Suspended sediment is a major water quality problem, yet few monitoring studies have been of sufficient scale and duration to assess the effectiveness of land-use change or conservation practice implementation at a watershed scale. Daily discharge and suspended sediment export from two 5,000-ha watersheds in central Iowa were monitored over a 10-year period (water years 1996-2005). In Walnut Creek watershed, a large portion of land was converted from row crop to native prairie, whereas in Squaw Creek land use remained predominantly row crop agriculture. Suspended sediment loads were similar in both watersheds, exhibiting flashy behavior typical of incised channels. Modeling suggested that expected total soil erosion in Walnut Creek should have been reduced 46% relative to Squaw Creek due to changes in land use, yet measured suspended sediment loads showed no significant differences. Stream mapping indicated that Walnut Creek had three times more eroding streambank lengths than did Squaw Creek suggesting that streambank erosion dominated sediment sources in Walnut Creek and sheet and rill sources dominated sediment sources in Squaw Creek. Our results demonstrate that an accounting of all sources of sediment erosion and delivery is needed to characterize sediment reductions in watershed projects combined with long-term, intensive monitoring and modeling to account for possible lag times in the manifestation of the benefits of conservation practices on water quality.