Modeling individual leaf area of ginger (Zingiber officinale Roscoe) using leaf length and width

Abstract

Leaf area estimation is an important biometrical observation one has to do for comparing plant growth in field and pot experiments. In this study, a leaf area estimation model was developed for ginger (Zingiber officinale Roscoe), using linear measurements of leaf length (L) and maximum width (W). Leaves from five ginger varieties (Varada, Rejatha, Mahima, Maran and Himachal) were used to develop the model in 2006–2007. The actual leaf area (LA) was measured with a leaf area meter (LI-3100, LI-COR, Lincoln, NE, USA) and taken as reference LA. The linear measurements were used to build linear (LA = a + b × L × W) and power models (LA = α × (L × W)^β) for each variety, as the modeling among variety were not different from each other, data for all five varieties have been pooled and compared with earlier models by graphical procedures and statistical criteria such as Mean Square Error (MSE), Root Mean Square Error (RMSE) and Chi-square (χ²). The selected model was validated during 2007–2008. The validation data set was used to produce a validation model for each variety by re-estimating the model parameters to develop the estimation model and the models were compared for consistency. The predicted LA (PLA) was compared with observed LA (OLA) by graphical procedures and lack of agreement was evaluated by calculating the relative bias, estimated by the mean of differences (d) and the standard deviation (SD) of the differences. Normality test was carried out by Spearman's rank correlation coefficient (r_s) and residuals were normally distributed. Finally, the proposed model for leaf area estimation of ginger is LA = −0.0146 + 0.6621 × L × W, R² = 0.997. This model can be reliably used for estimating leaf area of ginger non-destructively. The same equation can be extrapolated to all varieties and land races of ginger as it is vegetatively propagated crop with narrow genetic variability.

Introduction

Ginger (Zingiber officinale Roscoe), a monocotyledon belonging to family Zingiberaceae and in the natural order Scitamineae, is herbaceous perennial, usually grown as annually for its pungent rhizome. It is native of South East Asia and one of the earliest oriental spices known to Europe (Kandiannan et al., 1996, Parthasarathy et al., 2003, Ravindran and Nirmal Babu, 2005). Commercially more than 25 countries in the tropics and subtropics are producing it, however, India is the largest producer, and it is an export oriented crop (Singh and Tamil Selvan, 2003).

Leaves are important organs of the plant. Leaf area (LA) is a key variable for most agronomic and physiological studies involving plant growth, light interception, photosynthetic efficiency, evaporation, and responses to fertilizers and irrigation (Blanco and Folegatti, 2005). Therefore, LA strongly influences growth and productivity; estimating LA is a fundamental component of crop growth models (Lizaso et al., 2003). However, the measurement of the surface area of a large number of leaves is often costly, time consuming and destructive. The total LA of the plant can be obtained by either direct or indirect methods. The direct method consists of removing and measuring all leaves in plant. This method is destructive and requires expensive equipment. Indirect, non-destructive methods are user-friendly, less expensive, and can provide accurate LA estimates (Norman and Campbell, 1989) and help in in situ LA estimation. A modeling approach is rapid, reliable and involves linear relationships between LA and one or more dimension of the leaf (L – leaf length and W – maximum width) and it is an alternative for accurately measuring LA (Kandiannan et al., 2002, Williams and Martinson, 2003, Lu et al., 2004, Cho et al., 2007, Peksen, 2007, Antunes et al., 2008).

Although LA is an important parameter in growth studies, only a few workers from India, China and Australia reported for ginger (Gowda and Melanta, 2000, Xizhen et al., 2001, Ajithkumar et al., 2002, Ajithkumar and Jayachandran, 2003, Smith et al., 2004, Xizhen et al., 2005). Earlier methods, employing leaf dimension for estimating the area of ginger leaf was proposed LA = −1.7362 + 0.7153 × L × W (Jayachandran and Sethumadhavan, 1979) with coefficients estimated using least-square linear regression analysis. Similar equations were also developed later LA = −0.7608 + 0.6695 × L × W (Ancy and Jayachandran, 1994) and LA = −9.358 + 0.8549 × L × W (Reddy and Reddy, 1995). It was also proposed an equation LA = k × L × W, k = 0.666 (Reddy and Reddy, 1995), a coefficient ‘k’ derived as the ratio of LA to the product of L and W. Nwachukwu and Ene (1987) compared different methods of estimating ginger LA from Nigeria and reported that no differences were noted between grid and punch method. Punch method was not recommended because of the time and labour required for its use and other calculation methods based on leaf length and width gave unrealistic results for ginger. The above equations have been selected based only on values obtained for the coefficient of determination (R²) and without assessing their prediction accuracy by validation. They also have used limited number of samples for building equations and from only one variety Rio-de-Janeiro (Jayachandran and Sethumadhavan, 1979, Ancy and Jayachandran, 1994). Hence, no information is available on whether or not such models can be successfully used to other genotypes. Moreover, a graph paper method has been employed to find LA that might significantly underestimate the actual LA. Reddy and Reddy (1995) also used limited number of leaves from five genotypes and they have not validated their equation. In these studies, the adequacy of the model assumptions for estimating LA has not been carefully examined. A simple and effective method for detecting model deficiencies in regression analysis is the examination of residual (Bland and Altman, 1986, Rangaswamy, 1995). Therefore, it is necessary to have a simple and validated accurate model for LA estimation of ginger. In this study, we aimed to evaluate the current models (Jayachandran and Sethumadhavan, 1979, Ancy and Jayachandran, 1994, Reddy and Reddy, 1995), as well as to propose a reliable and accurate model using measurements of L and W for estimating the LA of ginger by non-destructive method.