Quantifying phosphorus loads and streambank erosion in the Ozark Highland Ecoregion using the SWAT model

Abstract

Phosphorus (P) and sediment are major pollutants of waterbodies in the United States. Recent research has shown that future management practices must consider legacy P, especially from streambanks. Identifying and quantifying legacy P sources are necessary to select the most cost-efficient conservation practices. The overall objectives of this research were to (1) develop and apply a mass balance and uncertainty analysis to determine the quantity and location of legacy P stored in the Eucha-Spavinaw and Illinois River watersheds, (2) apply the SWAT model to determine the current sources of P reaching Lakes Eucha and Tenkiller and the management practices required to meet Oklahoma numeric water-quality standards, (3) test and improve a proposed streambank- erosion routine for the SWAT model and (4) apply the improved streambank-erosion routine to determine the significance of streambank-derived P for the Barren Fork Creek watershed. The results of the P mass balance and uncertainty analysis found that an average of 7.0 kg ha-1 yr-1 P were added to the watersheds from 1925 to 2015 and approximately 80% was retained, predominantly in the soils and stream systems. SWAT proved to be a capable tool in the water-quality standard evaluation process and illustrated how a watershed-scale model, such as SWAT, can be used to provide critical information when developing watershed-based plans. At current conditions, the water- quality standard is being met for Lake Eucha and the Illinois River subwatershed above the point source discharge at Tahlequah, Oklahoma, but is exceeded below the point source and for the Flint Creek and Barren Fork Creek subwatersheds. To meet the water-quality standard, the proper management of cattle production and elevated soil test P are essential. Modifications to the streambank-erosion routine for the SWAT model improved model predictions. Although the process-based applied shear stress equation was the most influential modification, incorporating the top width, streambank depth and area-adjustment factor more accurately represented the measured irregular cross-sections. The improved routine predicted streambank erosion on the Barren Fork Creek with a relative error of 34% before calibration. During the study period, 47% of the total P entering the Barren Fork Creek derived from streambanks.