Comparison of different models for nondestructive leaf area estimation in taro

Leaf area is a valuable index in identifying taro [Colocasia esculenta (L.) Schott] growth and development. Several models for estimating the area of a taro leaf by using nondestructive measurements of length and/or width have been proposed. We evaluated these models as well as some modified versions of the models and analyzed variations in leaf area coefficients (K) across leaf developmental stages, crop seasons, and cultivars. Data were collected from experiments of two taro cultivars at three leaf developmental stages grown in two crop seasons at Taiwan Agricultural Research Institute. The models using both leaf length [maximum leaf length (L-X), length from the sinus base to the apex of leaf (L-SA), or length from petiole-attaching point to the apex of leaf (L-PA)] and leaf width (maximum width, leaf width passing the petiole-attaching point and perpendicular to L-PA, or leaf width passing the sinus base and perpendicular to LSA) factors (LW1-LW9) provided the most accurate estimations of taro leaf area. Using these models, the mean squared deviation (in the range 8874862), percentage of deviation for total leaf area (<3%), and mean percentage of deviation for individual leaf area (<8%) were smaller than those from other models. However, model L1, which used only the length factor L-SA, could be extended to the area estimation of nonexpanded taro leaves or leaves at initial wilting stage. Models LW1 to LW9 and L1 gave consistent K values across leaf developmental stages, crop seasons, and cultivars. With these models, estimating taro leaf area in large quantities could be done without the use of any expensive instruments.