Large plots
The larger units used in agroforestry studies mean that sampled areas tend to
be large, both to satisfy the need to sample from each stratum and to obtain
sufficient precision in the presence of large variation within the units. It is not
possible to recommend a fixed sampling fraction (sample area/area of unit).
Generally it is the absolute area of the sample, rather than the sampling
fraction, that determines precision. Thus a 100 m 2 sample from a 1000 m 2
plot can, other things being equal, be expected to give about the same
precision as a 100 m 2 sample from a 2000 m 2 plot.
If the study involves the comparison of different agroforestry systems, or a
comparison with a sole crop, it is not necessary to use the same sample area
in each system. Many statistical analysis procedures assume equal variance of
all observations. If the within-plot variation in the crop yields is greater in the
agroforestry system than in the annual crop system, a larger sample area is
required in the agroforestry system to give observations from the two systems
of equal plot-to-plot variance. A similar need for increased sample areas and
plot sizes has been found when comparing annual intercrops with annual sole
crops [Davis et al., 1981].