Experiments are much more worthwhile if they have predictive value applicable to large areas or regions. However, this basic requirement is commonly either overlooked or given so little thought, that many experiments are sometimes not truly representative. It is dangerous to extrapolate results to conditions that are very different from those of the study area. This often occurs because experiments are located in accessible or convenient areas, a practice usually advisable only for exploratory research (Wadsworth 1997).
The interpretation of research results should consider all the important values such as wood quality and non-wood products. Plantation productivity, for example, is usually expressed in terms of volume (or aboveground biomass) per unit of area and time. However, under some circumstances, the maximum volume yield per unit investment or per unit of employment may be most meaningful (Wood 1974, Wadsworth 1997).
Excessive site heterogeneity and strong border effects can results unless these are considered during the experimental design. Long-term (> 10 years) research, needed to provide reliable recommendations, requires:
1) A diversified agenda that includes a “basket of research alternatives”, including extreme treatments, to respond to unpredictable conditions, a varied clientele and changing perceptions;2) Replicated experiments with good error control, located over a range of sites. This allows for losses due to natural disasters, human interference or shifts in motivation, plus sound application of results and perhaps future modelling;
3) Large plots (e.g. 36 × 36 m) to allow for possible thinning, with adequate treated borders (Powers et al. 1994, Somarriba et al. 2000).
