Assessing potassium environmental losses from a dairy farming watershed with the modified SWAT model

Potassium (K) was intensively used to optimize agricultural crop yield. Potassium losses should be accurately quantified for efficient nutrient management. However, no hydrologic model had been developed yet to quantify daily K losses at watershed scale. The Soil and Water Assessment Tool (SWAT) model was modified (named SWAT-K) by including the main K dynamic processes (solid–liquid distribution in soil and stream, plant uptake, and transportation with water flow and soil erosion) to simulate stream K load and K budget at the watershed scale. The SWAT-K was tested on the dairy farming Shibetsu River Watershed (672km2), Eastern Hokkaido, Japan. The solid–liquid distribution coefficient for K in suspended sediment was 10mlg−1. Langmuir equation was fitted to describe the solid–liquid distribution of K in soil, which derived an affinity constant of 0.046LmgK−1 and was used directly in SWAT-K. The fitted Langmuir equation also derived an adsorption maximum for K in soil. The adsorption maximum for K in soil normalized for clay content ranged from 4 to 20gKkg−1, and the fitted value of 15.5g Kkg−1 was used in SWAT-K. The SWAT-K satisfactorily predicted the daily dissolved K load at watershed outlet, and estimated an annual dissolved K load of 27.3kgKha−1year−1. The simulated pasture K yield of 36.1 (±2.5) kgKha−1year−1 was close to the observed data of 38.0 (±3.1) kgKha−1year−1. Then the model was used to quantify K budget at watershed scale. The simulated dissolved K leaching was 15.1kgKha−1year−1, and simulated soil K surplus of 75.1kgKha−1year−1 was much higher than plant uptake of 28.4kgKha−1year−1. The large amount of leaching and soil storage indicated that agricultural K input might be excessive and reducing the K application was recommended.