We elaborated and tested a novel operative framework for sampling and analysing fine-scale pattern of plant composition and biomass. We combined presence/absence sampling of plant species with non-destructive biomass estimation. In an open perennial sand grassland, we used 46 m long circular transects consisting of 0. 05 m by 0. 05 m adjoining elementary sampling units. This arrangement allows us to scale across a range of 0. 05 to 20 m. For measuring aboveground green biomass, we applied digital camera sensitive to red and near infrared parts of light spectrum, and we calculated normalised differential vegetation index (NDVI). We used information statistics proposed by Juhász-Nagy to study the association between spatial patterns of production and species composition. Since information statistical functions applied require binary data, we transformed NDVI data into one or several binary variables. We found that not only dominant species but subordinate gap species were also associated to high biomass, although the strength of association varied across scales. Most of the significant associations were detected at fine scales, from 0. 05 to 0. 25 m. At the scales commensurable with quadrat sizes usually applied in grasslands, i. e., from 0. 5 to 2. 0 m, we could hardly find any significant associations between species composition and biomass. We concluded that the novel methods applied proved reliable for studying fine-scale relationships between species composition and biomass.