Quantitative assessment of soil erosion and deposition rates using fallout radionuclides (FRNs), including Beryllium-7 (7Be), requires establishment of a reliable reference inventory i.e. the inventory of a non-eroding point. Little information, however, is currently available on the microscale spatial variability of 7Be inventory within reference sites. This is important information to inform sample design and replication, and in addition, to evaluate the uncertainty of derived soil redistribution data. In this study, soil samples were taken systematically at grid points on a 5 m x 12 m experimental reference plot with a bare soil surface, at two sampling occasions (2019 and 2021) in southwest China. 7Be activities were measured to explore the potential variability of 7Be inventory at the microscale. To determine possible causes of 7Be inventory variation, physicochemical charac-teristics including organic matter content (OM), pH, cation exchange capacity (CEC) and grain size compositions were analyzed at each sample location. 7Be inventories for the two periods were estimated at 211.1 +/- 20.0 and 456.1 +/- 43.8 Bq m- 2 (mean +/- 2 SEM, n = 44), with coefficients of variation of 31.4 and 31.9% for the 2019 and 2021 sampling cases, respectively. No significant correlations were observed between 7Be activity and the measured soil compositional properties, suggesting observed spatial variability is primarily a result of random variation due to rainsplash and other processes, although sampling and measuring processes may contribute some uncertainties. Using the traditional method, ca. 40 independent reference samples are required to estimate the mean 7Be inventory, i.e. to represent input across the site, with an allowable error of 10% at 95% confidence, while application of a bootstrap approach suggests that ca. 28 would be adequate under similar accuracy. Overall, results of this study emphasize that the simple assumption of uniform distribution of 7Be across the reference area needs detailed examination on a case-by-case basis, if this radionuclide is to be used effectively to assess patterns and rates of soil redistribution from field to hillslope scale.