1991 • S2, sowing 5 Mar. 1992 • VE, emergence
|
INTRODUCTION
|
|---|
MEASURING LINEAR DIMENSIONS of leaves (e.g., length and breadth) is an established and successful method of nondestructive estimation of leaf area. For sunflower, this method has been used to estimate area of individual leaves or leaf area of individual plants (Rawson et al., 1984), leaf area index (Goyne et al., 1978), and rates of leaf expansion (Rawson and Constable, 1980). Various combinations of measurements (e.g., recording the length and breadth together or measuring the length or breadth only) and various models relating linear dimensions to area have been utilized (Table 1)
.
View this table:
[in this window]
[in a new window]
|
Table 1 Published models for sunflower using linear leaf dimensions to estimate individual leaf area. B, leaf breadth; L, leaf length; LA, individual leaf area; LB, leaf length times leaf breadth; a and d are coefficients; c is a constant
|
|
The method chosen depends on the time available to take measurements and the level of precision desired. While measurements of both breadth and length can be more precise than estimates based on one dimension (Schneiter, 1978), this method requires twice the number of measurements. Similarly, a calibration could be derived for each genotype or sampling time, but this requires additional independent data sets.
No published reports indicate that a one-dimension approach is consistently superior to a two-dimension approach. Furthermore, published reports fail to show that one specific approach for deriving leaf area of sunflower is superior across different varietal and environmental conditions. Schneiter (1978) derived a single function that could calculate leaf area with relative accuracy across different genotypes, row spacings, and plant populations, but did not explicitly test for differences in the capacity of various models to accommodate these factors.
As evident in Table 1, most published models are functions of a single variable (i.e., L, B, or LB), even when two dimensions have been measured (e.g., Pereyra et al., 1982). The practicality of using more complex functions and analyses by these authors may have been limited by the availability of technology. It is now possible to consider using more complex models than those presented in Table 1 in the nondestructive assessment of leaf area in the practical applications.
Our first objective was to compare models to determine the most appropriate and precise model to predict individual leaf area of sunflower. Our second objective was to use a selected model and test its usefulness across treatment differences between two sowing times, among three genotypes differing in maturity and stature, and among different sampling times during the growth of the crop.
|
Materials and methods
|
|---|
Cultural Details
A field experiment was conducted from 1991 to 1992 at The University of Queensland Gatton College (27°33' S; 152°17' E; altitude 100 m) in the Lockyer Valley, a subtropical location in southeast Queensland. The site has an alluvial-prairie Fluventic Haplustoll soil (Stace et al., 1968), with a mean annual rainfall of 784 mm, and mean maximum and minimum temperatures of 26.4 and 13.1°C, respectively.
Three genotypes (AgSeed 8640, Suncross 41, and 8793) differing in maturity and stature were sown in a randomized complete block design with split-plot arrangement and three replicates. Main plots were sowing time and subplots (12 by 22 m) were the three genotypes. Suncross 41 is a commercial genotype that has an inclined pendulous head with medium to slow maturity; Genotype 8640 has similar maturity, but is taller; Genotype 8793 has a tall stature and erect head angle, but matures later than both 8640 and Suncross 41. The two sowing times were 13 Sept. 1991 (S1, spring) and 5 Mar. 1992 (S2, fall).
An initial population of 140000 plants ha-1 was sown on a 70-cm row spacing and thinned to 70000 plants ha-1 before the V2 stage (Schneiter and Miller, 1981). Fertilizer was applied at sowing at a rate of 60 kg N ha-1, 17.2 kg P ha-1, 45.2 kg K ha-1, and 54.4 kg S ha-1. A second application of N at a rate of 60 kg ha-1 was applied 20 d after sowing. Irrigation was scheduled twice weekly and applied with overhead sprinklers to ensure that water was nonlimiting. Treated sewage effluent water was used, resulting in some additional nutrient input. The concentration of nitrate in the effluent water was 1.1 mL L-1. Plots were maintained so that weed and pest influences were minimal.
Daily maximum and minimum temperatures and solar radiation were measured throughout the experimental period (summarized in Table 2) . Additional climatic and experimental details are reported in Bange et al. (1997).
View this table:
[in this window]
[in a new window]
|
Table 2 Average temperature and radiation for the different phenological stages of sunflower for both sowing dates. Sowing 1 is 13 Sept. 1991; Sowing 2 is 5 Mar. 1992; S to VE is sowing to emergence; VE to R1 is emergence to bud visible; R1 to R5.1 is bud visible to anthesis; and R5.1 to R9 is anthesis to physiological maturity
|
|
Measurements
Fifty plants per plot were tagged and monitored daily for estimating developmental events. Emergence (VE), bud visible (R1), and first anthesis (R5.1) stages (Schneiter and Miller, 1981) were taken to be when 50% of plants reached the specific stage. Date of the R9 stage (Schneiter and Miller, 1981) was determined when 90% of plants reached the stage (Dubbelde, 1990).
Destructive plant samples for growth measurements were taken from each plot at intervals varying from 2 to 5 d for the duration of crop growth. The total number of sampling dates was 25 for S1 (VE–R1, five samples; R1–R5.1, eight samples; and R5.1–R9, 12 samples) and 18 for S2 (VE–R1, five samples; R1–R5.1, six samples; and R5.1–R9, seven samples). Sampling for leaf area determination consisted of selecting four to five leaves at random from nine adjacent plants in a row at least 1 m from any previous samples. For each sampled leaf, length in millimeters was measured as the distance from the apex to the base of leaf blade (point of attachment of petiole), and leaf breadth in millimeters was measured across the widest portion of the blade at a right angle to the measurement for length (Schneiter, 1978). The area in square millimeters of each individual leaf was then measured using a calibrated electronic planimeter (Paton Scientific, Adelaide, Australia1)
. A total of 1108 leaves were measured for S1 and 810 leaves were measured for S2.
Data Analysis
Our first objective was to determine the most precise model for the relationship between the leaf dimension(s) and the area of the leaf. Table 3
presents the nine models compared. These were either chosen from the literature or derived for the present study based on their logical mathematical relation to the published models.
View this table:
[in this window]
[in a new window]
|
Table 3 Models for the prediction of sunflower leaf area from linear dimension that were compared in this study. B, leaf breadth; L, leaf length; LA, individual leaf area; LB, leaf length times leaf breadth; a, d, and e are coefficients; c is a constant
|
|
Each model was fitted to six groups of data (samples collected between VE and R1, R1 and R5.1, and R5.1 and R9 for all genotypes at each of two sowing dates bulked across genotypes). For each group, the fitted models were ranked according to their residual mean squares (RMS; Cousens, 1985). The relative merits of defining LA with models with different degrees of freedom were tested using F-tests based on the RMS of each model. The model with the lowest RMS most frequently across all groups was regarded as the best model.
Having selected the best model, we analyzed differences in regression coefficients in the model between the sowing dates, genotypes, and times of leaf sampling. Forward stepwise linear regression analyses using Genstat Version 5 (Lawes Agricultural Trust, IACR, Rothamsted, UK) were used to test for significant differences among these treatments. Significant differences are expressed at the 95% (P < 0.05) confidence level unless otherwise stated.
|
Results
|
|---|
Comparison of Models
The RMS of the fitted models show that Model 9 fitted the data best in three groups, Model 6 was best in two groups, and Model 3 was best in one group (Table 4)
. As expected, there was a strong correlation (r = 0.895) between L and B for leaves of different sizes; however, the best models chosen for each group of data required the inclusion of the LB interaction term.
View this table:
[in this window]
[in a new window]
|
Table 4 Residual mean squares (RMS) for fitted models of individual sunflower leaf area compared in this study. Sowing 1 was 13 Sept. 1991, and Sowing 2 was 5 Mar. 1992; VE, emergence; R1, bud visible; R5.1, anthesis; R9, physiological maturity. Model no. corresponds to those given in Table 3
|
|
There were also marked differences within groups in the relative accuracy of models based on L or B, particularly before anthesis, and one was not consistently better than the other. In addition, no one model utilizing a single dimension (breadth or length) was best for any group of data. In five of the six groups, the quadratic term in unidimensional models (Models 5 and 6) significantly reduced the RMS compared with the relevant one-term model.
Comparison of Treatments
Model 9 and forward stepwise linear regression analysis were used to test for significant differences in coefficients derived for the different sowings, genotypes, and times of leaf sampling. As a first step, Model 9 was fitted to all the data to derive an overall relationship between individual leaf area (in square millimeters) vs. linear leaf dimensions (in millimeters):
 | (1) |
Separating sowing times led to a significant improvement (P < 0.001) in the relationship for the pooled data due to changes in the slope and intercept. The relationship between individual leaf area vs. leaf dimensions for Sowing 1 was described by
 | (2) |
For Sowing 2, individual leaf area vs. leaf dimensions was described by
 | (3) |
For all three relationships, the constant (c) was not significantly different from zero.
When genotype was added as a factor, no significant improvement over the original relationship was found. Sampling date was then added as a factor to each of the sowing dates individually, and significantly improved the relationships for both sowing dates (P < 0.001). In S1, the L coefficient (a; -16.0), the B coefficient (d; 28.1), and the LB coefficient (e; 0.64) did not vary between the times of sampling. Only the constant (c) changed between times of sampling (Table 5)
. For S2, with the exception of the LB coefficient (e; 0.67), coefficients for L (a), B (d), and constant (c) varied among sampling times (Table 6) . Figure 1
shows the relative precision of using Model 9 with sowing time and time of sampling taken into account.
View this table:
[in this window]
[in a new window]
|
Table 5 Fitted constants (c) for the relationship between individual leaf area (mm2) vs. leaf dimensions of sunflower for each sampling in Sowing 1 (13 Sept. 1991)
|
|
View this table:
[in this window]
[in a new window]
|
Table 6 Fitted coefficients and constants for the relationship between individual leaf area (mm2) vs. leaf dimensions of sunflower for each sampling in the second sowing (5 Mar. 1992). B, leaf breadth; L, leaf length; a, and d are coefficients; c is a constant
|
|
View larger version (19K):
[in this window]
[in a new window]
|
Fig. 1 Actual vs. predicted values of leaf area (in mm2) of sunflower for (a) Sowing 1 (13 Sept. 1991), and (b) Sowing 2 (5 Mar. 1992) using model LA = aL + dL + eLB + c and accounting for variation in sowing time and time of leaf sampling, where B is leaf breadth, L is leaf length, LA is individual leaf area, and a, c, d, and e are constants
|
|
|
Discussion
|
|---|
The most appropriate model for the relationship between linear dimensions and area of an individual leaf included both the length and breadth dimensions. While simpler models (Models 3 and 6) were best for some data sets (Table 3), Model 9 was considered the best overall because it was ranked highly in three of the six data sets and embraces the terms in the other three. In each case, the model that proved the best included the LB interaction term. While these are more complex models, they have a distinct advantage in that they are able to accommodate the potential effects of changing leaf shape during crop development. This is also supported by the literature, in that models with the LB factor were most commonly used (Table 1). Similar results also were found in the study by Schneiter (1978).
The relationship between linear dimensions and area of an individual leaf was significantly improved when sowing time (Eq. [2] and [3]; Fig. 1) and the times of leaf sampling were considered (Tables 5 and 6). Singh et al. (1995) had considered the effect of leaf sampling times and had separate models for pre- and postanthesis. In our study, the different temperature and radiation environments in which these leaves were grown (Table 2) may have caused the differences found between the sowing dates and leaf sampling times. In S1, leaf initiation and development experienced increasing average temperature and radiation, while S2 experienced the converse.
No significant differences were found among the genotypes evaluated. This suggests some stability of the relationship between individual leaf area and linear dimensions across closely related genotypes. While these cultivars differed in developmental rates and other morphological traits, whether the relationship will hold across more genetically divergent cultivars is unknown. With the exception of Goyne et al. (1978), all of the other models outlined in Table 1 were used across different genotypes; however, none of these models had been specifically tested for differences in measuring individual leaf area across genotypes.
The approach selected by an individual researcher is determined by the time available to take measurements and the level of precision required. While measuring both breadth and length has been found to be significantly more precise than using one dimension, it requires twice the number of measurements to be taken. In fitting the best single-dimension models for L (Model 7) and B (Model 8) to all the data, the r2 values were 0.85 and 0.90, respectively. This compares with an r2 value of 0.94 when fitting Model 9 (LA = aL + dL + eLB + c) to all the data and an r2 value of 0.96 when time of sowing and sampling date is taken into consideration. Thus, a high level of precision will be attained by at least collecting a random sample of leaf sizes at different sampling dates and at different times of sowing.
|
Conclusions
|
|---|
More accurate determination of individual leaf area required measuring both the dimensions of leaf length and breadth. In addition, using a model that accounted for changes in leaf shape during crop development added to precision. Neither length nor breadth appeared consistently better than the other as a basis estimating leaf area. Both length and breadth measurements were needed to attain precision for all leaf sizes, and given the availability of computer-based statistical packages, the use of relatively complex models in the assessment of individual leaf areas would appear to be a practical option.
|
ACKNOWLEDGMENTS
|
|---|
The assistance of the staff of Gatton College, APSRU, and particularly the support of Karen Healey and Shane Campbell, are gratefully acknowledged.
|
NOTES
|
|---|
1 Product names are provided for the reader's information and are not meant to imply superiority to other similar products. 
Received for publication July 28, 1999.
|
REFERENCES
|
|---|
- Bange M.P., Hammer G.L., Rickert K.G. Environmental control of potential yield of sunflower in the sub-tropics. Aust. J. Agric. Res. 1997;48:231-240.
- Cousens R. An empirical model relating crop yield to weed and crop density and a statistical comparison with other models. J. Agric. Sci. (Cambridge) 1985;105:513-521.
- Dubbelde E.A. Sunflowers in a semi-arid environment. Armidale, Australia: The Univ. of New England, 1990 Ph.D. diss..
- Goyne P.J., Woodruff D.R., Churchett F.D. Environmental causes of yield variation in raingrown sunflower in Central Queensland. Aust. J. Exp. Agric. Anim. Husb. 1978;18:129-134.
- Karanja D.R. The effect of nitrogen on growth and development of three cultivars of sunflower. Lawes, Australia: The Univ. of Queensland, 1990 Undergraduate diss..
- Pereyra, V.R., C. Farizo, and F. Cardinali. 1982. Estimation of leaf area on sunflower plants. p. 21–23. In Proc. 10th Int. Sunflower Conf., Surfers Paradise, Australia. 14–18 Mar. 1982. Int. Sunflower Assoc., Paris.
- Rawson H.M., Constable G.A. Carbon production of sunflower cultivars in field and controlled environments. I. Photosynthesis and transpiration of leaves, stems and heads. Aust. J. Plant Physiol. 1980;7:555-573.[Web of Science]
- Rawson H.M., Dunstone R.L., Long M.J., Begg J.E. Canopy development, light interception and seed production in sunflower as influenced by temperature and radiation. Aust. J. Plant Physiol. 1984;11:255-265.
- Schneiter A.A. Non-destructive leaf area estimation in sunflower. Agron. J. 1978;70:141-142.[Abstract/Free Full Text]
- Schneiter A.A., Miller J.F. Description of sunflower growth stages. Crop Sci. 1981;21:901-903.[Abstract/Free Full Text]
- Singh D., Singh S., Rao V.U.M. Alternate methods for leaf area measurement of sunflower. Ann. Arid Zone 1995;34:145-146.
- Stace, H.C.T., G.D. Hubble, R. Brewer, K.H. Northcote, J.R. Sleeman, M.J. Mulcahy, and E.G. Hallsworth. 1968. A handbook of Australian soils. CSIRO and ISSS, Rellim Tech. Publ., Glenside, SA, Australia.
- Trehan K.B., Chand H., Mehta S.K. Measurement of leaf area estimation in sunflower (Helianthus annuus L.). Sci. Cult. 1975;41:238-239.
This article has been cited by other articles:
|
|
|
|
 
H. Tewolde, K. R. Sistani, D. E. Rowe, A. Adeli, and T. Tsegaye
ESTIMATING COTTON LEAF AREA INDEX NONDESTRUCTIVELY WITH A LIGHT SENSOR
Agron. J.,
June 17, 2005;
97(4):
1158 - 1163.
[Abstract]
[Full Text]
[PDF]
|
|
|
TOP
ABSTRACT
INTRODUCTION
