per acre, Nevens (1945) found that subtract- 
ing the yield of the grazed plots for the pre- 
vious period from the current yield of the 
caged plot gave the best results. 
The Joint Committee of the American Soci- 
ety of Agronomy, American Dairy Science As- 
sociation, and the American Society of Animal 
Production (1943) recommend two movable- 
cage procedures: (1) Paired caged and grazed 
plots are both clipped to ground level at the 
end of each grazing period to obtain consump- 
tion by difference, or (2) the caged plot is hand 
plucked at the end of each grazing period to 
match use on the grazed plot. Both procedures 
require a new set of caged and grazed plots for 
each grazing period. Usually the first plot of 
each pair is chosen at random; and a second 
plot, nearby and similar to the first, is then 
chosen as its mate; the plot to be caged is then 
determined by flipping a coin (Klingman et al. 
1943). 
The reliability of caged plots, paired with 
grazed plots, for determining herbage yield 
was evaluated by Klingman et al. (1948). On a 
12-acre pasture where the average yield of for- 
age was 650 to 700 pounds per acre, 308 singly 
placed pairs of 4-foot-square plots were needed 
to estimate dry matter with a standard error 
of 50 pounds. Only 12 pairs of plots were 
needed for a standard error as large as 250 
pounds. These values suggest the intensity of 
sampling needed in intensive grazing studies to 
estimate gross herbage production or consump- 
tion. The variability involved in most range 
management and wildlife habitat research 
probably is greater than that in Klingman’s 
study. 
Grazed-Plant Methods 
Various adaptations of the grazed-plant 
method are used for determining grass utiliza- 
tion on rangeland. Usually 100 tufts are ob- 
served, and a previously developed regression 
curve is used to convert the number not grazed 
to an estimate of percent utilization. One 
shortcoming of the grazed-plant method is that 
cattle may graze without changing the percent- 
ages of grazed and ungrazed plants, by regraz- 
ing some plants close to the ground, and by 
leaving ungrazed plants of the same species 
untouched (Zemo 1968) .? 
Another deficiency of the grazed-plant method 
is that, to estimate consumption, it must be 
used in conjunction with some other method 
that measures production. Consequently, pro- 
*Zemo, Tesfay. 1968. Behavior and grazing prefer- 
ence of fistulated steers on a desert grassland. (Unpub- 
lished master’s thesis on file at the Univ. Ariz., Tucson, 
Ariz.) 
94 
duction and utilization are not measured on the 
same plants. 
The counts in plant-count methods are bino- 
mially distributed. For example, a sample of 
250 tufts is needed for a confidence interval 
within 20 percent of the mean if 30 percent of 
the plants are ungrazed. If fewer than 30 per- 
cent of the plants are ungrazed, more than 250 
plants must be observed for the same precision. 
Sampling Errors 
Green (1952) found the coefficient of varia- 
tion for herbage production to be between 20 
and 35 percent when the yield lay between 
1,000 and 2,500 pounds of dry matter per acre 
and sample units were 1 or 2 square yards. 
With yields as small as 500 pounds, as on 
grazed areas, the coefficient rose to 60 or 80 
percent. Sampling variation associated with 
grazed residues was high, partly because of low 
yield. Where estimates of dry matter consumed 
within 2 to 3 weeks ranged from 3,000 to 200 
pounds, the maximum standard deviations 
ranged from 700 to 500 pounds per acre at 
the highest and lowest mean values, respec- 
tively. Thus, while the coefficient of variation 
was 28 percent at 3,000 pounds and 250 per- 
cent at 200 pounds per acre, the absolute value 
of the standard deviation was almost as great 
where consumption was 200 pounds per acre as 
it was at 3,000 pounds. Standard deviations of 
500 pounds per acre are greater than we like 
—perhaps too great to be useful—on ranges of 
low productivity. Larger sample units will 
sometimes reduce plot-to-plot variations to an 
acceptable level. 
Use of permanent plots is supposed to elimi- 
nate the additional sampling error encountered 
when new sample plots are selected. This ad- 
vantage is partly offset because the mean from 
a set of randomly selected permanent plots 
may differ greatly from the pasture mean. The 
combined data for several short grazing pe- 
riods, using new plots for each period, may 
give more accurate averages than would be ob- 
tained from repeat measurements on a single 
set of plots. Nevertheless, repeated measure- 
ments on permanent plots provide an index to 
relative changes in the amount of forage on a 
given range unit with less effort than it can be 
obtained by choosing new plots each time. 
A major deficiency of permanent plots is 
that they cannot be clipped to measure herbage 
yields. Therefore, forage production and utili- 
zation must be estimated. Ocular estimates are 
both tedious and subjective. However, instru- 
ments that accurately and rapidly measure dry 
matter without disturbing the vegetation are 
being developed. Such instruments may elimi- 
nate ocular estimates. They may even eliminate 
the need for permanent plots. 
