Root length and nitrate under Sesbania sesban:: Vertical and horizontal distribution and variability

Measurements of root and nutrient distributions in agroforestry systems have often been obtained from unreplicated plots. We, therefore, measured the vertical and horizontal distribution of root length and soil nitrate in replicated plots of sesbania [Sesbania sesban (L.) Men.] to assess plot-to-plot variation and identify appropriate sampling schemes. Sampling was on profile walls exposed from pits at two on-farm sites (Ochinga and Muange) in Kenya. At Ochinga, soil was sampled at 27 0.15-m depth intervals and eight 0.15-m-wide locations perpendicular to 2.25-m-wide rows of 19-month-old sesbania. At Muange, sampling was 18 months after establishment of sesbania from seedlings with a 1 m by 1 m spacing. Soil was sampled at 17 0.15-m depth intervals and nine 0.15-m-wide sampling locations - one exactly below the tree and four each on opposite sides of the tree. Root length density generally decreased with distance from sesbania and with soil depth. Soil nitrate, however, followed different trends in the four replications at Ochinga and the three replications at Muange. At Ochinga, nitrate above 2.5-m depth clearly increased with distance from the tree in two replications but followed no distinct trend in another replication. At Muange, nitrate above 1.2-m depth decreased with distance from the tree in one replication, increased with distance from the tree in another replication and followed no clear trend in another replication. The suitability of various sampling schemes for measuring root length and nitrate was accessed from percent bias between measured values at sampling locations in the scheme and actual values for an entire transect of sampling locations. Bias associated with selection of sampling locations tended to be higher within the 2.25-m-wide sesbania rows than the 1 m by 1 m sesbania spacing. Soil sampling in systems with rows of young trees should be at several locations along a transect perpendicular to tree rows in order to minimize bias.