Long-term monitoring is an integral part of land management and biodiversity conservation. Sampling grain, a key component of monitoring design, can impact conclusions about spatial patterns in composition, but less is known about how sampling grain influences our ability to detect temporal compositional dynamics.

To evaluate how sampling grain affects conclusions about temporal dynamics, we analyzed vegetation data from permanently marked transects in a sagebrush-steppe ecosystem. Each transect was monitored for 9 y at three sampling grains: point-intercept, 15 × 15 cm sampling frame, and 60 × 60 cm sampling frame. We investigated grain-dependent patterns in diversity, functional group representation, and multivariate compositional change.

Inferred community dynamics were strongly affected by sampling grain. Large grains had the greatest richness and detection frequencies for nearly all species. Almost all additional species detected at larger grains were forbs, the life form that comprises most of the species richness in this system. Mean compositional change was lower and temporal compositional change was more pronounced when based on data acquired at larger sampling grains.

Small sampling grains result in undersampling, which biases measurements of community dynamics. These grains can provide reasonable estimates for coarse metrics such as the abundance of dominant species, but often fail to capture changes in nondominant species. We recommend that natural resource managers sample multiple grains based on the scales of the processes they are interested in monitoring.