Calculation of Sample Number to Accurately Measure Available Pasture Forage

J. L. Moyer, Southeast Agricultural Research Center, Kansas State University, Parsons 67357; and J. J. Higgins, Department of Statistics, Kansas State University, Manhattan 66506

Measurement of available forage dry matter is a fundamental practice in well-managed grazing systems. Sanderson and Rotz (3) evaluated the importance of accurately estimating available forage dry matter in pastures to dairy-cow graziers using a computer simulation model. The model indicated that estimating available forage within 10% would result in substantial profit increases. Tools for estimating available forage were described, and collection of 30 to 50 readings per 1- to 2-acre paddock was recommended for accurate estimation of available forage.

The number of readings required to obtain a designated level of precision depends on the variability of measurements within the paddock. Taking an appropriate number of readings could increase the efficiency of data collection while providing reliable results. We describe a method to estimate appropriate sample number as readings are collected based on a simple statistical formula used with a laptop computer or hand-held PDA.

If the pasture variation is uniform and the necessary precision of the mean estimate is specified, the number of readings required to accurately estimate available forage can be determined as readings are taken using a simplified formula from Kitchens (2):

N = 4 ×	S²

	D²

where N is the number of readings needed to produce the desired precision with 95% confidence, S² is the sample variance, and D is the desired precision expressed as a proportion (e.g., 5, 10, or 20%) of the mean. A laptop or PDA with Excel can be programmed to generate running estimates as readings are added.

Sample calculations from two pastures where available forage differed were shown in an Excel spreadsheet which also includes formulas used to calculate variance and desired sample number. The sample spreadsheet is available for download.

Data for available forage were collected from tall fescue (Festuca arundinacea Schreb.) pastures under grazing using disk meters constructed of plexiglass (Fig. 1). Readings were taken in random, triangular triplets; i.e., three drops 10 ft apart by two operators at the end of the grazing season. Disk height was measured to the nearest 1/8 inch. Readings taken by two operators differed, but a clear trend emerged for each after only a few readings.

Fig. 1. Details of disk meter construction.

Using a precision of 10% of the mean, Operator 1’s readings resulted in values for N that steadily declined to 25 after three triads (nine readings). Operator 2 obtained N values of 23 and 31 after six and nine readings, respectively, so he could have underestimated the number required had he quit calculating too soon. At the precision level of 20%, N values of seven and eight were obtained for Operators 1 and 2, respectively. A precision of 5% might be beyond the capability of the method, since 5% of the mean was less than 1/8 inch, the smallest measurement increment. In the second pasture, variances and N values were lower, as in Fig. 2, but the disparity between operators was greater.

In an earlier study, high nitrogen fertilization resulted in increased variance due to poor utilization and spot-grazing, as in Fig. 3, so a precision of 20% would have required an N of nearly 100. Aiken and Bransby (1) also noted difficulty in obtaining representative measurements of forage mass in underutilized areas.


	Fig. 2. Disk meter reading being taken in a uniformly grazed tall fescue pasture.		Fig. 3. Disk meter reading being taken in a tall fescue pasture with spot-grazing.

We recommend the following steps to estimate the number of readings needed to accurately predict available forage:

• Select a range of acceptable levels of precision, noting that variability is affected by management and season.

• Take measurements in uniform areas that are representative of available forage in the paddock in a random fashion. If there are areas that are distinctly different, take separate readings in each and weight results according to their relative prevalence.

• Take nine readings, calculate N for each level of precision, and continue until the results are stable or decreasing.

Literature Cited

1. Aiken, G. E., and Bransby, D. I. 1992. Observer variability for disk meter measurements of forage mass. Agron. J. 84:603-605.

2. Kitchens, L. J. 1998. Exploring Statistics, 2nd Ed. Brooks/Cole, Pacific Grove, CA.

3. Sanderson, M. A., and Rotz, A. 2004. Precision pays. The Forage Leader 9:14. Amer. Forage Grassl. Counc., Abilene, TX.