Reducing measurement error in riverbank litter sampling

Rivers play a key role in the global distribution of anthropogenic litter. Accurate and reliable monitoring data are essential to design effective litter reduction and mitigation strategies. One common approach used to monitor macro- and mesolitter (>0.5 cm) in rivers is through visual riverbank litter sampling, in which observers manually collect, count and categorize items deposited on riverbanks. While monitoring efforts are scaling up to meet growing demand for data, it is key to quantify and understand uncertainties in these data, as these insights can be used to design improved monitoring strategies. Such quantitative analysis has not yet been undertaken for visual riverbank litter sampling methods to date.We conducted a series of experiments to quantify the measurement error of visual riverbank litter sampling. Our findings demonstrate that inter-observer variability can be substantial with a mean coefficient of variation of 22.4%. Statistical analysis indicates no significant effect of the assessed litter concentration, total item count, or sampling area size. In contrast, we did find that both size and colour significantly affect the item detectability by observers. Smaller items, especially those that are transparent or black, showed substantially lower recovery rates (below 50% for items