Spatial and temporal replicability of meteoric and in situ Be-10 concentrations in fluvial sediment

Cosmogenic isotopes, short‐lived radionuclides, elemental concentrations and thermochronometric indicators are measured in river sand to quantify erosion rates and trace sediment sources, and/or infer erosional processes. Interpretations of detrital sediment analyses are often based on the rarely tested assumption of time‐invariant tracer concentration. A better understanding of when and where this assumption breaks down and what sampling strategies minimize temporal and small‐scale spatial variance will improve science done using detrital river sediment. Here, we present new and previously published spatial and temporal replicates measured for in situ and meteoric 10Be (10Bei and 10Bem, respectively). Our new data include 113 replicate pairs, taken from agricultural and/or tectonically active watersheds in China months to millennia apart and spatial replicates taken up to 2 km apart on the same day. The mean percentage difference is 10% (−122% to 150%) for both systems considered together; the mode is close to 0% for both systems; and 36% of pairs of samples replicate within our analytical accuracy at 2σ. We find that 10Bei replicates better than 10Bem (p \u3c 0.01). 10Bei replicability is worse in steeper basins, suggesting that stochastic processes (i.e. landslides) affect reproducibility. 10Bem replicability is worse in larger basins, suggesting non‐conservative behavior of 10Bem as sediment moves downstream. Our results are consistent with the few previously published replicate studies. Considering all replicate data in a wide range of landscapes, in areas with deep erosional processes, replicability is poor; in other areas, replicability is good. This suggests that, in steep, tectonically active, and/or agricultural landscapes, individual detrital sediment measurements do not represent upstream rates as well as larger populations of samples. To ensure that measurements are representative of the upstream watershed, our data suggest that samples be amalgamated either over time or from several places close by in the same channel