Soil information is one of the crucial inputs needed to assess the impacts of existing and alternative agricultural management practices on water quality. Therefore, it is important to understand the effects of spatial scale at which soil database is developed. In the United States, STATSGO and SSURGO are most commonly available soil databases. This study attempts to quantify the effect of scale by employing STATSGO (1:250,000) and SSURGO (1:24,000) soil databases by predicting and comparing flow, sediment, nitrate and phosphorus losses for High Island Creek, a minor agricultural watershed located in south-central Minnesota. For this purpose, a water quality model was calibrated for flow, sediment, nitrate and phosphorus losses for two years (2001-2002) using STATSGO and SSURGO soil databases. Further, the calibrated model was used to evaluate alternative tillage and fertilizer management practices such as adoption of rate of conservation tillage, rate, timing and method of N- and P-fertilizer applications. Statistical comparison of calibration results with observed data indicated excellent agreement for both (STATSGO with r 2 of 0.95, 0.97, 0.77 and 0.92 and SSURGO with r 2 of 0.90, 0.97, 0.82 and 0.99 for flow, sediment, nitrate and phosphorus losses, respectively) soil databases. However, evaluation of alternative management practices indicated that STATSGO based predicted annual nitrate losses are consistently higher than that for SSURGO data and vice-versa for predicted phosphorus losses. This brings up an important issue in developing TMDLs for impaired watersheds where conflicting interests of stakeholders may opt for soil database that support their interests.