Abstract: Automatic surface stations in China feature high spatial density, and their hourly relative humidity data serve as important basic data for refined climate monitoring, urban meteorological research and numerical simulation. However, insufficient instrument stability, and limited operation and maintenance conditions, unattended stations exhibit obvious data quality problems, including spurious zero values, internal logical inconsistencies, constant stagnant readings (stagnant values), abrupt hourly jumps, and slow sensor response due to complex observation environments, insufficient instrument stability and limited operation & maintenance support. Conventional operational quality control (QC) schemes adopt fixed global thresholds, which fail to distinguish meteorologically reasonable low humidity from instrumental faults and tend to miss concealed errors. This study integrates hourly relative humidity records of 49,910 automatic stations across China, including 2,413 manned national stations and 47,497 automatic stations during 2008–2024, and systematically analyzes the frequency, temporal and spatial distribution characteristics of various anomalies, and develops a multi-layer QC framework consisting of climate-constrained zero-value discrimination for humid regions, multi-factor collaborative stagnant-value screening, hourly abrupt change inspection and spatial background field verification to identify both obvious and concealed observational anomalies. Based on this QC system, an hourly relative humidity dataset for China’s automatic stations is compiled and comprehensively evaluated from three dimensions: data completeness, data quality and climatic rationality. Results show that the overall data availability of all 47,497 unattended stations reaches 81.09%, with a valid data accuracy rate of 93.38%. The frequency distribution and mean values of valid observations after QC are consistent with those of national manned stations. Compared with CRA and ERA5 reanalysis datasets, the average biases are reduced by 0.67% and 0.68%, respectively. The dataset can objectively reproduce the spatial disparity and seasonal evolution of relative humidity across China. Taking dense stations in Haidian District of Beijing as an example, it can finely depict small-scale humidity gradients among built-up areas, green spaces and suburban zones.