CALIBRATION OF 2 TECHNIQUES FOR ESTIMATING HERBAGE MASS

Two non-destructive techniques for estimating herbage mass (dry matter yield) were evaluated during the growing season of 1988, from April to September. Samples were collected from four sites, with ten observations made on twenty-five days at each site, giving a total of 612 sample units. Observations were obtained from each unit, either with a plastic rising-plate meter (335) or with a metal rising-plate meter (227). In addition, measurements with a single-probe capacitance meter (Pasture Probe) were also made on all units. The usefulness of the non-destructive methods for predicting the actual yield was evaluated by calibrating them by means of regression on the yield determined by mowing 1.5-m2 quadrats, weighing, subsampling and drying the grass. The yield measurements were logarithmically transformed to correct for the proportionality of the residual variance with the mowed dry matter yield. Logarithmic transformation of the sensor measurements improved the linearity of the calibration relationships. The combination of one rising-plate meter and the Pasture Probe resulted in a linear and additive model with a larger adjusted R2 than models with just one of these two sensors. The addition of the predictors 'external moisture' (e.g. dew), 'day number after mowing' and 'site number' significantly increased the goodness of fit of many models. In some models addition of the logarithm of the 'dry matter percentage' was significant. Other predictors in these models such as 'Julian day number' and 'grazing' proved not to be significant. The adjusted R2s in the best models for the metal and the plastic discs were 89.7% and 87.3% respectively. The corresponding coefficients of variation were 26.2% and 26.1%. The results show that further research is necessary to refine the techniques or to develop new techniques to estimate the dry matter yield, before the sensors can be usefully integrated into a management system.