Nutrient transfer from soils to surface waters is associated with large, hydrologically induced fluctuations. Consequently, stream-based estimation of long-term changes in nutrient leaching is masked by variations of stream discharge. Using high-resolution discharge and soluble reactive phosphorus (SRP) data from three small catchments (up to 42 km2), SRP loads are calculated by applying two different rating curves. Monte Carlo simulations are carried out to determine monitoring strategies for optimizing the number of water samples, their distribution between periods of low and high discharge, and the duration of composite sample collection. Trends in SRP load are isolated from natural variations by applying the discharge time series of 1 year each to annually changing rating curves. By applying this approach to various monitoring data sets, collected over the past 15 years, downward trends in SRP leaching of up to -3% yr-1 are detected. We describe how to determine the number of annual samples required to detect trends in nutrient load, depending on monitoring duration, available resources, and the magnitude of the expected trend.