Js. Ebdon et al., EVALUATION OF DISCRIMINANT-ANALYSIS IN IDENTIFICATION OF LOW-WATER AND HIGH-WATER USE KENTUCKY BLUEGRASS CULTIVARS, Crop science, 38(1), 1998, pp. 152-157
Identification of water conserving Kentucky bluegrass (Pen pratensis L
., KEG) is an important objective in turfgrass breeding programs. Warm
-season turfgrass selections with low evapotranspiration (ET) rates ha
ve been successfully identified by components of canopy resistance to
ET and leaf area. The objective of this study was to determine the eff
ectiveness of discriminant analysis in distinguishing water conserving
KEG on the basis of canopy resistance and leaf area from a population
of 61 KEG cultivars. By means of cluster analysis, the 61 KEG cultiva
rs were categorized as either low-or high-water use cases based on ET
rate evaluated in the growth chamber at three VPD environments (1.263,
1.664, and 2.261 kPa). Fourteen morphological and growth characterist
ics were assessed in the greenhouse with unmowed, spaced plants and mo
wed turfgrass (2O-cm-diam. lysimeters), Eased on single plant morpholo
gy, five-and seven-variable discriminant functions were identified tha
t correctly classified eases into their actual water use groups with a
n estimated 70.5% actual error rate from cross-validation with the lea
ve-one-out method (LOER), Compared with single plant morphology, turfg
rass morphology was more efficient in requiring fewer predictors (henc
e fewer measurements) to perform classification. Based on turfgrass mo
rphology, two-and three-variable functions were identified that correc
tly classified an estimated 75.4% of the cases into their true water u
se groups. A 75.4% correct classification was the best achieved and wa
s as good as that obtained with all 14 original variables in the analy
sis simultaneously. Leaf angle, a component of canopy resistance, was
the most important discriminator of water use group, predicting actual
group membership in 72.1% of the cases. Correct classification was im
proved only slightly over leaf angle alone by incorporating a single l
eaf area component such as leaf width or leaf extension rate, These re
sults show that discriminant analysis may be an efficient and useful t
ool for predicting the water use patterns of new cultivars on the basi
s of a few plant measurements that are routinely assessed by breeders.