IN-STREAM NONPOINT-SOURCE NUTRIENT PREDICTION WITH LAND-USE PROXIMITYAND SEASONALITY

This research developed multiple regression models relating land use t o in-stream concentrations of total nitrogen (TOTN) and total phosphor us (TOTP) in eight, low-order watersheds on the coastal plain of South Carolina, The study area (4860 km(2)) included dominant land-use cate gories of agriculture, forest, urban, and wetland comprising the lower portion of the Lake Marion drainage. Land-use data were obtained from a pre-existing GIS database derived by classification of satellite im ages. The models partitioned land-use categories according to distance from stream channels using a series of buffer zones around each strea m, Effects of point source contributions were removed from observed in -stream concentrations so that nonpoint source effects could be more c learly delineated. All models except two were significant at P < 0.05, The models for TOTN (r(2) from 0.25-0.63) explained more variability of stream nutrient concentrations than those for TOTE (r(2) from 0.16- 0.39), Greater predictive strength for TOTN than TOTP likely reflects differing pathways from terrestrial to aquatic systems, Land close to the stream channel (<150 m) was a better predictor of nutrient concent rations than land away from the channel (>150 m), Land-use change scen arios (converting forest and wetland to agriculture) support the concl usion that management of stream water quality will be most effective w ith emphasis on riparian and adjacent lands. Seasonal models were gene rally significant (P < 0.05) and demonstrate that the seasonal profile of stream nutrient concentrations is dependent on the mosaic of land uses in a specific subbasin.