Published online by Cambridge University Press: 27 March 2009
Modifications to the New Zealand meter (Campbell, Phillips & O'Reilly, 1962) are briefly described and the results obtained with the modified meters presented.
Water was shown to have the dominant influence on meter reading—dessication of the plant material reduced readings, application of water increased readings.
The relation between yield of water (W) and meter reading (MR) depended on the circuitry of the meter used and was similar to that between capacitance change induced by a variable capacitor and meter reading. With the new meters this relation was strongly linear. For individual species the simple correlation of W on MR was generally superior to that for weight of dry matter. Regression coefficients for weight of water on meter reading differed between species and/or days but were correlated with the relative water content of the species (RW). Inclusion of RW in addition to that of W increased the variation accounted for (V%) when all species were included from 76·0 to 86·9%. However, the multiple regression did not eliminate bias. For individual species, there was no benefit in including the RW term.
The inclusion of height of herbage (H) in a multiple regression equation with weight of water greatly improved the V% when compared with the regression of meter reading on weight of water alone for oats dressed with three levels of nitrogen. The V% increased from 71·3% for W alone to 90·0% when W, H and H2 were included in the regression. Further testing using height of herbage over a wide range of species will be required to confirm this improvement for use in an overall regression.
Emphasis is placed on the very strong linear relation between yield (of water or dry matter) and meter reading within any one pasture at any one time. Use is made of this relation in the presentation of a new method, which overcomes the bias associated with prediction equations, for estimating pasture yields. This entails establishing the mean reading for any pasture and cutting for yield estimation three samples having such a mean. The limitations of the method for use with mixtures is discussed.