a b s t r a c tTo investigate above-and belowground biomass allocation patterns and the relationship of root distribution between different functional groups (shrubs vs. herbs) along a groundwater depth gradient, an experiment was carried out in a Haloxylon persicum woodland. Four sites were selected along the groundwater depth gradient. At each site, five sample plots (20 Â 20 m) and 15 subplots (1 Â 1 m) were established for a vegetation inventory of Haloxylon persicum and the herbaceous layer. The vegetation inventory included plant density, canopy coverage, height, and basal stem diameter of H. persicum. Allometric equations were established for H. persicum based on the biomass of 22 excavated trees, which were then applied to inventory data to provide stand estimates of aboveground biomass (AGB), belowground biomass (BGB), and total biomass (TB) of H. persicum. The AGB of the herbaceous layer was measured in subplots (1 Â 1 m) and BGB was excavated and measured to a depth of 100 cm at 10-cm intervals. Our results indicated that the BGB-to-AGB ratio at the community level in the herbaceous layer increased notably along the gradient of declining groundwater depth. The depth of the soil horizon that contained 65% of the root surface area and 85% of the root biomass was shallower (0-2.5 m) in the deep end of the groundwater depth gradient than in the shallow end (0-3 m) of the gradient. H. persicum adjusted root distribution to capture water in the upper soil layers in response to the decline in groundwater depth. However, this adjustment was not sufficient to compensate for the decline in the water table: the BGB-to-AGB ratio showed no significant increase as AGB declined. On the contrary, herbaceous plants seemed better adapted to changes in the water table by effectively adjusting their BGB-to-AGB ratio. Use of BGB-to-AGB ratios that are specific to groundwater tables will significantly improve our estimates of BGB carbon stocks in these woodlands. Considering the scarcity of information on biomass stocks and partitioning in desert woodlands, our results might be applicable to other desert regions in Central Asia dominated by H. persicum.