Evaluation of the whole unit is difficult in annual crop systems and even harder in agroforestry systems, which require much larger plots than annual crops (see later). Sampling is inevitable in most situations. This allows greater control of the measurement processes, which often results in estimates of higher precision than those taken from entire units. If we are harvesting and weighing a sample of just 50 m 2 from a 1000 m 2 unit, it is not difficult to ensure that only the correct sample is harvested, the correct components are weighed and the weighing is accurate and recorded correctly. Harvest of the complete plot is difficult to manage, with losses, confusion and mistakes common. In sampling we introduce sampling error (due to only measuring part of the unit) but more than compensate for this by the reduction in measurement error (all those factors that result in a recorded yield that is not what was actually harvested).