Reduced sampling frequency is known to increase the error associated with estimates of stream solute load. However, the extent to which the magnitude of error differs among commonly measured solutes and across seasons is unclear. In this study, a high sampling frequency data set from two forested streams (one upland-draining and one wetland-draining stream) in south-central Ontario was systematically sub-sampled to simulate weekly, fortnightly and monthly fixed frequency sampling regimes for 12 stream solutes. We found that solutes which had a higher degree of temporal variation in concentration (i.e. higher %RSD) had poorer precision (Cv) in estimates of annual load relative to solutes with a lower %RSD. In addition, the magnitude and direction of bias varied considerably among solutes and were related to differences in spring concentration-discharge relationships (m[spring QvsC]) among the 12 solutes. Solutes which decreased in concentration with increases in spring flow (i.e. m[spring QvsC] <0) exhibited positive bias in annual load while solutes which increased in concentration with increases in spring flow (i.e. m[spring QvsC] >0) were negatively biased. In terms of differences between seasonal and annual load errors, precision was generally lower for estimates of seasonal load relative to annual load while bias varied in both magnitude and direction among seasons. When the root mean square error (RMSE) of load estimates was compared to a threshold of acceptable error (<15%), the proportion of solutes attaining acceptable levels of uncertainty ranged from 11/12 for annual load estimates at a weekly sampling frequency to only 4/12 at a monthly frequency when both annual and seasonal loads were considered. Our results demonstrate that commonly measured solutes do not behave uniformly in response to changes in sampling frequency and that estimates of seasonal loads are often less accurate than estimates of annual load. These findings provide important insights into the design of stream monitoring programs and the evaluation of existing long-term data sets. Copyright © 2015 John Wiley & Sons, Ltd.