Estimation of pollutant fluxes through river systems calls for accurate and precise load estimation. However, considerable uncertainty is associated with these estimates due to diffuse loading, which sets high requirements not only on sampling frequencies but also on calculation methods. The aim was to examine the variation in load calculations and the reliability of the load estimates of total phosphorus, total nitrogen and suspended solids in 24 Baltic rivers varying in size and land-use characteristics. Reliability of the load estimates was tested by simulation experiments in the river Paimionjoki using a Monte Carlo procedure.

The estimates calculated by the most reliable method were compared to the loads estimated by five other methods. The general reliability (RMSE) for P and SS was best by the correlation method and for N by the periodic method. Load calculations varied greatly depending both on the characteristics of the rivers and the calculation method. The flow-stratified method overestimated the P and SS loads by about 20% in large low-lake rivers. In small low-lake rivers, the overestimation was 10% and over 14% for P and SS, respectively. By contrast, the averaging method underestimated P and SS loads by 10% and 21% in small agricultural low-lake rivers. All the methods produced rather similar results for N in each of the river types.