Belowground vertical community composition and maximum rooting depth of the Edwards Plateau of central Texas were determined by using DNA sequence variation to identify roots from caves 5-65 m deep. Roots from caves were identified by comparing their DNA sequences for the internal transcribed spacer (ITS) region of the 18S-26S ribosomal DNA repeat against a reference ITS database developed for woody plants of the region. Sequencing the ITS provides, to our knowledge, the first universal method for identifying plant roots. At least six tree species in the system grew roots deeper than 5 m, but only the evergreen oak, Quercus fusiformis, was found below 10 m. The maximum rooting depth for the ecosystem was Ϸ25 m.18 O isotopic signatures for stem water of Q. fusiformis confirmed water uptake from 18 m underground. The availability of resources at depth, coupled with small surface pools of water and nutrients, may explain the occurrence of deep roots in this and other systems.Plant rooting depth influences the hydrology, biogeochemistry, and primary productivity of terrestrial ecosystems (1-7). Progress in determining the maximum rooting depth of species and in identifying the resources taken up at depth is limited by several factors. Access to the soil is difficult, particularly in rocky soils and in deeper layers. In addition, no universal method exists for identifying roots obtained from the soil, especially when only fine roots are available (8-10). There is considerable variation in maximum rooting depth and root biomass distributions, which affects the functioning of ecosystems (11-15). For example, in eastern Amazonia, water uptake from 2-to 8-m soil depths contributes to more than threefourths of the transpiration of evergreen forest in the dry season and helps maintain an evergreen canopy on Ͼ1 million km 2 of tropical forest (1, 16). Characteristics of roots and the soil are also needed in models of biosphere-atmosphere interactions (17). A comparison of 14 land surface parameterizations concluded that rooting depth and vertical soil characteristics were the most important factors explaining scatter among models for simulated transpiration (18, 19), determining the amount of water available to plants and partitioning its uptake from different layers. Conclusions were similar for global soil-moisture dynamics (20,21). Global simulations of net primary productivity and transpiration increased 16% and 18%, respectively, when optimized rooting depths incorporated soil water deeper than 1 m (22).We developed a method for identifying roots based on DNA sequence variation and applied this method to roots collected from caves 5-65 m deep to determine the belowground community structure and maximum rooting depth of the 100,000-km 2 Edwards Plateau of central Texas. The Edwards Plateau and other karst regions in Texas cover one-fifth of the state, with Ͼ3,000 caves identified to date (23, 24). Karst systems, in general, cover 7-10% of land surface area globally and supply a quarter of the earth's population ...