NONDESTRUCTIVE ESTIMATION OF DRY WEIGHT AND LEAF AREA OF PHALAENOPSIS LEAVES

Transcription

1 NONDESTRUCTIVE ESTIMATION OF DRY WEIGHT AND LEAF AREA OF PHALAENOPSIS LEAVES C. Chen, R. S. Lin ABSTRACT. A nondestructive dry weight estimation for growth models of Phalaenopsis was developed which allowed the area and dry weight of Phalaenopsis leaves to be estimated without removing leaves from the plants of four cultivars. Four dimensions of physical properties of Phalaenopsis leaves were measured in this study. These dimensions were length, maximum width, area, and a new parameter, K, which was calculated as the product of maximum width and length. A linear equation excluding the intercept was a good model for leaf area and K value. The linear function between logarithmic dry weights and logarithmic K values gave the best predictive equation. No significant difference in prediction model was found among data sets of three Grandiflora cultivars. However, significant differences existed between the Grandiflora and Multiflora cultivars. The leaf dry weight and area of Phalaenopsis can be estimated indirectly by the measurement of leaf length and maximum width, and then calculated by an empirical equation. Keywords. Orchids, Phalaenopsis, Dry weight, Regression equation. Phalaenopsis is a member of the orchidaceae. Because of the richness of cultivars and favorable climate, this orchid has become the most economically important plant in the flower industry in Taiwan. In 001, 0 million units were exported. To meet the increased demand, the production of Phalaenopsis needs industrial management (Su et al., 001). Crop growth mathematical models can be used to manage practical production in the horticultural industry (Challa, 1997; Jones et al., 1991). Growth models can predict crop growth and corresponding development influenced by various cultivating factors. A rose growth model was developed successfully (Lieth and Pasian, 1990; 1991; 1993) and was applied to optimize production (Lieth, 1996). Understanding crop dry matter partitioning is fundamental to a production system for obtaining the optimum culture technique. The traditional way to study crop dry matter partitioning patterns is to harvest and dry the plant materials. This technique is time consuming, parts of plants are destroyed and the growth rate is adversely affected. Destructive or nondestructive techniques can be used to measure the dry matter of plant parts. Until now, no other standard methods can be found. With destructive methods, samples are dried and weighed. Nondestructive methods are called indirect methods. Ondark (1971) suggested regression analysis to establish the linear relationship between plant part dry weight and various dimensions of length or Article was submitted for review in August 003; approved for publication by the Soil & Water Division of ASAE in February 004. The authors are Chiachung Chen, Professor, Department of Bio Industrial Mechatronics Engineering, and Ruey Song Lin, Professor, Department of Horticulture, National ChungHsing University, Taichung, Taiwan. Corresponding author: Chiachung Chen, Department of Bio Industrial Mechatronics Engineering, National ChungHsing University, 50 Kuokuang Road, Taichung, Taiwan, 407; phone: ; fax: ; e mail: ccchen@dragon.nchu.edu.tw. diameter. This technique can estimate the dry weight of plant parts without damaging plant structure. Mohsenin (1970) has proposed a linear model for area (A) and weight (M) of agricultural products: A = c 1 LW, and M = c LW, where L is the product length, W is the product width, and c 1 and c are constants. The surface area (A) of tobacco leaves were estimated by the equation A = b 0 LW, where b 0 is a constant, L is the leaf length, and W is the leaf width. To improve the predictive ability, Oh et al. (001) proposed a new model to estimate the area of tobacco leaves by an asymptotic regression equation. Daughtry (1990) reviewed the relationship between leaf areas and dry weights of leaves. The author indicated that cultivars, growing seasons, nutrition supply, and environmental variables such as temperature and solar radiation all affected leaf area. The dry weight of plant more meaningfully represents the growing conditions than that of leaf area. Pasian and Lieth (1994) used polynomial regressions to develop a set of equations that approximated the dry weights of rose leaves, stem sections, and flowers with simple measurements of plant dimensions. They computed the dry weight of a leaf with leaf length, dry weight of a stem with a stem diameter, and a flower bud dry weighs with flower bud diameter. Leaves are the major source of photosynthesis for Phalaenopsis and are an important parameter to consider in development of a growth model. The objective of this study was to develop a leaf dry weight and leaf area model specific to four cultivars of Phalaenopsis using simple leaf length and width measurements and regression analysis. METHODS AND MATERIALS Phalaenopsis plants were potted into 0.6 L plastic containers, diameter of 8.9 cm, filled with New Zealand sphagnum moss. The photograph of a Phalaenopsis plant is show in figure 1. All plants were placed on benches in a greenhouse with a photosynthetic photon flux (PPF) ranging Vol. 0(4): Applied Engineering in Agriculture 004 American Society of Agricultural Engineers ISSN

2 where Y is the measured value, Y is the value predicted by the model, and df is the degrees of freedom of the regression model. The validity of the model was evaluated using residual plots. Data points in the residual plots should tend to be a horizontal band centered around zero for an adequate regression model. Data sets were compared using an F test at the 95% significance level to determine if significant differences existed (Weisberg, 1986). Independent data sets, measured from the red hybrid and white hybrid, were used to evaluate the estimated ability of models. The predictive data was independent and was used to establish these adequate models. Figure 1. A Phalaenosis plant. between 10 (Jan.) and 55 (July) µmolm s 1. Greenhouse air temperature ranged from 18 C to 30 C. The plants were fertigated with Hyponex 5N 5P 0K soluble fertilizer (Hyponex Corp., Marysville, Ohio). When the medium surface appeared dry, plants were irrigated with 50 ml water. Deionized water without fertilizer was used every two weeks to avoid salt accumulation in the medium. The plants were grown for 11 months before leaf sampling. Four popular cultivars planted in Taiwan were selected. Red: Grandiflora cultivar, Dtps. (Ton Jy Pecan Ruey Lin Stripe) TH Semi alba: Grandiflora cultivar, Dtps. (New Grad Dalin Beauty) 88 White: Grandiflora cultivar, P. (Taisuco Kochdian) KH# Waxy yellow: Multiflora cultivar, P. (Taipei Gold ) GS Leaf area, length, and maximum width were measured with a LI 300A area meter (L1 COR, Inc., Lincoln, Nebr.). Before measuring, this area meter was calibrated with 1 and 5 cm standard area plates. The precision of this meter was 1 mm after calibrating Phalaenopsis samples had four or five leaves. Debris and soil were washed from the leaves. After measurement, samples were placed in an aluminum box and put in a drying oven. After drying, samples were removed and placed in a desiccator until they cooled to room temperature. Samples were then weighed. The dry weights of Phalaenopsis leaves were measured after air oven drying at 75 C for 7 h. No change of weight was found after a longer drying time. The final weights were measured with an electronic balance (Mettler PM400, HP, Palo Alto, Calif.). The accuracy of this balance was ±0.001 g. Data were analyzed by statistical software (SigmaStat, SPSS Inc., Chicago, Ill.) to evaluate the adequacy of the estimated model. Two quantitative standards, R squared values (R ) and standard error of the estimated value(s), were used to compare the fitting agreement of the empirical model. s = Y Y ) df (1) RESULTS AND DISCUSSION A new parameter, K, was defined as the product of maximum width (W max ) and length (L). The leaf physical characteristics of four Phalaenopsis cultivars are listed in table 1. THE RED HYBRID There were 5 data sets for the red hybrid cultivar. Coefficient of correlation values (r) were higher than 0.94 for the relationship of dry weight and K, dry weight and leaf area, and K and leaf area (table ). The relationship between dry weight and K values is presented in figure. The scattering data distribution was found for larger K values. Figure 3 shows the relationship between leaf area and K value. Comparing r value in table and the data distribution of figures and 3, the product of length and maximum width, the K value, was selected as the parameter to estimate the leaf areas and dry weights. The regression equation for leaf area (A) and K value: linear model including the intercept A = K, s = 3.66, R = () linear model excluding intercept A = K, s = (3) Table 1. Leaf characteristics of four Phalaenopsis cultivars. Red White Semi alba Waxy Hybrid Hybrid Hybrid Yellow Hybrid No. of data Length (cm) Min Max Avg S.D Maximum Min width Max (cm) Avg S.D Area (cm ) Min Max Avg S.D Dry weight Min (g) Max Avg S.D APPLIED ENGINEERING IN AGRICULTURE

3 Table.Correlation coefficients of five leaf characteristics of red hybrid Max. Width Dry Weight K [a] Area Length K 1 Area Length Max. width Dry weight [a] K = L W max From the comparison of s (standard error of the estimated values), it was found that the fitting agreement of equation 3 was better than equation. Uniform distribution of residual plots was found for each of the two linear equations. It indicated the adequacy of this model. The regression model for dry weights and K values was: D w = K, R = 0.856, s = 0.16 (4) The R value of this equation was high, however, the residual plot (fig. 4) indicated a funnel type distribution. The error distribution was not uniform. As the predicted values increased, the errors also increased, indicating that the Figure. Relationship between dry weights and K values of red hybrid Figure 4. Residual plots of linear model for the relationship between dry weights and K values of red hybrid variance of errors did not remain constant (Weisberg, 1986). This result revealed that the linear model could not be used to represent the relationship between dry weights and K values. Non variance problems can be solved for regression analysis by transforming the data. The logarithmic form of the parameter was selected to reduce the data variance. The relationship between logarithmic dry weights and logarithmic K values was stable (fig. 5) and was used to develop the following: ln Dw = ln K, R = (5) The uniform distribution of residual plots indicated that transformation of parameters stabilized the error distribution due to sample variance (fig. 6). Dependent variable (D w ) and independent variable (K) of equation 5 can be transformed back to the original form. D w = K 1.71 (6) The standard error of the estimated value(s) was calculated by originally measured data and predicted values obtained from equation 6. The s value was Figure 3. Relationship between leaf area and K values and red hybrid Phalaenopsis leaves. Figure 5. Relationship between logarithmic dry weights and logarithmic K values of red hybrid Vol. 0(3):

4 Figure 6. Residual plots of linear model for the relationship between logarithmic dry weights and logarithmic K values of red hybrid Phalaenopsis leaves. Figure 7. Relationship between dry weights and K values of semi alba hybrid THE SEMI ALBA HYBRID Two regression equations of leaf area for this cultivar were shown below: linear model including the intercept: A = K, R = 0.931, s = 4.35 (7) linear model excluding intercept: A = K, s = 4.05 (8) The linear model without intercept had a better fitting ability than that of the linear model including intercept when the s values were compared. The relationship between dry weight and K value is shown in figure 7. As the K values increase, the data become more scattered. The residual plots of the linear model for dry weights and K values also had a funnel type distribution. After the original values of dry weights and K values were logarithmically transformed, the relationship between logarithmic dry weights and logarithmic K values is presented in figure 8. The regression model was shown as: ln Dw = ln K, R = (9) After the logarithmic values were transformed into their original form: Dw = K (10) The s value calculated from the original form of the measured data and predicted values was THE WHITE HYBRID The adequate regression model for dry weights and K values of the White hybrid cultivar had the same form as the Red hybrid and Semi alba hybrid: A = 0.779K, and D w = K THE WAXY YELLOW HYBRID The waxy yellow hybrid is a Multiflora cultivar. From the values in table 1, the dimension of this cultivar was smaller than that of the other Grandiflora cultivars. The adequate equations obtained by regression analysis for leaf area or dry weights and K values had the same form as the three Figure 8. Relationship between logarithmic dry weights and logarithmic K values of semi alba hybrid Grandiflora cultivars. The adequate model was: A = 0.707K, and D w = K The data sets for the three Grandiflora cultivars were pooled. Figure 9 indicates the relationship between the leaf area and K values of all data sets. The adequate regression model is shown as: A = 0.71K, s = 4.65 (11) Figure 10 shows the relationship between dry weights and K values. When K values were larger than 150 cm, the data points were more scattered. The relationship between logarithmic dry weight and logarithmic K values is presented in figure 11. The regression for the two logarithmic parameters was: ln Dw = ln K, R = 0.93 (1) After transforming back into the original unit and calculating the s value, Dw = K, s = (13) No significant difference between the three data sets of the Grandiflora cultivars could be found from the F test at the 95% significance level. It indicated that equation 11 could be 4 APPLIED ENGINEERING IN AGRICULTURE

5 Figure 9. Relationship between leaf area and K value of three Grandiflora applied to the three Phalaenopsis cultivars in this study. The same conclusion can be drawn for the dry weight model. Equation 13 could also be applied to these three Grandiflora Phalaenopsis cultivars. Four data sets of leaf area and K were analyzed with the same statistical procedure. There was a significant difference among these data sets. Four data sets of dry weights and K were analyzed and we found that the difference among these data sets was statistically significant. The empirical equations 11 and 13 only can be applied for Grandiflora cultivars. They cannot be used for the Multiflora cultivar. Figure 11. Relationship between logarithmic dry weight values and logarithmic K values of three Grandiflora APPLICATION OF MODELS In this study, physical properties of four cultivars of Phalaenopsis leaves were measured. Linear regression was usually applied to find the relationship between areas or dry weights of leaves and other dimensions. However, the PREDICTIVE ABILITY OF MODELS The K values of predicted data sets were fitted to equation 11 and 13. The actual values of leaf area and dry weight for independent samples and their predicted values by the models are shown in figures 1 and 13. Linear distribution of these data indicated the validity of these models. The average absolute difference between measured values and predicted values was 4.4 cm for leaf area and g for dry weights, respectively. Figure 1. Relationship between actual and predicted values of leaf areas. Figure 10. Relationship between dry weight and K value of three Grandiflora Figure 13. The relationship between actual values and predicted values of leaf dry weights. Vol. 0(3):

6 residual plots indicated the inadequacy of the linear model. A transformed model improved the predictive ability and reduced the variance of model errors. An adequate model for dry weights of three Grandiflora cultivars is D w = K 1.013, and for area is A = 0.71K. An adequate model for a Multiflora cultivar for leaf dry weights is D w = K 1.19, and leaf area is A = 0.707K. The standard error of the estimated value(s) of regression equation can be used to express the uncertainty of measurement (ISO, 1995). For the Grandiflora cultivars, the uncertainty of leaf area from equation 11 was 4.65 cm. The range of actual leaf area was from 10 to 190 cm. The uncertainty of dry weight from equation 13 was g. The range of actual dry weight was from 0.1 to.0 g. For Multiflora cultivars, the uncertainty of leaf area was 4.05 cm. The range of actual leaf area was from 16 to 170 cm. The uncertainty of dry weight was g. The range of actual dry weight was from 0.1 to 1.8 g. Because the uncertainty had a fixed numerical value, the relative predicted model error was small for larger and more mature CONCLUSIONS In this study, the empirical equations of leaf area and dry weight were established. An adequate model for dry weights of three Grandiflora cultivars is D w = K 1.013, and for area is A = 0.71K. An adequate model for a Multiflora cultivar for leaf dry weights is D w = K 1.19, and leaf area is A = 0.707K. As the length and maximum width of Phalaenopsis leaf were measured, leaf area and dry weight were easy to calculate without destroying samples. The electronic digital caliper for measuring the dimension of leaf was very easy to use and inexpensive. This method is simple, rapid, inexpensive, easy to apply, and beneficial for developing the optimum culture technique. REFERENCES Challa, H On model for plant growth, environmental control and farm management in protected cultivation. Acta Hort The Netherlands: ISHS. Daughtry, C. S. T Direct measurements of canopy structure. Remote Sensing Reviews 5(1): ISO Guide to the Expression of Uncertainty in Measurement. Geneva: ISO. Jones, J. W., E. Dayan, L. H. Allen, H. Van Kelen, and H. Challa A dynamic tomato growth and yield model (TOMGRO). Transaction of the ASAE 34(3): Lieth, H Modeling roses for optimum production. Grower Talks 60(1): Lieth, J. H., and C. C. Pasian A model for net photosynthesis of rose leaves as a function of photosynthetically active radiation, leaf temperature, and leaf age. J. Amer. Soc. Hort. Sci. 115(3): Lieth, J. H., and C. C. Pasian A simulation model for the growth and development of flowering rose shoots. Scientia Horticulturae 46: Lieth, J. H., and C. C. Pasian Development of a crop simulation model for cut flower roses. Acta Hort. 38: Mohsenin, N. N Physical Properties of Plants and Animal Materials. New York: Gordon and Breach Science Publishers. Oh, I. H., S. H. Jo, and K. S. Rhim Measuring tobacco leaf area by numerical integration of asymptotic regression equations. Applied Engineering in Agriculture 17(1): Ondark, J. P Indirect estimation of primary values used in growth analysis. In Plant Photosynthetic Production: Manual of Methods, , eds. J. Sedeak, J. Castshy, and P. G. Jarvis. The Hugue, The Netherlands: Dr. Junk Publishers. Pasian, C. C., and J. H. Lieth Nondestructive dry matter estimation of rose leaves, stems, and flower buds using regression models. HortScience 9(3): Su, V., B. Hsu, and W. Chen The photosynthetic activities of bare rooted Phalaenopsis during storage. Scientia Horticulturae 87: Weisberg, S Applied Linear Regression. Boston, Mass.: Pws Publisher. 6 APPLIED ENGINEERING IN AGRICULTURE