Background and aims Lateral tree-scale variability in plantations should be taken into account when scaling up from point samples, but appropriate methods for sampling and calculation have not been defined. Our aim was to define and evaluate such methods. Methods We evaluated several existing and new methods, using data for throughfall, root biomass and soil respiration in mature oil palm plantations with equilateral triangular spacing. Results Three ways of accounting for spatial variation within the repeating tree unit (a hexagon) were deduced. For visible patch patterns, patches can be delineated and sampled separately. For radial patterns, measurements can be made in radial transects or a triangular portion of the tree unit. For any type of pattern, including unknown patterns, a triangular sampling grid is appropriate. In the case studies examined, throughfall was 79 % of rainfall, with 95 % confidence limits being 62 and 96 % of rainfall. Root biomass and soil respiration, measured on a 35-point grid, varied by an order of magnitude. In zones with steep gradients in parameters, sampling density has a large influence on calculated mean values. Conclusions The methods defined here provide a basis for representative sampling and calculation procedures in studies requiring scaling up from point sampling, but more efficient methods are needed.