Soil erosion is a serious global problem requiring effective modeling for accurate assessment of sensitive areas and related erosion rates. The outcome of soil erosion models depends strongly on the estimation of sediment transport capacity. In most of the existing spatially distributed soil erosion models sediment transport capacity of overland flow is often estimated using stream flow transport capacity functions. The applicability of stream flow functions to overland flow conditions is questionable because hydraulic conditions like flow depth, slope steepness and surface roughness under overland flow are substantially different from stream flow conditions. Hence, the main objectives of this study were i) to check the suitability of five existing well known and widely used transport capacity functions (Yalin 1963; Low, 1989; Govers, 1990; modified Engelund and Hansen (Smith et al., 1995); and Abrahams et al., 2001) for use under overland flow conditions, and ii) to derive a new function based on unit stream power by dimensional analysis to quantify transport capacity for overland flow. To accomplish the objectives, experiments in a 3.0 m long and 0.5 m wide flume were carried out using four different sands (0.230, 0.536, 0.719, and 1.022 mm). The unit discharges used for experimentation ranged from 0.07 to 2.07 x 10(-3) m(2) s(-1) and slopes ranged from 5.2 to 17.6%. In this study, none of the predictions with the existing functions was in good agreement with measured results over the whole range of experimental conditions, especially at low flow intensities. The percentages of observations in which the discrepancy ratio ranged between 0.5 and 2.0 were: 65% (Yalin 1963), 74% (Low, 1989), 57% (Govers, 1990), 54% (modified Engelund and Hansen (Smith et al., 1995)), and 25% (Abrahams et al., 2001). The results showed that the selected functions reasonably estimate transport capacities only under those ranges of conditions for which they were formulated. Although the excess.shear stress concept based function (i.e. Low's function) produced excellent results, the degree of accuracy of the results varied substantially with grain size (P.O.(0.5-2.0): 53-100%). In contrast, the performance of the Govers' function, which is based on the unit stream power concept, was quite similar for all the selected sands (P.O.(0.5-2.0): 50-63%). Based on the unit stream power concept, a new function for low flow intensities was derived by dimensional analysis using the data gained from the flume experiments. (c) 2012 Elsevier B.V. All rights reserved.